« first day (1202 days earlier)      last day (4115 days later) » 

00:00
@TedShifrin Yes, yes. Sorry about that.
$\star \omega$ is a two form.
Flux is what goes across normal to a curve in a surface or normal to a surface in $\Bbb R^3$.
ROFL. I miss my cat :(
@TedShifrin I couldn't finish this one the other day, but I am guessing it is because I ended up with a big calculation I didn't want to do.
00:04
Consider $\omega =(y+\sin x)dx+\frac 3 2z^2+\cos y)dy+2x^3dz$
And the curve $(\cos t,\sin t,\sin 2t)$.
$t\in [0,2\pi]$.
The task is to find $$\int_C\omega$$
Now, the hint is to use the curve is on $z=2xy$.
So $\omega$ can be transformed using $dz=2(xdy+ydx)$ and so on, right?
You want to consider $d\omega$ on the surface.
can some one tell me the steps to find a cricles center x:y from 2 points and its radius... i calculated the straight line distance of point1 and point2 but don't know what i have to calculate next?
@TedShifrin OK. SO $$\omega = \left( {4{x^3} + 1} \right)ydx + \left( {4{x^4} + 6{x^2}{y^2} } \right)dy + \sin xdx + \cos ydy$$ on the surface.
@Dave: Draw circles of the given radius centered at your two points.
im using computer programming so i cant do it that way
00:09
@TedShifrin And $d\omega=(4x^3+1)dy\wedge dx+(16x^3+12xy^2)dx\wedge dy=(12x^3+12xy^2-1)dx\wedge dy$
Solve equations? @Dave?
hm ?
confused now xD
@Pedro: so what region in the $xy$-plane parametrizes the surface? Can you do the integral pain-free?
So, I was playing around with a question in game theory, and ended up proving a statement about automorphisms of compact metric spaces which I can't think of an elementary way of proving.
Let $M$ be a compact metric space. $M$ can not admit both an order-$2$ automorphism with a unique fixed point, and an order-$2$ automorphism with no fixed points.
assuming I haven't made any mistakes with my game theory proof anyway
@TedShifrin Hmm.
00:11
What can you do with "computer programming," @Dave?
@TedShifrin well i can show you what ive done so far: jsfiddle.net/3kA34/17
@TedShifrin A disk.
So $f(x,y)=2xy$ over the unit circle?
@Pedro: Yup. Now do the integral in your head.
click "launch" when the two objects are apart a bit and it should draw a curved line joining them but im calculating the red circle's center in the wrong place
@TedShifrin LOL, OK.
00:14
@Dave: I'm on an iPad here. I can't look.
Interesting @Daniel.
ah ok well basically ive got distance of two points that are on the circle's circumfrence and the circles radius and in order to draw the circle on the screen i need to know the circle's center x:y
i dont know what step i do once i calculate the distance between the two points
@TedShifrin I cannot do it in my head. =P
I need to change to polar coordinates.
You need the perpendicular bisector of that segment and then Pythagoras.
Also, I didn't pay attention to orientation, so the sign may be off.
NO @Pedro. Use symmetry!
I teach my students "Exploit symmetry!"
00:17
what should i put in google to refine my results to find a good explaination
@TedShifrin Ok, $2xy$ is symmetric about $x=y$ and $x=-y$...?
Think even/odd @Pedro.
@TedShifrin Yes, $f(x,y)=f(-x,-y)$ and $f(\mp x,\pm y)=-f(x,y)$.
@daniel: Too bad it's not a simplicial complex.
@Ted I guess automorphisms of simplicial complexes are simplicial maps?
finite-type simplicial complexes anyway
00:21
Not $f$, @Pedro. What your double integral is ...
@TedShifrin I'm not sure. I have $$\iint_{D}(12x^3+12xy^2-1)dxdy$$ where $D$ is the unit disk.
Right. The $-1$ you can handle. Think about the rest.
gah im so confused lol
So does Lefschetz get us anywhere, @Daniel?
@TedShifrin Oh, well, I have $$12 \iint_D x(x^2+y^2)dxdy$$
So go polar coordinates! =D
00:23
Symmetry @Pedro.
@TedShifrin Hmmm, homology groups of arbitrary metric spaces can be quite gnarly.
@TedShifrin OK, it is zero.
