Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am new to this site so I am not sure if this is the right place to be asking this question but I will try anyway. I am reading an economics paper for my research and the author does the following: $$\frac{\partial}{\partial C_t(j)} \int_0^1 P_t(j) C_t(j) dj = P_t(j)$$

I feel that this derivative is not properly defined but I am probably missing something obvious because the author knew what he was doing. Could someone please tell me if this is a legitimate derivative?


share|cite|improve this question
up vote 4 down vote accepted

There is a mathematically serious notion of a functional derivative, but it requires distributions in order to obtain a rigorous foundation. I'll shoot for a hand-wavy view of the situation.

We define the Dirac delta function $\delta$ to be such that $\displaystyle \int_{-\infty}^{+\infty}\delta(x)f(x)dx=f(0)$. Such a thing doesn't technically exist, but if you play by certain rules you can "act" like it does and remain consistent; you'll have to dive more into the theory of distributions if you want a more satisfying explanation.

Additionally, while a function $f$ maps numbers to numbers typically (for example $\mathbb{R}\to\mathbb{R}$), we introduce what's called a functional, which maps functions to numbers. Now we define a derivative

$$\frac{\delta \mathcal{F}}{\delta \phi}=\lim_{\epsilon\to0}\frac{\mathcal{F}(\phi(\cdot)+\epsilon\,\delta(\;\cdot\;-y))-\mathcal{F}(\phi(\cdot))}{\epsilon}$$

Here $\phi(\cdot)$ is a function, $\mathcal{F}(\;)$ is a functional, $\epsilon$ is restricted to actual numbers, and the derivative here is actually a function of $y$. So if we have a linear functional

$$\mathcal{F}(f)=\int_a^b K(x)f(x)dx$$

then our derivative is given by

$$\lim_{\epsilon\to0}\frac{\int_a^b K(x)[f(x)+\epsilon\,\delta(x-y)]dx-\int_a^b K(x)f(x)dx}{\epsilon}=\lim_{\epsilon\to0}\frac{\int_a^b \epsilon\, K(x)\delta(x-y)dx}{\epsilon}=K(y).$$

(Actually it would evaluate to $0$ if $y$ is outside of $[a,b]$.) This is consistent with what the author wrote, with the caveat that there should be a non-$j$ variable on the right-hand side. But as I said, you will need to go at least knee-deep into the theory of distributions in order to fully make sense of these concepts and see how they are rigorously defined. Hope that helps.

share|cite|improve this answer
Thanks, that was very helpful. – Jonathan Jan 29 '12 at 8:24

That derivative can be properly defined if and only if

there exists an appropriate pair of values $t$ and $j$ such that $\: C_t(j) = 0 \:$


all appropriate values of $t$ such that
[there exists an appropriate value of $j$ such that $\: C_t(j) = 0 \:$]
give the same value of the integral


$0$ is a limit point of the set of values taken by $C_t(j)$ for appropriate values of $t$ and $j$

("appropriate values" are those associated with the definition of $C_t(j)$, and even if the derivative "can be properly defined", it may still fail to exist, as is the case with non-differentiable functions).

If either of the first two fail there is not a unique "base value" to plug into the difference quotient.
If the last one fails, it is not clear what the definition of "limit" would be.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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