-
Notifications
You must be signed in to change notification settings - Fork 13
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Bad structure for 2LPT fields #63
Comments
Hi @cramirezpe ok, interesting. Reminds me of this problem: #10 Basically, the default way in which we bias sources in colore is by doing 1+delta_qso = (1+delta_M)^b_qso There are two ways to check if this is what's happening.
Let me know if this doesn't help and we'll delve deeper. |
Hi @damonge - I would also bet in this direction. I think we should try running a few boxes with different schemes to populate quasars, and use the one that is "better behaved" (although not clear how to define this metric). I'd try to not smooth the field too much, though, since it would also remove the power in the density skewers that we need for Lyman alpha. Let's start first with different biasing models. |
Yes, exactly. I think the alternative bias models curbing the exponential dependence on b will be the answer here. I guess the definition of "better behaved" depends on how you want to use these mocks. In all cases it probably means "whatever method can recover a realistic level of power down to the smallest possible scale"? |
Right, the metric should be something like that. I just didn't want to get to the point of doing a comparison to data, or doing a comparison to a particular theory (linear theory should not work that well either). We just don't want something that is clearly wrong, like these spikes in the density field. I'd just try a couple of settings, and hopefully one will be clearly better even by eye. I guess the linear bias model would need to be truncated to delta_q) >= -1, while the others do that by construction. The documentation does not mention the linear bias, but I saw in the code that this is BIAS_MODEL_3, and that indeed it already enforces positive densities of galaxies. @cramirezpe , are you using BIAS_MODEL_1? If so, let's try the other two and see what you get. |
Yeah, the bias models are shown in |
We have been working with 2LPT CoLoRe boxes while trying to improve current DESI Ly--alpha mocks (which currently use lognormal boxes).
We've analyzed the Ly-alpha auto-correlation, the Ly-alpha-qso cross-correlation and the qso auto-correlation so far. The Ly-alpha auto does depend on the density field and it looks similar to the results obtained with the lognormal boxes.
However, when computing the Ly-alpha-qso and the qso auto, we got weird results. Therefore we wanted to inspect the differences in the quasars obtained with and without the 2LPT option set in CoLoRe.
What we observe is a few spots in the sky with a very high density of objects. I'll give a couple of examples of this. Focusing into a single pixel, let's start with the 2LPT box. In the left panel it is shown a whole healpix pixel while in the right panel it is shown a small region with an "unusual high density". The middle panel shows the location of this region inside the healpixel.
![2lpt1](https://user-images.githubusercontent.com/58682644/103948325-e3b1d400-5139-11eb-8c54-9f3ee19e460c.png)
Then for the same configuration but just using a lognormal transformation:
![lognormal1](https://user-images.githubusercontent.com/58682644/103948338-e9a7b500-5139-11eb-9943-6892c45cada0.png)
The get a better intuition of the separation between objects in both cases, I've computed distances between objects for a subsample inside the small region:
![histogram1](https://user-images.githubusercontent.com/58682644/103948355-f1675980-5139-11eb-8f2f-bbdf96b7a6ae.png)
The second example shows more clearly that there is something wrong happening as the structure is box-shaped:
![2lpt2](https://user-images.githubusercontent.com/58682644/103948372-f88e6780-5139-11eb-8422-02a44e02a6cc.png)
![lognormal2](https://user-images.githubusercontent.com/58682644/103948376-faf0c180-5139-11eb-8996-e049d3fd5c42.png)
![histogram2](https://user-images.githubusercontent.com/58682644/103948381-fd531b80-5139-11eb-8764-ffe9c49de55e.png)
The
param.cfg
used for this box is the following:maybe there is something in this configuration not convenient for 2LPT that we don't understand.
And the relevant run parameters are:
![Screenshot 2021-01-07 022201](https://user-images.githubusercontent.com/58682644/103948432-0d6afb00-513a-11eb-99ce-f924abef8951.png)
I tried to reproduce this same problem in smaller boxes (
n_grid
< 4096) without success.Any idea of what could be happening here, or any analysis that we can perform to shed some light on this?
The text was updated successfully, but these errors were encountered: