The asymptotic equivalence of the solutions of certain linear systems of differential equations
I. N. Zboychik
Submitted 1967-01-01 | RussiaRxiv: ru-196701.35033 | Translated from Russian

Abstract

The question of the equivalence of systems of two differential equations \cite{4}
\begin{gather}\frac{dx}{dt}=X[R(t)+Q(t)],\label{1}\\frac{dx}{dt}=YR(t)\label{2}\end{gather}
in the sense of $x=YA(t)$, where $A(t)\to A=\operatorname{const}$ as $t\to\infty$, is considered. Here $R+Q=P_0+P_1t^{-1}+P_2t^{-2}+\cdots$, where $P_k$ are constant second-order matrices, and $R(t)$ is a segment of this series. Bibliography: 7 items.

Full Text

Preamble

In this section, we investigate the asymptotic behavior of the solutions to the system of differential equations as $t \to \infty$. We consider the relationship between the matrices $R(t)$, $Q(t)$, and $X$, specifically focusing on the case where $X' = X[R(t) + Q(t)]$ and $Y' = YR(t)$. We assume that $Y^{-1} X = A(t)$, where $A(t) \to A = \text{const}$ as $t \to \infty$.

§ 1. Case of Distinct Real Eigenvalues

Let the matrix $P_0$ have distinct real eigenvalues $\alpha_1$ and $\alpha_2$, with $\alpha_1 > \alpha_2$. Suppose the matrix $R(t)$ can be expanded in a series of the form:
$$R(t) = P_0 + P_1 t^{-1} + P_2 t^{-2} + \dots = \sum_{k=0}^{\infty} P_k t^{-k} \qquad \text{(1.1)}$$
where $P_k$ are constant matrices. Following the methods established in \cite{5}, the fundamental matrix $Y$ can be represented as:
$$Y = e^{\text{diag}[\alpha_1, \alpha_2]t} Z(t) \qquad \text{(1.2)}$$
where $Z(t) = I + O(t^{-1})$ as $t \to \infty$. If we define $A(t) = Y^{-1} X$, and assume $A(t) \to A = \text{const}$, then the matrix $A$ must be diagonal, $A = \text{diag}[b_{11}, b_{22}]$. If $b_{11} b_{22} \neq 0$, then the matrix $X$ exhibits the same asymptotic growth as $Y$.

§ 2. Case of Purely Imaginary Eigenvalues

Consider the case where $P_0$ has purely imaginary eigenvalues $P_0 = \text{diag}[i\alpha, -i\alpha]$. If the first-order perturbation term $P_1$ vanishes ($P_1 = 0$), the asymptotic behavior is determined by higher-order terms. Under these conditions, the transformation $Y = e^{\text{diag}[i\alpha, -i\alpha]t} Z(t)$ yields a matrix $A(t)$ that approaches a constant diagonal matrix $B = \text{diag}[b_{11}, b_{22}]$. If $b_{11} b_{22} \neq 0$, the system remains stable, and the solutions are bounded.

§ 3. Case of Complex Conjugate Eigenvalues

When $P_0$ has complex conjugate eigenvalues of the form $\alpha \pm i\alpha$, we utilize the representation:
$$Y = e^{\alpha t} e^{\text{diag}[i\alpha, -i\alpha]t} Z(t) \qquad \text{(3.1)}$$
The analysis follows a similar logic to the previous sections. The matrix $A(t)$ converges to a constant matrix $A$ as $t \to \infty$. The specific form of $A$ depends on the coefficients $P_k$ of the expansion of $R(t)$. If $P_1 = 0$, the convergence rate is determined by $O(t^{-2})$.

§ 4. Case of Multiple Eigenvalues

In the case where $P_0$ has a multiple eigenvalue $\alpha$ with a non-trivial Jordan block:
$$P_0 = \begin{pmatrix} \alpha & 1 \ 0 & \alpha \end{pmatrix} \qquad \text{(4.1)}$$
the asymptotic expansion of the solution $X$ involves logarithmic terms. Specifically, the matrix $Y$ can be expressed as:
$$Y = e^{\alpha t} \begin{pmatrix} 1 & t \ 0 & 1 \end{pmatrix} Z(t) \qquad \text{(4.4)}$$
where $Z(t) = I + O(t^{-1})$. If $A(t) = Y^{-1} X \to A$, then $A$ must commute with the Jordan form of $P_0$.

For the general case where $R(t) = P_0 + \sum_{k=1}^{\infty} P_k t^{-k}$, we define the matrices $L_k$ based on the coefficients $P_k$. If the determinant $D(S) \neq 0$ and the structural conditions on the elements $s_{21}, s_{22}$ are met, the solution $X$ can be represented as:
$$X = t^C Z(t) S^{-1} \text{diag}[t, 1] e^{\alpha t} \qquad \text{(4.18)}$$
where $C$ is a constant matrix related to the eigenvalues of the perturbation.

If the leading perturbation term $P_1$ satisfies certain nullity conditions, the logarithmic growth is suppressed, and $A(t)$ converges to a constant matrix $B$ more rapidly. Specifically, if $b \neq 0$ and the resonance conditions are avoided, the solution maintains the form:
$$A(t) = \begin{pmatrix} t^{-1} & -\ln t \ 0 & 1 \end{pmatrix} Z^{-1}(t) B X Z(t) S^{-1} \begin{pmatrix} t & 1 \ 0 & 1 \end{pmatrix} \qquad \text{(4.37)}$$
This indicates that the interaction between the Jordan structure of the unperturbed system and the power-law decay of the perturbation $R(t) - P_0$ leads to a variety of asymptotic regimes, including pure power-law growth, logarithmic corrections, or convergence to a steady state.

References

  1. Bellman, R. Stability Theory of Differential Equations. McGraw-Hill, 1953.
  2. Gel'fand, I. M. Lectures on Linear Algebra. Moscow, 1951.
  3. Coddington, E. A., and Levinson, N. Theory of Ordinary Differential Equations. McGraw-Hill, 1955.
  4. Erugin, N. P. Linear Systems of Ordinary Differential Equations. Minsk, 1963.
  5. Harris, W. J. Asymptotic Expansions for Ordinary Differential Equations. 1966.
  6. Rapoport, I. M. On Some Asymptotic Methods in the Theory of Differential Equations. Kiev, 1954.
  7. Shchelkanovtsev, N. M. Journal of Mathematics, Vol. 10, No. 6, 1966.

Submission history

The asymptotic equivalence of the solutions of certain linear systems of differential equations