In the book Calculus of variations by I. M. Gelfand, the differential of a functional is defined in the following way, and its uniqueness is proven:

enter image description here

enter image description here

However, it seems strange to me that the expression $\Delta J[h]=\varphi[h]+\epsilon||h||$ shouldn't be written as $\Delta J[h]=\varphi[h]||h||+\epsilon||h||$ instead. As far as I understand, a similar expression holds for the derivative of a function of a single real variable:


where $\Psi(h)$ is a function such that $lim_{h\to 0}\Psi_1(h)=0$. But here there is an $h$ multiplying $f'(a)$, why shouldn't there be a $||h||$ multiplying $\varphi[h]$ in the case of a functional?

And there is something else that doesn't make sense to me either in the proof of Theorem 1. Taking as hypothesis that

$\Delta J[h]=\varphi_1[h]+\epsilon_1||h||$ $\Delta J[h]=\varphi_2[h]+\epsilon_2||h||$

why can we conclude that $\varphi_1[h]-\varphi_2[h]=\epsilon_2||h||$? Wouldn't it be $\epsilon_1-\epsilon_2$ instead of $\epsilon_2$? Maybe there is a typo and they meant to define a new infinitesimal $\epsilon_0=\epsilon_1-\epsilon_2$, or am I interpreting this totally wrong?


In the real variable case, the linear functional is $h \mapsto f'(a) h$. It is not $f'(a)$ itself. Accordingly, in the general case, the linear functional is $h \mapsto \varphi(h)$; $\varphi$ includes both the "$f'(a)$" and the "multiplication by $h$" from the real variable setting.

I agree with you that at the end there is a problem, that $\varepsilon_2$ should not be the same $\varepsilon_2$ from earlier (though this "re-use" of symbols for quantities that we care little about is common in functional analysis).

  • $\begingroup$ Thank you! But the $\varphi[h]$ doesn't literally have $||h||$ as a factor in whatever its expression might be, right? Because then it wouldn't be a linear functional, as the norm of the sum does not equal the sum of the norms... It's just that it is defined that way, isn't it? $\endgroup$ – Wild Feather Feb 4 '18 at 18:15
  • 1
    $\begingroup$ @WildFeather That's right. You can relate this to the notion of directional derivative that you probably learned in multivariable calculus by noticing that for $h \neq 0$ you have $\varphi(h)=\varphi(h/\| h \|) \| h \|$. $\endgroup$ – Ian Feb 4 '18 at 18:17

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.