6

Consider the continuous functions $f,g:\mathbb{R}\rightarrow\mathbb{R}$.

Show that $F:\mathbb{R}\rightarrow\mathbb{R}$ with $x\mapsto \max\{f(x),g(x)\}$ is continuous using the $\epsilon - \delta$ definition of continuity.

I know there must be four cases.

If $f(x)\leq g(x)$ and $f(x_0)\leq g(x_0)$ or

if $g(x)\leq f(x)$ and $g(x_0)\leq f(x_0)$ it is easy.

However, assuming $f(x_0)\neq g(x_0)$, what if

$g(x)\leq f(x)$ and $f(x_0)\leq g(x_0)$ or

$f(x)\leq g(x)$ and $g(x_0)\leq f(x_0)$?

For example:

$|f(x)-g(x_0)|$... how do I get from here to $|x-x_0|$?

Arthur
  • 1,557
  • Define $d(x) = f(x) - g(x)$. If $f(x_0) < g(x_0)$, then $d(x_0) < 0$. Since $d$ is continuous, it follows that $d(x) < 0$ (and hence $f(x) < g(x)$) for all $x$ in some neighborhood of $x_0$. –  Dec 07 '15 at 21:10
  • @Bungo: Why does $d(x)<0$ follow from the continuity of $d$? – Arthur Dec 07 '15 at 21:15
  • Any reference to this question,like its an excercise of which text? – BAYMAX Sep 29 '17 at 13:55

4 Answers4

13

Hint: The following identity may make the calculation more familiar. $$\max(a,b)=\frac{1}{2}\left(a+b+|a-b|\right).$$

André Nicolas
  • 507,029
5

Given $\epsilon>0$, find $\delta_f$ so that when $|x-x_0|<\delta_f$, $|f(x)-f(x_0)|<\epsilon$. Similarly find $\delta_g$ for $g$.

Then define $\delta = \min(\delta_f,\delta_g)$. Prove this is good enough for $\max(f,g)$.

Thomas Andrews
  • 177,126
4

Trying to stick as close as possible to the definition of the maximum.

Take a point $x_0 \in \mathbb R$. You want to prove that $h(x)=\max (f(x),g(x))$ is continuous at $x_0$.

In fact you only have 3 cases:

  1. $f(x_0) < g(x_0)$
  2. $f(x_0)=g(x_0)$
  3. $f(x_0) > g(x_0)$

$x$ will come into play afterwards.

Case 1. As $g(x_0)-f(x_0) > 0$, you can find $\delta$ such that for $\vert x - x_0 \vert < \delta$ you have $\vert g(x) - g(x_0) \vert < \frac{g(x_0)-f(x_0)}{2}$. Hence for $x \in (x_0-\delta,x_0+\delta)$ you have $h(x)=g(x)$ and $h$ is therefore continuous at $x_0$.

Case 3. Is similar to case 1.

Case 2. $f(x_0)=g(x_0)$

For $\epsilon > 0$, you can find $\delta_f$ such that $\vert f(x)-f(x_0) \vert < \epsilon$ for $\vert x - x_0 \vert < \delta_f$ and $\delta_g$ such that $\vert g(x)-g(x_0) \vert < \epsilon$ for $\vert x - x_0 \vert < \delta_g$.

Now for $\vert x- x_0 \vert < \inf(\delta_f,\delta_g)$, you have $f(x),g(x) \in (f(x_0)-\epsilon,f(x_0)+\epsilon)=(g(x_0)-\epsilon,g(x_0)+\epsilon)$ hence $h(x)=\max(f(x),g(x)) \in (f(x_0)-\epsilon,f(x_0)+\epsilon)$ which allows to conclude.

mathcounterexamples.net
  • 70,018
2

First verify that for general $a,b,c,d\in\mathbb R:$

$$|\max(a,b)-\max(c,d)|<|a-c|+|b-d|.$$

Then apply that to show that, given explicit deltas for $f$ and $g$ separately that satisfy half a given epsilon, the appropriate delta for $\max(f,g)$ is the smallest of the individual deltas.

In general, if $f$ and $g$ are continuous on the same domain, then their cartesian product $f\times g$ is a continuous mapping from that domain to the cartesian product of the original two target sets.

The function "max" from $\mathbb R^2$ to $\mathbb R$ is continuous.

Your function is then the composition of two continuous mappings.

Justpassingby
  • 10,029
  • 1
    Did you read the question? Seeking $\epsilon-\delta$, so probably not allowing results in multi-dimensional calculus. As a rule, when writing an answer that will probably not meet the OPs needs, it is only polite to start with "This is not what you asked for but the (advanced/general/fun/illuminating) way to do this..." – Thomas Andrews Dec 07 '15 at 21:07
  • 1
    This is a nice approach, but may beyond those who are just learning analysis and have not learned how to consider the topology of the product of metric spaces. – Ben Grossmann Dec 07 '15 at 21:07
  • OK I'll give it a try. – Justpassingby Dec 07 '15 at 21:08