I am wondering if there is a mean value theorem for line integral. For example, let $f(x):\mathbb{R}^n\rightarrow \mathbb{R}$ be a continuous (not necessarily monotonic) function defined on smooth curve $C$, do we have the following theorem:

$F(x)=\int_Cf(x)dx=f(x_0)\cdot L_C$

where $x_0$ is some point on the curve $C$ (i.e., $x_0\in C$ ) and $L_C$ is the length of the curve $C$?

In addition, if $f(x)=\nabla F(x)$, do we have $F(a)-F(b)=||f(x_0)||\cdot L_C$, where $||\cdot||$ is some norm (e.g., $\ell_2$ norm)?



The answer is yes for continuous functions $f$ on continuous curves $C$. The reason for this is that the line integral (with respect to arc length) satisfies the inequality

$$ \min(f) \cdot L_C \leq \int_C f(x_1, \ldots x_n) \,ds \leq \max(f) \cdot L_C $$

which can be rearranged as

$$ \min(f) \leq \cfrac{1}{L_C} \int_C f(x_1, \ldots x_n) \,ds \leq \max(f) $$

We know that the minimum and maximum of the continuous function $f$ on $C$ exist because $C$ is compact, being the continuous image of $[0, 1]$. But $C$ is also connected for the same reason, so the range of $f$ must be an interval. The only option is that the range of $f$ must be the interval $[\min(f), \max(f)]$.

This is important because this number $\cfrac{1}{L_C} \int_C f(x_1, \ldots x_n)$ was just shown to be in that interval, and therefore is in the range of $f$. So, pick a value $\vec{c} \in C$ with $f(\vec{c}) = \cfrac{1}{L_C} \int_C f(x_1, \ldots x_n)$ and the theorem is proved.

EDIT: The added part of your question about the integral of a gradient vector field $f(x) = \nabla F(x)$ will have a negative answer. For example, a line integral of a gradient across a loop (initial point = endpoint) will give you zero, but the gradient vector field you are integrating might never be zero.

Example: Integrate the gradient of $F(x, y) = x + y$ along the counterclockwise oriented unit circle.

  • $\begingroup$ Thanks for the answer, @Ben Passer! How about if $f(x)=\nabla F(x)$? Does the statement still hold: $F(a)-F(b)=||\nabla f(x_0)||\cdot L_C$? $\endgroup$ – user3138073 Dec 29 '14 at 4:15
  • $\begingroup$ I think there may have been some confusion. The line integral above is the integral of a real-valued function with respect to arc length, not the integral of a vector field. You should not expect to see mean value theorems for integrals of vector fields. Especially take a look at the right hand side of your edit. That quantity is always a positive number! See the edit to my answer. $\endgroup$ – Ben Passer Dec 29 '14 at 5:35
  • $\begingroup$ you are right. Thanks! $\endgroup$ – user3138073 Dec 29 '14 at 13:36

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.