Take the 2-minute tour ×
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.

This is the total differential $$df=dx\frac {\partial f}{\partial x}+dy\frac {\partial f}{\partial y}.$$

How do we take higher orders of total differential, $d^2 f=$?

Suppose I have $f(x,y)$ and I want the second order total differential $d^2f$?

share|improve this question
Product and chain rule just how you did to get the first one. Now imagine that $df$ is a function $g(x,y)$, which is formed by the sum of two products, and get the differential of $g$. –  MyUserIsThis Jun 30 '13 at 14:03
Like $d^2f=dx\wedge(\frac{\partial^2f}{\partial x\partial x}dx+\frac{\partial^2f}{\partial x\partial y}dy)+dy\wedge(\frac{\partial^2f}{\partial y\partial x}dx+\frac{\partial^2f}{\partial y\partial y}dy)=dx\wedge dy(\frac{\partial^2f}{\partial x\partial y}-\frac{\partial^2f}{\partial y\partial x})$? –  awllower Jun 30 '13 at 14:05

3 Answers 3

$d=dx\frac{\partial}{\partial x}+dy\frac{\partial}{\partial y}$ is a differential operator that when applied to $f$ gives $df=dx\frac{\partial f}{\partial x}+dy\frac{\partial f}{\partial y}$

$d^2=\left(dx\frac{\partial}{\partial x}+dy\frac{\partial}{\partial y}\right)\left(dx\frac{\partial}{\partial x}+dy\frac{\partial}{\partial y}\right)$

$d^2f=\left(dx\frac{\partial}{\partial x}+dy\frac{\partial}{\partial y}\right)\left(dx\frac{\partial}{\partial x}+dy\frac{\partial}{\partial y}\right)f=\left(dx^2\frac{\partial^2}{\partial x^2}+dy^2\frac{\partial^2}{\partial y^2}+2dx dy\frac{\partial^2}{\partial x\partial y}\right)f=dx^2\frac{\partial^2 f}{\partial x^2}+dy^2\frac{\partial^2 f}{\partial y^2}+2dx dy\frac{\partial^2 f}{\partial x\partial y}$

share|improve this answer
Do you mean, when $f$ is $C^2$, $d^2f=dx^2\frac{\partial^2 f}{\partial x^2}+dy^2\frac{\partial^2 f}{\partial y^2}+2dx dy\frac{\partial^2 f}{\partial x\partial y}$? Even if so, you must have forgotten that $dxdy+dydx=0$. Regards. –  awllower Jul 1 '13 at 2:41
@awllower You mean $dx dy\wedge dy dx=0$ not $dx dy+dy dx=0$ –  whatever Jul 1 '13 at 7:28
No, I mean $dx\wedge dy+dy\wedge dx=0$, as the wedge product is anti-symmetric. –  awllower Jul 1 '13 at 12:28
Yes, but as you saw, I avoided the use of the wedge product because I think @dan only want to use elementary calculus. –  whatever Jul 1 '13 at 16:34
Then a question arises: what is your definition of $dxdy$? And could you please explain how you obtained the equalities in the answer? Thanks in advance. Otherwise this answer is not so elementary as maybe supposed. –  awllower Jul 2 '13 at 4:20

Edit: I interpret your $d^2 f$ as $d^2f=d(df)$.

If $f$ is $C^2(\Omega)$, with $(x,y)\in\Omega$, then

$$d(df)=dy\wedge dx \frac{\partial^2 f}{\partial y\partial x}+dx\wedge dy \frac{\partial^2 f}{\partial x\partial y}=\left(\frac{\partial^2 f}{\partial x\partial y}- \frac{\partial^2 f}{\partial y\partial x} \right)dx\wedge dy=0,$$

as $dx \wedge dy=-dy\wedge dx$ and $\frac{\partial^2 f}{\partial x\partial y}= \frac{\partial^2 f}{\partial y\partial x}$ on $\Omega$.

If $f$ is not $C^2$ at $(x,y)$, then we arrive at

$$d(df)=\left(\frac{\partial^2 f}{\partial x\partial y}- \frac{\partial^2 f}{\partial y\partial x} \right)dx\wedge dy. $$

share|improve this answer
But OP did not assume that $f\in C^2$, so that the result is not necessarily $0$, right? –  awllower Jun 30 '13 at 14:22
@awlower, right. I used the Schwarz lemma. If $f$ is not $C^2$ then we have just the first equality. –  Avitus Jun 30 '13 at 16:19
@awlower I add this comment in my answer. –  Avitus Jun 30 '13 at 16:48

I will assume that you are referring to the Fréchet derivative. If $U\subseteq\mathbb{R}^n$ is an open set and we have functions $\omega_{j_1,\dots,j_p}:U\to\mathbb{R}$, then $$ D\left(\sum_{j_1,\dots,j_p} \omega_{j_1,\dots,j_p} dx_{j_1}\otimes\cdots\otimes dx_{j_p}\right) = \sum_{j_1,\dots,j_p}\sum_{j=1}^n \frac{\partial\omega_{j_1,\dots,j_p}}{\partial x_{j}} dx_j\otimes dx_{j_1}\otimes\cdots\otimes dx_{j_p}. $$

Here $dx_{i}$ is the projection onto the $i$th coordinate, and if $\alpha,\beta$ are multilinear forms then $\alpha\otimes\beta$ is the multilinear form defined by $(\alpha\otimes\beta)(x,y)=\alpha(x)\beta(y)$.

For example, let $f(x,y)=x^3+x^2 y^2+y^3$. Then \begin{align} Df&=(3x^2+2xy^2)dx+(2x^2y+3y^2)dy; \\ D^2f&=(6x+2y^2)dx\otimes dx+4xy(dx\otimes dy+dy\otimes dx)+(2x^2+6y)dy\otimes dy; \\ D^3f&=6dx\otimes dx\otimes dx+4y(dx\otimes dx\otimes dy+dx\otimes dy\otimes dx+dy\otimes dx\otimes dx)\\ &\qquad+4x(dx\otimes dy\otimes dy+dy\otimes dx\otimes dy+dy\otimes dy\otimes dx) \\ &\qquad+6dy\otimes dy\otimes dy. \end{align}

Since $D^p f(x)$ is always a symmetric multilinear map if $f$ is of class $C^p$, you might want to simplify the above by using the symmetric product (of tensors).

share|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.