Definition
Let
. The Chebychev polynomial of degree
is the map defined as follows:
![T_n:[-1,1] \rightarrow \mathbf{R},\quad x \longmapsto \cos (n\mbox{Arccos } x)
T_n:[-1,1] \rightarrow \mathbf{R},\quad x \longmapsto \cos (n\mbox{Arccos } x)](local/cache-vignettes/L265xH31/cb3a2f61e1312c46cf97078b2551bf13-4bd24.png)
Explicit computation of Chebychev Polynomials
Let
and
.
![\begin{array}{ccl}
T_n(x)&=&\Re((\cos n\alpha +i\sin n\alpha))\\
&=&\Re((\cos \alpha +i\sin \alpha)^n)\\
&=&\Re((\cos \mbox{Arccos } x +i\sin \mbox{Arccos } x)^n)\\
&=&\Re(( x +i\sqrt{1-x^2})^n)\\
&=&\Re(\displaystyle\sum_{k=0}^n C_{n}^{k}x^{n-k} i^{k}(\sqrt{1-x^2})^k)\\
&=&\displaystyle\sum_{k=0}^{E[n/2]} C_{n}^{2k}x^{n-2k} (i)^{2k}(1-x^2)^k\\
&=&\displaystyle\sum_{k=0}^{E[n/2]} C_{n}^{2k}x^{n-2k}(x^2-1)^k
\end{array}
\begin{array}{ccl}
T_n(x)&=&\Re((\cos n\alpha +i\sin n\alpha))\\
&=&\Re((\cos \alpha +i\sin \alpha)^n)\\
&=&\Re((\cos \mbox{Arccos } x +i\sin \mbox{Arccos } x)^n)\\
&=&\Re(( x +i\sqrt{1-x^2})^n)\\
&=&\Re(\displaystyle\sum_{k=0}^n C_{n}^{k}x^{n-k} i^{k}(\sqrt{1-x^2})^k)\\
&=&\displaystyle\sum_{k=0}^{E[n/2]} C_{n}^{2k}x^{n-2k} (i)^{2k}(1-x^2)^k\\
&=&\displaystyle\sum_{k=0}^{E[n/2]} C_{n}^{2k}x^{n-2k}(x^2-1)^k
\end{array}](local/cache-vignettes/L301xH230/938d66d3366488012ba407fc72f68251-3427c.png)
where

is the floor function and

is the real part.
The following table gives the first Chebychev polynomials:

Recurrence relation between Chebychev polynomials
Proposition.
,
and for any number

Proof. Let
![x\in[-1,1] x\in[-1,1]](local/cache-vignettes/L70xH31/3fc4488933571bb861da5bf45c51cbc8-bc3b3.png)
and

. By means of trigonometry formulae, we have the following two equalities:


Adding these two equalities, one obtains:

Therefore, for any

in
![[-1,1] [-1,1]](local/cache-vignettes/L44xH31/d060b17b29e0dae91a1cac23ea62281a-4c645.png)
:

Proposition.The coefficient of the

-degree term of

is

.
Proof. We shall proceed by induction. The coefficient of
is
. Assume that the coefficient of the
-degree term of
is
. Then, given the previous recurrence relation, one can see that the coefficient of the
-degree term is twice that of
(
. Thus,
.
Orthogonality of Chebychev polynomials
Chebychev polynomials are pairwise
-orthogonal; that is, they are orthogonal with regard to a weighted function defined by:

In particular:

Proof. To compute the previous integral, we use the following substitution:

We thus have:

The conclusion is therefore obvious.
Roots of Chebychev polynomials
Proposition. Let
.
has exactly
simple roots defined as follows:

Proof. Let

and

:

Since

has degree

, the

are precisely the roots of

. They are simple roots since for all

, we have
Extrema of Chebychev polynomials
Proposition. Let
.
has exactly
extrema defined by:

Proof. Let

and

. The first derivative of

is:
![\forall x\in [-1,1],T'_n(x)=\frac{n}{\sqrt{1-x^2}}\sin (n\mbox{Arccos } x)
\forall x\in [-1,1],T'_n(x)=\frac{n}{\sqrt{1-x^2}}\sin (n\mbox{Arccos } x)](local/cache-vignettes/L293xH43/9eeaac61f71eff0f4d5af46a2018d790-b344d.png)
Therefore:

Proposition.

Proof.

The maximum reached by Chebychev polynomials
A forthwith consequence of the previous propositions is that:
![\max_{[-1,1]} |T_n(x)|=1
\max_{[-1,1]} |T_n(x)|=1](local/cache-vignettes/L104xH38/63dfd85b818554546774cfede4061425-afaba.png)
Best choice of points of interpolation of Lagrange polynomial
We have seen that if
and
, then:
![\forall x\in [a,b],\quad |f(x)-P_n(x)|\leq
\displaystyle\frac{\displaystyle|(x-x_0)(x-x_1)\ldots(x-x_n)|}{(n+1)!}\sup_{x\in[a,b]}|f^{n+1} (x)| \forall x\in [a,b],\quad |f(x)-P_n(x)|\leq
\displaystyle\frac{\displaystyle|(x-x_0)(x-x_1)\ldots(x-x_n)|}{(n+1)!}\sup_{x\in[a,b]}|f^{n+1} (x)|](local/cache-vignettes/L486xH53/8f29f55f9200b8f8d95501c49731528c-355c3.png)
The aim is to determine the points of interpolation

such that
![\sup_{x\in[a,b]}|(x-x_0)(x-x_1)\ldots(x-x_n)|\leq \sup_{x\in[a,b]}|Q(x)|
\sup_{x\in[a,b]}|(x-x_0)(x-x_1)\ldots(x-x_n)|\leq \sup_{x\in[a,b]}|Q(x)|](local/cache-vignettes/L323xH44/d7f79ee8beeda20eca915701b1ebde6f-de984.png)
for all