Right, @Daniel. But what about the nice caee?
Hurrah @Pedro, and don't ever forget symmetry! :)
whats the correct term for the center point of a circle? is it origin ?
No, @Dave, center.
00:26
@TedShifrin I have another one.
hm
@TedShifrin The line integral of $F=(2xy+z^2,x^2-2yz,2xz-y^2)$ across the line on the intersection of the sphere $x^2+y^2+z^2=1$ and the plane $x=y$.
I got that $dF=2(x-z)dx\wedge dz$.
Well, not $dF$ ....
@TedShifrin ?
$F$ is the vector field. You mean the corresponding $1$-form.
00:35
@TedShifrin Bah, what's the difference...?
Again, iso, not same.
damn im miles off =/
I have a book that basically says "has anyone ever been as far as to decide how not what to look like"
@TedShifrin Anyhow, I have no clever way of killing this one.
Even if you don't help somebody reply "lol" so I at least know I'm not the only person who had to read it 4 times
00:37
@Pedro: Are you sure about your exterior derivative?
While a horrible integral @TedShifrin and @PedroTamaroff can't you just paramaterise the resulting circle and integrate over that
Sure, with orthogonal change of coordinates. But the problem is trivial.
@TedShifrin Heh, it is zero.
I mean it's trivial, the circle is something of the form ( $\frac{\sqrt{2}}{2}cos....$
I got $x,z$ swapped somewhere.
00:40
No, I mean trivial in the technical sense.
@TedShifrin What does that mean?
Yeah, I usually use trivial to mean "I don't know how to prove it"
What you discovered: The integral of $0$.
LOL @Alec. I mean trivial in the sense of the trivial solution in linear algebra.
Oh!
Also @TedShifrin if I write "and the result follows" it means "I think I've done enough to earn the marks, lets not prove the rest shall we?"
@TedShifrin Ah, but the integral is not over a closed curve.
It is from $(1/\sqrt 2,1/\sqrt 2,0)$ to the north pole.
00:44
I'll remember that when I grade your exams, @Alec :D
"we see from earlier" = "I can imagine a vague connection to a previous question, but I'll assume you - the marker - have made a formal connection"
Huh?@Pedro We've got a great circle on the sphere.
"I wouldn't want to live in a world where it wern't true" = lost all hope
@PedroTamaroff seriously, unit sphere and plane tilted 45deg up (use your imagination) = circle
@TedShifrin I mean the integral is through that segment.
@AlecTeal Thanks, Sherlock!
What does that mean?
00:46
@TedShifrin The integral is through the piece of the circle joining those points.
From $(1/\sqrt 2,1/\sqrt 2,0)$ to $(0,0,1)$.
@PedroTamaroff I hate to be insulting but I follow this question
Why it so hard?
@AlecTeal It is not hard.
I'm just trying to tell Ted the whole problem.
Admitting there's a problem is the first step.....
What doesn't Ted get?
@TedShifrin At any rate, what I can do I just change the path as I see fit.
Ugh. Try again. Use the FTC: Your $\omega$ is exact.
00:51
@TedShifrin Yes, I know it is exact.
Then use the FTC.
Easiest version of Stokes's.
@TedShifrin I have to find a primitive for that.
@TedShifrin Oh, well...
I got ${y{x^2} - z{y^2} + x{z^2}}$.
@TedShifrin
By the usual $\int_0^1 \omega(t{\bf x})\cdot {\bf x}dt$ thingy.
@TedShifrin Why did you say "Easy."? What method did you have in mind?
That's fine, but it'e easier to solve by integrating $\partial f/\partial x$ and continuing. See section 8.3 of my book.
00:59
@TedShifrin Hmm, I think the above is simpler. Specially in this case, since all is homogeneous.
I disagree. I can do it in my head, and so can you.
@TedShifrin I am not a big fan of doing things in my head. I like writing things down.
I guess I can try...
Well, I am getting the integral is $\sqrt 2 /4$.
Write it down, indeed, but it's straightforward. Your way is ok, too, but I save it for $k$-forms for $k>1$.
Hm, let's see if I can find something non trivial and non easy to solve...
