Maths WorkshopMaths problems with full solutions
Exercise 2075Calculus › Single-variable calculus › Differentiation › Mean Value TheoremL2Added 2 Nov 2026

Legendre polynomials have all their roots in (-1, 1)

#Rolle's theorem#polynomials#induction#Legendre

Statement

Let n≥1n\ge1 and Qk=dkdxk((x2−1)n)Q_k=\dfrac{d^k}{dx^k}\bigl((x^2-1)^n\bigr). Prove that QnQ_n (the polynomial that defines the Legendre polynomial PnP_n up to a constant) has nn distinct real roots, all in (−1,1)(-1,1).

Hint

Show by induction on k≤nk\le n that Qk=(x2−1)n−kRkQ_k=(x^2-1)^{n-k}R_k where RkR_k has kk distinct roots in (−1,1)(-1,1). Rolle's theorem creates one new root between two consecutive roots.

Show solutionHide solution

Solution

Claim. For 0≤k≤n0\le k\le n, Qk=(x2−1)n−kRkQ_k=(x^2-1)^{n-k}R_k, where RkR_k is a polynomial of degree kk with kk distinct roots in (−1,1)(-1,1).

Degree check. QkQ_k has degree 2n−k2n-k, so deg⁡Rk=2n−k−2(n−k)=k\deg R_k=2n-k-2(n-k)=k. Hence the kk roots are all the roots of RkR_k, and they are simple.

Start. k=0k=0: Q0=(x2−1)nQ_0=(x^2-1)^n and R0=1R_0=1 has 00 roots.

Inductive step. Suppose the claim holds for some k<nk<n. Then Qk=(x−1)n−k(x+1)n−kRkQ_k=(x-1)^{n-k}(x+1)^{n-k}R_k vanishes at the k+2k+2 distinct points
−1<r1<⋯<rk<1,-1<r_1<\dots<r_k<1,
because n−k≥1n-k\ge1. By Rolle's theorem applied to QkQ_k on each of the k+1k+1 intervals between consecutive points, Qk+1=Qk′Q_{k+1}=Q_k' vanishes at some point in each of them: that gives k+1k+1 distinct roots in (−1,1)(-1,1).

Moreover, differentiating a product (x2−1)mR(x^2-1)^{m}R with m=n−k≥1m=n-k\ge1 leaves a factor (x2−1)m−1(x^2-1)^{m-1}, so Qk+1=(x2−1)n−k−1Rk+1Q_{k+1}=(x^2-1)^{n-k-1}R_{k+1} for a polynomial Rk+1R_{k+1}. By the degree count, Rk+1R_{k+1} has degree k+1k+1, so the k+1k+1 roots found are all its roots.

Conclusion. For k=nk=n, Qn=RnQ_n=R_n has nn distinct roots in (−1,1)(-1,1).

Remark. A degree nn polynomial cannot have more than nn roots, so these are all of them. This is the key property of the Legendre polynomials in numerical integration (Gauss quadrature).