10

I need to find out the maximum possible number of real roots of the equation:

$$e^x=mx^2$$

where m is a real parameter.

I'm interested in some easy approaches. Moreover, is it possible to solve it without using derivatives at all? Thanks.

user 1591719
  • 44,216
  • 12
  • 105
  • 255
  • Probably easiest to use Rolle's Theorem. – Gerry Myerson Jun 21 '12 at 09:57
  • @Gerry Myerson: right. This is a good option. – user 1591719 Jun 21 '12 at 09:58
  • It seems highly related to monotonicity. I believe that the growth rate must be used, and this is naturally linked to derivatives. – Siminore Jun 21 '12 at 10:06
  • Here's what WA gives for value 1.8473 and for 1.8472. It seems to be an interesting questions whether there is $m$ such that there are 2 real solutions and what is the threshold value between 1 and 3 solutions. – Martin Sleziak Jun 21 '12 at 10:29
  • 3
    A bit of easy calculus shows that taking $m = e^2/4$ gives the critical case where the curve $y=mx^2$ meets the curve $y=e^x$ at a double point, but I wouldn't think it likely that it can be found without calculus. – Old John Jun 21 '12 at 10:33
  • @OldJohn Did you get $e^2/4$ like this? We want $e^x=mx^2$ (it is intersection) and $e^x=2mx$ (the slope is the same.) This leads to $mx^2=2mx$ and $x=2$. Hence we have $e^2=m2^2$, i.e. $m=e^2/4$. – Martin Sleziak Jun 21 '12 at 10:41
  • Yes, that is basically the approach I used, but it makes non-trivial use of calculus, as we use the derivative of $e^x$. – Old John Jun 21 '12 at 10:45

4 Answers4

3

After the umpteenth edit I decided to rewrite the solution in a concise way.

Consider the function $f:\mathbb{R} \setminus \{0\}\to \mathbb R$ with $f(x)=\frac{e^x}{x^2}$. Then the number of solutions of the equation $e^x=mx^2$ for a given $m$ is precisely the number of elements in $f^{-1}(m)$. Since $f$ is strictly positive there is no solution for $m\leq 0$. Since the restriction $f:\mathbb R^-\to \mathbb R^+$ is bijective, there is always one negative solution. And since the restriction $f:\mathbb R^+\to \mathbb R^+$ is convex with $f(x)\to \infty$ for $x\to 0^+$ and $x\to \infty$ there are either $0$, $1$ or $2$ positive solutions.

It is easy to see that $f$ restricted to $\mathbb R^+$ is convex using calculus as the second derivative is $\frac{e^x(x^2-4x+6)}{x^4}>0$ but it should also be possible to show this by foot.

Siminore
  • 35,136
Simon Markett
  • 10,636
  • 1
  • 22
  • 32
2

If $m>0$, there would always be a solution for $x<0$. If m is sufficiently large such that $m*x^2$ overcomes $e^x$ for some $x>0$ then $e^x$ will meet $m*x^2$ again as $e^x$ grows faster than any polynomial giving total of $3$ solutions. If $m$ is such that when $e^x$ meets $m*x^2$ for some $x>0$ and their slopes are equal at that point, then there would be $2$ solutions.If $m>0$ is small, curves would intersect only for some $x<0$ giving only $1$ solution while for $m<0$, they would never intersect giving no real solutions.

Aang
  • 14,672
1

It is not completely elementary, but you could try to change variables: $\mathrm{e}^x=t$, so that your equation becomes $$\frac{1}{m}t = \left( \log t \right)^2 .$$ The behavior of the right-hand side should be guessed easily (monotonicity, convexity, and so on). The advantage is that you can probably exploit convexity. But I believe that some differentiation is anyway needed.

Siminore
  • 35,136
1

Provided $m > 0$ there is always a negative solution. We now turn our attention to solutions on $(0,\infty)$.

Put $$g(x) = {e^x\over x^2}$$ for $x > 0$. Differentiating, you get $$g'(x) = {(x-2)e^x\over x^3}.$$ If you draw the sign chart for g', you will see it is negative if $x < 2$ and positive if $x > 2$. Therefore there is a global minimum at $x = 2$. We conclude that $${e^x\over x} \ge {e^2\over 4}$$ for $x > 0$.

So, if $m < e^2/4$, no solution on the $(0,\infty)$ exists. If $m > e^2/4$, two solutions exist, one to the left of 2 and one to the right. If $m = e^2/4$, there is one solution.

ncmathsadist
  • 49,383