Mathematics
V. V. IVANOV
Submitted 1958-01-01 | RussiaRxiv: ru-195801.76659 | Translated from Russian

Abstract

Full Text

Mathematics

V. V. IVANOV

SOME PROPERTIES OF SINGULAR INTEGRALS OF CAUCHY TYPE AND THEIR APPLICATIONS

(Presented by Academician I. N. Vekua on 14 IV 1958)

Let us take a function \(f(t)\), defined on a simple closed Lyapunov curve \(L\), situated in the complex plane with the origin inside \(L\), and consider the integral \(Sf=\dfrac{1}{\pi i}\int_L \dfrac{f(t)\,dt}{t-t_0}\), \(t_0\in L\), understood as an improper integral in the sense of the Cauchy principal value. Introduce the classes of continuous functions \(H(p+\alpha,L)\), where \(p\) is an integer, \(p\ge 0\), \(0\le \alpha\le 1\). A function \(f(t)\in H(p+\alpha,L)\) if its \(p\)-th derivative \(f^{(p)}(t)\) satisfies on \(L\) the condition

\[ f^{(p)}(t_1)-f^{(p)}(t_2)=O\left(|t_1-t_2|^\alpha\right),\qquad t_1,\ t_2\in L. \]

The well-known theorem of Plemelj–Privalov \((^1)\) states that from \(f\in H(p+\alpha,L)\) it follows that \(Sf\in H(p+\alpha,L)\), and conversely, if \(0<\alpha<1\). For \(\alpha=0,1\) this is no longer so.

The present note is devoted to the study of the limiting cases of the Plemelj–Privalov theorem \((\alpha=0,1)\). The results obtained are then applied to the question of the structural properties of the solution of the Riemann boundary-value problem and of the singular integral equation of normal type.

Let \(Z(p,L)\) be the class of functions defined on \(L\) whose \(p\)-th derivative satisfies the relation

\[ f^{(p)}[t,(s+h)]+f^{(p)}[t(s-h)]-2f^{(p)}[t(s)]=O(h), \]

where \(s\) is the arc abscissa of the point \(t(s)\in L\). Denote by \(\rho_n(f,L)\) the magnitude of the best approximation of the function \(f\), given on \(L\), by means of polynomials of degree \(n\) in the positive and negative powers of \(t\):

\[ \rho_n(f,L)=\sup_{P_n\in H_n}\inf_{t\in L}|f(t)-P_n(t)|, \]

where \(H_n\) is the totality of all polynomials of the form \(\sum_{-n}^{n} a_k t^k\).

Theorem 1. If \(f\in Z(p,L)\), then \(Sf\in Z(p,L)\), and conversely.

Theorem 2. If \(f\in Z(p,L)\), then \(\rho_n(f,L)=O(n^{-p-1})\), and conversely.

Theorem 3. If \(\rho_n(f,L)=O(n^{-p-\alpha})\), then \(\rho_n(Sf,L)=O(n^{-p-\alpha})\), \(0<\alpha\le 1\), and conversely.

Theorem 4. If \(f\in C\), i.e., is continuous on \(L\), then \(\exp(Sf)\in L_p\), \(p>0\), i.e., is summable on \(L\) with any power \(p>0\).

The proof of all these theorems is based on the representation

\[ Sf=\frac{1}{\pi i}\int_{|w|=1}\frac{f[\psi(w)]}{w-w_0}\,dw +\frac{1}{\pi i}\int_{|w|=1} f[\psi(w)] \left[ \frac{\psi'(w)}{\psi(w)-\psi(w_0)}-\frac{1}{w-w_0} \right]dw, \]

in which \(t=\psi(w)\) are the boundary values of a function conformally mapping the exterior of the circle \(|w|=1\) onto the exterior of \(L\), \(t_0=\psi(w_0)\), and also to the corresponding (when \(L\) is the circle: \(|w|=1\)) or similar results of the authors \((^{2-4})\).

Let us now consider the Riemann boundary-value problem \((^{1,5,6})\): to find a piecewise-holomorphic function of the form \(\dfrac{1}{2\pi i}\int_L \dfrac{\varphi(\tau)\,d\tau}{\tau-z}\), whose angular boundary values \(\varphi^+\) and \(\varphi^-\), respectively from inside and from outside \(L\), satisfy the condition

\[ \varphi^+ = G\varphi^- + f \tag{1} \]

almost everywhere on \(L\), where \(G\) is continuous, and \(f\in L_r,\ r>1\), on \(L\), the index \(\varkappa=\dfrac{1}{2\pi}[\arg G]_L=0\).

We shall regard two functions as identical if their values coincide almost everywhere on \(L\). Following the reasoning of F. D. Gakhov \((^5)\), represent \(G\) in the form \(G=\psi^+/\psi^-\), where

\[ \psi^\pm=\exp\left[\pm \frac{1}{2}\ln G+\frac{1}{2}S(\ln G)\right] \tag{2} \]

and \((\psi^\pm)^{-1}\), by virtue of Theorem 4, belongs to \(L_p\) for every \(p>0\). After this, relation (1) is given the form

\[ \frac{\varphi^+}{\psi^+}-\frac{\varphi^-}{\psi^-}=\frac{f}{\psi^+}. \tag{3} \]

By Hölder’s inequality \(\dfrac{f}{\psi^+}\in L_{r-\varepsilon}\), where \(\varepsilon\) is any arbitrarily small positive number. Solving problem (3) (see \((^6)\)), we obtain

\[ \varphi^\pm=\psi^\pm\left[\pm \frac{1}{2}\frac{f}{\psi^+}+\frac{1}{2}S\left(\frac{f}{\psi^+}\right)\right]. \tag{4} \]

Thus, under the assumptions made, there exists a unique solution of the Riemann problem belonging to \(L_{r-\varepsilon}\).*

By modifying and generalizing the above arguments, one can similarly investigate the Riemann problem when the index \(\varkappa\ne0\) and when \(f(t)\) is a piecewise-continuous function with a finite number of discontinuities of the first kind. The results obtained can also be applied to the study of a general singular integral equation of normal type.

From Theorem 3 and formulas (2), (4) it follows: if \(\rho_n(G,L)\) and \(\rho_n(f,L)\) have order \(n^{-p-\alpha}\) (this will be the case if, for example, \(G\) and \(f\) belong to \(H(p+\alpha,L)\)), then \(\rho_n(\varphi^\pm,L)\) also have order \(n^{-p-\alpha}\). This last assertion may find application in investigating the rate of convergence of an approximate solution of the singular integral equation \((^7)\).

Rostov-on-Don Institute
of Agricultural-Machinery Construction

Received
14 IV 1958

CITED LITERATURE

\({}^1\) N. I. Muskhelishvili, Singular Integral Equations, Moscow–Leningrad, 1946.
\({}^2\) A. Zygmund, Duke Math. J., 12, 47 (1945).
\({}^3\) V. I. Smirnov, Math. Ann., 107, 313 (1932).
\({}^4\) S. Ya. Alper, Izv. Akad. Nauk SSSR, Ser. Math., 9, No. 6 (1955).
\({}^5\) F. D. Gakhov, Matem. sbornik, 2, No. 4 (1937).
\({}^6\) B. V. Khvedelidze, Tr. Tbilisi Math. Inst., 23 (1956).
\({}^7\) V. V. Ivanov, DAN, 114, No. 5 (1957).

* I. B. Simonenko showed that in the case \(r=2\) in the conclusion formulated, \(\varepsilon\) can be omitted.

Submission history

Mathematics