cardinal number




Type theory

natural deduction metalanguage, practical foundations

  1. type formation rule
  2. term introduction rule
  3. term elimination rule
  4. computation rule

type theory (dependent, intensional, observational type theory, homotopy type theory)

syntax object language

computational trinitarianism =
propositions as types +programs as proofs +relation type theory/category theory

logiccategory theorytype theory
trueterminal object/(-2)-truncated objecth-level 0-type/unit type
falseinitial objectempty type
proposition(-1)-truncated objecth-proposition, mere proposition
proofgeneralized elementprogram
cut rulecomposition of classifying morphisms / pullback of display mapssubstitution
cut elimination for implicationcounit for hom-tensor adjunctionbeta reduction
introduction rule for implicationunit for hom-tensor adjunctioneta conversion
logical conjunctionproductproduct type
disjunctioncoproduct ((-1)-truncation of)sum type (bracket type of)
implicationinternal homfunction type
negationinternal hom into initial objectfunction type into empty type
universal quantificationdependent productdependent product type
existential quantificationdependent sum ((-1)-truncation of)dependent sum type (bracket type of)
equivalencepath space objectidentity type/path type
equivalence classquotientquotient type
inductioncolimitinductive type, W-type, M-type
higher inductionhigher colimithigher inductive type
-0-truncated higher colimitquotient inductive type
coinductionlimitcoinductive type
completely presented setdiscrete object/0-truncated objecth-level 2-type/preset/h-set
setinternal 0-groupoidBishop set/setoid
universeobject classifiertype of types
modalityclosure operator, (idempotent) monadmodal type theory, monad (in computer science)
linear logic(symmetric, closed) monoidal categorylinear type theory/quantum computation
proof netstring diagramquantum circuit
(absence of) contraction rule(absence of) diagonalno-cloning theorem
synthetic mathematicsdomain specific embedded programming language

homotopy levels




The cardinal numbers (or just cardinals) constitute a generalisation of a natural numbers to numbers of possibly infinite magnitudes. Specifically, cardinal numbers generalise the concept of ‘the number of …’. In particular, the number of natural numbers is the first infinite cardinal number.


Naïvely, a cardinal number should be an isomorphism class of sets, and the cardinality of a set SS would be its isomorphism class. That is:

  1. every set has a unique cardinal number as its cardinality;
  2. every cardinal number is the cardinality of some set;
  3. two sets have the same cardinality if and only if they are isomorphic as sets.

Then a finite cardinal is the cardinality of a finite set, while an infinite cardinal or transfinite cardinal is the cardinality of an infinite set. (If you interpret both terms in the strictest sense, then there may be cardinals that are neither finite nor infinite, without some form of the axiom of choice).

Taking this definition literally in material set theory, each cardinal is then a proper class (so one could not make further sets using them as elements). For this reason, in axiomatic set theory one usually defines a cardinal number as a particular representative of this equivalence class. There are several ways to do this:

  • The cardinality of a set SS is the smallest possible ordinal rank of any well-order on SS. In other words, it is the smallest ordinal number (usually defined following von Neumann) which can be put in bijection with SS. A cardinal number is then any cardinality, i.e. any ordinal which is not in bijection with any smaller ordinal.

    • On well-orderable sets, this cardinality function satisfies (1–3), but one needs the axiom of choice (precisely, the well-ordering theorem) to prove that every set is well-orderable. This approach is probably the most common one in the presence of the axiom of choice.

    • In the absence of excluded middle, when the “correct” definition of well-order is different from the usual one (and so “the least ordinal such that …” may not exist), a better definition of the cardinality of SS is as the set of all ordinal numbers less than the ordinal rank of every well-order on SS.

  • Alternatively, we can define the cardinality of a set XX to be the set of all well-founded pure sets that are isomorphic as sets to XX and such that no pure set of smaller hereditary rank (that is, which occurs earlier in the von Neumann hierarchy) is isomorphic to XX.

