0

I've been working with vectors in einstein notation for a while and I believe I have a good intuition on how indices and transformations work in this notation. But for some reason, I can't figure out how to add 2 vectors in this notation.

Normally, if I have vectors: $\vec{v}$ and $\vec{u}$

Their sum is:

$$\vec{v} + \vec{u} = v^i \vec{e_i} + u^i\vec{e_i} = (v^i+u^i)\vec{e_i} = (v+u)^i \vec{e_i}$$

so $(v^i + u^i) = (v+u)^i$

but that's not how we add vectors in, say, polar coordinates. I.e: the vector $[r,\theta]^T+ [r',\theta']^T \neq [r+r', \theta + \theta']$.

What am I doing wrong?

Kurt G.
  • 17,136
Habouz
  • 386
  • 2
    What are you doing wrong? You are misunderstanding what the symbol $+$ represents in different contexts. Yes, we might use the same symbol... but for "adding two polar-coordinate vectors together" we are talking about adding the complex numbers together that those polar coordinates represent... which is not coordinate-wise addition of their polarcoordinate representations. – JMoravitz Sep 01 '23 at 14:19
  • 1
    We also use the symbol $+$ for adding two matrices together, or adding two real numbers together... adding two sequences together... adding two times together on a clock... none of these are interchangeable and they all have different definitions and work with different objects in different ways. – JMoravitz Sep 01 '23 at 14:21
  • 5
    It makes no sense to add points of a manifold, only tangent vectors at the same point. – Ted Shifrin Sep 01 '23 at 14:39
  • To expand on Ted's comment: (i) When we work with vectors in the plane there are really four "coordinates" in play, two representing location in the plane, and two representing displacement from an arbitrary point. In fancy terms we're in the tangent bundle of the plane. (ii) In Cartesian coordinates, it's possible to be sloppy about positions versus displacements because Euclidean parallel transport corresponds to "constant components". (iii) In non-affine coordinates, the sloppiness of (ii) does not work, and we must take care whether ordered pairs represent points or vectors. – Andrew D. Hwang Sep 12 '23 at 14:34

2 Answers2

3

Expanding on Ted Shifrin's comment:

It is shown in this answer that the relationship between the components of vector fields $V$ and $U$ in Cartesian and polar coordinates are:

\begin{align} V&=v^x\partial_x+v^y\partial_y\\ &=\underbrace{(v^x\cos\color{red}\varphi\,+v^y\sin\color{red}\varphi)}_{\displaystyle =:v^r}\,\partial_r+\underbrace{\frac{-v^x\sin\color{red}\varphi+v^y\cos\color{red}\varphi}{\color{red}r}}_{\displaystyle=:v^\varphi}\,\partial_\varphi\,,\\ U&=u^x\partial_x+u^y\partial_y\\ &=\underbrace{(u^x\cos\color{red}\varphi\,+u^y\sin\color{red}\varphi)}_{\displaystyle =:u^r}\,\partial_r+\underbrace{\frac{-u^x\sin\color{red}\varphi+u^y\cos\color{red}\varphi}{\color{red}r}}_{\displaystyle=:u^\varphi}\,\partial_\varphi\,. \end{align}

  • $v^x,v^y,u^x,u^y$ are the Cartesian components.

  • $v^r,v^\varphi,u^r,u^\varphi$ are the polar components.

  • In both systems the upper indices are just labels.

  • In contrast, the highlighted $\color{red}{r,\varphi}$ above are the coordinates of the point $\color{red}p$ on the manifold $\mathbb R^2$ at which both vectors $V,U$ must reside in order to be additible.

