2
$\begingroup$

I face an infinite system of linear algebraic equations. What conditions must the matrix $\bf A$ and the vector $\bf b$ satisfy for the existence of the bounded solution of the linear system $$ {\bf A} {\bf x} = {\bf b}? $$

Discussions Cramer's rule and Physics forum do not help.

$\endgroup$
1
$\begingroup$

The answer seems to be given in the well-known monograph "Approximate methods of higher analysis" by Kantorovich and Krylov (pp. 26-27).

The infinite system $$ {\bf A}{\bf x} = {\bf b} $$ is called

  • regular if $$ \sum_{j = 1}^\infty |a_{i j}| < 1, $$
  • fully regular if for a positive constant $\theta$, $$ \sum_{j = 1}^\infty |a_{i j}| \leq 1 - \theta < 1. $$ Here $a_{i j}$ are the elements of the matrix ${\bf A}$.

Denote by $$ c_i = 1 - \sum_{j = 1}^\infty |a_{i j}|. $$ Then, the following theorem holds.

Theorem. If the infinite system of linear equations $$ {\bf A}{\bf x} = {\bf b} $$ is regular, and the inequality $$ |b_i| \leq K c_i $$ holds for a constant $K > 0$, then the bounded solution $$ |x_i| \leq K $$ to the infinite system exists.

Remark If the infinite system is fully regular, and $$ |b_i| \leq K \theta $$ holds, then the bounded solution $$ |x_i| \leq K $$ to the infinite system exists.

$\endgroup$
0
$\begingroup$

There are only 3 possibilities when you solve linear systems: 0 solutions, 1 solution or infinite amount of solutions. This can be shown by proving that if there are 2 solutions, there must be an infinite amount of solutions -- how do you do that?

Once that is proven, the only possibility for the unique solution to exist is that $b$ is in the column space of $A$ (so at least one solution exists) and $A$ has full rank (so at most one solution exists).

$\endgroup$
  • $\begingroup$ So the Cramer's rule is true also for infinite systems of LAE? $\endgroup$ – Asatur Khurshudyan Apr 24 '18 at 3:05
  • $\begingroup$ @AsaturKhurshudyan I am not sure you can extend Cramer's rule to infinite systems, but you can possible define the determinant of a linear operator as a product of its eigenvalues, although I am not sure where that will lead you exactly $\endgroup$ – gt6989b Apr 24 '18 at 3:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.