ASTRO/PHYS 545 · Week 1
Donghui Jeong · Department of Astronomy & Astrophysics · Penn State
Start here
It is easy to mistake them for the subject. Consider what the precision is hiding.
| 1% | We know the dark matter density this well. We do not know what dark matter is. |
| better | We know the dark energy density better than that. We have no account of it whatsoever. |
| 20 yr | We knew the baryon density to a percent for twenty years while unable to say where most of the baryons were. |
| 1990 | "The CMB is a blackbody" is a bound measured in 1990, never improved, sitting three orders of magnitude above the distortion $\Lambda$CDM predicts. |
Every one of these is a statement about what the universe contains or how it began. Precision and ignorance sit together here.
The premise
So producing that material is no longer the scarce thing. What is scarce is carrying it the whole way: choosing what is worth asking, telling whether what comes back is true, and getting from there to something nobody knew before.
Carrying a parameter without its physical content is exactly the part of the work an AI already does better than you.
What we practice
Think → Ask AI → Inspect → Challenge →
Verify → Revise → Synthesize → Ask the next question.
The loop never changes. What changes is how well defined the material is, and how much of it is yours.
Consequence
Read the notes. Solve eight problems. Use AI freely.
Four names are drawn. Those four work the problems at the blackboard. Everyone else enhances what goes up there. I ask what each line costs.
I take both meetings this week. It is the only week I do.
75 minutes
Four names drawn at the start of class, one to each problem.
So you prepare all four.
Every class
Reproduce the argument on the board by hand. State where each assumption enters. Mark what you could not resolve.
Three kinds of enhancement earn a mark. Securing it: an assumption used but not stated. Continuing it: picking up a stalled derivation. Widening it: another route, a connection, a physical reading.
Four names, every class. Every draw is independent: presenting on Tuesday does not take you out of Thursday's. Everyone else is challenger and alternate at once, on duty every class.
Both are recorded: a circle for taking the board, a check for an enhancement. That record is the participation grade.
Say this out loud once
The room is working out together whether the argument holds. The presenter is the one holding the chalk.
When someone stalls: thirty seconds on their own, then the floor, then me. Picking up somebody else's stalled derivation is among the most valuable things anyone does here, and it earns a check.
Errors made and worked through at the board are not penalised. Reciting a memorised solution is.
Two things about how this room works
You will be stuck at that blackboard, in front of everyone and me, more than once. Leave the fixed picture of yourself at the door; it was never accurate.
Evaluating your performance is the last thing I do here. The TA marks the written work. So when I press a line, nothing is at stake for your grade: answer it the way you would answer a colleague.
And what I am actually here for: that you leave a productive scientific commander of the AI army. Choosing the objective stays yours.
The week
| Mon noon | Solutions due, all eight problems |
| Tue | Class · problems 1–4 |
| Wed | No class. Repair what Tuesday exposed |
| Thu | Class · problems 5–8 |
| Fri | Final version, carrying its repair history |
The final version shows what changed and why you had not caught it.
That history is the graded part. A solution that was wrong and correctly diagnosed scores above one that was right and unexamined. Official solutions go up after Friday, never before.
Written without AI
The point in your solutions you were least able to verify.
Where AI was wrong, incomplete, or confidently vague, and how you found out. If it was not, say so.
These two lines are the only written record this course keeps of how you work. I read every opening line before Tuesday.
Three times · weeks 2, 4, 6
A board segment is eight minutes, so the weekly problems never ask you to sustain a calculation across an afternoon. One deep problem does.
The reflection is what is assessed. The solution is not. The ten minutes before AI exist so the reflection has something to compare against.
Weeks 2–6
Five, not fourteen. Recombination, nucleosynthesis, dark matter and structure formation are not skipped; you meet them in the map weeks, as things people argue about rather than things you derive.
The arc
The course lecture notes: supposed to be correct, comparable line by line. I wrote them and nobody refereed them. Finding an error there is worth more than any solved problem.
Papers. Authoritative in a weaker sense, and they disagree with each other.
Your own map, and nothing else.
What is not withdrawn is the standard in this room. It is the same in week 15 as in week 1.
Weeks 7–9 and 12–13
photons · baryons · neutrinos · dark matter · dark energy · initial conditions
In order of how well we know them, so the sessions open as they go.
four things we still want to know about each
DESI, HETDEX · Roman, Euclid, LSST · LVK, LISA, PTA · SPHEREx
What will sharpen those claims, or break them.
the purpose · the chain · the systematics · the rest of the data
Four segments each, and the two blocks ask different questions: a constituent is something the universe contains, a probe is an instrument for finding out. The board fills into an actual map. One rule: any claim you attribute to a paper, you can point to in that paper.
The project
| Week 8, Friday | The corner you are working in, one paragraph |
| Week 9, Thursday | The question, revised against my reply. I approve it before I leave |
| Week 10 | Clinic with the TA. The room cuts your plan to size |
| Week 11 | Back at the board with what you built |
| Weeks 12–13 | Refine it while the class keeps mapping |
| Week 14, Monday | The report. Everyone, same night |
| Weeks 14–15 | Fifteen minutes, and ten of questions |
The prototype is small by design. Reproduce one published figure. Check one approximation numerically. Analyse a subset of public data. The question is never whether the code runs.
You hand in a report on the Monday of week 14. Nobody is expected to turn it into a paper (although it will be great!). If that happens later, good.
The question the project actually asks
You will have answered it four times by then, against the Friedmann equations, three distance measures, the thermal history, and slow roll. The difference in the project is that nobody knows the answer.
Grading, and one rule about showing up
| Weekly problems and maps | 25% |
| Participation: board, enhancement, attendance | 30% |
| Problem sets (3) | 15% |
| Final project | 30% |
If you are not here you are not in the draw and you are enhancing nothing, and neither shows up in the record. Both halves of the participation grade are earned in this room and nowhere else.
Attendance is not encouraged. It is required.
Attendance credit
Illness, research travel and comparable obligations are arranged absences. The distinction is notice, not justification. If someone is out and their segment is open, pick it up and take the circle; you prepared all four anyway.
AI policy, in full
Solutions, derivations, code and literature analysis may be AI-generated in whole or in part. This course is not interested in which route you took. It is interested in whether you can defend where you arrived. Say in your closing line which problems you did which way.
Two exceptions: the opening and closing lines, and the reflection in each problem set. Those are where your judgment rather than your output is visible.
Before Thursday
Estimate the temperature at which the universe became neutral.
Two numbers are available to you: the binding energy of hydrogen, 13.6 eV, and the ratio of photons to baryons, roughly $10^9$. Use both, and say what each one does and where it comes from.
Then say what you would need in order to do better than an estimate.
Submit it by noon on Wednesday. Open and close it with the two lines. On Thursday I will work it at the board under the rules you just heard, and you will get to take it apart.