Monthly Archives: October 2013

Exercise 22.1

Guess what: the in-class exercise loop has winding number 1 now! That means I’m up again.

The exercise was to prove that \( \|M_g\| = m = \sup \left\{ {|g(x)| \colon x \in [0,2 \pi] } \right\} \). In class we proved that \( \|M_g\| \le m \), so I’ll prove the converse. Also, I apologize for the strange formatting of the line breaks, but I don’t know how to insert linebreaks in MathJax, and MathJax doesn’t detect that there’s a border so it doesn’t automatically add the break for me.

Let \( E_\epsilon = \left\{ {x \colon x \in [0,2 \pi], g(x) \ge m- \epsilon } \right\} \). As \(m \) is the supremum, this set is nonempty, so the following function is nonzero on a set of nonzero measure. Define \[ f(x) = \left\{ \begin{array} {lr} 1 \colon x \in E_\epsilon \\ 0 \colon x \in (E_\epsilon)^c \end{array} \right\} \]

\[ \|M_g f\|^2 = \frac{1}{2 \pi} \int_0^{2 \pi} | M_g f(x) |^2 dx = \frac{1}{2 \pi} \left ( \int_{E_\epsilon} | g(x) f(x) |^2 dx + \int_{(E_\epsilon)^c} | g(x) f(x) |^2 dx \right ) \]

\[= \frac{1}{2 \pi} \int_{E_\epsilon} | g(x) f(x) |^2 dx \ge (m- \epsilon)^2 \frac{1}{2 \pi} \int_{E_\epsilon} | f(x) |^2 dx = (m- \epsilon)^2 \|f\|^2 \]

Hence \( \|M_g\| \ge \frac{\|M_g f \|} {\|f\|} \ge m-\epsilon \), and as epsilon was arbitrary, \( \|M_g\| \ge m \).

Exercise 21.1

Proof:Let \(T:U\to V\) and \(S: V\to W\) be Fredholm operators, where \(U\),\(V\) and \(W\) are Hilbert Spaces. Then consider \(ST: U\to W\):

If \(u\in \textrm{Ker} ST\)(i.e. \(ST(u)=0\)), then we have \(T(u)=0\)(i.e.\(u\in \textrm {Ker} T\) ) or \(T(u)\in \textrm {Ker} S\), thus \(\textrm {dim (Ker} ST)\leq \textrm {dim(Ker}T)+\textrm {dim(Ker} S)\).

If \(w\in \textrm {Complement of Range} ST\), then we have \(w\in \textrm {Complement of Range} S\) or \({S}^{-1}(w)\in \textrm {Complement of Range} T\), thus \(\textrm {Corank} ST\leq \textrm {Corank} T+\textrm {Corank} S\).

So \(\textrm {Ker} ST\) and \(\textrm {Complement of Range} ST \) all have finite dimensions, the composition ST is a Fredholm operator.

Next let’s consider the index. First consider the operator \({A}_{t}: V\oplus U\rightarrow W\oplus V\), and \({A}_{t}\) is defined by:

\[{A}_{t}= \begin {pmatrix}  \cos {(t)}S & -\sin {(t)}ST \\ \sin {(t)}I & \cos {(t)}T \\ \end{pmatrix}\]

for \(0\leq t \leq \pi\), \({A}_{t}\) is continuous about \(t\). And \({A}_{t}\) can also be written as

\[{A}_{t}=\begin {pmatrix}S & 0 \\ 0 & I \\ \end {pmatrix}\begin {pmatrix}\cos {(t)}I &  -\sin {(t)}I \\ \sin {(t)}I & \cos {(t)}I \\ \end{pmatrix}\begin {pmatrix}I & 0 \\ 0 & T \\ \end{pmatrix}\]

Since \(S\) and \(T\) are Fredholm operators, and \(\begin {pmatrix} \cos {(t)}I & -\sin {(t)}I \\ \sin {(t)}I & \cos {(t)}I \\ \end {pmatrix}\) is a rotation from Hilbert space to itself\((\textrm {Nullity} = \textrm {Corank}=0)\), which is also a Fredholm operator, so \(\begin {pmatrix}S & 0 \\ 0 & I \\ \end {pmatrix}\), \(\begin {pmatrix}\cos {(t)}I &  -\sin {(t)}I \\ \sin {(t)}I & \cos {(t)}I \\ \end{pmatrix}\), \(\begin {pmatrix}I & 0 \\ 0 & T \\ \end{pmatrix}\) are all Fredholm operators. Thus \( {A}_{t}\) is a composition of Fredholm operators \(\Longrightarrow\) \( {A}_{t}\) is a Fredholm operator for all \(t \in [0, \pi ]\).

