1

Can anyone tell me of a general form of a quartic function that has one extremum and two inflection points?

I am looking for a general form, not a specific function.

Edit: It doesn't have to be as general as possible, actually; just a form which generates solutions. Ideally with one variable other than the variable of the polynomial.

Arj
  • 35

2 Answers2

2

Completely general form

First let's find the general form of all such quartics. Let $Q(x)$ be the quartic polynomial in question. Then $Q'(x)$ must have only one real root, and $Q''(x)$ must have two real roots. We are looking for the conditions under which this is true.

Let's start with $$ Q''(x) = a(x-r)(x-s) = ax^2 - a(r+s)x + a(rs) $$ for real $a,r,s$, $r \ne s$, $a \ne 0$.

Now we can let $$ Q'(x) = \frac13 ax^3 - \frac{1}{2} a (r+s)x^2 + a(rs)x + c, $$ which is a cubic whose derivative has two changes of sign, so there is both a strict local max and a strict local min, i.e. two "turning points". The cubic $Q'(x)$ thus has only one real root iff the values of $Q'(r)$ and $Q'(s)$ have the same sign (see this answer). Finally we integrate one more timme, $$ Q(x) = \frac{1}{12} ax^4 - \frac{1}{6}a(r+s)x^3 + \frac{1}{2} a(rs)x^2 + cx + d, $$ and there are no additional conitions on $d$. Therefore, the entire solution set is given as follows:

Complete solution set:

All polynomials $$ Q(x) = \frac{1}{12} ax^4 - \frac{1}{6}a(r+s)x^3 + \frac{1}{2} a(rs)x^2 + cx + d $$ where $$ a,r,s,c,d \in \mathbb{R} $$ and $$ a \ne 0, r \ne s $$ and $$ Q'(r) Q'(s) > 0 $$ i.e. $Q'(r)$ and $Q'(s)$ have the same sign.


Response to edit

Since you've now indicated you prefer a formula (not necessarily completely general) for quartics with only one variable other than $x$, note that we can use the above general form to get many such forms:

  • The easiest thing is to vary the parameters $a$ or $d$, which don't matter. For instance we can set $r = -1$, $s = 1$, $c = 36$, $a = 12$ generating the set of solutions

    $$x^4 - 6x^2 + 12x + d\quad \text{ for all } d \in \mathbb{R}$$

    And setting $r = -1$, $s = 1$, $d = 0$, $c = a$ similarly gives the set of solutions (after scaling by $12$)

    $$ax^4 - 6ax^2 + 12ax\quad \text{ for all } a \ne 0$$

  • These solution sets seem a bit trivial, though, so fix $a = 12$ and $d = 0$. We're left with $$ Q(x) = x^4 - 2(r+s)x^3 + 6(rs)x^2 + cx $$

    We may now take $r = 0$, $s > 0$; $c = Q'(0)$ has to be the same sign as $Q'(s) = 4s^3 - 6s^3 + c$, which equals $c - 2s^3$. We can guarantee this e.g. if $c = -1$. So another solution set is

    $$Q(x) = x^4 - 2sx^3 - x \quad \text{ for all } s > 0.$$

  • If we want to vary $c$ instead, let's take $r = 0$, $s = 1$, so that $Q'(r) = c$, $Q'(s) = c - 2$. This will work for example for any $c < 0$. So we can have (switching sign of $c$)

    $$Q(x) = x^4 - 2x^3 - cx \quad\text{ for all } c > 0.$$

There are many other such forms we could come up with, but this should give you something to work with. The completely general form can easily be used in this way to give any number of specific forms with only one varying parameter.

  • @Arj OK. I updated your question to better suit what you want. It can't be changed completely, because your question was already answered in its original form, but I specified that you were ideally looking for a general form with only one parameter other than the variable of the polynomial. – Caleb Stanford Aug 06 '16 at 21:14
  • @Arj I added four (!) options for a specific form with only one variable other than $x$. – Caleb Stanford Aug 06 '16 at 21:45
  • Yes, this is exactly what I was looking for, thank you so much.

    Also, would you be able to mention a few more forms like the first set? By that, I mean, forms similar to $\ x^4-6x^2+12x+d$ since it has one extremum and two points of inflection. All other sets you mentioned either have more than one extremum or no inflection points. I am specifically looking for forms with one extremum and two points of inflection, so if you could mention one or two more, that would be great.

    Thank you.

    – Arj Aug 07 '16 at 18:38
  • 1
    @Arj I'm sorry, I did understand the question (one extremum and two inflection points) but there was a calculation error. Answer should be correct now, with all four equations at the end generating the kind of quartic you want. – Caleb Stanford Aug 08 '16 at 05:16
  • No worries mate. Thank you so much for your help. – Arj Aug 09 '16 at 09:53
  • @Arj you're welcome. By the way, it is courteous to upvote helpful answers and also to accept the answer that best solved your question. – Caleb Stanford Aug 09 '16 at 10:14
  • 1
    Done! Sorry, a little new here, so don't know how things work. Thanks once again. =) – Arj Aug 14 '16 at 01:54
1

An extremum at point $x_k$ has first derivative zero: $$f'(x_k) = 0$$ An inflection point is a point $x_k$ where $$f''(x_k) = 0$$

Given you know three such points, $x_1,x_2,x_3$ you can build a system of equations:

$$\cases{f'(x_1) = 0\\f''(x_2) = 0\\f''(x_3) = 0}$$ Where the coefficients $c_l$ are what we are solving for: $$\cases{f(x) = \sum_{l=0}^4c_lx^l\\\\ f'(x) = \sum_{l=0}^4l\cdot c_lx^{l-1}\\\\f''(x) = \sum_{l=0}^4l\cdot(l-1)\cdot c_lx^{l-2}}$$

Since differentiation is linear on polynomials this will be a homogeneous linear equation system with three equations and five unknowns. Unless conditions create a degenerate system there will be a linear subspace of dimensionality $5-3=2$ (why?) of coefficients which fulfill the criteria.

mathreadler
  • 25,824
  • The general form will be the solutions to the equation system. I don't know how to express it easier than that without removing components crucial for the understanding of the problem. – mathreadler Aug 06 '16 at 20:50