Given nonempty set $A$ of positive real numbers, and define $$\frac{1}{A}=\left\{z=\frac{1}{x}:x\in A \right\}$$ Show that $$\sup\left(\frac{1}{A}\right)=\frac{1}{\inf A}$$

let $\sup\left(\frac{1}{A}\right)=\alpha$ and $\inf A = \beta$. Apply the definition of supremum, $z<\alpha$, then there exist $z'\in 1/A$ such that for every $\epsilon>0$, $z'>\alpha-\epsilon$

And the definition of infimum, $x>\beta$ and there exists $x'\in A$ such that for every $\epsilon>0$, $x'<\epsilon+\beta$.

At this step, I don't see how to relate $\beta$ with $\alpha$ which is $\alpha=\frac{1}{\beta}$, can anyone give me a hit or suggestion. Thanks.

  • 1
    $\begingroup$ $1/x$ has a property that it reverses inequalites: $x \leq y$ if and only if $1/x \geq 1/y$. So you can look at the definition of sup and inf and see that they're the same, but with all the inequalities reversed. hope this helps $\endgroup$
    – user40167
    Jun 21 '15 at 3:58

The crucial facts here are that if $x \in A$, then $x>0$ and the function $x \mapsto {1 \over x}$ reverses order in $A$, that is, if $x,y \in A$, then $x<y$ iff${1 \over x } > {1 \over y}$.

We have $\sup_{x' \in A} {1 \over x'} \ge {1 \over x}$ for all $x \in A$. Now let $x_n \in A$ such that $x_n \to \inf A$. Then this gives $\sup_{x' \in A} {1 \over x'} \ge {1 \over \inf A}$.

Since $x \mapsto {1 \over x}$ reverses order in $A$, we have ${1 \over \inf A} \ge {1 \over x}$ for all $x \in A$. Taking the $\sup$ yields the desired answer, ${1 \over \inf A} \ge \sup_{x \in A} {1 \over x}$.


Let Inf$(A)=\beta(\neq0) \implies \forall\ a\in A,\ a\geq\beta$ and $\alpha=\sup\left(\frac{1}{A}\right)$. Since, $\frac{1}{A}$ is defined which means that $\forall\ a\in A,\ a\neq0. $

Claim: $\alpha=\frac{1}{\beta}$

$(i)\frac{1}{\beta}$ is an UB of $\frac{1}{A}.$

Since, $\beta $ is LB of $A\ \implies$ $$ \forall\ a\in A,\ a\geq\beta \implies \forall\ a\in A,\ \frac{1}{a}\leq \frac{1}{\beta}\implies \frac{1}{a} $$ So, $\frac{1}{\beta}$ is an UB of $A.$

Now we have to prove that $\frac{1}{\beta} $ is the LUB of $\frac{1}{A}\ i.e.\ \forall\ \epsilon>0\ \exists\ b\in \frac{1}{A}$ s.t. $\frac{1}{\beta}-\epsilon>b.$ Without loss of generality we can assume that $\epsilon $ is such that $\beta\epsilon<1.$

Let $\delta=\frac{\beta}{1-\beta\epsilon}-\beta$. Clearly, $\delta>0.$ Since, $\beta $ is the GLB of $A\ \implies \beta+\delta$ is not a LB of $A \implies \exists\ a\in A$ s.t $$\beta+\delta>a\geq\beta.$$ $$\implies \beta+\frac{\beta}{1-\beta\epsilon}-\beta>a\geq\beta. $$ $$\implies \frac{\beta}{1-\beta\epsilon}>a\geq\beta. $$ $$ \implies\frac{1-\beta\epsilon}{\beta}<\frac{1}{a}\leq \frac{1}{\beta} .$$ Since, $\frac{1}{a}\in \frac{1}{A} \implies \frac{1}{\beta}$ is the LUB of $\frac{1}{A}.$

  • $\begingroup$ It is possible that $\inf A = 0$. Take $A= (0,\infty)$ for example. $\endgroup$
    – copper.hat
    Jun 21 '15 at 16:21
  • $\begingroup$ Then the sup($A$) does not exist. Because according to the definition of LUB "a non-empty set which is bounded above has the LUB"$\in \mathbb{R}$". So, inf $A$ must be non-zero, as I think. $\endgroup$ Jun 22 '15 at 5:24
  • $\begingroup$ A fairly standard convention is that if $A$ is unbounded above, then we write $\sup A = \infty$. $\endgroup$
    – copper.hat
    Jun 22 '15 at 5:58
  • $\begingroup$ Okay, thanks for the information. $\endgroup$ Jun 22 '15 at 6:02
  • $\begingroup$ Please can you say if we take $\epsilon \beta \geq 1$ , then what will happen here? please say $\endgroup$ Aug 24 '17 at 20:11

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.