Try   HackMD

Functors presented by operations and equations

(under construction … up )

Example: Functors on Distributive Lattics

  1. DL is a variety and, as all varieties, has a presentation by operations and equations.

  2. Presentations by operations and equations of functors

    L on a variety were introduced by (Bonsangue and Kurz, 2006, Definition 6), generalising the notion of presentation for
    Set
    -functors as quotients of polynomial functors. The reason to introduce this notion was to show that

    • the category of L-algebras has a presentation by operations and equations that is obtained in modular way from the presentation of the variety and of the functor (Theorem 5);
    • presentations of functors L on a variety correspond, via Stone duality, to logics for coalgebras (Diagram (4), Theorem 27).
  3. Varieties are, in particular, locally finitely presentable (lfp). (The finitely presentable objects are exactly the finite

    DLs.) That
    DL
    is lfp means that every
    DL
    is a filtered colimit of finite
    DL
    s. This result can also be phrased by saying that the inclusion of finitely presentable
    DL
    s
    DLfp→DL

    is dense and that filtered colimits are a density presentation of this inclusion.

    • Theorem (Kelly and Power, 1993): Filtered colimit preserving functors on
      DL
      are precisely those that have a presentation as a coqualiser of polynomial functors.
    • The theorem is actually much more general as it applies not only to
      DL
      but to all categories that are lfp, or even those that are lfp in the enriched sense.
    • While this theorem is beautiful and very general, it has, for our purposes, the disadvantage that the category of arities, that is, the category
      DLfp
      , is too big: We are not interested in presenting functors
      L
      by operations that have distributive lattices as arities. We are only interested in operations that have arities in the usual sense, that is, arities that are finite sets, or, briefly, discrete arities.
    • The solution, then, is to work with a density presentation of the free functor
      Setfin→DL
      , or, equivalently, with a density presentation of the functor
      DLfgf→DL
      , where
      DLfgf
      is the category of finitely generated free algebras.
  4. Let

    DLfgf be the category of finitely generated free
    DL
    s. Then the (ordinary) inclusion
    DLfgf→DL

    is dense and filtered colimits and reflexive coequalisers are a density presentations.

  5. The results in (Velebil and Kurz, 2011) also apply when we consider

    DL as an ordered variety, that is, we turn our attention from the forgetful functor
    DL→Set
    to the forgetful functor
    DL→Pos
    . This has the following consequences.

    • Representing a functor
      L:DL→DL
      now means to represent it by monotone operations (because we work order enriched).
    • DLfgf
      refers to the category of distributive lattices that are free over finite posets (because we consider
      DL→Pos
      ).
    • In other words, being finite posets, arities are not necessarily discrete anymore. This would mean going beyond operations and equations in the sense of universal algebra, which we want to avoid here.

    What modifications to the setting of (Velebil and Kurz, 2011) are needed to get an account of functors that have presentations by finitary monotone operations in discrete arities?

  6. (Kurz and Velebil, 2017) combines the approaches of 4. and 5. above by considering the inclusion

    DLfgf→DL

    as a Poset-enriched functor. This corresponds to restricting to discrete arities (as in 4.) and to monotone operations (as in 5.).

    • Theorem: (Kurz and Velebil, 2017, Proposition 6.8) A functor between ordered varieties has a presentation by monotone operations (in discrete arities) and equations iff it is finitary and preserves surjections.

Question: The two theorems now look the same on ordered varieties where the order is equationally definable: Preservation of filtered colimits and surjections … but one should give all presentations and the other only the monotone ones.

  1. (Balan etal) …

  2. (Dahlqvist and Kurz) …