6
$\begingroup$

Approximate the solutions of $$\epsilon x^4 + (x-1)^3=0$$

I can't perform a singular perturbation because if I let $\epsilon=0$ then I lose a root. My professor suggests The Method of Dominant Balance for problems like this, but I don't really know how to proceed with that method.

If someone could just get me started, then I could do the rest in maple.

Thanks

$\endgroup$
  • $\begingroup$ Is $\epsilon$ a small quantity? $\endgroup$ – mike Sep 12 '14 at 4:38
  • $\begingroup$ See this. $\endgroup$ – Mhenni Benghorbal Sep 12 '14 at 4:52
12
+100
$\begingroup$

If we naively just send $\epsilon \to 0$ then we get the equation

$$ (x-1)^3 = 0, $$

so we can deduce that we have three roots tending to $x=1$ as $\epsilon \to 0$. We'll suppose they have asymptotic series of the form

$$ x \approx 1 + \sum_{k=1}^{\infty} a_k \delta_k(\epsilon), $$

where

$$ 1 \gg \delta_1(\epsilon) \gg \delta_2(\epsilon) \gg \delta_3(\epsilon) \gg \cdots $$

as $\epsilon \to 0$.

If we substitute just the first two terms $x \approx 1 + a_1 \delta_1(\epsilon)$ into the original equation

$$ \epsilon x^4 + (x-1)^3 = 0 \tag{$*$} $$

and expand it we get

$$ a_1^4\epsilon\delta_1(\epsilon)^4 + 4a_1^3\epsilon\delta_1(\epsilon)^3 + a_1^3\delta_1(\epsilon)^3 + 6a_1^2\epsilon\delta_1(\epsilon)^2 + 4a_1\epsilon\delta_1(\epsilon) + \epsilon \approx 0. $$

Now we'll apply the method of dominant balance. First, certain terms are, by assumption, smaller than others, so they could not possibly be part of a dominant balance and may be ignored. Specifically we know that

$$ a_1^4\epsilon\delta_1(\epsilon)^4 \ll 4a_1^3\epsilon\delta_1(\epsilon)^3 \ll 6a_1^2\epsilon\delta_1(\epsilon)^2 \ll 4a_1\epsilon\delta_1(\epsilon) \ll \epsilon, $$

so if we ignore all but the largest of these then our equation becomes

$$ a_1^3\delta_1(\epsilon)^3 + \epsilon \approx 0. $$

From this we see that

$$ \delta_1(\epsilon) = \epsilon^{1/3} \quad \text{and} \quad a_1^3 = -1, $$

and with three choices for $a_1$---namely the three cube roots of $-1$---we obtain approximations for each of the three roots of the original equation with tend to $x=1$:

  • $x \approx 1 - \epsilon^{1/3}$,
  • $x \approx 1 + e^{i\pi/3} \epsilon^{1/3}$, and
  • $x \approx 1 + e^{i5\pi/3} \epsilon^{1/3}$.

This suggests that we might be able to take

$$ \delta_k(\epsilon) = \epsilon^{k/3}, \quad k \geq 1, $$

and indeed if we substitute

$$ x \approx 1 + \sum_{k=1}^{\infty} a_k \epsilon^{k/3} $$

into $(*)$ and collect like powers of $\epsilon$ then we obtain equations for the coefficients $a_k$,

  • $a_1^3 + 1 = 0$,
  • $3 a_2 a_1^2+4 a_1 = 0$,
  • $3 a_3 a_1^2+6 a_1^2+3 a_2^2 a_1+4 a_2 = 0$, etc.

It just remains to find the last root of $(*)$. If we simply expand it out we get

$$ \epsilon x^4 + x^3 - 3x^2 + 3x - 1 = 0. \tag{$**$} $$

We will again use the method of dominant balance. By ignoring the $\epsilon x^4$ term in the very beginning we essentially assumed it would not be part of a dominant balance, so to find the last root we must assume the opposite. As such we'll be looking for balances between $\epsilon x^4$ and the remaining four terms, so we'll consider the cases

  1. $\epsilon x^4 \asymp 1$,
  2. $\epsilon x^4 \asymp 3x$,
  3. $\epsilon x^4 \asymp 3x^2$, and
  4. $\epsilon x^4 \asymp x^3$.

In case 1 we have $x \asymp \epsilon^{-1/4}$, but then the term $3x$, being $\asymp \epsilon^{-1/4}$, dominates the terms $\epsilon x^4$ and $1$ in the balance, so the balance is not dominant.

In case $2$ we have $x \asymp \epsilon^{-1/3}$, but then the term $-3x^2$, being $\asymp \epsilon^{-2/3}$, dominates both terms in the balance, so the balance is again not dominant.

In case $3$ we have $x \asymp \epsilon^{-1/2}$, but then the term $x^3$, being $\asymp \epsilon^{-3/2}$, dominates both terms in the balance, namely $\epsilon x^4$ and $x^2$, which are $\asymp \epsilon^{-1}$.

In case $4$ we have $x \asymp \epsilon^{-1}$, and this balance is dominant since

$$ \epsilon x^4 \asymp x^3 \asymp \epsilon^{-3} \gg 3x^2 \gg 3x \gg 1. $$

Ignoring the terms outside of the balance we then have

$$ \epsilon x^4 + x^3 \approx 0, $$

so that

$$ x \approx - \epsilon^{-1}. $$

We might then suspect that this root has an asymptotic series of the form

$$ x \approx -\epsilon^{-1} + \sum_{k=0}^{\infty} a_k \epsilon^k, $$

and if we substitute this into $(*)$ or $(**)$ and collect like powers of $\epsilon$ we obtain equations for the coefficients $a_k$,

  • $a_0+3 = 0$,
  • $3 a_0^2+6 a_0-a_1-3 = 0$,
  • $-3 a_0^3-3 a_0^2+6 a_1 a_0+3 a_0+6 a_1-a_2-1 = 0$, etc.

We've thus accounted for all four roots of the polynomial equation and can calculate as many terms of their asymptotic series as we wish.

$\endgroup$
  • $\begingroup$ Further reading: Miller, Asymptotic Methods in Analysis, section 1.5, and Hinch, Perturbation Methods, sections 1.1-1.3. $\endgroup$ – Antonio Vargas Sep 14 '14 at 22:10
-2
$\begingroup$

Let $\epsilon=\tau^3$, then we can solve the equation

$$0=\tau^3 x^4 + (x-1)^3$$

There are 4 roots. When $\tau \to 0$, they are given by:

$$x_1=-3-\frac{1}{\tau^{3}}$$ $$x_2=1-\tau$$ $$x_3=1+\left(\frac{1}{2}-i\frac{1}{\sqrt{3}}\right)\tau$$ $$x_4=1+\left(\frac{1}{2}+i\frac{1}{\sqrt{3}}\right)\tau$$

$\endgroup$

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.