-degree monic polynomials

.
Via an affine substitution, the equivalent problem is:
![\sup_{x\in[-1,1]}|(x-x_0)(x-x_1)\ldots(x-x_n)|\leq \sup_{x\in[-1,1]}|Q(x)|
\sup_{x\in[-1,1]}|(x-x_0)(x-x_1)\ldots(x-x_n)|\leq \sup_{x\in[-1,1]}|Q(x)|](local/cache-vignettes/L339xH44/13290c93e0199239e1663042a0d8fe5d-b3d72.png)
We shall prove that the points of interpolation that verify this property are precisely te roots of the Chebychev polynomial

.
Theorem.
![\sup_{x\in[-1,1]}|(x-x_0)(x-x_1)\ldots(x-x_n)|\leq \sup_{x\in[-1,1]}|Q(x)|
\sup_{x\in[-1,1]}|(x-x_0)(x-x_1)\ldots(x-x_n)|\leq \sup_{x\in[-1,1]}|Q(x)|](local/cache-vignettes/L339xH44/13290c93e0199239e1663042a0d8fe5d-b3d72.png)
where

Proof.
Let

be the monic Chebychev polynomial associated to

. The roots of

are those of

and are defined by:

We
a fortiori have:

Additionally:
![\begin{array}{ccl}
\displaystyle \sup_{x\in[-1,1]}|(x-x_0)(x-x_1)\ldots(x-x_n)|&=&\displaystyle \sup_{x\in[-1,1]} |\bar{T}_{n+1}(x)|\\
&=&\displaystyle\frac{1}{2^n}\sup_{x\in[-1,1]} |T_{n+1}(x)|\\
&=&\displaystyle\frac{1}{2^n}
\end{array}
\begin{array}{ccl}
\displaystyle \sup_{x\in[-1,1]}|(x-x_0)(x-x_1)\ldots(x-x_n)|&=&\displaystyle \sup_{x\in[-1,1]} |\bar{T}_{n+1}(x)|\\
&=&\displaystyle\frac{1}{2^n}\sup_{x\in[-1,1]} |T_{n+1}(x)|\\
&=&\displaystyle\frac{1}{2^n}
\end{array}](local/cache-vignettes/L400xH113/3d482f1e63ddcb573f9f652832024b97-fe129.png)
Since

where

We thus have to show that:
![\frac{1}{2^n}\leq \sup_{x\in[-1,1]}|Q(x)|
\frac{1}{2^n}\leq \sup_{x\in[-1,1]}|Q(x)|](local/cache-vignettes/L125xH49/04ea9dab136483e9923af53f2be52f2c-f92b7.png)
We shall proceed by contradiction by assuming that:
![\frac{1}{2^n} > \sup_{x\in[-1,1]}|Q(x)|.
\frac{1}{2^n} > \sup_{x\in[-1,1]}|Q(x)|.](local/cache-vignettes/L126xH49/85dc503d657bf3ef628b20e0292409e4-0de0f.png)
Since

and

are monic polynomials:

Also

If

is an even number:

If
is an odd number:

Finally:

This means that there are

sign changes for the map

, and consequently,

has

roots. But,

has degree

. Therefore:


Finally:
![\sup_{x\in[-1,1]}|Q(x)|=\sup_{x\in[-1,1]}|\bar{T}_{n+1}(x)|=\frac{1}{2^n}.
\sup_{x\in[-1,1]}|Q(x)|=\sup_{x\in[-1,1]}|\bar{T}_{n+1}(x)|=\frac{1}{2^n}.](local/cache-vignettes/L246xH49/05cefa0816ee637805eec70690b29ca4-0bae4.png)
which contradicts our assumption. We finally conclude that:
![\begin{array}{ccl}
\displaystyle\frac{1}{2^n}&=&\displaystyle\sup_{x\in[-1,1]}|\bar{T}_{n+1}(x)|\\
&=&\displaystyle\sup_{x\in[-1,1]}|(x-x_0)(x-x_1)\ldots(x-x_n)|\\
&\leq&\displaystyle \sup_{x\in[-1,1]}|Q(x)|
\end{array}
\begin{array}{ccl}
\displaystyle\frac{1}{2^n}&=&\displaystyle\sup_{x\in[-1,1]}|\bar{T}_{n+1}(x)|\\
&=&\displaystyle\sup_{x\in[-1,1]}|(x-x_0)(x-x_1)\ldots(x-x_n)|\\
&\leq&\displaystyle \sup_{x\in[-1,1]}|Q(x)|
\end{array}](local/cache-vignettes/L298xH113/f927f915ccacd1d84e9291ca2560bbaa-a89f8.png)