0
$\begingroup$

I am always confused with the algebra of differentiating the index notation, and have browsed many other posts but still confused. There must be details I have been missing. It would be really appreciated if anyone could answer my questions.

My questions are:

  1. What is the differentiation rule for $\partial^2$?

  2. What is the rule for differentiating terms with indices $\mu$ with respect to a different index, e.g., $\nu$?

  3. What's the difference between $\frac{\partial}{\partial x^\mu}$ and $\frac{\partial}{\partial x_\mu}$? What do we do with the differentiations with respect to $\partial x_\mu$?

Using the Klein-Gordon field as an example: $$ L = \frac{1}{2}(\partial_\mu\phi)(\partial^\mu\phi) - \frac{1}{2}m^2 \phi^2, $$ I wonder why is $$ \frac{\partial L}{\partial (\partial_\mu\phi)} = \partial^\mu \phi? $$ How do we compensate the coefficient of $\frac{1}{2}$? It is clear if thinking of the terms as the linear combination of $\phi^i$'s. But from the usual differentiation rule, it seems to be $$ \frac{\partial L}{\partial (\partial_\mu\phi)} = \frac{1}{2}\partial^\mu \phi. $$

$\endgroup$
4
  • $\begingroup$ There's no magic here. Write $2L = (\partial_1 \phi)^2 + \dots (\partial_4 \phi)^2$ and differentiate with respect to $\partial_1 \phi$. $\endgroup$ Commented Apr 11, 2023 at 17:37
  • $\begingroup$ @ConnorBehan Thanks. I wonder if there's a way of reading off the derivative from the equation? Or we need to go through this thinking process every time when we meet such differentiations. Also curious what would happen if we have $\frac{\partial L}{\partial (\partial_\nu\phi)}$? $\endgroup$
    – user174967
    Commented Apr 11, 2023 at 17:50
  • $\begingroup$ If the arbitrary symbol $\mu$ is changed to $\nu$ on the left side then the same happens on the right side. $\endgroup$ Commented Apr 11, 2023 at 18:00
  • $\begingroup$ Remember back to the basics of index notation, $x^{\mu}=(x^0,x^1,x^2,x^3)$ which in cartesian coordinates it's given by $(t,x,y,z)$. Thus $$\partial_{\mu}=\frac{\partial}{\partial x^{\mu}} =(\partial_t,\partial_x,\partial_y,\partial_z)$$ then you just use the metric tensor to move indices up and down when necessary. $\endgroup$
    – Triatticus
    Commented Apr 11, 2023 at 18:05

1 Answer 1

0
$\begingroup$

If you are confused by the Einstein summation convention as well as the use of the metric $g_{\mu\nu}$ and its inverse $g^{\mu\nu}$ to move up and down indices, you can write everything without using those. When you do this, each greek letter $\mu,\nu,\ldots$ which appear is an integer labeling the $4$ coordinates on your $4$-dimensional space time.

So, if you call your coordinates $x^0 = t$, $x^1 = x$, $x^2 = y$ and $x^3 = z$, then $\partial_0 \varphi = \frac{\partial \phi}{\partial x^0} = \frac{\partial \phi}{\partial t}$ and likewise for $\mu =1,2,3$. Therefore, we have, for any values of $\mu$ and $\nu$ : $$\frac{\partial}{\partial (\partial_\mu \phi)} \big[\partial_\nu \phi\big] = \delta^\mu_\nu= \left\{\begin{array}{cc} 1 &\text{if } \mu = \nu \\ 0 & \text{if } \mu\neq \nu \end{array}\right.$$

The Klein-Gordon Lagrangian (with $m=0$ for simplicity) is : $$L = \sum_{\mu = 0}^4\sum_{\nu = 0}^4 \frac 12 \frac{\partial \phi}{\partial x^\mu} \frac{\partial \phi}{\partial x^\nu} g^{\mu\nu} $$

If you want to compute $\frac{\partial L}{\partial (\partial_\alpha\varphi)}$, for any given value of $\alpha$, you have : \begin{align} \frac{\partial L}{\partial (\partial_\alpha\varphi)} &= \frac{\partial }{\partial (\partial_\alpha\varphi)}\left[\sum_{\mu = 0}^4\sum_{\nu = 0}^4 \frac 12 \frac{\partial \phi}{\partial x^\mu} \frac{\partial \phi}{\partial x^\nu} g^{\mu\nu}\right] \\ &=\frac 12\sum_{\mu = 0}^4\sum_{\nu = 0}^4 \left( \left[\frac{\partial }{\partial (\partial_\alpha\varphi)}\frac{\partial \phi}{\partial x^\mu}\right] \frac{\partial \phi}{\partial x^\nu} g^{\mu\nu} + \frac{\partial \phi}{\partial x^\mu} \left[\frac{\partial }{\partial (\partial_\alpha\varphi)}\frac{\partial \phi}{\partial x^\nu}\right] g^{\mu\nu}\right)\\ &= \frac 12 \left(\sum_{\nu = 0}^4 \frac{\partial \phi}{\partial x^\nu} g^{\alpha\nu} + \sum_{\mu = 0}^4\frac{\partial \phi}{\partial x^\mu} g^{\mu \alpha}\right) \\ &= \sum_{\mu = 0}^4g^{\alpha \mu}\frac{\partial \phi}{\partial x^\mu} \end{align} where on the last line we used the fact that the metric is symmetric and that we can rename dummy indices.

Rewriting this with the Einstein summation convention we get : $$\frac{\partial L}{\partial (\partial_\alpha \phi)} = g^{\alpha\mu}\partial_\mu\phi$$ and if we raise the indices using the (inverse) metric, we get : $$\frac{\partial L}{\partial (\partial_\alpha \phi)} = \partial^\alpha \phi$$

$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.