Since \({A}_{t}\) is continuous about \(t\) and is a Fredholm operator for every \(t \in [0, \pi]\), \({A}_0\) and \({A}_{\frac{\pi}{2}}\) are in the same component of \(\textrm {Fred}({V\oplus U}, {W\oplus V}\), i.e. \(\textrm {Index} {A}_0=\textrm {Index} {A}_{\frac{\pi}{2}}\).

\[{A}_{0}=\begin{pmatrix} S & 0\\ 0 & I \\ \end {pmatrix} \begin {pmatrix} I & 0 \\ 0 & I \\ \end {pmatrix}\begin{pmatrix} I & 0 \\ 0 & T \\ \end {pmatrix}=\begin{pmatrix} S & 0 \\ 0 & T \\ \end {pmatrix} \]

\[{A}_{\frac{\pi}{2}}=\begin{pmatrix} S & 0\\ 0 & I \\ \end {pmatrix} \begin {pmatrix} 0 & -I \\ I & 0 \\ \end {pmatrix}\begin{pmatrix} I & 0 \\ 0 & T \\ \end {pmatrix}=\begin{pmatrix} 0 & -ST \\ I & 0 \\ \end{pmatrix} \]

\[\textrm {Index} {A}_{0}= \textrm {Index}  S + \textrm {Index}  T \]

\[\textrm {Index} {A}_{\frac{\pi}{2}}= \textrm {Index} -ST + \textrm {Index} I=\textrm {Index} ST\]

Thus, \(\textrm {Index} S + \textrm {Index} T=\textrm {Index} ST\)

 

Alphabetical list

Thought I would repost this here so we didn’t have to hunt it down everyday back at the beginning of the blog:

1. Aaron Calderon

2. Scott Conrad

3. Nick Dyszel

4. Sam Fairchild

5. Leah Frederick

6. Max Goering

7. Andrew Hanlon

8. Wilson Jarrell

9. Joey Kraisler

10. Steven Metallo

11. Justin Miller

12. Mike Miller

13. Minh Nguyen

14. Juan Vargas-Molina

15. Radoslav Vuchkov

16. Charles Walker

17. Liuquan Wang

18. Zheyi Xu

19. Zhiying Xu

20. Zhaoning Yang

21. Hongyuan Zhan

Triangulation

We were talking in class the other day about the triangulation (or cellulation) of manifolds.  My colleague Frank Quinn from Va Tech just published a survey about triangulation, including the very latest results from this year, and also including a lot of interesting history and philosophy… you might be interested to take a look at it

Quinn, Frank. 2013. “The Triangulation of Manifolds”. ArXiv e-print 1310.7644. http://arxiv.org/abs/1310.7644.

Solution to Exersice in Lecture 19

Exercise 19.1 Recall \(S\) is a smooth surface we defined in class, prove that:

(a) the assertion just made, that \({{r}_{x}},{{r}_{y}}\) are linearly independt.

(b) Also, prove that the definition of tangent plane is chart independent.

Solution.

(a) 

Let \(r:V\to U\) be a chart such that \(r({{x}_{0}},{{y}_{0}})=p\), here \(V\subset {{\mathbb{R}}^{2}}\) is an open set, \(U\subset S\) is an open set which contains \(p\). We can choose \(r\) so that \(r(0,0)=p\), otherwise we can replace \(r(x,y)\) by \(r(x’+{{x}_{0}},y’+{{y}_{0}}).\)

From the graph property of the surface \(S\), we can represent other \(N-2\) coordinates by smooth functions \({{r}_{1}}(x,y),\cdots ,{{r}_{N-2}}(x,y)\), so we have

                               \(r(x,y)=(x,y,{{r}_{1}}(x,y),\cdots ,{{r}_{N-2}}(x,y)).\)

Hence

\({{r}_{x}}=(1,0,{{r}_{1}}_{x},\cdots ,{{r}_{(N-2)x}}),{{r}_{y}}=(0,1,{{r}_{1y}},\cdots ,{{r}_{(N-2)y}})\)

They are obviously independent since \((1,0)\) and \((0,1)\) are independent.

(b)

Suppose \(({{U}_{1}},{{V}_{1}},\varphi )\) and \(({{U}_{2}},{{V}_{2}},\psi )\) are two graph charts such that \(p\in {{U}_{1}},{{U}_{2}}\), \({{V}_{1}}\subset {{\mathbb{R}}^{2}},{{V}_{2}}\subset {{\mathbb{R}}^{2}}\) are open sets, \(\varphi ,\psi \) are smooth functions.

