I made this question here Show $a e^x=1+x+\frac{x^2}{2}$ has exactly one real root but I wrote the equation wrong. I reproduce the post with the correct equation:

I'm struggling with the following problem:

Show that the equation $a e^x=1+x+\frac{x^2}2$, where $a$ is a positive constant, has exactly one real root.

Looking at the graphs of both $ae^x$ and $1+x+\frac{x^2}2$, they will intersect on the left half of the plane when $a>1$ and on the right half when $a<1$ (because a makes $e^x$ grow faster or slower). So if I could find a point $x_1$ where $a e^{x_1}-(1+x_1+\frac{{x_1}^{2}}2)<0$ and a point $x_2$ where $a e^{x_2}-(1+x_2+\frac{{x_2}^2}2)>0$, by the intermediate value theorem, the equation would have a real root. Also,since the derivative of $a e^x-(1+x+\frac{x^2}2)$ is always positive, I could colclude that there can't exist another real root.

So my question is, how can I find $x_1$ and $x_2$? I can't figure out how to express $x$ in terms of $a$.

  • $\begingroup$ Consider what happens when $|x|$ is very large- but try $x$ both positive and negative. $\endgroup$ – Geoff Robinson Jan 5 '13 at 18:26
  • 2
    $\begingroup$ Note the derivative of $g(x)=ae^x-(1+x+{x^2\over2})$ is always positive for $a>1$; but for $0<a<1$, this is not the case... $\endgroup$ – David Mitra Jan 5 '13 at 18:31
  • $\begingroup$ @DavidMitra You're right, I did not pay attention to that $\endgroup$ – user55286 Jan 5 '13 at 18:35

I will deal with the case $0<a<1$ as the other one is much simpler:

Let $$f(x)=ae^x-1-x-\frac{x^2}{2}$$ Observe $f^{\prime}(x)=ae^x-1-x$, $f^{\prime\prime}(x)=ae^x-1$.

We have $f^{\prime\prime}(x)<0$ for $x<-\ln a$ and $f^{\prime\prime}(x)>0$ for $x>-\ln a$. Therefore, $f^{\prime}$ decreases in $(-\infty,-\ln a]$ and then increases in $[-\ln a,+\infty)$. The minimum is $f^{\prime}(-\ln a)=-a^2+\ln a-1<0$ as $a<1$. Therefore $f^{\prime}$ has only two roots $\xi\in (-\infty,-\ln a)$ and $\eta\in (-\ln a,+\infty)$.

Therefore, $f^{\prime}(x)>0$ in $(-\infty,\xi)$, $f^{\prime}(x)<0$ in $(\xi,\eta)$ and $f^{\prime}(x)>0$ in $(\eta,+\infty)$. This means that $f$ increases in $(-\infty,\xi)$, has a local maximum in $\xi$, decreases in $(\xi,\eta)$, has a local minimum in $\eta$ and finally increases in $(\eta,+\infty)$. Because $f(\xi)<0$ (*), $f$ has only one root in $(\eta,+\infty)$

(*) $f^{\prime}(\xi)=0\implies ae^{\xi}-1-\xi=0$ and so $f(\xi)=\frac{-\xi^2}2<0$ ($\xi\neq 0$ as $a<1$)

  • $\begingroup$ Can (*) follow from $ln(a)<0$ when $a<1$? Also, how do we know $f'$ has two roots? $\endgroup$ – user55286 Jan 5 '13 at 20:50
  • $\begingroup$ @user55286 I hadn't seen that. Yes. $f^{\prime}$ has two roots because it goes from $+\infty$ to $f^{\prime}(-\ln a)<0$ (while strictly decreasing) and then from $f^{\prime}(-\ln a)<0$ to $+\infty$ (while strictly increasing). It is also continuous $\endgroup$ – Nameless Jan 5 '13 at 20:54
  • $\begingroup$ Oh, now I get it! $\endgroup$ – user55286 Jan 5 '13 at 20:55

With $f(x)=a e^x-(1+x+\frac{x^2}2)$ we have $f'(x)=ae^x-(1+x)$, $f''(x)=ae^x-1$ and $f'''(x)=ae^x$. Assume $f(x_1)=f(x_2)=0$ with $x_1<x_2$. Then $f'(\xi)=0$ for some $\xi\in(x_1,x_2)$. From $f'(x_1)=f(x_1)+\frac{x_1^2}2\ge 0$ and $f'(x_2)=f(x_2)+\frac{x_2^2}2\ge 0$, we conclude that $f'$ has a local minimum in the interior of $[x_1,x_2]$, hence $f''(\zeta)=0$ for some $\zeta\in(x_1,x_2)$. This implies $\zeta=-\ln a$. Therefore, $e^{x_1}<\frac1a<e^{x_2}$ and $ x_1+\frac{x_1^2}2<0<x_2+\frac{x_2^2}2$. This implies $0<x_1<2$ and then $x_2>2$. Consequently, $f$ has at most two zeroes (because otherwise two of them would be $\ge 2$ or two of them would be $\le 2$). Also, $f$ can change sign only at its zeroes. We have $f(x)\to \pm\infty$ as $x\to\pm\infty$, hence $f$ changes signs an odd number of times. Therefore, $f$ has either exactly one zero, or it does not change signs at one of the zeroes. But the latter is impossible as ist would imply $f'(x_i)=0$, hence $f'(x_i)-f(x_i)=\frac{x_i^2}2=0$ and finally $x_i=0$, contradicting $0<x_1<2<x_2$.


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.