Upper limit to the number of real and distinct roots an equation can have.

70 Views Asked by At

An equation is formed using ONLY polynomials (including fractional powers), exponentials and logs. One such equation is shown below; a,b,c,d are real constants. $$ ae^{x+b}-3x^2\left(\ln x^5-c\right)+\left(x^{\frac{6}{d}}-\ln\left(10x-d^8\right)\right)+\frac{ax^{-1}+b}{b^4-c^3}=0\qquad a,b,c,d,x\in\mathbb{R}. $$ Changing the constants $a, b, c, d$ clearly gives rise to a different number of roots and there is obviously an upper limit to how many roots are obtainable in this way.

I am trying to show how the number of roots will have an upper limit (without necessarily finding this limit) but not having any luck. Eg. no matter what $a, b, c, d$ are, the equation above cannot have more than $3$ roots.

I appreciate any pointers, advice and help. Thank you.

1

There are 1 best solutions below

2
On BEST ANSWER

Yes, this is true, except when it's trivially false.

To make this question more precise, let $f(x,y_1,\ldots,y_n)$ be an expression formed by composition of exponentials, natural logarithms and rational functions.

Then there is a number $N$ such that for every parameter tuple $(b_1,\ldots,b_n)\in \mathbf R^n$, if the solution set $f(x,b_1,\ldots,b_n)$ does not contain any interval, then it has at most $N$ points.

Note that it can contain an interval, for example if $f=b_1x/(x-1)-b_2x/(x-1)$, then when $b_1=b_2$, $f=0$ everywhere except $x=1$, where it is undefined. (Note that in this case, $f(x,b_1,b_2)$ is zero as an element of the function field; I think that may be the only way we have an interval of solutions, but I don't quite see the argument right now.)

The reason why this is true is that every $f$ of this form is definable in the exponential field $(\mathbf R, +, \cdot, \exp)$ which is known to be o-minimal by a celebrated result of Wilkie. This means that the solution set of $f(x,\bar y)=0$ (in $\mathbf R^{1+n}$) admits an o-minimal cellular decomposition, and the number of cells in this decomposition gives $N$ (this is a weak bound, in fact a much smaller bound is more typical).

For some more background on o-minimality and cellular decomposition, see e.g. these notes by Sergei Starchenko (I'm afraid the details are beyond the scope of an answer here).