# A general condition on polynomial roots

1. Mar 12, 2012

### lunogled

Consider a polynomial of the following type:

A_n w^n + A_{n-1} w^{n-1}k + A_{n-2} w^{n-2} k^2 + ... + A_1 k^n =0

What are the general conditions on {A_i} in order for the roots w(k) to be EITHER real OR functions with even imaginary parts, Im[w[k]]=Im[w[-k]]?

I would be interested in whether anyone has ever worked on this problem, or if this problem was ever shown to be unsolvable (perhaps there is a trivial connection to, e.g., Galois theory I missed).

Thanks, and best

2. Mar 12, 2012

### chiro

Hey lunogled and welcome to the forums.

One way to look at this problem is in terms of the derivatives. If you order your turning points in order and you find that a pair of ordered points occurs on the same side of the x-axis, then you will have an imaginary root.

Think of it this way: if you have a real root, then the function will cross the x-axis at some point between the two-turning points and since one turning point will happen at a positive y value and another turning point will happen at a negative y value then if the function is continuous, it has to have a real root because it has to touch the x-axis and not only that, it will touch it only once due to some theorems in analysis/calculus but you can use an intuitive argument if you wish.

So easiest way to tell for a standard y = f(x) function, if you have this kind of condition, you should get an imaginary root if this happens, and to check this you need to solve the information for your first and possibly second derivatives (for inflection points) of your function.

3. Mar 12, 2012

### morphism

I'm confused by your setup. Where do the A_n's live? Is w the indeterminate? If so, why do you have k's in there (i.e. why are you homogenizing)? If w(k) is a root, what does it mean for w(-k) to be a root? Is the k in w(k) a subscript or are you somehow viewing w as a function of k?