Difference between revisions of "Polynomial least deviating from zero"
(TeX) |
m (dots) |
||
Line 6: | Line 6: | ||
P.L. Chebyshev [[#References|[1]]] proved that, among all polynomials of the form | P.L. Chebyshev [[#References|[1]]] proved that, among all polynomials of the form | ||
− | $$Q_n(x)=x^n+a_1x^{n-1}+\ | + | $$Q_n(x)=x^n+a_1x^{n-1}+\dotsb+a_n,\tag{1}$$ |
there is exactly one, viz. | there is exactly one, viz. |
Revision as of 12:05, 14 February 2020
polynomial deviating least from zero
An algebraic polynomial of degree $n$, with leading coefficient 1, having minimal norm in the space $C[a,b]$ or $L_p[a,b]$.
P.L. Chebyshev [1] proved that, among all polynomials of the form
$$Q_n(x)=x^n+a_1x^{n-1}+\dotsb+a_n,\tag{1}$$
there is exactly one, viz.
$$T_n(x)=2\left(\frac{b-a}{4}\right)^n\cos n\arccos\left(\frac{2x-a-b}{b-a}\right),$$
of minimal norm in $C[a,b]$, and that norm is
$$\|T_n\|_{C[a,b]}=2\left(\frac{b-a}{4}\right)^n.$$
The polynomial
$$U_n(x)=2\left(\frac{b-a}{4}\right)^{n+1}\frac{\sin((n+1)\arccos(2x-a-b)/(b-a))}{\sqrt{(b-x)(x-a)}}$$
is the unique polynomial deviating least from zero in $L_1[a,b]$ (among all polynomials \ref{1}), and its norm is
$$\|U_n\|_{L_1[a,b]}=4\left(\frac{b-a}{4}\right)^{n+1}.$$
In $L_p[a,b]$, $1<p<\infty$, there also exists a unique polynomial deviating least from zero; various properties of this polynomial are known (see [2], [5]).
The integral
$$\int\limits_a^bQ_n^2(x)\rho(x)dx,\quad\rho(x)>0,\tag{2}$$
considered for all polynomials \ref{1}, is minimal if and only if $Q_n(x)$, with respect to the weight function $\rho(x)$, is orthogonal on $(a,b)$ to all polynomials of degree $n-1$. If
$$a=-1,\quad b=1,\quad\rho(x)=(1-x)^\alpha(1+x)^\beta,$$
where $\alpha,\beta>-1$, then the integral \ref{2} is minimized by the Jacobi polynomial (cf. Jacobi polynomials) (if $\alpha=\beta=0$ by the Legendre polynomial; cf. Legendre polynomials) of degree $n$ with leading coefficient 1.
Among all trigonometric polynomials of the form
$$a\cos nx+b\sin nx+\sum_{k=0}^{n-1}(a_k\cos kx+b_k\sin kx),$$
where $a$ and $b$ are fixed, the polynomial of minimal norm in any of the spaces $C[0,2\pi]$ and $L_p[0,2\pi]$ (for an arbitrary $p\geq1$) is
$$a\cos nx+b\sin nx.$$
References
[1] | P.L. Chebyshev, "Complete collected works" , 2 , Moscow-Leningrad (1947) pp. 23–51 (In Russian) |
[2] | V.M. Tikhomirov, "Some problems in approximation theory" , Moscow (1976) (In Russian) |
[3] | A.F. Timan, "Theory of approximation of functions of a real variable" , Pergamon (1963) (Translated from Russian) |
[4] | G. Szegö, "Orthogonal polynomials" , Amer. Math. Soc. (1975) |
[5] | S.M. Nikol'skii, "Quadrature formulas" , Hindushtan Publ. Comp. , London (1964) (Translated from Russian) |
[6] | P.K. Suetin, "Classical orthogonal polynomials" , Moscow (1976) (In Russian) |
Comments
The polynomials $T_n$ and $U_n$ are called (normalized) Chebyshev polynomials of the first, respectively second, kind (cf. Chebyshev polynomials).
References
[a1] | I.P. Natanson, "Constructive function theory" , 2 , F. Ungar (1964–1965) pp. Chapt. 6 (Translated from Russian) |
[a2] | T.J. Rivlin, "The Chebyshev polynomials" , Wiley (1974) |
[a3] | M.J.D. Powell, "Approximation theory and methods" , Cambridge Univ. Press (1981) |
Polynomial least deviating from zero. Encyclopedia of Mathematics. URL: http://encyclopediaofmath.org/index.php?title=Polynomial_least_deviating_from_zero&oldid=33288