On the Algebraic Complexity of Optimal Polynomial Approximation Constants

Abstract

We investigate the algebraic nature of constants arising from Chebyshev equiripple (minimax) polynomial approximation of $L_p$ norms on $[0,1]$. For the Euclidean case $\sqrt{1+t^2}$, we compute equiripple solutions from degree~1 through~8 and determine exact minimal polynomials and Galois groups for degrees~1 and~2 in both absolute and relative error formulations. We find a sharp phase transition: the degree-1 constants are solvable by radicals (Galois groups $C_4$, $D_4$), while the degree-2 constants provably are not (Galois groups $S_{12}$, $S_{10}\times C_2$). We explain this transition by a structural dichotomy between decoupling and coupling of critical points, and extend the analysis to $L_3$ norms, where the minimal polynomial degree jumps to~246. A general impossibility result follows from Hilbert's irreducibility theorem. We also develop a theory of piecewise equiripple approximation with jointly optimized breakpoints, proving that each doubling of the number of subintervals gains $n+1$ bits of accuracy at no additional arithmetic cost. These results establish a previously unobserved connection between Chebyshev approximation theory and the non-solvability of algebraic equations.