If $f(x)=e^{-x}x^n$, $n\in \mathbb{Z}$, find the values of $n$ such that $f(x)$ has $0,1, 2$ or $3$ inflection points. (That is, give the set of $n$ for each possible number of inflection points (0,1,2,3).)

This is a question that I found in a past exam paper for a math subject in Australia.

The previous parts of the question ask to find $f''(x)$ and to find $x: f''(x)=0, x\neq 0$

The answers are $f''(x)= e^{-x} x^{n - 2} (n^2 - n (2 x + 1) + x^2)$, and for the next part, $x=n\pm \sqrt{n}$ ($\neq0$).

So for the question I've asked, I actually did solve it, but in a fairly long winded way. As an inflection point is defined if the following conditions are met: $f''(x)=0, \frac{f''(x+a)}{f''(x-a)}<0$, (the last condition saying that the value of $f''(x)$ in the neighbourhood of the x-intercept of $f''(x)$ are opposite signs immediately to the left and right of it.), I did the following:

For each of the factors in $f''(x)$, consider when they would satisfy the alternating sign condition, then I essentially intersected each set of n-values to arrive at the answer.

0 POIs: $n\in \mathbb{Z}^-\cup \{0\}$

1 POI: $n=1$

2 POIs: $n=2k, k\in \mathbb{Z}^+$

3 POIs: $n=2k+1, k\in \mathbb{Z}^+$

The problem though is that this took too long, and the question was only worth 2 marks which really means that it shouldn't take more than a few minutes to figure out.

My question: is there are nicer less complex way of solving this? My method involved intersecting a bunch of sets and considering certain 'edge' cases, which doesn't seem very efficient at all.

  • $\begingroup$ Note that if the first part specifically asked for no-zero roots, then the answer $n\pm\sqrt n$ sometimes (namely, for $n=0,1$) wrongly includes the to be excluded $x=0$. $\endgroup$ – Hagen von Eitzen Apr 14 at 10:19

Let's start from where you set the second derivative equal to zero: $e^xx^{n-2}(x^2 - 2nx + (n^2 - n)) = 0$. Notice that we can divide both sides by the $e^x$ because it's never zero, so $x^{n-2}(x^2 - 2nx + (n^2 - n)) = 0$, and now our equation is polynomial.

Now we can use one of the properties of the roots of a polynomial, which is that if we have a root with an even multiplicity, then the polynomial will have the same sign on either side, but if we have a root with an odd multiplicity, then the polynomial will have opposite signs on either side. So, we want to look for roots with odd multiplicities.

For the first term, $x^{n-2}$, if $n > 2$ we will have a root at $x = 0$. If $n$ is even, this will have even multiplicity, otherwise it will have odd multiplicity.

Let's look at the first term: $x^{n-2} = 0$. When $n \leq 2$, this trivially has no solutions. Otherwise, we get a root at $x = 0$ with even multiplicity if $n$ is even, otherwise odd multiplicity if $n$ is odd.

Now let's look at the second term: $x^2 - 2nx + (n^2 - n) = 0$. Let's look at the discriminant, which is the expression under the square root in the quadratic formula: $(-2n)^2 - 4(1)(n^2 - n) = n$. So when $n < 0$, we get no real roots, when $n = 0$ we get one double root, and when $n > 0$ we get two distinct real roots with multiplicity $1$.

There is a potential concern here with whether the roots from the two terms will match up, so let's consider if $x = 0$ is ever a root of the second term. If $x = 0$ is a root of the second term, then we must have $n^2 - n = 0$, so $n$ is $0$ or $1$. But when $n$ is $0$ or $1$, our expression is undefined at $x = 0$ due to division by zero, so we have to disregard this root.

So, for $n \leq 0$ we have no roots, and therefore no points of inflection.

When $n = 1$, the first term has no roots and the second term gives two roots with multiplicity $1$, but we have to ignore the root at $x = 0$ so we get one point of inflectin.

For $n \geq 2$, the second term gives us $2$ points of inflection for every value of $n$, and the first gives one additional point of inflection if $n$ is odd. This matches your answer.

  • $\begingroup$ The expression under the square root in the quadratic formula is called the "discriminant" , no "determinant". $\endgroup$ – jjagmath Apr 14 at 10:44
  • $\begingroup$ I did say determinant didn't I. My bad, I'll fix that now $\endgroup$ – Stephen Donovan Apr 14 at 10:45

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.