Fixed-point lemma for normal functions: Difference between revisions

Content deleted Content added
mNo edit summary
moving some general facts to normal function, shorting the proof considerably
Line 1:
The '''fixed-point lemma for normal functions''' is a basic result in [[axiomatic set theory]]; it states that any [[normal function]] has arbitrarily large [[fixed point (mathematics)|fixed point]]s and can often be used to construct [[ordinal number]]s with interesting properties. A formal version and proof (using the [[Zermelo-Fraenkel axioms]]) follow.
 
== Formal version ==
Line 7:
== Proof ==
 
FirstWe of all, it is clearknow that for any &alpha; &isin; Ord, ''f''(&alphagamma;) &ge; &alphagamma;. Iffor thisall wasordinals not&gamma;. theWe case, wenow coulddeclare choosean aincreasing minimalsequence &lt;&alpha; with <sub>''fn''(</sub>&alphagt;) (''n'' &lt; &alpha;omega;) then,by sincesetting ''f''&alpha;<sub>0</sub> is= normal&alpha;, and thus monotone, &alpha;<sub>''fn''(+1</sub> = ''f''(&alpha;)<sub>''n''</sub>) &lt;for ''fn''( &alphalt; &omega;), whichand isdefine a&beta; contradiction= tosup &lt;&alpha;<sub>''n''</sub>&gt;. beingClearly, minimal&beta; &ge; &alpha;. Since ''f'' commutes with [[supremum|suprema]], we have
# :''f''(&beta;) = 0. Then we have''f''(sup {&alpha;<sub>''n''</sub> = 0 for all: ''n'', and&lt; thus ''f''(&betaomega;}) = 0.
:&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; = sup {''f''(&alpha;<sub>''n''</sub>) : ''n'' &lt; &omega;}
:&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; = sup {&alpha;<sub>''n''+1</sub> : ''n'' &lt; &omega;}
:&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; = &beta;
(The last step uses that the sequence &lt;&alpha;<sub>''n''</sub>&gt; increases).
 
== Example application ==
We now declare a sequence &lt;&alpha;<sub>''n''</sub>&gt; (''n'' &lt; &omega;) by setting &alpha;<sub>0</sub> = &alpha;, and &alpha;<sub>''n'' + 1</sub> = ''f''(&alpha;<sub>''n''</sub>) for ''n'' &lt; &omega;, and define &beta; = sup &lt;&alpha;<sub>''n''</sub>&gt;. There are three possible cases:
 
# &beta; = 0. Then we have &alpha;<sub>''n''</sub> = 0 for all ''n'', and thus ''f''(&beta;) = 0.
# &beta; = &delta; + 1 for an ordinal number &delta;. Then there exists ''m'' &lt; &omega; such that for all ''n'' &ge; ''m'', &alpha;<sub>''n''</sub> = &delta; + 1. It follows that ''f''(&delta; + 1) = ''f''(&alpha;<sub>''m''</sub>) = &alpha;<sub>''m'' + 1</sub> = &delta; + 1, and thus ''f''(&beta;) = &beta;.
# &beta; is a [[limit ordinal]]. We first observe that sup &lt;''f''(&nu;) : &nu; &lt; &beta;&gt; = sup &lt;''f''(&alpha;<sub>''n''</sub>) : ''n'' &lt; &omega;&gt;. "&ge;" is trivial; for "&le;", we choose &nu; &lt; &beta;, then find an ''n'' with &alpha;<sub>''n''</sub> &gt; &nu;, and since ''f'' is monotone, we have ''f''(&alpha;<sub>''n''</sub>) &gt; ''f''(&nu;). Now we have ''f''(&beta;) = sup &lt;''f''(&nu;) : &nu; &lt; &beta;&gt; (since ''f'' is continuous), and thus ''f''(&beta;) = sup &lt;''f''(&alpha;<sub>''n''</sub>) : ''n'' &lt; &omega;&gt; = sup &lt; &alpha;<sub>''n''</sub> : ''n'' &lt; &omega; &gt; = &beta;.
 
== Notes ==
 
It is easily checked that the function ''f'' : Ord &rarr; Ord, ''f''(&alpha;) = &#1488;<sub>&alpha;</sub> is normal; thus, there exists an ordinal &Theta; such that &Theta; = &#1488;<sub>&Theta;</sub>. In fact, the above lemma shows that there are infinitely many such &Theta;.