Then \({{\psi }^{-1}}\circ \varphi :{{V}_{1}}\to {{V}_{2}}\) is a smooth function.  Suppose

                \(\varphi ({{x}_{0}},{{y}_{0}})=\psi ({{s}_{0}},{{t}_{0}})=p,\varphi (x,y)=\psi (s,t)\in {{U}_{1}}\cap {{U}_{2}}\),

Since the coordinate function \({{\pi }_{1}}:(s,t)\to s,{{\pi }_{2}}:(s,t)\to t\) are smooth, we have two smooth function \(S={{\pi }_{1}}\circ {{\psi }^{-1}}\circ \varphi \) and \(T={{\pi }_{2}}\circ {{\psi }^{-1}}\circ \varphi \) such that

                                                       \(s=S(x,y),t=T(x,y)\).

Now from \(\varphi (x,y)=\psi (s,t)=\psi (S(x,y),T(s,t))\) and the chain rule of differentiation, we have

\({{\varphi }_{x}}({{x}_{0}},{{y}_{0}})={{\psi }_{s}}(S({{x}_{0}},{{y}_{0}}),T({{x}_{0}},{{y}_{0}})){{S}_{x}}({{x}_{0}},{{y}_{0}})+{{\psi }_{t}}(S({{x}_{0}},{{y}_{0}}),T({{x}_{0}},{{y}_{0}})){{T}_{x}}({{x}_{0}},{{y}_{0}}),\)

\({{\varphi }_{y}}({{x}_{0}},{{y}_{0}})={{\psi }_{s}}(S({{x}_{0}},{{y}_{0}}),T({{x}_{0}},{{y}_{0}})){{S}_{y}}({{x}_{0}},{{y}_{0}})+{{\psi }_{t}}(S({{x}_{0}},{{y}_{0}}),T({{x}_{0}},{{y}_{0}})){{T}_{y}}({{x}_{0}},{{y}_{0}}).\)

Hence \({{\varphi }_{x}},{{\varphi }_{y}}\) can be represented as the linear combination of  \({{\psi }_{s}},{{\psi }_{t}}\),  similarly \({{\psi }_{s}},{{\psi }_{t}}\) can be represented as the linear combination of  \({{\varphi }_{x}},{{\varphi }_{y}}\), so

                          \(\textrm{span}\{{{\varphi }_{x}},{{\varphi }_{y}}\}=\textrm{span}\{{{\psi }_{s}},{{\psi }_{t}}\}\),

hence they produce the same 2-dimensional subspace in both cases, so the definition of tangent plane is chart independent. 

 This solution is a little long, but the idea is simple, maybe I just used too much words. If there are some mistakes or questions about the solution, please reply to this post, thank you.

Exercise 20.1

Let {\( T_i\)},  \(i=1,2,…,n\) be the linear transformations, \( T_i:  V_{i-1} \to V_i\), \(T_0:  0 \to V_1\), \(T_{n+1}:  V_n \to 0\).

Since \(\textrm{Im}   T_i = \textrm{Ker}   T_{i+1}, i=0,1,…,n\), we have \(\textrm{Rank}   T_i =\textrm{Nullity}   T_{i+1}\)

Since \(\textrm{Ker}   T_0 = 0, \textrm{Im}   T_{n+1} = 0\), we obtain \(\textrm{Nullity}   T_0 = 0,\textrm{Rank}   T_{n+1} = 0,\textrm{Rank}   T_0 = \textrm{dim}   0 – \textrm{Nullity}   T_0 = 0\)

Thus, \(\sum\limits_{k=0}^{n}{(-1)^k\textrm{dim}   V_k}\)

\(=\sum\limits_{k=0}^{n}{ (-1)^k(\textrm{Rank}   T_{k+1} + \textrm{Nullity}   T_{k+1}) }\)

\(=\sum\limits_{k=0}^{n}{ (-1)^k(\textrm{Rank}   T_{k+1} + \textrm{Rank}   T_k )}\)

\(=\sum\limits_{k=1}^{n+1}{-(-1)^k(\textrm{Rank}   T_k } + \sum\limits_{k=0}^{n}{(-1)^k\textrm{Rank}   T_k }\)

\(=(-1)^0\textrm{Rank}   T_0 – (-1)^{n+1}(\textrm{Rank}   T_{n+1} )\)

\(=0\)

Exercise 18.1

The exercise requires us to write and justify a formula for a vector field with an index +2 singularity. Note that I will be using the same notation of a vector field as given in definition 18.1.

Thus, let our vector field be given by \[X(z) = z(z-2),\]

where z is a complex number.

We can see that the vector field vanishes at 0 and 2. Also, each of these singularities is isolated. We consider the of index of the singularities at each point. Now construct (as in the lecture notes) the function: t \(\to\) \(X(a +\epsilon  e^{2\pi i t})\), where a is an isolated point, \(\epsilon\) is small, and t is between 0 and 1. For each isolated point 0 and 2, we create a smooth loop with our function. The winding number is +1 for each of these loops. Thus, 1+1 = +2, and the index of our vector field is then +2.