LOL at this one:
Let $S$ be the surface given the graph of $f(x,y)=(1+x^2+y^2)^{-1}$ over $\lVert (x,y)\rVert \leqslant 1$, and let $${\bf F}(x,y,z)=\left(\frac{zx}{x^2+y^2},\frac{zy}{x^2+y^2},0\right)$$ Find $\displaystyle \int_S {\bf F}d{\bf S}$. Think before acting.
(Verbatim)
So what's your approach?
01:12
@TedShifrin Let me think.
The derivative is zero.
Is it not?
@TedShifrin
Now you need the $2$-form, not the $1$-form. I'm not sure either has derivative $0$.
I'm off to bed now, night all and bump math.stackexchange.com/questions/571161/…
@TedShifrin Yes, I know I need the two-form.
The derivative w.r.t. $x$ of the first component is $z(y^2-x^2)/(x^2+y^2)^2$ and by symmetry the other is the opposite of this.
Night @Alec.
Ok, I'll trust you. Now what?
Find a sequence of closed, connected subsets $C_1, C_2, \dots$ of $\mathbb{R}^2$ such that $C_1$ contains $C_2$ which contains $C_3$ (and so on) and $\bigcap_{n = 1}^{\infty} C_n$ is not connected. Isn't this also proving a point isn't connected?
01:25
@TedShifrin I can take the integral of that along the disk of radius 1 and height 1/2 instead of the function I am given.
Awesome.
It would have to mean its a set that doesnt contain its limit point
@DonLarynx We're talking about connected sets.
And closed too, yes.
which is why its so hard
any hints?
@DonLarynx Sorry. Misread.
:12223606 Where did you get that exercise? What book?
01:28
My professor assigned it for hw and I am stuck. I would like: One hint.
You're saying $C_1\supset C_2\supset C_3\supset \cdots$
Well, you need to think unbounded closed sets.
If they are bounded, they are compact, and $\bigcap C$ consists of only one point, so it is connected.
Needn't be just one point.
@TedShifrin Sorry, confused an hypothesis.
So scratch that Don.
Nope. =)
@DonLarynx Ah?
01:33
I mean... $\mathbb{R}$ is unbounded and closed
I know I know
The set in question (the intersection from $1$ to $n$) cannot lie in the closure
Otherwise it is connected!
Because $\bar A \cap B \neq \emptyset$
ONLY if it is in the closure of $A$!
So the intersection $B$
must be outside of $\bar A$
@Pedro
In order for it to be separated
@TedShifrin, I looked at your site out of curiosity. You're teaching Riemannian Geometry this semester?
Yup @Mike. 5 classes left.
@DonLarynx hi don
How's it going?
Oops. That one's my phone. I guess I have two accounts.
01:42
LOL, fine for me. Actually ending with Grassmannians, curvature of vector bundles, Chern classes and Chern-Gauss-Bonnet.
Interesting problem, @Don. I don't have it yet.
@DonLarynx Consider two open circles centered at $(1,0)$ and $(-1,0)$ of radius 1. They are disconnected.
Can you approximate them by nested connected closed sets?
Connected sets
That's the picture I have, but I don't have it.
@TedShifrin Ah?
The problem you're discussing is the one about the intersection of closed, connected sets, right?
Of that I know Gauss-Bonet, hah. Sounds like you enjoyed teaching it!
01:45
Yup.
I can't use spoilers here.
Hm.
Gauss-Bonnet in high dimension
@TedShifrin Did the dimension get arrested?
Ok, I think I got it but I'd like to see if it can be done with compact, connected sets
Only when you ran the red light.
01:46
I assumed, @Ted
I'm taking a course in Riemann surface next quarter. Should be fun.
I suppose I could post my problem sets on the webpage ...
@PedroTamaroff: Are you on fb?
I've taught Riemann surfaces several times.
I don't want to post spoilers here
@FernandoMartin OK.
01:48
Decidedly different subjects, I assume, despite that one is just a special case
I'm taking it from an analyst rather than a geometer.
Hmm, I have no Argentinian FB friends.
@TedShifrin ha! I do!! take it
No, Riemann surfaces is either complex analysis or algebraic geometry (or both, when I've taught it). No Riemannian geometry in it.
Well, when I said a geometer I meant algebraic. But is a riemann surface not just a dimension one riemanian manifold?
$1$-dim complex manifold. The metric structure doesn't ordinarily enter.
Hodge theory can come in, in which case we want some Riemannian/Hermitian metric structure.
01:54
@Pedro: are you there?