    • On those sets that are isomorphic to some well-founded set, this cardinality function satisfies (1–3), but one needs some assumption to prove that every set is isomorphic to a well-founded set. (This will follow directly from the axiom of foundation; it will also follow from the axiom of choice, since then every set is isomorphic to a von Neumann ordinal.)

In the absence of the appropriate axioms, the definitions above can still be used to define well-ordered cardinals and well-founded cardinals, respectively.

From the perspective of structural set theory, it is evil to care about distinctions between isomorphic objects, and unnecessary to insist on a canonical choice of representatives for isomorphism classes. Therefore, from this point of view it is natural to simply say:

  • A cardinal is a set (that is, an object of Set).

Madeleine Birchfield: Wouldn’t a cardinal number be an object of the decategorification of the category Set, just as a natural number is an object of the decategorification of the category FinSet?

However, one still may need sets of cardinals, that is sets that serve as the target of a cardinality function satisfying (1–3) on any (small) collection of sets. One can construct this as a quotient set of that collection.

Lowercase Greek letters starting from κ\kappa are often used for cardinal numbers.

Roger Witte First of all sorry if I am posting in the wrong place

While thinking about graphs, I wanted to define them as subobjects of naive cardinal 2 and this got me thinking about the behaviour of the full subcategories of Set defined by isomorphism classes. These categories turned out to be more interesting than I had expected.

If the background set theory is ZFC or similar, these are all large but locally small categories with all hom sets being isomorphic. They all contain the same number of objects (except 0, which contains one object and no non-identity morphisms) and are equinumerous with Set. Each hom Set contains N NN^N arrows. In the finite case N!N! of the morphisms in a particular hom set are isomorphisms. In particular, only 0 and 1 are groupoids. I haven’t worked out how this extends to infinite cardinalities, yet.

If the background theory is NF, then they are set and 1 is smaller than Set. I haven’t yet worked out how 2 compares to 1. I need to brush up on my NF to see how NF and category theory interact.

I am acutely aware that NF/NFU is regarded as career suicide by proffesional mathematicians, but, fortunately, I am a proffesional transport planner, not a mathematician.

Toby: Each of these categories is equivalent (but not isomorphic, except for 0) to a category with exactly one object, which may be thought of as a monoid. Given a cardinal NN, if you pick a set XX with NN elements, then this is (up to equivalence, again) the monoid of functions from XX to itself. The invertible elements of this monoid form the symmetric group, with order N!N! as you noticed. Even for infinite cardinalities, we can say N NN^N and N!N!, where we define these numbers to be the cardinalities of the sets of functions (or invertible functions) from a set of cardinality NN to itself.

From a structural perspective, there's no essential difference between equivalent categories, so the fact that these categories (except for 0) are equinumerous with all of Set is irrelevant; what matters is not the number of objects but the number of isomorphism classes of objects (and similarly for morhpisms). That doesn't mean that your result that they are equinumerous with Set is meaningless, of course; it just means that it says more about how sets are represented in ZFC than about sets themselves. So it should be no surprise if it comes out differently in NF or NFU, but I'm afraid that I don't know enough about NF to say whether they do or not.

By the way, every time you edit this page, you wreck the links to external web pages (down towards the bottom in the last query box). It seems as if something in your editor is removing URLs.

Cardinal arithmetic


For SS a set, write |S|{|S|} for its cardinality. Then the standard operations in the category Set induce arithmetic operations on cardinal numbers (“cardinal arithmetic”):

For S 1S_1 and S 2S_2 two sets, the sum of their cardinalities is the cardinality of their disjoint union, the coproduct in SetSet:

|S 1|+|S 2||S 1⨿S 2|. {|S_1|} + {|S_2|} \coloneqq {|S_1 \amalg S_2|} \,.

More generally, given any family (S i) i:I(S_i)_{i: I} of sets indexed by a set II, the sum of their cardinalities is the cardinality of their disjoint union:

i:I|S i|| i:IS i|. \sum_{i: I} {|S_i|} \coloneqq {|\coprod_{i: I} S_i|} \,.

