\newcommand{\eqo}{\mbox{$=\!\!o$}}
\newcommand{\leqo}{\mbox{$\leq\!\!o$}}
\newcommand{\lesso}{\mbox{$<\!\!o$}}
\begin{abstract}
We develop a basic theory of ordinals and cardinals in Isabelle/HOL, up to the
point where some cardinality facts relevant for the ``working mathematician" become available.
Unlike in set theory, here we do not have at hand canonical notions of ordinal and cardinal.
Therefore, here an ordinal is merely a well-order relation and a cardinal is an
ordinal minim w.r.t. order embedding on its field.
\end{abstract}
\section{Introduction}
In set theory (under formalizations such as Zermelo-Fraenkel or Von Neumann-Bernays-G\"{o}del), an
{\em ordinal} is a special kind of well-order, namely one
whose strict version is the restriction of the membership relation to a set. In particular,
the field of a set-theoretic ordinal is a transitive set, and the non-strict version
of an ordinal relation is set inclusion. Set-theoretic ordinals enjoy the nice properties
of membership on transitive sets, while at the same time forming a complete class of
representatives for well-orders (since any well-order turns out isomorphic to an ordinal).
Moreover, the class of ordinals is itself transitive and well-ordered by membership as the strict relation
and inclusion as the non-strict relation.
Also knowing that any set can be well-ordered (in the presence of the axiom of choice), one then defines
the {\em cardinal} of a set to be the smallest ordinal isomorphic to a well-order on that set.
This makes the class of cardinals a complete set of representatives for the intuitive notion
of set cardinality.\footnote{The ``intuitive" cardinality of a set $A$ is the class of all
sets equipollent to $A$, i.e., being in bijection with $A$.}
The ability to produce {\em canonical well-orders} from the membership relation (having the aforementioned
convenient properties)
allows for a harmonious development of the theory of cardinals in set-theoretic settings.
Non-trivial cardinality results, such as $A$ being equipollent to $A \times A$ for any infinite $A$,
follow rather quickly within this theory.
However, a canonical notion of well-order is {\em not} available in HOL.
Here, one has to do with well-order ``as is", but otherwise has all the necessary infrastructure
(including Hilbert choice) to ``climb" well-orders recursively and to well-oder arbitrary sets.
The current work, formalized in Isabelle/HOL, develops the basic theory of ordinals and cardinals
up to the point where there are inferred a collection of non-trivial cardinality facts useful
for the ``working mathematician", among which:
\begin{itemize}
\item Given any two sets (on any two given types)\footnote{Recall that, in HOL, a set
on a type $\alpha$ is modeled, just like a predicate, as a function from $\alpha$ to \textsf{bool}.},
one is injectable in the other.
\item If at least one of two sets is infinite, then their sum and their Cartesian product are equipollent
to the larger of the two.
\item The set of lists (and also the set of finite sets) with element from an
infinite set is equipollent with that set.
\end{itemize}
Our development emulates the standard one from set-theory, but keeps everything
{\em up to order isomorphism}.
An (HOL) ordinal is merely a well-order. An {\em ordinal embedding} is an
injective and order-compatible function which maps its source into an initial segment
(i.e., order filter) of its target. Now, a {\em cardinal} (called in this work a {\em cardinal order})
is an ordinal minim w.r.t. the existence of embeddings among all
well-orders on its field. After showing the existence of cardinals on any given set,
we define the cardinal of a set $A$, denoted $|A|$, to be {\em some} cardinal order
on $A$. This concept is unique only up to order isomorphism (denoted $\eqo$), but meets
its purpose: any two sets $A$ and $B$ (laying at potentially distinct types)
are in bijection if and only if $|A|\;\eqo\;|B|$. Moreover, we also show that numeric cardinals
assigned to finite sets\footnote{Numeric cardinals of finite sets are already formalized in
Isabelle/HOL.}
are {\em conservatively extended} by our general (order-theoretic) notion of
cardinal. We study the interaction of cardinals with standard set-theoretic
constructions such as powersets, products, sums and lists. These constructions are shown
to preserve the ``cardinal identity" $\eqo$ and also to be monotonic w.r.t. $\leqo$, the ordinal
embedding relation. By studying the interaction between these constructions, infinite sets and
cardinals, we obtain the
aforementioned results for ``working mathematicians".
For this development, we did not follow closely any particular textbook, and in fact are not
aware of such basic theory of cardinals previously
developed in HOL.\footnote{After writing this formalization, we became aware of
Paul Taylor's membership-free development of the theory of ordinals \cite{taylor-ordinals}.} On
the other hand,
the set-theoretic versions of the facts proved here are folklore in set theory, and can be found,
e.g., in the textbook \cite{card-book}. Beyond taking care of some locality aspects
concerning the spreading of our concepts throughout types, we have not departed
much from the techniques used in set theory for establishing these facts -- for instance,
in the proof of one of our major theorems,
\textit{Card-order-Times-same-infinite} from Section 8.4,\footnote{This theorem states that, for any
infinite cardinal $r$ on a set $A$, $|A\times A|$ is not larger than $r$.}
we have essentially applied the technique described, e.g., in the proof of
theorem 1.5.11 from \cite{card-book}, page 47.
Here is the structure of the rest of this document.
The next three sections, 2-4, develop some
mathematical prerequisites.
In Section 2, a large collection of simple facts about
injections, bijections, inverses, (in)finite sets and numeric cardinals are proved,
making life easier
for later, when proving less trivial facts.
Section 3 introduces upper and lower
bounds operators for order-like relations and studies their basic properties.
Section 4 states some useful variations of well-founded recursion and induction principles.
Then come the major sections, 5-8.
Section 5 defines and studies, in the context of a well-order relation,
the notions of minimum (of a set), maximum (of two elements), supremum, successor (of a set),
and order filter (i.e., initial segment, i.e., downward-closed set).
Section 6 defines and studies (well-order) embeddings, strict embeddings, isomorphisms, and
compatible functions.
Section 7 deals with various constructions on well-orders, and with the relations
$\leqo$, $\lesso$ and $\eqo$ of well-order embedding, strict embedding, and isomorphism, respectively.
Section 8 defines and studies cardinal order relations, the cardinal of a set, the connection
of cardinals with set-theoretic constructs,
the canonical cardinal of natural numbers and finite cardinals, the successor
of a cardinal, as well as regular cardinals. (The latter play a crucial role in the development of
a new (co)datatype package in HOL.)
Finally, section 9 provides an abstraction of the previous developments on
cardinals, to provide a simpler user interface to cardinals, which in most of
the cases allows to forget that cardinals are represented by orders and use them
through defined arithmetic operators.
More informal details are provided at the beginning of each section, and also at the
beginning of some of the subsections.