# Jacobi polynomials

Orthogonal polynomials on the interval $[- 1, 1]$ with the weight function

$$h ( x) = ( 1 - x) ^ \alpha ( 1 + x) ^ \beta ,\ \ \alpha , \beta > - 1,\ \ x \in [- 1, 1].$$

The standardized Jacobi polynomials are defined by the Rodrigues formula:

$$P _ {n} ( x; \alpha , \beta ) = \ P _ {n} ^ {( \alpha , \beta ) } ( x) =$$

$$= \ \frac{(- 1) ^ {n} }{n! 2 ^ {n} } ( 1 - x) ^ {- \alpha } ( 1 + x) ^ {- \beta } \frac{d ^ {n} }{dx ^ {n} } [( 1 - x) ^ \alpha ( 1 + x ) ^ \beta ( 1 - x ^ {2} ) ^ {n} ],$$

and the orthonormal Jacobi polynomials have the form

$$\widehat{P} _ {n} ( x; \alpha , \beta ) =$$

$$= \ \sqrt { \frac{n! ( \alpha + \beta + 2n + 1) \Gamma ( \alpha + \beta + n + 1) }{2 ^ {\alpha + \beta + 1 } \Gamma ( \alpha + n + 1) \Gamma ( \beta + n + 1) } } P _ {n} ( x; \alpha , \beta ).$$

The polynomial $P _ {n} ( x; \alpha , \beta )$ satisfies the differential equation

$$( 1 - x ^ {2} ) y ^ {\prime\prime} + [ \beta - \alpha - ( \alpha + \beta + 2) x] y ^ \prime + n ( n + \alpha + \beta + 1) y = 0.$$

When $\alpha \geq - 1/2$ and $\beta \geq - 1/2$, the orthonormal Jacobi polynomials satisfy the following weighted estimate:

$$( 1 - x) ^ {( 2 \alpha + 1)/4 } ( 1 + x) ^ {( 2 \beta + 1)/4 } | \widehat{P} _ {n} ( x; \alpha , \beta ) | \leq c _ {1} ,$$

$$x \in [- 1, 1],$$

where the constant $c _ {1}$ does not depend on $n$ and $x$. At $x = \pm 1$ the sequence $\{ \widehat{P} _ {n} ( x; \alpha , \beta ) \}$ grows at a rate $n ^ {\alpha + 1/2 }$ and $n ^ {\beta + 1/2 }$, respectively.

Fourier series in Jacobi polynomials (cf. Fourier series in orthogonal polynomials) inside the interval $(- 1, 1)$ are similar to trigonometric Fourier series. But in neighbourhoods of the end points of this interval, the orthogonality properties of Fourier–Jacobi series are different, because at $x = \pm 1$ the orthonormal Jacobi polynomials grow unboundedly. The Fourier–Jacobi series of a function $f$ is uniformly convergent on $[- 1, 1]$ if $f$ is $p$ times continuously differentiable on this segment and $f ^ { ( p) } \in \mathop{\rm Lip} \gamma$ with $p + \gamma > q + 1/2$, where

$$q = \max \{ \alpha , \beta \} > - { \frac{1}{2} } .$$

Under these conditions the following inequality holds:

$$\left | f ( x) - \sum _ {k = 0 } ^ { n } a _ {k} \widehat{P} _ {k} ( x; \alpha , \beta ) \right | \leq \ \frac{c _ {2} }{n ^ {p + \gamma } } n ^ {( 2q + 1)/2 } ,$$

$$x \in [- 1, 1],$$

where the constant $c _ {2}$ does not depend on $n$ and $x$. On the other hand, when $\alpha \geq - 1/2$ and $\beta \geq - 1/2$, the remainder in the Fourier–Jacobi series for $f$ satisfies the following weighted estimate:

$$( 1 - x ^ {2} ) ^ {1/4} \sqrt {h ( x) } \left | f ( x) - \sum _ {k = 0 } ^ { n } a _ {k} \widehat{P} _ {k} ( x; \alpha , \beta ) \right | \leq$$

$$\leq \ c _ {3} E _ {n} ( f ) \mathop{\rm ln} n,\ \ x \in [- 1, 1],$$

where $n \geq 2$, the constant $c _ {3}$ does not depend on $n$ and $x$, and $E _ {n} ( f )$ is the best uniform approximation error (cf. Best approximation) of the continuous function $f$ on $[- 1, 1]$ by polynomials of degree not exceeding $n$.

The Jacobi polynomials were introduced by C.G.J. Jacobi  in connection with the solution of the hypergeometric equation. Special cases of the Jacobi polynomials are: the Legendre polynomials (when $\alpha = \beta = 0$); the Chebyshev polynomials of the first kind (when $\alpha = \beta = - 1/2$); the Chebyshev polynomials of the second kind (when $\alpha = \beta = 1/2$); and the ultraspherical polynomials (when $\alpha = \beta$).