1
$\begingroup$

If $f$ is injective and satisfies the Intermediate theorem, then $f$ must be continuous.
The original problem statement:

Suppose that f satisfies the conclusion of the Intermediate Value Theorem, and that f takes on each value only once. Prove that f is continuous.

Spivak's solution:
Firstly, suppose $f$ were not continuous at $a$, then $$\lim_{x \rightarrow a} f(x) = f(a)$$ is false. Thus, $$\exists \epsilon > 0: \forall \delta > 0: \exists x: |x - a| < \delta \land (f(x) -f(a) \geq \epsilon \lor f(x)-f(a) \leq -\epsilon)$$ So, there are four cases:

  1. $x > a$ and $f(x) - f(a)\geq \epsilon$
  2. $x > a$ and $f(x) - f(a) \leq \epsilon$
  3. $x < a$ and $f(x) - f(a) \geq \epsilon$
  4. $x < a$ and $f(x) - f(a) \leq \epsilon$

Consider $x > a$ and $f(x) \geq f(a) + \epsilon$. It is easy to see that there exists $x'$ between $x$ and $a$ such that $f(x') < f(a) + \epsilon$. Sure. Take any value between $f(a)$ and $f(a) + \epsilon$. It is still between $f(x)$ and $f(a)$. It makes sense to me.

But then, the solution concludes that there exists $y$ between $a$ and $x'$ with $f(y) > f(a) + \epsilon$. How this possible? Where does the $y$ come from?

enter image description here

$\endgroup$
10
  • 1
    $\begingroup$ You claim that "the solution concludes that..." How about writing out the rest of what the solution says, so that we don't have to trust that you've captured the sense of it correctly? BTW, I suspect that the "y" is the "x" in the large exists-forall-exists formula you've written, but I'd feel more confident saying that if I knew the remainder of the solution text. $\endgroup$ Commented Nov 5 at 15:09
  • $\begingroup$ I'm not looking at the text, but something's wrong here. The fucnction $f(x) = x$ satisfies the hypotheses of the problem (and is indeed continuous), but there doesn't exist such a $y$ for $x = a + \epsilon > a$. $\endgroup$
    – anomaly
    Commented Nov 5 at 15:10
  • $\begingroup$ @JohnHughes Edited it to include the original reasoning. $\endgroup$
    – Duck Gia
    Commented Nov 6 at 3:38
  • $\begingroup$ One approach is to first prove that $f$ is strictly monotone from the given conditions. Next monotone nature plus IVT gives continuity in an easy manner. $\endgroup$
    – Paramanand Singh
    Commented Nov 6 at 5:43
  • $\begingroup$ However if we accept the argument of the given solution then the existence of $y$ is coming from the assumption there are values of $x$ greater than $a$ and arbitrarily close to $a$ with $f(x) >f(a) +\epsilon$. One such value of $x$ is $y$ and $a<y<x'$. Instead of symbol $y$ it would be better to use $x''$. $\endgroup$
    – Paramanand Singh
    Commented Nov 6 at 5:49

1 Answer 1

1
$\begingroup$

Let us first analyze the statement $$\lim_{x\to a} f(x) \neq f(a) \tag{1}$$ which says that there exists an $\epsilon >0$ such that for every $\delta>0$ there is at least one corresponding value of $x$ satisfying both $0<|x-a|<\delta$ and $|f(x) - f(a) |\geq \epsilon$.

So we first have an existence of an $\epsilon $ and that means it is a fixed positive number. Next let us concentrate on the phrase "for every $\delta>0$ there is at least one corresponding value of $x$" and that means we can choose any arbitrary value for positive $\delta$ giving rise to infinitely many choices of $\delta$. It may however happen that for all these choices of $\delta$ we may just have a single $x$ which works. NO, this won't happen and in fact we can ensure that for each value of $\delta$ we get get a different corresponding $x$. Let us take one value of $\delta$, say $\delta_1$ arbitrarily and we are ensured a specific value of $x$, say $x_1$ corresponding to it such that $0<|x_1-a|<\delta_1$. Next we choose $\delta_2=|x_1-a|$ and then the corresponding $x_2$ will satisfy $$0<|x_2-a|<\delta_2=|x_1-a|$$ so that $x_2$ is nearer to $a$ than $x_1$ and hence different from $x_1$. Continuing in this fashion we get a sequence $x_n$ of distinct values of $x$ with each term being nearer to $a$ than the previous term. And further the starting value $x_1$ is based on an arbitrary choice $\delta_1$ which can be as small as we desire, it follows that there are infinitely many values of $x$ in every neighborhood of $a$ for which the inequality $|f(x) - f(a) |\geq \epsilon$ holds.

Out of these infinitely many values of $x$ some will satisfy $f(x) \geq f(a) +\epsilon$ and some will satisfy $f(x) \leq f(a) - \epsilon$. And then one of these cases must be satisfied by an infinitely number of values of $x$. And then Spivak handles one case implying that the reader can supply a similar argument for another case. Once a particular case is chosen, say $f(x) \geq f(a) +\epsilon$ we can observe that the infinitely many values of $x$ satisfying it are either less than or greater than $a$ and again there must be infinitely many values of $x$ satisfying at least one of the conditions $x<a$ or $x>a$. Spivak again chooses $x>a$ here and leaves $x<a$ for readers.

Thus Spivak handles the case when there are infinitely many values of $x$ arbitrarily near to and greater than $a$ for which $f(x) \geq f(a) +\epsilon$. Depending on the function one may have to handle remaining three cases as well, but the argument is similar and left for the reader to supply. And in his argument he finds three values of $x$, namely $x, x'$ and $y$ (it would have been better to call it $x''$ instead of $y$) and obtains a contradiction. A similar contradiction can be obtained by choosing three suitable values of $x$ in other remaining three cases as well. You can try other cases and note that argument requires only reversal of inequality signs in appropriate places.


In case this comes up in a test I will personally not require the students to write similar arguments for all four cases when the only difference between these cases is the sign of some inequality. The grading should be done by taking into account how the student uses the injective nature as well as intermediate value property of the function to arrive at the desired conclusion. But then I am not in academic profession and different professors may have different expectations from students in a test.

$\endgroup$
2
  • $\begingroup$ "And then one of these cases must be satisfied by an infinitely number of values of $x$". That is "weird", again. What can't it be one $x$ satisfying the condition $f(x) < f(a) - \epsilon$ and infinite many satisfying $f(x) > f(a) + \epsilon$? There's not contradiction $\endgroup$
    – Duck Gia
    Commented Nov 7 at 4:08
  • $\begingroup$ @DuckGia: check your comment again. It says that there are infinitely many satisfying $f(x) >f(a) +\epsilon $ and this becomes the one case where we have an infinity of values of $x$. Given any function satisfying $\lim_{x\to a} f(x) \neq f(a) $ there is *at least one* case (out of four discussed in the answer) satisfied by infinitely many values of $x$. It may also happen that more than one case is satisfied by infinitely many values of $x$. And the handling of each case is similar. $\endgroup$
    – Paramanand Singh
    Commented Nov 7 at 4:31

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .