# Chebyshev polynomials and Gaussian quadrature

help me please with this question: definition of Chebyshev's polynomials it's given by $T_n(x)=\cos(n\arccos(x))$

1. Find by Gauss Quadrature method $\displaystyle\int_{-1}^1 \sqrt{1-x^2} \; dx$ with null error

2. Estimate $\displaystyle\int_{-1}^1 \frac{e^x}{\sqrt{1-x^2}} \; dx$ accuracy of 5 decimal places

Thanks!

-
How are the Chebyshev polynomials related to your questions? – Antonio Vargas Apr 1 '12 at 21:15
@Antonio : They are used in numerical analysis to find the nodes of interpolation. OP seems confused, hence the confusing question, but a numerical analyst will probably understand the concept. This is behind be, I don't want to get back into it.. – Patrick Da Silva Apr 1 '12 at 21:23
Thanks @Patrick, I agree that the OP seems confused. I wanted to find out exactly where the confusion lies, and my question was meant to prompt some thought about that. – Antonio Vargas Apr 1 '12 at 21:25
What do you mean by "null error"? If you want the error to be $0$, don't use a numerical method. – Robert Israel Apr 1 '12 at 21:37
Ah: Chebyshev-Gauss quadrature. With $n=1$, the second case at en.wikipedia.org/wiki/Chebyshev%E2%80%93Gauss_quadrature gives the exact result for $\int_{-1}^1 \sqrt{1-x^2} f(x)\ dx$ where $f$ is a polynomial of degree $\le 1$. – Robert Israel Apr 2 '12 at 0:11

As Robert mentions, you do not want to use the Chebyshev polynomials of the first kind for your first integral; what you need are the Chebyshev polynomials of the second kind,

$$U_n(x)=\frac{\sin((n+1)\arccos\,x)}{\sqrt{1-x^2}}$$

which are orthogonal with respect to the inner product

$$\langle f,g\rangle=\int_{-1}^1 \sqrt{1-u^2}f(u)g(u)\mathrm du$$

Knowing that $\sin\,k\pi=0$ if $k$ is an integer ought to be a big hint on how to generate the nodes for Gauss-Chebyshev quadrature. From the theory, this quadrature rule is designed to give exact results for integrands of the form $\sqrt{1-x^2}p(x)$, where $p(x)$ is a polynomial, and since constant functions are effectively polynomials...

For the second one: just keep increasing the number of nodes until your error estimate (that I presume was mentioned in your textbook) gives something less than $10^{-5}$; one convenient thing about Gauss-Chebyshev is that the weights stay the same, and all one does is to change the nodes. (P.S. for what values of $u$ is $\cos\,u$ equal to zero?)

-
That may be the fifth unique spelling of Chebyshev I've seen. – Antonio Vargas Apr 3 '12 at 20:00
@Antonio: that, unfortunately, is a typo and not another permutation of old man Pafnuty's surname. I'll fix, and thanks for checking! – J. M. Apr 10 '12 at 3:44