0
$\begingroup$

For a sequence $X_n$ of real numbers with $\limsup\limits_n X_n = \inf\limits_n\sup\limits_{k\geqslant n} X_k$, $\liminf\limits_n X_n = \sup\limits_n \inf\limits_{k\geqslant n} X_k$,

(a) How to prove that $X_n$ tends to $x$ is equivalent to saying that $\limsup_n X_n = \liminf_n X_n = x$

(b) If $X_n \geqslant Y_n$ for all $n \geqslant 1$ and if one of them is convergent, how to prove that $\liminf_n X_n \geqslant \limsup_n Y_n$. In general, $\liminf_n X_n \geqslant \liminf_n X_n;\; \limsup_n X_n \geqslant \limsup_n Y_n$

$\endgroup$

1 Answer 1

1
$\begingroup$

b) is simpler than it looks. If $x_n \ge y_n$ for all $n$ then $\liminf x_n \ge \liminf y_n$ and $\limsup x_n \ge \limsup y_n$. If (for instance) $y_n$ is convergent, then $\liminf y_n = \limsup y_n$ so you get the inequality you need.

For a) suppose the liminf and limsup are equal to $x$. Let $y_n = \sup_{k \ge n} x_k$ and $z_n = \inf_{k \ge n} x_k$. Then $y_n \searrow x$ and $z_n \nearrow x$. Let $\epsilon > 0$ be given. There exist indices $N_y$ and $N_z$ with the property that $n \ge N_y \implies x \le y_n < x + \epsilon$ and $n \ge N_z$ implies $x-\epsilon < z_n \le x$. If $N = \max\{N_y,N_z\}$ then $$n \ge N \implies x-\epsilon < y_n \le x_n \le z_n < x + \epsilon.$$ It follows that $x_n \to x$. The converse is proven in a similar way.

$\endgroup$

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.