Then, \begin{align} V+U&=(v^x+u^x)\,\partial_x+(v^y+u^y)\,\partial_x\,\\[2mm] &=(v^r+u^r)\,\partial_r+(v^\varphi+u^\varphi)\,\partial_\varphi\,,\\[2mm] v^r+u^r&=(v^x+u^x)\cos\color{red}\varphi+(v^y+u^y)\sin\color{red}\varphi\,,\\[2mm] v^\varphi+u^\varphi&=\frac{-(v^x+u^x)\sin\color{red}\varphi+(v^y+u^y)\cos\color{red}\varphi}{\color{red}r}\,. \end{align}

  • Regardless in which coordinate system we are: addition of vectors at the same point $\color{red}p$ is equivalent to addition of components. Anything else would be a nightmare.

Since there seems a disagreement with the answer by TurlocTheRed I elaborate further.

  • The origin does not have polar coordinates so we cannot consider vectors that reside at $\color{red}p=(0,0)\,.$ Instead I shall consider $$ \color{red}p=(1,1)=(\sqrt{2}\,;\pi/4) $$ using the convention that the comma separates cartesian coordinates, and the semicolon separates polar coordinates. The same convention shall apply to vector components. This point has polar coordinates

\begin{align} \color{red}r&=\sqrt{2},\quad\color{red}\varphi=\pi/4\,,&\cos\color{red}\varphi &=\sin\color{red}\varphi=\frac{1}{\sqrt{2}}\,. \end{align} With the formulas above one can easily check that in both coordinates systems components add when we add the vectors: \begin{align} V_\color{red}p&=(1,1)=(\sqrt{2}\,;0)\,,&U_\color{red}p&=(0,1)=\textstyle(\frac{1}{\sqrt{2}};\frac12)\,,&V_\color{red}p+U_\color{red}p&=(1,2)=\textstyle(\frac{3}{\sqrt{2}};\frac12)\,. \end{align}

  • The polar components of these vectors could be a bit puzzling at first but note that $V_\color{red}p$ points from $(1,1)$ to $(2,2)\,.$ This means we don't go in an angular direction and only in the radial direction. That's why the polar components of $V_\color{red}p$ are $(\sqrt{2}\,;0)\,.$

enter image description here

Kurt G.
  • 17,136
-1

Basis vectors and their components mean different things in different coordinate systems. Further, the basis vectors vary from one point to another.

In Euclidean coordiantes, $\vec{i}(x_1,y_1,z_1)$ is the same $\vec{i}(x_2,y_2,z_2)$ It's the same basis vector anywhere in the coordinate plane.

In polar coordinates, $\vec{e_r}=(x/r)\hat{i}+(y/r)\hat{j}+(z/r)\hat{k}$ and $\vec{e_\theta}=(-y/r)\hat{i} + (x/r)\hat{j}$

There are different basis vectors at different positions, so you need different coordinates at different positions.

So $(1,1)$ in Cartesian coordinates is $(\sqrt{2}, \pi/4)$ in polar.

$(0,1)$ in Cartesian Coordinates is $(1, \pi/2)$ in polar.

$<1,1>+<0,1>=<1,2> $ in Cartesian. In polar that is $(\sqrt{5}, \arctan(2))$

The formula for adding cartesian coordinates is $<x_1,y_1>+<x_2,y_2> = <x_1+x_2, y_1+y_2>$

For polar, it's $<r_1,\theta_1> + <r_2, \theta_2> = <r,\theta>$

$r=\sqrt{r_1^2+r_2^2+2r_1r_2 \cos (\theta_1-\theta_2)}$

$\theta = \arctan (\frac{r_1\sin \theta_1+r_2\sin \theta_2}{r_1\cos\theta_1+ r_2\cos\theta_2})$

TurlocTheRed
  • 6,458
  • Then it looks like we disagree. I believe the reason for that is that the point $p=(1,1)$ in Cartesian coordinates is $(\sqrt{2},\pi/4)$ in polar coordinates but be careful: vectors transform as vectors, not via $(x,y)\mapsto (\sqrt{x^2+y^2},\arctan(y/x))$ which transforms points. The simple rule that must hold for the vector components is $v=v^x,\hat{i}+v^y,\hat{j}=v^r,\vec{e_r}+v^\theta,\vec{e_\theta},.$ – Kurt G. Sep 01 '23 at 18:38