Likewise, the product of their cardinalities is the cardinality of their cartesian product, the product in SetSet:

|S 1||S 2||S 1×S 2|. {|S_1|} \, {|S_2|} \coloneqq {|S_1 \times S_2|} \,.

More generally again, given any family (S i) i:I(S_i)_{i: I} of sets indexed by a set II, the product of their cardinalities is the cardinality of their cartesian product:

i:I|S i|| i:IS i|. \prod_{i: I} {|S_i|} \coloneqq {|\prod_{i: I} S_i|} \,.

Also, the exponential of one cardinality raised to the power of the other is the cardinality of their function set, the exponential object in SetSet:

|S 1| |S 2||Set(S 2,S 1)|. {|S_1|}^{|S_2|} \coloneqq {|Set(S_2,S_1)|} \,.

In particular, we have 2 |S|2^{|S|}, which (assuming the law of excluded middle) is the cardinality of the power set P(S)P(S). In constructive (but not predicative) mathematics, the cardinality of the power set is Ω |S|\Omega^{|S|}, where Ω\Omega is the cardinality of the set of truth values.

The usual way to define an ordering on cardinal numbers is that |S 1||S 2|{|S_1|} \leq {|S_2|} if there exists an injection from S 1S_1 to S 2S_2:

(|S 1||S 2|):((S 1S 2)). ({|S_1|} \leq {|S_2|}) \;:\Leftrightarrow\; (\exists (S_1 \hookrightarrow S_2)) \,.

Classically, this is almost equivalent to the existence of a surjection S 2S 1S_2 \to S_1, except when S 1S_1 is empty. Even restricting to inhabited sets, these are not equivalent conditions in constructive mathematics, so one may instead define that |S 1||S 2|{|S_1|} \leq {|S_2|} if there exists a subset XX of S 2S_2 and a surjection XS 1X \to S_1. Another alternative is to require that S 1S_1 (or XX) be a decidable subset of S 2S_2. All of these definitions are equivalent using excluded middle.

This order relation is antisymmetric (and therefore a partial order) by the Cantor–Schroeder–Bernstein theorem (proved by Cantor using the well-ordering theorem, then proved by Schroeder and Bernstein without it). That is, if S 1S 2S_1 \hookrightarrow S_2 and S 2S 1S_2 \hookrightarrow S_1 exist, then a bijection S 1S 2S_1 \cong S_2 exists. This theorem is not constructively valid, however.

The well-ordered cardinals are well-ordered by the ordering <\lt on ordinal numbers. Assuming the axiom of choice, this agrees with the previous order in the sense that κλ\kappa \leq \lambda iff κ<λ\kappa \lt \lambda or κ=λ\kappa = \lambda. Another definition is to define that κ<λ\kappa \lt \lambda if κ +λ\kappa^+ \leq \lambda, using the successor operation below.

The successor of a well-ordered cardinal κ\kappa is the smallest well-ordered cardinal larger than κ\kappa. Note that (except for finite cardinals), this is different from κ\kappa's successor as an ordinal number. We can also take successors of arbitrary cardinals using the operation of Hartog's number, although this won't quite have the properties that we want of a successor without the axiom of choice.


  • It is traditional to write 0{}_0 for the first infinite cardinal (the cardinality of the natural numbers), 1\aleph_1 for the next (the first uncountable cardinality), and so on. In this way every cardinal (assuming choice) is labeled μ\aleph_\mu for a unique ordinal number μ\mu, with ( μ)) += μ +(\aleph_\mu))^+ = \aleph_{\mu^+}.

  • For every cardinal π\pi, we have 2 π>π2^\pi \gt \pi (this is sometimes called Cantor's theorem). The question of whether 2 0= 12^{\aleph_0} = \aleph_{1} (or more generally whether 2 μ= μ +2^{\aleph_\mu} = \aleph_{\mu^+}) is called Cantor’s continuum problem; the assertion that this is the case is called the (generalized) continuum hypothesis. It is known that the continuum hypothesis is undecidable in ZFC.

