No label defined (Q3312646)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 3842414
Language Label Description Also known as
default for all languages
No label defined
    English
    No label defined
    scientific article; zbMATH DE number 3842414

      Statements

      0 references
      1978
      0 references
      divergent integrals
      0 references
      singular integrals
      0 references
      oscillatory integrands
      0 references
      anti-derivative
      0 references
      series transformation
      0 references
      An anti-derivative of the function \(\psi\) of a complex variable is a function \(J(\psi,\mu)\) for which \(\mathcal D J(\psi,\mu)=\psi(\mu)\), where \(\mathcal D\equiv d/d\mu\), for values of \(\mu\) in a domain over which \(\psi\) is analytic. Such functions \(J\) occur naturally in many problems of applied mathematics. For example, when \(\psi\) represents the complex velocity components of irrotational fluid flow vanishing at infinite, the real and imaginary parts of \(J\) normalized by the condition that \(J(\psi,\mu)\) tends to zero for large \(\mu\) are the velocity potential and stream function of the flow. Again, solutions of the differential equation \(p_ n(\mathcal D)f(\mu)=\psi(\mu)\), where \(p_ n\) is a polynomial with complex number coefficients, may be exhibited, by use of the Heaviside calculus, in the form \(\sum_{i=1}^m e^{\alpha_ i\mu}J^{(i)}(\psi_i,\mu)\), where \(J^{(i)}\) is a first or higher order anti-derivative: \[ J^{(i)}(\psi_i,\mu)={\mathcal D}^{- k(i)}\psi_ i(\mu),\quad \psi_ i(\mu)=e^{-\alpha_ i\mu}\psi(\mu), \] the \(k(i)\) and \(\alpha_i\) depending upon \(p_n\). The same method of treatment may be extended to certain differential equations of more complex structure; in this way such equations and systems of such equations may be treated by purely algebraic methods (see, for example, the excellent work of \textit{V. P. Maslov} [Operational methods. (Russian) Moskva: Nauka (1973; Zbl 0288.47042)]). Subject to appropriate restrictions upon \(\psi\), the integral \[ (\alpha) I(\psi,\mu)=\int^{\infty}_{\mu}\psi(t)\,dt \] satisfies the equation \((\beta)\mathcal D I(\psi,\mu)=-\psi(\mu): -I(\psi,\mu)\) is then an anti-derivative. When the integral in \((\alpha)\) does not exist in, for example, a Riemann sense, an anti-derivative may serve to define an singular or divergent integral. When the integral exists but the integrand is highly oscillatory upon the path of integration, so that numerical treatment by quadrature methods is ineffective, the anti-derivative may be used to assist computation. Four methods for constructing an anti-derivative using the value of \(\psi\) and those of its derivatives at the point \(\mu\) are considered. Algebraic properties of these methods are also given. The main thrust of the paper is towards the derivation of convergence theory applicable to functions \(\psi\) not having special forms. With the possible singularities and brunch cuts of \(\psi\) confined to a bounded convex open domain \(\mathfrak D\) in the complex plane, \(\psi(z)= O(z^{-2})\) for large \(z\), and \(I(\psi,\mu)=(2\pi i)^{-1}\int_{\mathcal L}\ln(\mu -z)\psi(z)\,dz\), \(\mathcal L\) being a semi-infinite loop lying in a half-plane over which \(\psi\) is analytic and encircling the point \(z=\mu\), regions in the \(\mu\)-plane \((\sigma\)-plane) for which \(\phi_ r(\sigma /\psi,\mu)\to I(\psi,\mu)\) for a fixed \(\sigma\) (fixed \(\mu)\) are derived. With \(\bar{\mathfrak