Hi, does anybody know what $\mathcal O(-3)$ is? It appears in the context of a bundle over the complex projective 2d space $\mathcal O(-3) \to \mathbb P^2$
This is called $E = \mathcal O(-1)$. The tensor products $\otimes_k E$ are called $\mathcal O(-k)$. Its inverse in the group of holomorphic line bundles is called $\mathcal O(1)$.
Yeah, @s.harp, the best way I know how to describe the tensor product is to tensor the transition functions to $GL_n$; for $GL_1$, tensor product is just multiplication
These are all of the holomorphic line bundles over $\Bbb P^2$ up to isomorphism. They're named such, because given a generic meromorphic section of $\mathcal O(-k)$, its zero/pole set gives a divisor, which defines an element of $H_2(\Bbb P^2;\Bbb Z)$
Given the canonical isomorphism $H_2(\Bbb P^2;\Bbb Z) \cong \Bbb Z$ (so that the generator is $\Bbb{P}^1 \subset \Bbb P^2$), the appropriate element is $-k$
well, one issue is that trig functions aren't always $>0$. so that could produce some problems. but I don't really see what you actually mean without an example
given a complex manifold/algebraic variety, there's its holomorphic cotangent bundle $T^*M$; this is a holomorphic vector bundle of rank $n$. you can define $K_X = \Lambda^n T^*M$. this is a holomorphic line bundle, called the "canonical bundle"
@JoeStavitsky i don't see an issue. for example, subbing $x=\tan\theta$ in $\frac{1}{(1+x^2)^{5/2}}$ gives $\frac{1}{(\sec^2\theta)^{5/2}} = \cos^5\theta$. nothing particularly strange there.
The definition of closed is that every limit point is contained in the set. In $\mathbb{R}$ for example $\infty$ is a limit point but it is not contained in $\mathbb{R}$ so how is it closed?
@user19405892 To talk about closed sets, you need to be living inside some bigger space. What space are we talking about when we say "$\Bbb R$ is closed"?
@MikeMiller I kind've wish I did understand what's going on in that 'stability conditions' paper, if only because I did run into Bridgeland's papers on stability conditions in triangulated categories when delving into quadratic/strebel differentials stuff
I wonder actually, in my mathematical physics seminars we have lots of strange arguments imported from physics and used to prove some mathematical theorems, like showing Gauß bonnet with path integral methods. Do mathematicians also use these physics methods or are they restricted to those that call themselves mathematical physicists?
Unfortunately, @user19405892, there are a few levels of confusion. There is a threshold after which one can try to bring someone out of confusion. But in this discussion (and in previous) you have tended to seem so confused about the very nature of what you're even asking (and then do not clarify for some minutes after someone prods) that I don't think I can do anything to help. So I'm not going to try to engage.
@s.harp I think there are a group of mathematicians that understand path integrals, but I don't, and that group tends to be filled with category theorists.
I think somehow the right notion of chern-simons theory on 3-manifolds and path integrals and whatnot is Reshitikhin-Turaev theory, which involves modular categories? One day someone will explain this to me.
when I heard it in QFT every second term in the derivation was a divergent quantity, the others were terms I did not know what their actual definition was
plus, as s.harp says, there's applications to nonequillibrium systems i.e. Keldysh formalism. my advisor has done a lot of work in that area, though i've stayed away from it
I think the book "Functional Integration And Quantum Physics" by Barry Simon is seen as the best rigorous treatment. Unfortunately it was very dense and I could not read it.
The path integral is like drugs, if you do it and like it everything is great but if you don't like it and everybody around is doing it you don't feel too happy
I was told that one of the most efficient tools (e.g. in terms of computations relevant to physics, but also in terms of guessing heuristically mathematical facts) that physicists use is the so called "Feynmann path integral", which, as far as I understand, means "integrating" a functional (actio...
"It is possible to rigorously define the path integral. What you can't rigorously define is the 'Lebesgue' measure that physicists write as $\mathcal{D}\phi$. When a physicist writes $\frac{1}{Z}e^{iS(\phi)}\mathcal{D}\phi$, you should understand this as a notational shorthand; they're telling you roughly what form the finite-dimensional approximations you use to define the path integral ought to take. "
So, I saw this presentation I didn't entirely understand, but you can redefine $D\phi$ without the notion of measure, and instead using cohomology, and then the path integral becomes well defined.
@Mambo The basis is all sets of the form $\{x:a<x<b\}$ for $a,b\in\Bbb R\cup\{\infty\}$. The neighborhoods of infinity are the sets containing $(a,\infty)\cup\{\infty\}$ for real $a$.
well, I take your point to be that while there's no notion of a probability measure on $\mathbb{R}$, you can consider the sequence of average values generated by the intervals $[-n,n]$
and while that may or may not converge as $n\to\infty$, it is at least a well-defined sequence
and i guess my point is that, if I have a function which decays to zero sufficiently fast at infinity, I can find an approximation of it over a sufficiently large interval $[-L,L]$ with periodic boundary conditions
@anon Hi anon, yes a center of symmetry $O$ of a closed, bounded figure is a point that such that for any collinear points $X,O,Y$ with $X,Y$ on the figure, $OX = OY$
and if $L$ isn't too small (large enough that the Gaussians don't overlap much), then I've got a function which is periodic on $[-L/2,L/2]$ and which approximates the original function on this interval to exponential precision
you keep saying average value as if it's clear that the right place to take the 'average value' is strictly the domain over which it's not zero; that's not clear to me
i could tell you that the actual average value there was $\pi/4$, or i could point out that because it's a continuous function and almost always less than 1, its integral is less than 2
to draw a connection between our two strands of conversation, the Gaussian will have 'average value of zero' according to Mike's prescription. But the periodized Gaussian won't.
the point being, of course, that the physicist would assume a certain abuse of notation, i.e. that $(x,y)$ isn't a label of inputs but rather Cartesian coordinates
so there's a certain amount of baggage that isn't being stated.
similarly, the path integral shouldn't be as a literal construction so much as a convenient way of implying a whole slew of assumptions about how such a construction should proceed.
Remember that theorem I mentioned earlier? About how $\Bbb R^n$ can't be written as the union of two connected open sets with disconnected intersection? And there's a quick LES proof. I think it's really illuminating to see what happens at the level of cycles, because the LES proof kinda obscures what's going on.
When I said "level of cycles" I meant "level of chains," I guess
So, first off, connectedness is the same as path-connectedness for open sets, and all the relevant sets are open
and path-connectedness is the same as saying that, for any two points $x$ and $y$, there's a 1-chain whose boundary is $x-y$
So, call the open sets $A$ and $B$; we want to show that $A\cap B$ is connected
So, choose arbitrary $x$ and $y$. Because $A$ and $B$ are connected, there's are paths (1-chains) $\alpha_1\in C_1(A)$ and $\beta_1\in C_1(B)$ whose boundaries are $x-y$
$\alpha_1-\beta_1$ is a 1-cycle in $\Bbb R^n$, so it's a boundary of some 2-chain. You can write that 2-chain as $\alpha_2+\beta_2$ for $\alpha_2\in C_2(A)$ and $\beta_2\in C_2(B)$, by subdivision
Then $\partial(-\alpha_2)+\alpha_1=\partial(\beta_2)+\beta_2$ is a chain in $A\cap B$, with boundary $x-y$
and so it's the desired 1-chain connecting $x$ and $y$. Since $x$ and $y$ were arbitrary, any two points can be connected by a 1-chain in $A\cap B$.