Suppose that ${\alpha,\beta,\gamma}$ are real numbers with ${\beta<\gamma}$.

(a) Prove that ${\alpha+\beta\leq\alpha+\gamma}$

(b) Prove that ${\alpha+\beta\neq\alpha+\gamma}$

I have gone about this various ways and I think it comes down to confusion as to what is meant by ${\alpha}$ and ${\beta}$. Because the definition of their sum is ${\gamma=\{a+b:a\in \alpha,b\in \beta\}}$

We are using the dedekind cut definition of real numbers and in this sense, these real numbers beta and gamma CONTAIN all numbers less than them. so why couldn't i pick a number in gamma that is less than the number i pick in beta. in which case, this isn't true. or is the definition of sum the sum of the supremums of both cuts. In which case, why is a sum defined like this - as if it is ANY element of beta plus ANY element of gamma. Clearly, this shouldn't be a hard proof but I am lacking some fundamental understanding here.

  • 1
    $\begingroup$ It helps to use the definition of Dedekind cut (and addition of them) to clarify what the hypotheses and desired conclusions are. We're assuming that $\beta\subset\gamma$ and that $\beta\ne\gamma$ (so that, for instance, there exists $x\in\gamma$ with $x\notin\beta$). We're trying to prove that $\{a+b\colon a\in\alpha,\,b\in\beta\}\subset\{a+c\colon a\in\alpha,\,c\in\gamma\}$ and that the two sets are not equal (so it would suffice to prove, for example, that there exists $y\in\{a+c\colon a\in\alpha,\,c\in\gamma\}$ such that $y\notin\{a+b\colon a\in\alpha,\,b\in\beta\}$. Can you go from there? $\endgroup$ – Greg Martin Feb 14 at 6:07
  • 1
    $\begingroup$ " so why couldn't i pick a number in gamma that is less than the number i pick in beta." That would only tell you information about one element. Nothing about the sets as a whole. "as if it is ANY element of beta plus ANY element of gamma" we aren't finding one sum. It is a set of all possible sums. $\endgroup$ – fleablood Feb 14 at 7:47
  • $\begingroup$ Example $\alpha =\{q\in\mathbb Q|q^3 < 5\}; \beta=\{q\in\mathbb Q|q\le$ or a circle will circumference of $q$ will have a diameter$< 1\};\gamma=\{q\in\mathbb Q|q^5<1023\}$. Then you must prove $\alpha + \beta=\{m+n| m\in \alpha;n\in \beta\}\subsetneq\alpha+\gamma=\{m+n|m\in\alpha;n\in \gamma\}$. Only we have no way to refer to identify the supremas. (In this example $\alpha\approx \sqrt[3]{5}$, $\beta\approx \pi$ and $\gamma\approx \sqrt[5]{1023}$. $\endgroup$ – fleablood Feb 14 at 8:04
  • $\begingroup$ greg martin yes, thank you. $\endgroup$ – mayalarson Feb 14 at 14:28
  • $\begingroup$ I guess one thing is that for cuts $\alpha \le \beta$ doesn't mean the values $a \in alpha;b\in \beta$ are such that $a \le b$ (which isn't true as you point out) but that $\alpha \subseteq \beta$. It's a different ordering. Some text us different notatation such $\alpha \underline{\prec}\beta$ means $\alpha \subseteq \beta$. This forms an ordering. $\endgroup$ – fleablood Feb 14 at 17:41

With the Dedekind cut definition is that $\beta$ or any other real number, is viewed as a subset of $\mathbb Q$; a subset with properties $\beta\ne \emptyset, \beta \ne \mathbb R$ and for any $q \in \beta$ then for an $r\in \mathbb Q$ so that $r < q$ then $r\in \beta$; and that $\beta$ has no largest element.

Now I assume you have proven already that if $\alpha$ and $\beta$ are such cuts that the set $\tau = \{a+b|a\in \alpha; b\in \beta\}$ is also a Dedekind cut.

So you need to prove that $\alpha, \beta, \gamma$ are cuts that if $\beta \subsetneq \gamma$ (that's what $\beta<\gamma$ means) that $\alpha + \gamma \subseteq \alpha + \beta$.

[In this post when I use a roman letter such as $a,b,$ or $c$ it will be assumed they are rational numbers. If I compare $a < b$ this is to be assumed to be the usual rational number order and not the Dedekind Cut order. $\beta < \gamma$ will mean $\beta \subsetneq \gamma$ and $\beta \le \gamma$ will mean $\beta \subseteq \gamma$]

So if $m\in \alpha +\beta$ then there are an $a\in \alpha$ and $b\in \beta$ so that $m = a+b$. Now $b \in \beta\subset \gamma$ so $b\in \gamma$ and so $a+b \in \alpha + \gamma$.

That's all there is to it.

And to prove $\alpha + \beta \ne \alpha + \gamma$ is actually harder.

The problem is you can't do: let $c\in \gamma; c\not \in \beta$ (that's legit) and so let $a\in\alpha$ so $a+c> a+b$ for any $b\in \beta$ (which is true); but there could be an $a' >a$ so that $a' + b = a + c$.

We need a lemma and the notion that a cut has no largest element.

Lemma: For any positive rational $d$ there is an $a\in \alpha$ so that $a + d\not \in \alpha$.

Pf: Let $a'\in \alpha;$ $c \not \in \alpha$ then $c > a'$. Let $e=c-a'$. Archimedian principal says there is a positive integer $n$ so that $nd > e$.

Consider $a' + kd$ for integers $k=0... n$. $a'+0*d\in \alpha$ and $a'+nd > a'+e=c \not \in \alpha$. So there is some first element $k$ so that $a'+kd\not \in \alpha$ and a last element $a'+(k-1)d \in \alpha$. Let $a=a'+(k-1)d$ and that is an $a\in \alpha$ so that $a+d \not \in \alpha$.

Okay. Let $c_1\in \gamma; c_1\not \in \beta$. There is a $c_2\in \gamma; c_2 > c_1$. Let $d=c_2 - c_1$. Let $a\in \alpha$ so that $a + d\not \in \alpha$.

$a + c_2 \in\alpha + \gamma$ but we can show that $a+c_2\not \in \alpha + \beta$.

Let $m=a'+b\in \alpha + \beta$ with $a'\in \alpha $ and $b\in \beta$. we will show that $a'+b$ can not equal $a+c_2$.


$b < c_1 < c_2$ so $b < c_2-d$. So $a'+b < (a'-d) +c_2$. $a'-d < a'$ so $a'-d\in \alpha$ and $a'=(a'-d)+d\in \alpha$.


But $a+d \not \in \alpha$ so $a+d >a'$ and $a > a'-d$.


So $a'+b < (a'-d)+c_2 < a+c_2$. So $a'+b$ can not be equal to $a+c_2$ so $a+c_2 \not \in \alpha + \beta$.

| cite | improve this answer | |

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.