This repository has been archived by the owner on Jan 27, 2023. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 0
/
slidenotes.html
128 lines (128 loc) · 4.89 KB
/
slidenotes.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
<h1>OU Symposium Slide Notes</h1>
<h2>General Format</h2>
<ul>
<li>5 minute overview/background of our science</li>
<li>5 minutes on our specific research</li>
<li>5 minutes on our general computation need</li>
<li>5 minutes on our specific analysis in relation to computation and what is our specific needs for computers</li>
<li>5 minutes for results and greater impacts</li>
<li>5 minutes for questions</li>
</ul>
<h2>Table of Contents</h2>
<ul>
<li>title slide: First Look at Protostars: High Fidelity Modeling and the Youngest of Stars with OSCER</li>
<li>collaboration: VANDAM, HOPS, misc</li>
<li>motivation<ul>
<li>questions we try to answer are generally: where did we come from, what else is out there, and where are we going to end up?</li>
<li>questions we (specifically) answer: Where did we come from, how did we get here, what else is out there, and are we unique</li>
</ul>
</li>
</ul>
<p>(not everyone is an astronomer much less expert of star/planet formation)</p>
<ul>
<li>background on star formation<ul>
<li>evolution of stars: MC -> Protostar -> Main Sequence (star) -> Giants -> WD/NS/BH -> MC ...<ul>
<li>What determines the specific track: to first order mass</li>
<li>whats interesting is once you hit MS, the only thing that matters (to first order) is mass</li>
<li>what determines MS and single/multiple star. </li>
<li>what about planets? Gas is gone so gas giant formation not possible at MS</li>
</ul>
</li>
<li>Look at Cloud -> Class III<ul>
<li>evolution</li>
<li>when does gas leave?</li>
</ul>
</li>
<li>pictures of orion/perseus HST</li>
<li>these are optically thick....big problem<ul>
<li>interesting, opacity is related to frequency so just go to longer wavelengths: IR/RADIO</li>
</ul>
</li>
</ul>
</li>
<li>Instruments<ul>
<li>problem, resolution is also frequency dependent theta = lambda/size<ul>
<li>so now we need large telescopes. We want <10AU scales and stuff is away so we need 0.1'' resolution</li>
<li>similar to saying 1/1000 of your fingernail width at arms length away from your eye.</li>
<li>which means we need telescopes on order of the size of km at least...</li>
<li>not doable</li>
</ul>
</li>
<li>interferometry<ul>
<li>lots of smaller telescopes across several miles acts like 1 big telescope(with limitations)</li>
</ul>
</li>
<li>fixed our problems<ul>
<li>alma, vla, noema, carma</li>
</ul>
</li>
<li>so what kind of data do we get?<ul>
<li>photometry<ul>
<li>just like a camera from your cellphone we can get an image</li>
<li>if we do this a lot we get the energy distribution of objet</li>
</ul>
</li>
<li>spectroscopy<ul>
<li>this gives us color, not just 3 color but thousands. Gives us information of atoms, molecules</li>
</ul>
</li>
<li>we get both simultaneously</li>
</ul>
</li>
<li>so to summarize we went from HST -> ALMA<ul>
<li>now we can look at the systems themselves and not just the environment or nulk properties</li>
<li>from these we can get stellar mass, disk mass, size, speeds, molecular abundance, temperatures, geometries, multiplicities, etc</li>
</ul>
</li>
</ul>
</li>
<li>Computation<ul>
<li>parameters<ul>
<li>we now have 20-40 parameters to fit with a lot of degeneracy</li>
<li>turning into a computation problem</li>
<li>we want to make a model, fit this model to the data, and then get the model parameters</li>
<li>sounds like a great problem for parallelization and MCMC</li>
</ul>
</li>
<li>so now our problem is one of scalability<ul>
<li>we want to reach numerical convergence of the models and statistical convergence of the model parameter sets</li>
<li>insert EMCEE,openmpi, pdspy</li>
<li>insert slide about scalability</li>
</ul>
</li>
<li>computation requirements<ul>
<li>generally models take 60k core-hours for a single system<ul>
<li>we have a few hundred of these</li>
</ul>
</li>
<li>our systems<ul>
<li>2x 28 physical core</li>
<li>so while good not enough (months of convergence)</li>
</ul>
</li>
<li>into oscer<ul>
<li>generally we ask for 5-15 nodes x 20 core x 48 hours x 15-5 separate runs x systems</li>
<li>some take more and some take less depending on complexity, how quickly they converge, etc</li>
</ul>
</li>
<li>currently not GPU ready so we just want max CPU core hours</li>
</ul>
</li>
<li>we are requesting a lot but why again?</li>
</ul>
</li>
<li>impacts and fundamentals<ul>
<li>we are trying to understand our universse and namely understand how did we form and what other things form in the universe</li>
<li>are we unique?</li>
<li>where/when/how do planets form and same for stars.</li>
<li>how do we get the tichness of chemicals</li>
<li>we want to answer these in a statistically significant way too so that means large observational surveys, lots of models</li>
<li>our group at OU + collaborations around the world are the forerunners of this field</li>
</ul>
</li>
<li>thanks<ul>
<li>none of this would happen without OSCER, OU, NSF, and the collaborators</li>
<li>special thanks to Henry, Horst, Oscer IT, and CAS IT</li>
</ul>
</li>
</ul>