0
$\begingroup$

I have a cubic function: \begin{equation*} h(x)\triangleq \eta+x-\frac{V(\eta-x)^3}{c\eta} \end{equation*} we know that $x\in[0,\eta)$ and all letters are positive and $V>c/\eta$. Hence we know that $h(0)<0$ and $h(\eta)>0$ and $h(x)$ is concave increasing for $x\in[0,\eta)$. So we can infer that there will be exactly one root of $h(x^*)=0$ for $x^*\in(0,\eta)$. Question is: how to characterize the monotonicity of $x^*$ as a function of $\eta$?

$\endgroup$
2
$\begingroup$

Differentiating $h(x, \eta) = 0$ implicitly, we get

$$ \dfrac{dx}{d\eta} = \dfrac{-c\eta^2 + V(2\eta+x)(\eta-x)^2}{\eta (3 V (\mu - x)^2 + c \eta)}$$

The denominator is always positive. On any interval where the numerator doesn't change sign, $x$ is monotone.

For the numerators of $dx/d\eta$ and $h(x,\eta)$ to both be $0$, we need $$ 108 v^2 - 36 v - 1 = 0$$ where $v = V \eta/c$. But the roots of that are approximately $-.0258$ and $0.3591$, so this won't happen if $V \eta/c > 1$.

$\endgroup$
  • $\begingroup$ Could you please elaborate a little bit on how you get $108v^2−36v−1=0$? And so in this case $x$ is monotone increasing in $\eta$? $\endgroup$ – FTXX Sep 28 '16 at 0:24
  • $\begingroup$ The resultant of $c \eta (\eta + x) - V (\eta - x)^3$ and $-c\eta^2 + V (2\eta+x)(\eta-x)^2$ with respect to $x$ is $V^2 \eta^5 c^2 (108 V^2 \eta^2 - 36 V c \eta - c^2) = V^2 \eta^5 c^4 (108 v^2 - 36 v - 1)$ where $v = V\eta/c$. $\endgroup$ – Robert Israel Sep 28 '16 at 20:45
1
$\begingroup$

In reply to the follow-up questions from FTXX, here is an argument. Using Israel's notations, suppose $h(x,\eta)=0$ and $\frac{dx}{d\eta}=0$. That means $$V(\eta-x)^3=c\eta(\eta+x) \tag1$$ and $$-c\eta^2+V(2\eta+x)(\eta-x)^2=0. \tag2$$ Multiply (2) by $\eta-x$ to get $$-c\eta^2(\eta-x)+V(2\eta+x)(\eta-x)^3=0.$$ Now, replace (1) into this last equation: $$-c\eta^2(\eta-x)+c\eta(2\eta+x)(\eta+x)=0.$$ Divide through by $c\eta$ and expand to get the quadratic equation $$\eta^2+4\eta x+x^2=0$$ which is impossible since $x\geq0$ and $\eta>0$.

$\endgroup$

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.