Following the tree of a single Subhalo[i] at Snapnum[j], I'm trying to deduce the total stellar mass (not affected by feedback) between episodes of star formation; I can't understand why very often there is star formation, but there's no stellar+wind mass produced over many redshift, then suddenly the stellar+wind mass appears, without any star formation (it then decreases for feedback episodes, I think). I can't get the mechanism behind this behavior. Can somebody help me?Thanks.
I post an example of what I've just said. This Subhalo doesn't experiment mergers.
Dylan Nelson
4 Jul '18
Hi Flaminia,
Note that this is quite a small halo, i.e. it only has a few tens of star particles in it (once M* becomes nonzero). I think if you look at something bigger, you will see more clearly the behaviors you are expecting. Then you can try to push back into this 'poor statistics' regime.
In particular, keep in mind that the process of star formation is stochastic. Just because a population of gas cells as a nonzero SFR, does not mean that a star particle will be formed, this is only true in the time average sense.
Also, you say there are no 'mergers', but I suppose you mean no other branches in the tree, i.e. no 'resolved mergers'. If there are very small objects (less than 20 particles) coming in, these would be mergers, but not included in the tree. Given that you are looking at such small numbers of stars, this could be the case. Since stars keep their unique IDs going backwards in time, you could track this handful of stars and figure out where (in space) they came from - if they formed inside the main progenitor branch of the galaxy you are considering, or not.
ok, thank you for the explanation, very useful. I have a question about what you wrote; you said:
"In particular, keep in mind that the process of star formation is stochastic. Just because a population of gas cells as a nonzero SFR, does not mean that a star particle will be formed, this is only true in the time average sense."
So (e.g. in the example I posted above, even if it's poor) I might think that the mass shown at z=2.58 is the result of the past SF (and mergers and feedback processes) at z=4.43, z=3.49 and z=2.73. Keep in mind that this subhalo has been selected as "quiescent" (sSFR<1e-11 yr-1), so I interpret episodes of SF as brief bursts. Does it make sense to calculate the effective final stellar mass at z=2.58 as the sum of the integral between zi (first episode of star formation) and z(i-1) of SFR*dt ? I hope I've made my point.
Thank you.
Dylan Nelson
5 Jul '18
Hi Flaminia,
Something you would want to try is to compare the "SFH" (star formation history) derived from the star particles themselves, to the "SFH" obtained by getting e.g. SubhaloSFR at each snapshot backwards following the tree.
The former will always be above the latter, under the assumption that no stars ever leave the galaxy (e.g. get thrown out somehow). The difference should be, I believe, largely the merger contribution.
The former also gives an 'exact' SFH without any ambiguity of tracking the galaxy through time, or this complication of the stochastic gas SFR. It can be computed using the GFM_StellarFormationTime field. I.e., just taking a histogram of this for all the stars in a galaxy gives a "stellar mass formed per time bin", which can normalized by the time bin size to get e.g. "SFR (msun/yr) vs. redshift".
Flaminia Fortuni
10 Jul '18
Ok, thanks for helping me, I'd like to do the last thing you suggest, so is there a simple way to link a subhalo found in SubLink file and star particles in Snapshot file (ID or something similar)?
Dylan Nelson
11 Jul '18
The SubfindID field in the tree, together with the SnapNum, gives the information needed to load all the particles belonging to that subhalo, using e.g. the loadSubhalo() function of the example scripts.
Flaminia Fortuni
12 Jul '18
No, I was not clear. I mean: is there a way to relate a Subhalo[i] in a Snapshot[j] of a SubLink tree (Appendix B, table B.6 of Data Release) with the particles given in the Snapshot file (Appendix A , table A.7 of the Data Release)? E.g. to make the histogram of GFM_StellarFormationTime for a single Subhalo indentified in a tree, as you suggested.
