@MatheinBoulomenos I should hopefully finish first 4 chapters + problems by July
@MatheinBoulomenos after Ravi Vakil I plan to move on to Shava both 2 volumes and finish Maybe Michael Atiyah and Eisenbudd and finally go back to hartshrone
Does anyone know of some resources that explain the basic manipulations of covariant and contravariant four vectors used in physics mostly? I'm trying to evaluate $\partial_\mu(x^2x_v)$ but I don't know exactly what to do. I see that I can do $\partial_\mu(x_v x^v x_v)$ but don't know how to continue
Given a locally Euclidean subset $X\subset\mathbb R^n$ and $p\in X$, let $\widetilde{\mathrm{T}}_pX$ denote the *tangent set* of $X$ at $p$, namely the set of derivatives of differentiable curves in $X$ based at $p$.
Suppose for all $p\in X$ we have that $\widetilde{\mathrm{T}}_pX$ is a linear subspace of $\mathbb R^n$ of dimension $\dim_pX$. Does it follow that $X\subset\mathbb R^n$ is an embbeded submanifold?
@Arrow Consider a homeomorphism $f : B \to \Bbb R^N$ which takes $X \cap B \subset B$ to $\Bbb R^n \subset \Bbb R^N$. Your statement about curves should mean $f^{-1}|_{\Bbb R^n}$ is differentiable.
Oh no. Oh-ho-ho. There are functions which have directional derivatives in all directions but does not have a derivative, right? What's a dumb example of such a continuous function?
> Moreno (1998) showed that the space of pairs of norm-one sedenions that multiply to zero is homeomorphic to the compact form of the exceptional Lie group G2. (Note that in his paper, a "zero divisor" means a pair of elements that multiply to zero.)
Okay, great. The graph is a topological submanifold of R^3 which has tangent space in the sense you said, I think, but should not be a differentiable submanifold
I don't think so: $f(x,x^2)$ is a tilted parabola in $\mathbb R^3$ and the tangent to this parabola at the origin lies in a non-horizontal plane. However there are also tangents to the graph at the origin in every single direction in the $xy$-plane. Particularly the tangent set cannot be a two-dimensional vector subspace.
So I think everything boils down to proving if $f : \Bbb R^2 \to \Bbb R$ is a function such that $Df_v(0)$ exists for all $v \in \Bbb R^2$ and $\{(v, Df_v(0)) \in \Bbb R^3 : v \in \Bbb R^2\}$ form a vector subspace of $\Bbb R^3$ then $f$ is differentiable at $0$.
@BalarkaSen that's good to know :) I am not sure I understand your reduction yet - it may be that a graph is an embedded submanifold yet the function is not differentiable (if there's a vertical tangent).
Hi @MikeMiller, could you please help me out with the linked question? I'm trying to understand whether a locally Euclidean subset $X\subset \mathbb R^n$ with tangent spaces of the correct dimension is necessarily an embedded differentiable submanifold.
@Arrow (cc @Akiva) I was trying to write an answer and immediately realized I am using local flatness. So that makes me wonder how the tangent sets of the Alexander horned sphere at the "bad points" look like.
Even if that's true they may not have the right dimension. For instance the points over the axes (except the origin) in the graph of $f(x,y)=|xy|$ have tangent lines but not tangent planes.
@BalarkaSen hmm... what about looking for a natural differentiable structure on $X$? That is, a sheaf of commutative rings making $(X,\mathcal O_X)$ locally isomorphic to an open subset of Euclidean space locally ringed with its differentiable functions?
@Eric Oh, by the way. Suppose $\{e_1, \cdots, e_n\}$ is a frame on a chart $U$ on a manifold. This is integrable, i.e., there's local coordinates where $e_i = \partial/\partial \mathbf{x}_i$ iff $[e_i, e_j] = 0$ by Frobenius, right?
Suppose $\omega_1, \cdots, \omega_n$ are dual to the frame. $d\omega_k(e_i, e_j) = e_i \omega_k(e_j) - e_j \omega_k(e_i) - \omega_k([e_i, e_j])$. If $k \neq i, j$ everything is zero. If $k = i$ the first and last thing vanishes, and the middle vanishes because $\omega_i(e_i) = 1$ is constant function; similar if $k = j$.
So $d\omega_i = 0$...
We chose the frame such that our symplectic 2-form $\omega = \omega_1 \wedge \omega_2 + \cdots + \omega_{n-1} \wedge \omega_n$ (ie omega is standard on that frame). Since $d\omega_i = 0$, $d\omega = 0$. Just to explicitly establish it as a integrability criterion
What we'd need to show is that local sections of the subbundle of $F(TM)$ consisting of frames $\{e_1, \cdots, e_n\}$ such that $\omega$ is standard on that frame are integrable. It's not obvious to me how one goes about doing that.
My argument goes two ways until that last step. $d\omega = 0$ does not necessarily imply $d\omega_k$'s are zero
@BalarkaSen an idea for the submanifold problem - perhaps we can construct an exponential map giving a diffeomorphism berween a neighborhood about $p$ of the tangent plane and an open neighborhood of $p$ in $X$? Would this be enough?
I thought about somehow laying down straight lines in the plane onto geodesics. This is described in 4-6 of do Carmo's Differential Geometry of Curves and Surfaces but not quite in this generality.
I am 90% sure the ideas I gave solves your problem. Pick a ball $B$ such that $B \cap X$ is homeo to $\Bbb R^k$ by $f : \Bbb R^k \to B \cap X$. If $\gamma$ is any smooth curve through origin in $\Bbb R^k$, $f(\gamma)$ is a smooth curve in $\Bbb R^n$ and it's derivative at $p$ is well-defined. Then $f$, as a map to $\Bbb R^n$, has directional derivatives well-defined: $D_{\gamma'}f(0) = (f \circ \gamma)'(0)$
Since it's differentiable in every direction, and the directional derivative is a linear map in the vector component, it should be differentiable at $0$ as a multilinear map
Well, the problem is $f(\gamma)$ might not be a smooth curve in $\Bbb R^n$. Somehow the dimension condition says for any $v \in \Bbb R^k$, there's a curve $\gamma$ through origin such that $\gamma'(0) = v$ and $f(\gamma)$ is a smooth curve
Unless I misunderstood something it also doesn't seem that having a directional derivative linear in the vector component implies differentiability (see here).
No, the tangent set is not a 2-plane, which is why the function isn't differentiable, but all directional derivatives are zero, so taking a direction to the directional derivative along it is the zero map which is linear. (Sorry if I'm being stupid, it's not on purpose)
The question itself says $\partial_vf(x)=L(v)\,\forall v\in\mathbb{R}^n$ where $L$ is linear.
The tangent set is not a 2-plane because directional derivatives cannot detect the derivative of the composite of $f$ with a "curved" curve i.e not a straight line. (If I understand correctly)
@Arrow I didn't suggest an approach per se, just that it's an easier and a more doable question that if a function's graph has a tangent space at a point then the function is differentiable at that point.
Given a differential 2-form w on a mfld M, then on a nbhd of a point m in M I can write it in local coords. I might have more than one local coord chart covering the same point, and I read that the compatibility condition we want is $\phi_{ab}^*w_b=w_a$ where \phi_ab is the transition function between local chart a and b. The question is: point-wise is this process just applying a change of variable to the matrix of coefficients of w?
In this paper1 on the history of functional analysis, the author mentions the following example of an infinite system of linear equations in an infinite number of variables $c_i = A_{ij} x_j$:
\begin{align*}
\begin{array}{ccccccccc}
1 & = & x_1 & + & x_2 & + & x_3 & + & \dots \\
1 & = & & &...