UDC 517.43
MATHEMATICS
Submitted 1967-01-01 | RussiaRxiv: ru-196701.41876 | Translated from Russian

Full Text

UDC 517.43

MATHEMATICS

S. I. OSIPOV

ON A TRANSFORMATION OF OPERATOR SUMS

(Presented by Academician A. A. Dorodnitsyn, 3 X 1966)

Let \(D = d/dt\) be the differentiation operator, \(\theta_\alpha = t^\alpha D\), \(\alpha\) any complex number,* and

\[ \theta_{\alpha_k}\theta_{\alpha_{k-1}}\cdots\theta_{\alpha_1} \tag{1} \]

denote the superposition of the operators \(\theta_{\alpha_1}, \theta_{\alpha_2}, \ldots, \theta_{\alpha_k}\) \((k = 1, 2, \ldots)\), and let \(G_n\) be the set of all functions of \(t\) differentiable \(n\) times.

The operator sum

\[ \sum_{p=1}^{n} a_{n-p}(\theta_{\alpha_k}\theta_{\alpha_{k-1}}\cdots\theta_{\alpha_1})^p \tag{2} \]

with arbitrary constant coefficients \(a_{n-p}\) can be transformed to the form

\[ \sum_{\nu=1}^{nk} Q(n,k,\nu,\alpha_i,t)D^\nu, \tag{3} \]

where \(Q(n,k,\nu,\alpha_i,t)\) are certain completely determined functions. Let

\[ Q(n,k,\nu,\alpha_i,t) = \sum_{p=\mathrm{E}\left(\frac{\nu+k-1}{k}\right)}^{n} C(k,\nu,p,\alpha_i)a_{n-p}t^{-\mu\nu}, \tag{4} \]

where

\[ \mu = k - \sum_{i=1}^{k}\alpha_i, \]

\[ C(k,\nu,p,\alpha_i) = \frac{1}{(\nu!)^2} \sum_{m=\nu}^{pk} \frac{m!}{(m-\nu)!}\,S(k,p,m,\alpha_i), \]

\[ S(k,p,m,\alpha_i) = \sum_{r=0}^{m} \frac{(-1)^{r+pk}m!}{r!(m-r)!}\, N(p,k,\alpha_i,r), \]

\[ N(p,k,\alpha_i,r) = \prod_{j=1}^{p} R(p,k,\alpha_i,r,j), \]

\[ \begin{aligned} R(p,k,\alpha_i,r,j) &= [r+(p-j+1)\mu+\alpha_k] [r+(p-j+1)\mu+\alpha_k+\alpha_{k-1}-1] \\ &\quad {}\times \cdots \times [r+(p-j+1)\mu+\alpha_k+\alpha_{k-1}+\cdots+\alpha_1-(k-1)], \end{aligned} \]

\(\mathrm{E}\left(\dfrac{\nu+k-1}{k}\right)\) is the integer part of the number \(\dfrac{\nu+k-1}{k}\).

Theorem. On the set \(G_{nk}\), the operators (2) and (3)—(4) are equivalent.

The proof is obtained by generalizing our method \((^1)\).

* The function \(t^\alpha\), as a general power function, is generally multivalued, and it is necessary to fix some single-valued branch of it.

Relying on the theorem, one can solve the following problems.

Problem 1. A linear differential operator is given,

\[ q_0(t)D^s+q_1(t)D^{s-1}+\cdots+q_{s-1}(t)D, \]

where \(q_0(t), q_1(t), \ldots, q_{s-1}(t)\) are linear combinations of generalized power functions. It is required to determine the possibility (or impossibility) of representing it in the form of a polynomial in the operator (1).

This problem is of interest in connection with methods for integrating ordinary linear differential equations supplied by the operational calculus of V. A. Ditkin \((^2)\) and its generalizations.

Problem 2. Let \(\sigma\) be any complex number,

\[ \delta(n,m)= \begin{cases} 0, & \text{if } n<m,\\ 1, & \text{if } n\geq m, \end{cases} \]

for \(p=1,2,\ldots\) and \(\xi=1,2,\ldots,(p+1)k\) the numbers \(A_{\xi,p}\) satisfy the recurrence equation

\[ A_{\xi,p+1}= \sum_{\nu=1}^{\min(\xi,pk)} \sum_{r=1}^{k} \frac{\delta(\nu+r,\xi)\,r!\,\Gamma(\nu-p\sigma+1)} {(\nu+r-\xi)!(\xi-\nu)!\,\Gamma(\xi-p\sigma-r+1)} A_{r,1}A_{\nu,p} \]

under the conditions \(A_{r,1}=c_r\), where the \(c_r\) are arbitrarily prescribed constants with the sole restriction \(c_k\ne 0\), and it is required to find \(A_{\xi,p}\) in the form of some analytic expression.

In the solution of this problem an important role is played by \(C(k,\nu,p,\alpha_i)\).

The Lah numbers \((^3)\) and the classical Stirling numbers of the second kind are special cases of \(A_{\xi,p}\).

In conclusion, let us note the formula

\[ (xy)^{-\lambda/2}e^{-y}I_\lambda(2\sqrt{xy}) = \sum_{k=0}^{n} \frac{L_k^\lambda(x)}{\Gamma(\lambda+k+1)}(-y)^k + \frac{(-y)^{n+1}}{\Gamma(\lambda+n+2)}e^{x-\xi^2}L_{n+1}^{\lambda}(\xi^2), \tag{5} \]

\[ x\geq y\geq 0,\qquad \lambda\geq -\tfrac12,\qquad \xi=\sqrt{x}+\theta\sqrt{y},\qquad |\theta|<1. \]

It can be obtained by relying on a result of B. M. Levitan \((^4)\) and on our representation \((^5)\) of the Laguerre polynomials \(L_n^\lambda(x)\) by means of the operator (1) for \(k=2\), \(\alpha_1=1+\lambda\), \(\alpha_2=-\lambda\). Formula (5) is closely related to the well-known formula of N. Ya. Sonin \((^6)\) and differs by the remainder term.

Computing Center
Academy of Sciences of the USSR

Received
27 IX 1966

REFERENCES

  1. S. I. Osipov, Zhurn. vychislit. matem. i matem. fiz., 4, No. 1, 149 (1964).
  2. V. A. Ditkin, DAN, 116, No. 1, 15 (1957).
  3. J. Riordan, An Introduction to Combinatorial Analysis, IL, 1963, p. 56.
  4. B. M. Levitan, DAN, 73, No. 2, 269 (1950).
  5. S. I. Osipov, Zhurn. vychislit. matem. i matem. fiz., 3, No. 1, 191 (1963).
  6. N. Ya. Sonin, Investigations on Cylindrical Functions and Special Polynomials, Moscow, 1954, p. 68.

Submission history

UDC 517.43