Dylan Nelson
13 Jul '18
Hi Flaminia,
Yes that's right, the loadSubhalo() function loads the particles (of a given type, e.g. 4 for stars) belonging to one particular subhalo.
Hello everybody
Following the tree of a single Subhalo[i] at Snapnum[j], I'm trying to deduce the total stellar mass (not affected by feedback) between episodes of star formation; I can't understand why very often there is star formation, but there's no stellar+wind mass produced over many redshift, then suddenly the stellar+wind mass appears, without any star formation (it then decreases for feedback episodes, I think). I can't get the mechanism behind this behavior. Can somebody help me?Thanks.
I post an example of what I've just said. This Subhalo doesn't experiment mergers.
Hi Flaminia,
Note that this is quite a small halo, i.e. it only has a few tens of star particles in it (once M* becomes nonzero). I think if you look at something bigger, you will see more clearly the behaviors you are expecting. Then you can try to push back into this 'poor statistics' regime.
In particular, keep in mind that the process of star formation is stochastic. Just because a population of gas cells as a nonzero SFR, does not mean that a star particle will be formed, this is only true in the time average sense.
Also, you say there are no 'mergers', but I suppose you mean no other branches in the tree, i.e. no 'resolved mergers'. If there are very small objects (less than 20 particles) coming in, these would be mergers, but not included in the tree. Given that you are looking at such small numbers of stars, this could be the case. Since stars keep their unique IDs going backwards in time, you could track this handful of stars and figure out where (in space) they came from - if they formed inside the main progenitor branch of the galaxy you are considering, or not.
ok, thank you for the explanation, very useful. I have a question about what you wrote; you said:
"In particular, keep in mind that the process of star formation is stochastic. Just because a population of gas cells as a nonzero SFR, does not mean that a star particle will be formed, this is only true in the time average sense."
So (e.g. in the example I posted above, even if it's poor) I might think that the mass shown at z=2.58 is the result of the past SF (and mergers and feedback processes) at z=4.43, z=3.49 and z=2.73. Keep in mind that this subhalo has been selected as "quiescent" (sSFR<1e-11 yr-1), so I interpret episodes of SF as brief bursts. Does it make sense to calculate the effective final stellar mass at z=2.58 as the sum of the integral between zi (first episode of star formation) and z(i-1) of SFR*dt ? I hope I've made my point.
Thank you.
Hi Flaminia,
Something you would want to try is to compare the "SFH" (star formation history) derived from the star particles themselves, to the "SFH" obtained by getting e.g.
SubhaloSFRat each snapshot backwards following the tree.The former will always be above the latter, under the assumption that no stars ever leave the galaxy (e.g. get thrown out somehow). The difference should be, I believe, largely the merger contribution.
The former also gives an 'exact' SFH without any ambiguity of tracking the galaxy through time, or this complication of the stochastic gas SFR. It can be computed using the
GFM_StellarFormationTimefield. I.e., just taking a histogram of this for all the stars in a galaxy gives a "stellar mass formed per time bin", which can normalized by the time bin size to get e.g. "SFR (msun/yr) vs. redshift".Ok, thanks for helping me, I'd like to do the last thing you suggest, so is there a simple way to link a subhalo found in SubLink file and star particles in Snapshot file (ID or something similar)?
The
SubfindIDfield in the tree, together with theSnapNum, gives the information needed to load all the particles belonging to that subhalo, using e.g. theloadSubhalo()function of the example scripts.No, I was not clear. I mean: is there a way to relate a Subhalo[i] in a Snapshot[j] of a SubLink tree (Appendix B, table B.6 of Data Release) with the particles given in the Snapshot file (Appendix A , table A.7 of the Data Release)? E.g. to make the histogram of GFM_StellarFormationTime for a single Subhalo indentified in a tree, as you suggested.
Hi Flaminia,
Yes that's right, the
loadSubhalo()function loads the particles (of a given type, e.g. 4 for stars) belonging to one particular subhalo.