For whatever reason I'd thought every 1d complex manifold had a Riemannian metric. Dunno why.
Well, every manifold has lots. That doesn't make it natural.
Good point, thanks.
What kind of topics are usually taught from an analyst's point of view?
I dunno what he'll do. Otto Forster has a beautiful book with a lot of analysis and also sheaf theory.
What's the title?
(I've read a good chunk of Gunning's first volume, but considering a main goal is to prove Riemann-Roch, I'd think it more an algebraic geometry text.)
02:03
@PedroTamaroff, I already proved every set lying in the closure of a connected set must be connected
Can I use that?
I also will look at ur example in a sec
@DonLarynx My example is wrong, sorry.
@Don: you're working with all closed sets, so I don't see the relevance. @Pedro is hinting tgat you want to keep pushing the intersections further and further out.
But @FernandoMartin has a nice solution. Think about $\Bbb R^2\setminus ((-1,1)\times\Bbb R)$. This is closed and disconneted.
If we consider those two open circles, we're essentially considering a line of length two with the point $\{1\}$ missing @Pedro
That's not a connected set
Can you obtain it from the plan by chopping off open rectangles?
02:06
I have an example like you were envisioning, @Pedro.
What about a closed circle and an open circle that are disjoint
I think that works
Which problem are you discussing, @Don?
What are you doing with them, @Don?
Ok, what's your chain of closed sets?
02:08
@TedShifrin Nothing, that example goes nowhere. I was thinking I could find a point that isn't in $\bar A$ or $B$ but thats a contradiction
nvm
Hint: Try to make the intersection two parallel lines (or rays).
@DonLarynx No, because that is not closed.
$\bigcap C$ is closed, since each $C_n$ is closed.
@Pedro: I still can't manage to do it with compact sets
I tried earlier, @Fernando. Can we prove it's impossible?
I'm not sure
02:19
I'm pretty sure it is, but I haven't thought long enough to give a proof
Maybe it follows from some sort of ÄŒech homology argument.
@FernandoMartin I think it's impossible.
Pray tell @Pedro.
No, I have a simpler proof if you want to hear it. No need for homologous, thank the lord
The no was @Ted, not @Pedro
02:26
Let's hear, guys.
You'd start by showing that the intersection of such a sequence of compact sets is none pry
nonempty
We know that.
Pick A,B disjoint and closed so that the intersection is their union. Pick A', B' disjoint and open that contain the respective closed sets (we agree you can do this?)
Open in what?
R^2. Pardon for not TeXing
One second. Let me fiddle with this, it might no work after all
02:31
So nothing to do with our $C_j$?
Ah, I got it. And no.
@TedShifrin ?
Hmm. I was going to say let $D_i = C_i - (A' \cup B')$, and this intersection is empty, so $D_i$ is empty for large enough $i$...
But we can write $C_i = (C_i \cap A') \cup (C_i \cap B')$. This contradicts connectedness.
Huh?
Oh ...
Finite intersection property ... Very clever.
I'm pleased. I had to pause because I tripped myself up on the last step for a minute.
I wonder if this is true more generally? A sequence of closed, connected subsets of a compact space intersects to a connected set?
02:38
What do you mean by a sequence?
I don't see where your argument isn't general.
I assumed Hausdorfness when I picked A' and B'.
Well, you also need Hausdorff for closed subset of compact to be compact, so sure.
Actually, you used normality to choose those open sets.
Ahh. Good point.
So if we take Hausdorff but not normality...
Then there may be a weird counterexample ...
02:42
@Sanchez now that I'm on a computer and not artificially shortening what I write: I meant to say "Given a descending sequence of closed, connected subsets of a compact space, i.e., $C_1 \supset C_2 \supset ... \supset C_i \supset ...$, do we have that $\cap_{i=1}^\infty C_j$ is connected?"
Oh, @TedShifrin: All compact Hausdorff spaces are normal
Who says the ambient space is compact?
Oh, I did.
You are using normality of the ambient space ... Or maybe you don't need $A'$ open in the ambient space?
@Mike, what context is this from? How lenient can you be in adding conditions?
Oh, my question was the $C_j$ compact connected, not the ambient space :)
The original question was subsets in $\Bbb R^2$ :)
02:53
Augh. Yeah.