D}\) being the smallest convex closed domain containing the singularities of \(\psi\), the optimal value of \(\sigma\) inducing the most rapid convergence of the sequence \(\{\phi_r(\sigma /\psi,\mu)\}\) for a fixed \(\mu\) is derived. The special case in which \[ \psi(\mu)=\sum^{\infty}_{\tau =0}\int^{\beta(\tau)}_{\alpha(\tau)}(\mu -t)^{-\tau - 2}\,d\xi_\tau(t), \] the real and imaginary parts of the \(\xi_\tau\) being of bounded variation over \([\alpha (\tau),\beta (\tau)]\subseteq [\alpha,\beta]\subset(-\infty,\infty)\), is considered in detail. In particular, it is shown that if \(f\) is an entire function, \(\mu\neq \alpha\) and \(\psi(\mu)=(\mu -\alpha)^{-2}f\{(\mu -\alpha)^{-1}\},\quad | \phi_ r(\alpha | \psi,\mu)-I(\psi,\mu)|\) tends to zero faster than the terms of any geometric progression. It is shown that with \[ \psi(\mu)=\int^{\beta}_{\alpha} (\mu -t)^{-2}\,d\xi(t), \quad I(\psi,\mu)=\int^{\beta}_{\alpha}(\mu -t)^{-1}\,d\xi(t),\tag{*} \] \(\xi\) being a nondecreasing real valued function of bounded variation over \([\alpha,\beta]\subset(-\infty,\infty)\). \(\rho_{2r}(\psi,\mu)\to I(\psi,\mu)\) for \(\mu\in(-\infty,\infty)\backslash [\alpha,\beta]\). With \(\mathcal M\) a certain point set not containing the origin, and \(\psi(\mu)=\int_{\mathcal M} e^{-\mu z}\,d\xi(z)\), \(I(\psi,\mu)=\int_{{\mathcal M}}z^{-1}e^{-\mu z}\,d\xi(z)\), the real and imaginary parts of \(\xi\) being of bounded variation over \(\mathcal M\), regions in the \(\mu\)-plane \((\sigma\)-plane) for which \(\gamma_r(\sigma | \psi,\mu)\to I(\psi,\mu)\) for a fixed \(\sigma\) (fixed \(\mu)\) are derived. With \[ \psi(\mu)=e^{-\lambda \mu}\int^{\beta}_{\alpha} e^{\mu t}\,d\xi(t),\quad I(\psi,\mu)=e^{-\lambda \mu} \int^{\beta}_{\alpha}(\lambda -t)^{-1}e^{\mu t}\,d\xi(t) \] where \(\xi\) is as in (*) over \([\alpha,\beta]\subseteq [- \infty,\infty]\) and \(\lambda\not\in [\alpha,\beta]\), \(\varepsilon_{2r}(\psi,\mu)\to I(\psi,\mu)\) for all \(\mu\in(- \infty,\infty)\) for which the derivatives of \(\psi\) are defined. (In all of the above cases a priori error bounds are obtained.) Simple recursive algorithms for the construction of the four methods used in paragraph C are given. The above work is extended to the estimation of integrals over a finite range, of the form \(\int^{\mu +h}_{\mu}\psi(x)\,dx\), in terms of the derivatives of \(\psi\) at the endpoints \(\mu\) and \(\mu +h\) of the range of integration. Favorable comparisons with the use of the Euler-Maclaurin series, which makes use of the same derivatives, are given. An extension of the theory to the estimation of the value of \(\lim F(z)\) \((z\to \infty)\) in terms of the values of \(F\) and its derivatives for a finite argument value \(\mu\) may be made. Letting \({\mathcal D}F(\mu)=\psi(\mu)\), it follows that \(I(\psi,\mu)=F(\infty)-F(\mu):\) an approximation to \(F(\mu)+I(\psi,\mu)\) is an approximation to \(F(\infty)\). Convergence theories for the modified processes are obtained by considering functions F whose derivatives \(\psi\) are dealt with in paragraph E. Further extension, in the opposite direction, to the estimation of higher order anti-derivatives is also possible. The paper under review extends earlier work by the author [in particular, that described in Arch. Ration. Mech. Anal. 28, 83--148 (1968; Zbl 0162.37202); Calcolo 9, 197--234 (1972; Zbl 0248.65007); J. Reine Angew. Math. 285, 181--208 (1976; Zbl 0326.40005)].
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references