Let $f,g:D \rightarrow \mathbb R$ be multivariable functions over the open set $D$. Assume that $f,g$ both have a global maximum point inside $D$. Can we be sure that $f+g$ also has a global maximum point inside $D$? I know that the maximum point if exists must satisfying $\max(f+g) \leq \max f + \max g$. But will the maximum point exist in general? If not, is there any condition that is needed for the existence of the global maximum?
Is maximum of sum exists if each function has a maximum
analysismaxima-minimamultivariable-calculusoptimization
Best Answer
Let $D$ be a bounded simply connected open set, and $\vec{v}, \vec{w}$ be any two points in $D$. Then define $h$ as $$h:D\rightarrow \mathbb{R}:\vec{x}\mapsto \|\vec{x}-\vec{v}\|\cdot\|\vec{x}-\vec{w}\|$$ This function reaches a maximum on $\overline{D}$. If the set $D$ is sufficiently large, this maximum will be achieved on the boundary, rather than in the local extremum $\frac{\vec{v}+\vec{w}}{2}$. Let this maximal value be $M$, then define $f$ and $g$ in the following way.
$$f(\vec{x}) = \begin{cases} h(\vec{x}) &\text{if } \vec{x}\neq\vec{v}\\ M &\text{if } \vec{x} = \vec{v} \end{cases}$$
And similarly for $g$ and $\vec{w}$. Then the functions $f$ and $g$ will achieve their maximal values in $\vec{v}$ respectively $\vec{w}$, but the sum $f+g$ has a limit point in $\overline{D}$ where the limit is equal to $2M$, which is not achieved in any of the interior points of $D$.
This answer the first question negatively.
To figure out a criterion in which the maximum is achieved, first the open set needs to be simply connected. Otherwise any $f$ and $g$ that satisfy the condition on $D$ where $f+g$ reaches its maximum in $\vec{x}$, will fail to achieve a maximum on $D\setminus \{\vec{x}\}$.
Any kind of differentiability criterion also doesn't work. The example is discontinuous, but by replacing the discontinuity with a sufficiently small peak of the form $\frac{M}{c\|\vec{x}-\vec{v}\|^2+1}$ for some large positive value of $c$ you can make even an analytic counterexample. $$ \begin{align} f &= h + \frac{M}{c\|\vec{x}-\vec{v}\|^2+1}\\ g &= h + \frac{M}{c\|\vec{x}-\vec{w}\|^2+1} \end{align} $$
You can require both $f$ and $g$ to be concave and the set $D$ to be convex, which implies $f+g$ is concave as well, and the function $f+g$ reaches its maximum on an interior point of $D$. This seems to be the simplest criterion, albeit a very restrictive one.