A Problem from Tutorial

Recently the following question came up in tutorial as I had a student looking at ordinal arithmetic.  Basically, you want to show that if an ordinal is divisible on the right by 2 and by 3, then it is divisible on the right by 6.   It caught me off guard, because how often are we called on to prove non-immediate facts about ordinal arithmetic?  Here is the solution I finally came up with.

Lemma:  Suppose \(\gamma\) and \(\delta\) are ordinals. Then

\[\gamma+\delta+\gamma =\delta+\gamma+\delta \Longleftrightarrow \gamma=\delta.\]

Proof.   We work on the non-trivial implication, so suppose by way of contradiction that \(\gamma<\delta\) but

(*)   \(\gamma+\delta+\gamma = \delta+\gamma+\delta.\)

Our first observation is

(**)   \(\delta+\gamma<\gamma+\delta.\)

 If not, then \(\gamma+\delta\leq\delta+\gamma\) and using our assumption that \(\gamma<\delta\) we obtain

\[(\gamma+\delta)+\gamma\leq(\delta+\gamma)+\gamma<(\delta+\gamma)+\delta,\]

contradicting (*).    

From (**) we have

\[\gamma+(\delta+\gamma)<\gamma+(\gamma+\delta)\leq \delta+\gamma+\delta\]

contradicting (*) again.   The same argument shows that assuming \(\delta<\gamma\) leads to a contradiction, and so the lemma holds.  \(_\blacksquare\)

 

Problem: Suppose \(\epsilon\) is divisible on the right by both 2 and 3.  Show \(\epsilon\) is divisible on the right by 6.

Solution. We may assume that \(\epsilon\) is non-zero (or else the conclusion is immediate), so suppose \(\alpha\) and \(\beta\) are non-zero ordinals for which

\[\epsilon = \alpha+\alpha = \beta+\beta+\beta.\] 

Clearly \(\beta<\alpha\) and so we can fix a non-zero ordinal \(\gamma\) such that

 \[\alpha=\beta+\gamma.\]

Combining these together, we find

 \[\beta+\gamma+\beta+\gamma = \beta+\beta+\beta\]

and so

(1)  \[\gamma+\beta+\gamma = \beta+\beta.\]

We now claim that  \(\gamma\) is strictly less than \(\beta\).  If not and \(\beta\leq\gamma\) then from our previous work we obtain

\[\gamma+\beta+\gamma\leq\gamma+\beta\]

and therefore

\[\beta+\gamma\leq\beta.\]

But this is absurd because \(\gamma\) is non-zero. Thus \(\gamma<\beta\) as claimed.

Give this, there is a non-zero ordinal \(\delta\) such that

\[\beta=\gamma+\delta.\]

Again, by (1) we have 

\[\gamma+\beta+\gamma=\gamma+\gamma+\delta+\gamma=\gamma+\delta+\gamma+\delta\]

and so

 \(\gamma+\delta+\gamma = \delta + \gamma+\delta.\)

Now we apply our lemma to conclude that \(\gamma=\delta\).  In particular, 

\(\beta = \gamma+\gamma\)

and

\(\alpha = \beta+\gamma = \gamma+\gamma+\gamma\)

Putting all this together, we obtain

\[\epsilon = \alpha + \alpha = \gamma\cdot 6 = \beta+\beta+\beta\]

and so \(\epsilon\) is divisible on the right by 6.