On the extrapolation of perturbation series

DOI10.1023/A:1006493830706MaRDI QIDQ1591304zbMATH OpenOpenAlexFDO

Publication date 29 November 2001

Published in Acta Applicandae Mathematicae (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/physics/0212086

algebraic approximants convergence acceleration eigenvalues extrapolation numerical examples perturbed Hamiltonian power series expansions

Mathematics Subject Classification ID

Convergence and divergence of series and sequences (40A05) Approximation to limiting values (summation of series, etc.) (40A25) Extrapolation to the limit, deferred corrections (65B05) Numerical summation of series (65B10) Numerical methods for eigenvalue problems for boundary value problems involving PDEs (65N25)

Abstract: We discuss certain special cases of algebraic approximants that are given as zeroes of so-called "effective characteristic polynomials" and their generalization to a multiseries setting. These approximants are useful for the convergence acceleration or summation of quantum mechanical perturbation series. Examples will be given and some properties will be discussed.

Review text

The paper deals with algebraic approximants deriving from a number of power series. Let nonnegative integers \( {\xi } \), \( {\tau }\), degrees \( D(\mu) (0 \leq \mu \leq xi) \) and contact orders \( C(\nu) (0 \leq \nu \leq \tau) \) satisfying the condition \[ \xi + \Bigl\{\sum D(\mu) |0 \leq \mu \leq \xi \Bigr\} = \tau + 1 + \Bigl\{\sum C(\nu) |0 \leq \nu \leq \tau \Bigr\} \] be prescribed, and let \( {\mathcal F}(\nu|x) (0 \leq \nu \leq \tau) \) be \(\tau+1 \) series in ascending nonnegative powers of \( x \). Determine the coefficients \( P \) in the polynomials \[ p(\mu|x) = \Bigl\{\sum P(\mu,\nu) x^{\nu} |0 \leq \nu \leq D(\mu) \Bigr\} \] for \( 0 \leq \mu \leq \xi \) by setting \[ h(\xi,\tau|x,y) = \Bigl\{ \sum p(\mu|x) y^{\mu} |0 \leq \mu \leq \xi \Bigr\} \] and imposing the condition that the coefficients of \( x \) up to and including the \( C(\nu) \)'th in \( h(\xi,\tau|x, {\mathcal F}(\nu|x)) \) are all zero for \( 0 \leq \nu \leq \tau \), and setting \( P(\xi|0) = 1 \) (a system of linear algebraic equations results). Having derived the coefficients \( P \), evaluate the approximant \( r({\mathcal F}||\xi,\tau,C,D|x) \) by solving the equation \[ h(\xi,\tau|x,r({\mathcal F}||\xi,\tau,C,D|x)) = 0 \] for each \( x \) in this range (alternatively, after having solved an algebraic equation for the constant term, use a recursion to derive that power series associated with \( r \) corresponding to each root obtained). (To provide a simple example, take \( \xi = 1 \), \(\tau = 0 \), so that \( C(0) = D(0) + D(1) \). The coefficients \( P \) in the two polynomials \( p(0|x) \) and \( p(1|x) \) with \( p(1|0) = 1 \) are derived from the single power series \( {\mathcal F}(x) = {\mathcal F}(0|x) \) by imposing the condition that in the ascending power series expansion of \( p(0|x) + p(1|x) {\mathcal F}(x) \), the coefficients of all powers of \( x \) up to and including \(x^{D(0) + D(1)} \) should be zero. Thereafter \( r(x) = r({\mathcal F} ||1,0,C,D|x) \) is obtained by solving \( p(0|x) + p(1|x) r(x) = 0 \) for all \( x \) of interest. \( r \) is the approximating fraction with numerator and denominator degrees \( D(0) \) and \( D(1) \) respectively derived from \( {\mathcal F} \).) The author gives a conservation law for a variant of the above approximants. The allocation \( p(\xi|x) = 1 \) is forced, and the choices \( \tau = \xi - 1 \) and \( D(\nu) = C(\nu) = \gamma \) for \( 0 \leq \nu < \xi \) are adhered to, so that \( h(\xi,\gamma|x,y) \) may be written as \( h'(\xi,\gamma|x,y) \) and the approximant derived from \( { \mathcal F } \) as \( r'({\mathcal F}||\xi, \gamma|x) \). The first \( \gamma + 1 \) coefficients in the ascending power series associated with \( r'({\mathcal F}||\xi, \gamma|x)\) are preserved in \( r'({\mathcal F}||\xi, \gamma+1|x) \). (To give a simple example, set \(\xi = 1 \). The single polynomial \(p(0|x) \), of degree \( \gamma \), is derived from the condition that the first \( \gamma +1 \) terms in \( p(0|x) + {\mathcal F}(x) \) should be zero and is formed from the first \( \gamma+1 \) terms of \( {\mathcal F}(x) \) (with sign reversed). The approximant \( r'({\mathcal F}||\xi, \gamma|x) \) is derived from the condition \( p(0|x) + r'({\mathcal F}||\xi, \gamma|x) = 0 \): the first \(\gamma+1 \) terms of \( {\mathcal F } \) are recovered: they are indeed preserved in \( r'({\mathcal F}||\xi,\gamma+1|x) \).) Some numerical examples relating to power series expansions of eigenvalues of perturbed Hamiltonians are given.

Recommendations

Cited in

(11)

This page was built for publication: On the extrapolation of perturbation series

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1591304)