Canonical: https://knowlpedia.com/real-analysis/zero-count-for-distinct-powers/

# Zero count for distinct real powers

A nonzero combination of m distinct powers has at most m minus one distinct positive zeros.

Let \(\alpha_1<\cdots<\alpha_m\) be real. A nonzero [linear combination](https://knowlpedia.com/convex-analysis/linear-combination/) \(f(x)=\sum_{j=1}^m c_jx^{\alpha_j}\) has at most \(m-1\) distinct zeros in \((0,\infty)\).

## Inductive proof

The claim is immediate for one nonzero term. Divide by the smallest power actually present. The resulting function has a nonzero constant term; its derivative is a combination of at most \(m-1\) distinct powers. If the original function had \(m\) distinct positive zeros, Rolle's theorem would give at least \(m-1\) zeros of this derivative, contradicting the inductive bound \(m-2\). A derivative that vanished identically would leave a nonzero constant and hence no zeros.

Consequently the evaluation matrix \([x_j^{\alpha_i}]\) is invertible at distinct positive nodes. Otherwise a nonzero combination would vanish at all \(m\) nodes. This is the generalized-power version of Vandermonde independence.