I was going to say we can just take $C_1$ compact with the subspace topology but I'm sure that causes issues with the $A'$
Anyway, your claim is true for compact Hausdorff spaces.
hey guys, just going over some notes and my prof used an inequality that I can't seem to prove. The inequality is as follows: $$ |x|^{2N} \leq n^N \sum_{\gamma = N}|x^\gamma|^2$$ where $\gamma$ is a multiindex, and $x \in \mathbb{R}^n$.
Yeah - but now Ted has me wondering about non-compact Hausdorff spaces.
So what exactly is the question?
Ambient space non-compact, and a sequence of closed connected sets?
It's hard to see at what generality you should talk about this without the context though.
A sequence of compact connected sets.
The context was just a homework problem. Exhibit a sequence of closed connected sets in $\mathbb{R}^2$ whose intersection is disconnected.
03:00
Huh, but then everything lies in $C_1$. So how is it different from the compact case?
I see.
Not necessarily - when we pick those open $A'$, $B'$, open sets that contain them in the Hausdorff space might not be disjoint
What are $A',B'$?
Sanchez, I proved the statement above by picking some disconnected open sets that contained $A$ and $B$ where $A \cup B$ is the intersection of that sequence (and $A \cap B = \emptyset$)
I doubt you would need "openness" in the whole space.
I'm not sure what your proof is, but it's likely that you only need to assume the intersection = $A \cup B$, $A,B$ being disjoint closed sets
03:13
dis joint -> I____\ \ \W ~
How do we form the minimalist structure such that a polynomial $f(x,y,z)$ is trilinear on it?
@EnjoysMath Come again? "Minimalist structure?"
most minimal
that's what she said
Pedro,
you know how you can form a tensor product of two modules using the free group quotient of $M \times N$?
$F/H$ where $H$ is generated by elements satisfying some form
they call that the tensor product over $R$
Yes, but that's all I know.
well couldn't you find some structure such that a polynomial $f(x)$ is linear
You mean we're at the same page on that topic?
@EnjoysMath Well, finite fields is an option.
I think.
03:21
How do you take the tensor product of a field?
Ah, nah.
In a finite field of $p$ elements you do have $(x+y)^p=x^p+y^p$ though.
Oh, ic
How do you extend that to multiple variables though :\
You mean (x+y+z)^p? Exercise.
In fact.
A polynomial over the finite field of $p$ elts is linear iff it is a polynomial in $x^p$.
I mean.
That wouldn't include $x$.
03:27
Wait.
@Zibadawa There.
Good.
@Zibadawa Bad again.
$$\sum_{i=0}^na_i x^{p^i}$$
There.
My wording above is completely bad.
Yes.
@EnjoysMath So, there you have it. That's an exercise in Ireland and Rosen.
What are other structures that your know of @PedroTamaroff that are constructed as a quotient group of the free group on something?
03:38
@EnjoysMath I can't tell you much, but you can construct a lot of things with the free group, cannot you? ;)
@TedShifrin
I think I got the compact case
Yes, @Pedro?
@TedShifrin When one has a surface and a vector field, the "flux" is defined as $\int_S F\cdot{\bf n}dS$ with $n$ the outwards pointing normal, yes?
That is, if $\omega$ is the corresponding $1$-form, it is $$\int_S \star\omega$$
03:44
Oriented surface :)
Yes.
@TedShifrin OK.
I have a problem. =P
That nested connected sets problem was intriguing!
Wherefor art thy problem?
@TedShifrin One second.
OK
Consider the upper cap of a sphere of radius $R$.
I am given $(xz-x\cos z,y\cos z-yz,4-x^2-y^2)$ and I need to find $R$ that makes the flux maximum.
But the derivative of $\star \omega$ is zero if I am not crazy.
That's right. Go on.
@TedShifrin Well, isn't the flux always zero then?
03:54
It's like the problem you did cleverly earlier!
@TedShifrin But it is all zero here, isn't it?
The compact case is proved there as well; nice!
I like my example and @Mike's proof in the compact case.
What was your example?
(Nice timing - just came back)
03:59
I think I missed Mike's proof
Let's check
@Pedro. You lack a closed surface!
The one I thought if was $\mathbb{R}^2 \setminus \{(-n, n) \times \{0\}\}$

« first day (1202 days earlier)      last day (4115 days later) »