Though the following seems intuitive, is it correct precisely?

Suppose $T:\mathbb{R}^2\to\mathbb{R}$ is a smooth function. Then $$\Big({\partial\over\partial y}{\partial T\over\partial x}\Big)(0, y_0) = {d\over dy}\Big({\partial T\over\partial x}(0, y)\Big)(y_0)$$ for all $y_0\in\mathbb{R}$.

What I expect is to "see" the mathematical machinery that gets me from the LHS to the RHS.


Note: I'm a physics undergraduate, who is trying to understand the math (as precisely as possible) that the physics textbooks often sweep under the rug.

Also, to prevent abuse of notation, I think that RHS is better written as $\mathcal{T}'(y_0)$, where $\mathcal{T}:\mathbb{R\to R}$ is a function defined by $\mathcal{T}(y):={\partial T\over\partial x}(0, y)$


The meaning of the LHS is as follows: we first define $f: \Bbb{R}^2 \to \Bbb{R}$ to be the function $f:= \dfrac{\partial T}{\partial x}$. In other words, for every $(\alpha, \beta) \in \Bbb{R}^2$, \begin{align} f(\alpha, \beta) := \dfrac{\partial T}{\partial x}(\alpha, \beta) \end{align} So, we now want to express $\dfrac{\partial f}{\partial y}(0,y_0)$ in a different way. Well, by definition, \begin{align} \dfrac{\partial f}{\partial y}(0,y_0) &:= \dfrac{d}{dy} \bigg|_{y=y_0}\bigg( y \mapsto f(0,y)\bigg) \tag{$*$}\\ &:= \dfrac{d}{dy} \bigg|_{y=y_0}\bigg(y \mapsto \dfrac{\partial T}{\partial x}(0, y) \bigg) \end{align}

So, really, this equality is true by definition of a partial derivative. Now, at this point, you might find my first equality a little suspicious and wonder if it is really true by definition. Well, ok depending on how you look at it, it's either a definition or a simple theorem with a one-line-proof (so really, I just consider it a definition).

You're probably more familiar with this definition:

Let $f: \Bbb{R}^2 \to \Bbb{R}$ be a function, and $(\alpha, \beta) \in \Bbb{R}^2$ be a point. If the limit \begin{align} \lim_{h \to 0} \dfrac{f(\alpha, \beta + h) - f(\alpha, \beta)}{h} \end{align} exists, we denote it as $\dfrac{\partial f}{\partial y}(\alpha, \beta)$, or $(\partial_2f)(\alpha, \beta)$, and call it the partial derivative of $f$ with respect to the second variable, evaluated at the point $(\alpha, \beta)$.

But if you look carefully, what is this definition saying? Well, the existence of that limit is precisely equivalent to the differentiability of the function $y \mapsto f(\alpha, y)$ at the point $\beta$. In other words, the partial derivative (if it exists) is exactly equal to any of the following: \begin{align} \dfrac{\partial f}{\partial y}(\alpha, \beta) &:= \bigg( y \mapsto f(\alpha, y)\bigg)'(\beta) \\ &\equiv \dfrac{d}{dy} \bigg|_{y = \beta} \bigg( y \mapsto f(\alpha, y)\bigg)\\ &= \dfrac{d}{ds} \bigg|_{s = 0} \bigg( s \mapsto f(\alpha, \beta + s)\bigg)\\ &= \dfrac{d}{ds} \bigg|_{s = 0} \bigg( s \mapsto f\left((\alpha, \beta) + s e_2 \right)\bigg) \end{align} where the $\equiv$ means "same thing expressed in different notation", and the third line is true by a simple application of the single variable chain rule (or just look at the limit definition, it's trivially true), and $e_2 = (0,1)$.

BTW, you're perhaps right to introduce a new function $\mathcal{T}$ to avoid any abuse of notation, but I didn't feel like coming up with new names for each new function I defined along the way, which is why I used the stopped arrow $\mapsto$ notation to indicate precisely what the function is, and I was also careful to distinguish between the bound and free variables, and a function vs the function's values at a point so hopefully it's clear exactly how to read the above notation.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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