Theorem about deformations of positively oriented vectors in $\mathbb{R}^{n}$

56 Views Asked by At

I'm reaching out to you because of theorem 1.17 in the textbook "Multivariable calculus with applications". The theorem states the following:

Theorem Every positively oriented ordered list of n linearly independent vectors $\vec{V_{1}},...,\vec{V_{n}}\in \mathbb{R}^n$ can be deformed into a list of unit vectors $\vec{E_1},\dots,\vec{E_n}$.

A proof outframe is given but it is kind of vague. So the problem I'm having is with the statement that there is a rotation that takes the vector $V_n$ into $pE_n$. yes that may be true for some matrix $M_{n\times m}$ but later on, it says that the mentioned rotation "carries the vectors $V_j$ in the vectors $W_j,j<n$" which I assume is true because $V_{n-k}M_{n-k\times n-k}=pE_{n-k}=W_{n-k}$, however it later says that we "shrink" the n-th component of $W_j$ to zero by adding $W_n=pE_n$, which is something of the form $W_j(0,\dots, p, \dots, 0)+W_n(0, \dots, p)=W'(0,\dots, p, \dots, p)$ I suppose? Then it says that the resulting vectors in the hyperplane $x_n=0$ are linearly independent which is true, yes but I don't see the relevance of shrinking and doing everything else that I've mentioned.

You can find the whole thing here page 41, Theorem 1.17.

1

There are 1 best solutions below

0
On BEST ANSWER

(the author) after careful consideration, I've reached several conclusions. The first one is the following:

We choose one vector namely the $\vec{V_n}$ which we multiply by the rotation matrix in such a way to reach the form $\vec{V_n}M_{n\times n}=p\vec{E_n}$, we choose the rotation matrix in such a way as to bring the vector back to the basis of $\mathbb{R}^{n}$ for example:

The vector $\vec{V_2}(1,1)$ makes a 45 degrees angle with the x-axis, which means that it will take him 45 degrees to get back to the y-axis and be normalized. Essentially we have: \begin{equation} \begin{pmatrix} 1\\ 1 \end{pmatrix} \begin{pmatrix} \cos(-\frac{\pi}{4}) & \sin(-\frac{\pi}{4}) \\ -\sin(-\frac{\pi}{4}) & \cos(-\frac{\pi}{4}) \end{pmatrix} = \begin{pmatrix} 1\\ 1 \end{pmatrix} \begin{pmatrix} \frac{\sqrt{2}}{2} &-\frac{\sqrt{2}}{2}\\ \frac{\sqrt{2}}{2} &\frac{\sqrt{2}}{2} \end{pmatrix}= \begin{pmatrix} \frac{\sqrt{2}}{2} -\frac{\sqrt{2}}{2} \\ \frac{\sqrt{2}}{2} +\frac{\sqrt{2}}{2} \end{pmatrix}= \begin{pmatrix} 0 \\ \sqrt{2} \end{pmatrix} \end{equation} The resulted vector is of the form $\sqrt{2}E_2$ which is what we were aiming for. From here we can write $V_1$ as $V_1(a,b)$, to obtain the vector mentioned in the theorem (namely $W_1$) we multiply it by the same rotational matrix: \begin{equation} \begin{pmatrix} a\\ b \end{pmatrix} \begin{pmatrix} \cos(-\frac{\pi}{4}) & \sin(-\frac{\pi}{4}) \\ -\sin(-\frac{\pi}{4}) & \cos(-\frac{\pi}{4}) \end{pmatrix}= \begin{pmatrix} \frac{\sqrt{2}}{2}(a -b) \\ \frac{\sqrt{2}}{2}(a +b) \end{pmatrix} \end{equation} We now create the determinant: \begin{equation} \begin{vmatrix} \frac{\sqrt{2}}{2}(a-b)&\frac{\sqrt{2}}{2}(a +b)\\ 0 & \sqrt{2} \end{vmatrix} =(-1)^{2+2}\sqrt{2}\left|\frac{\sqrt{2}}{2}(a -b)\right| \end{equation}

As far as I understand the determinant $\left|\frac{\sqrt{2}}{2}(a -b)\right|$ should be positive due to the positive orientation of the vectors taken.

The same is with the theorem. We don't aim for all vectors but we are deforming only the n-th one ($V_n$) which bring us the following determinant: \begin{equation} \begin{vmatrix} a_{11} & a_{12} &\dots &a_{1n}\\ \dots &\dots & \dots&\dots\\ 0 & 0 &\dots &p\\ \end{vmatrix} \end{equation} When we multiply the n-th row by $-\frac{a_{jn}}{p}$ and add it to the j-th row we get the n-th column of the j-th row to be zero, so we obtain something of the form: \begin{equation} \begin{vmatrix} a_{11} & a_{12} &\dots &0\\ \dots &\dots & \dots&0\\ 0 & 0 &\dots &p\\ \end{vmatrix} \end{equation} Factor out p so to get: \begin{equation} (-1)^{n+n}p\begin{vmatrix} a_{11} & a_{12} &\dots &a_{1,n-1}\\ \dots &\dots & \dots&\dots\\ a_{n-1,1} & a_{n-1,2} &\dots &a_{n-1,n-1}\\ \end{vmatrix} \end{equation} And you do this inductively for all the rest of the $n-1$ cases to obtain $det(Z)>0$. The final question is to determine whether $n+n$ is odd or even. If n-even then $2k+2+2k+2=2(k+k)+4$ which is even, if odd $2k+1+2k+1=2(k+k)+2$ which is even as well. Either way our determinant has a positive sign.