I am currently studying Calculus of Variations and have come up with this problem.

What functions $f:\Bbb R^n\to\Bbb R$ satisfy $$\sum\limits_i\frac{\partial f}{\partial x_i}=0\tag1$$ with $f$ of differentiability class $C^\infty$?

It can be shown that if the total derivative is zero; that is, if $$\frac{dF}{dt}=\sum_i\frac{\partial f}{\partial x_i}\cdot\frac{dx_i}{dt}=0$$ with $F(t)=f(x_1(t),\cdots,x_n(t))$, then $f$ is constant. However, I cannot see how the same approach can be used for $(1)$.

The trivial function $f(x)=c$ satisfies $(1)$. For even $n>1$, a non-trivial solution is the function $$f(x_1,\cdots,x_n)=\exp\left(\sum_i(-1)^{i+1}x_i\right)$$ since consecutive partial derivatives of $f$ cancel each other out. I suspect there are more solutions that invoke trigonometric expressions.

Is it possible to derive the entire family/families of functions $f$ that satisfy $(1)$?


Note that your equation is equivalent to $$ (1,1,1,\dots,1)\cdot\nabla f=0 $$ This means that $f$ is constant on lines parallel to $(1,1,1,\dots,1)$; these lines are the characteristics of the partial differential equation.

Thus, we can define $f$ freely on $x_n=0$ and then $$ f(x_1,x_2,x_3,\dots,x_n)=f(x_1-x_n,x_2-x_n,x_3-x_n,\dots,0) $$


Define $$y_i:=x_i-x_n$$ for every $i=1,2,\ldots,n-1$, and $$y_n:=x_n\,.$$ Then, we have $$x_i=y_{i}+y_n$$ for each $i=1,2,\ldots,n-1$, and $$x_n=y_n\,.$$ Observe that $$\frac{\partial}{\partial y_n}=\sum_{k=1}^n\,\left(\frac{\partial x_k}{\partial y_n}\right)\,\frac{\partial}{\partial x_k}=\sum_{k=1}^n\,\frac{\partial}{\partial x_k}\,.$$ Hence, the given partial differential equation is equivalent to $$\frac{\partial \phi}{\partial y_n}(y_1,y_2,\ldots,y_{n-1},y_n)=0\,,$$ where $$\phi(y_1,y_2,\ldots,y_{n-1},y_n):=f\left(y_1+y_n,y_2+y_n,\ldots,y_{n-1}+y_n,y_n\right)\,.$$ Consequently, $\phi(y_1,y_2,\ldots,y_{n-1},y_n)$ is a function of $y_1,y_2,\ldots,y_{n-1}$. In other words, there exists a smooth function $\Phi:\mathbb{R}^{n-1}\to \mathbb{R}$ such that $$f\left(x_1,x_2,\ldots,x_{n-1},x_n\right)=\Phi\left(x_1-x_n,x_2-x_n,\ldots,x_{n-1}-x_n\right)\,.$$ We can check that such $f$ satisfies the condition. Your nontrivial example for even $n$ arises from taking $$\Phi(t_1,t_2,\ldots,t_{n-1}):=\exp\left(\sum_{i=1}^{n-1}\,(-1)^{i+1}\,t_i\right)\,.$$

In general, let $\alpha_1,\alpha_2,\ldots,\alpha_n\in\mathbb{R}$ be arbitrary with $\alpha_n\neq 0$. Then, all differentiable functions $f:\mathbb{R}^n\to\mathbb{R}$ that satisfy the partial differential equation $$\sum_{i=1}^n\,\alpha_i\,\frac{\partial f}{\partial x_i}(x_1,x_2,\ldots,x_n)=0$$ for every $x_1,x_2,\ldots,x_n\in\mathbb{R}$ take the form $$f(x_1,x_2,\ldots,x_n)=\Phi\left(\alpha_n\,x_1-\alpha_1\,x_n,\alpha_n\,x_2-\alpha_2\,x_n,\ldots,\alpha_n\,x_{n-1}-\alpha_{n-1}\,x_n\right)\,,$$ where $\Phi:\mathbb{R}^{n-1}\to\mathbb{R}$ is a differentiable function.


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.