Consider $f(x)=x^{\frac {5}{7}}$, is it $x=0$ an inflection point? $$f'(x)=\frac {5}{7}x^{\frac {-2}{7}}$$ $$f''(x)=\frac {-10}{49}x^{\frac {-9}{7}}$$ As far as I know, the inflection point is the point in which $f''(x)=0$ or $f''(x)$ does not exist and $f''(x)$ also changes sign at that point.
However, here $f'(0)=+\infty $ from both sides, so it should have verical tangent at point $x=0$.
My calculation shows that $f_+''(0)=-\infty$ and $f_\_''(0)=-\infty$ so the $f''(x)$ changes sign around the point $x=0$. I'm not sure whether I'm correct, please tell me the safe steps to find the inflection point.

  • $\begingroup$ $f$ is not real on the negative $x$-axis. Or do you mean its antisymmetric extension? $\endgroup$ – frog Oct 15 '14 at 6:42
  • $\begingroup$ Of course $f$ is real on negative $x$-axis, since the domain of $f$ is $R$. $\endgroup$ – FreeMind Oct 15 '14 at 6:44
  • $\begingroup$ This is no argument. But I guess I see what you mean: You mean probably $f(x):=\operatorname{sgn}(x)\exp\left( \frac{5}{7}\ln|x|\right)$... $\endgroup$ – frog Oct 15 '14 at 7:00
  • $\begingroup$ @frog Please, read my question carefully, what I'm simply asking is, whether $x=0$ is a point of inflection for $f$ or not and why? $\endgroup$ – FreeMind Oct 15 '14 at 7:33

For an inflexion point, the only thing you need to find is whether concavity changes, i.e. if f"(x)=0, and you specifically test if there is a change in sign in f"(x)=0.

To test if it is a horizontal point of inflexion, f'(x) must also equal 0 at the point where there is an inflection.

This should be enough to test a point of inflexion.

  • $\begingroup$ Could you please answer example? $\endgroup$ – FreeMind Oct 15 '14 at 6:45

I think there is an issue with what is exactly meant by an inflection point. I think it is more convenient to work with the inverse function of $f$ since so the tangent at zero is simply the $x$-axis. (Inflection points are preserved under symmetries of the space).

So you look at $g(x)=x^{7/5}$ and you find $g'(0)=0$. But the second derivative is not defined in $0$. However, $g''(x)>0$ for $x>0$ and $g''(x)<0$ for $x<0$. Hence the curvature changes sign in $0$ but is undefined in $0$.

You can of course define such a point to be an inflection point too, but the usual definition says that $g''$ must vanish in an inflection point. Since it is not even defined here, I would say, $0$ is no inflection point.


I think that some of the confusion comes from what exactly we call an inflection point. Unfortunately, I don't know that any of this terminology is standard. Here is how I would describe it.

By analogy, we call a critical point of a function $f(x)$ a point such that one of the following is satisfied

  • We have that $f'(c) = 0$
  • We have that the derivative of $f$ does not exist at $c$
  • We have that the function is not defined at $c$, but is on both sides.

Then if we are looking for extrema of a function, we search among all of these. However, a critical point may not be an extrema!

Now, for inflection points the case is a little different. The way I define inflection points (and I'm pretty sure that this is how the Stewart texts do) is much simpler. We say that $c$ is an inflection point if

  • $f''(x)$ exists on both sides of $c$ (although not necessarily at $c$!)
  • $f''(x)$ changes signs from one side to the other

So in your case, with the function $f(x) = x^{5/7}$ you get $f''(x) = -\frac{10}{49}x^{-9/7}$. This obviously doesn't exist at 0, but it is well defined away from zero. Moreover, the concavity changes signs from one side to the other. As such, I would call it an inflection point.

It should further be remarked that if $f''(x)$ is defined at $c$ and the concavity changes, then we would of course have $f''(c) = 0$. But this is not necessary, and may not occur if the derivative or second derivative does not exist at that point.


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.