  • For every transfinite cardinal π\pi we have (using the axiom of choice) π+π=π\pi + \pi = \pi and ππ=π\pi \cdot \pi = \pi, so addition and multiplication are idempotent.

Properties of cardinals

  • A transfinite cardinal π\pi is a regular cardinal if no set of cardinality π\pi is the union of fewer than π\pi sets of cardinality less than π\pi. Equivalently, π\pi is regular if given a function PXP \to X (regarded as a family {P x} xX\{P_x\}_{x\in X}) such that |X|<π{|X|} \lt \pi and |P x|<π{|P_x|} \lt \pi for all xXx \in X, then |P|<π{|P|} \lt \pi. Again equivalently, π\pi is regular if the category Set <π\Set_{\lt\pi} of sets of cardinality <π\lt\pi has all colimits of size <π\lt\pi. The successor of any infinite cardinal, such as 1\aleph_1, is a regular cardinal.

  • A cardinal is called singular if it is not regular. For instance, ω= n n\aleph_\omega = \bigcup_{n\in \mathbb{N}} \aleph_n is singular, more or less by definition, since n< ω\aleph_n \lt \aleph_\omega and ||= 0< ω{|\mathbb{N}|} = \aleph_0 \lt \aleph_\omega.

  • A limit cardinal is one which is not a successor of any other cardinal. Note that every infinite cardinal is a limit ordinal (in the picture where cardinals are identified with certain ordinals).

  • A strong limit cardinal is a cardinal π\pi such that if λ<π\lambda \lt \pi, then 2 λ<π2^\lambda \lt \pi, for any cardinal λ\lambda. Since λ +2 λ\lambda^+ \le 2^\lambda, any strong limit is a limit. Conversely, assuming the continuum hypothesis, every limit is a strong limit. Since 2 λ2^\lambda is the cardinality of the power set P(λ)P(\lambda), a cardinal π\pi is a strong limit iff the category Set <π\Set_{\lt\pi} is an elementary topos.

  • An inaccessible cardinal is any (usually uncountable) regular strong limit cardinal. A weakly inaccessible cardinal is a regular limit cardinal.

  • A cardinal κ\kappa is a measurable cardinal if some (hence any) set of cardinality κ\kappa admits a two-valued measure which is κ\kappa-additive, or equivalently an ultrafilter which is κ\kappa-complete.


The original article is

The book

  • Peter J Cameron, Sets, Logic and Categories (ISBN: 1-85233-056-2 )

contains a very readable account of ZFC and the definitions of both ordinal and cardinal numbers.

Any serious reference on set theory should cover cardinal numbers. The long-established respected tome is

there are also some references listed at

A good introduction to infinite cardinals is:

  • Frank R. Drake, Set Theory: An Introduction to Large Cardinals, Studies in Logic and the Foundations of Mathematics, vol. 76, Elsevier, 1974.

For a much deeper treatment, which assumes most of the material in Drake as a prerequisite, see:

  • Akihiro Kanamori, The Higher Infinite: Large Cardinals in Set Theory from Their Beginning, Springer, 2003.

This is a very readable and freely available historical introduction:

  • Akihiro Kanamori and M. Magidor, The evolution of large cardinal axioms in set theory, in Higher Set Theory, Springer Lecture Notes in Mathematics 669, 1978. (pdf)

Standard approaches start with a material set theory, such as ZFC, whereas the approach here uses structural set theory as emphasized above. Since cardinality is isomorphism-invariant, it is easy to interpret the standard material structurally, although the basic definitions will be different. There does not seem to be a standard account of cardinality from within structural set theory.

For a critical discussion of the history of the meaning of Cantor’s “Kardinalen”, see

which argues that Cantor’s original meaning of set was more like what today is cohesive set and that his Kardinalen refer to the underlying set (see at flat modality).

Last revised on April 2, 2020 at 02:27:57. See the history of this page for a list of all contributions to it.