I'm doing this thread because I have some ideas about how to define self-similarity in algebra, but I don't know if this is known at all. Any critics, comments and references are more than welcomed. The philosophy behind this is to see algebra as the study of relations, that is, of at least binary laws. We therefore impose that there is no $0-ary$ or $1-ary$ relations in its signature (if we want to use this concept to formalize what's a structure). For simplicity, I'll only focus on binary laws.

Let $A$ be some set and let $.$ be some binary law $.$ : $A \times A \rightarrow A$. We say that an element $1_A$ of $A$ is a unit relative to $A$ if for all $x$, in $A$, $1_A.x = x.1_A = x$. A structure $(A,.)$ that admits a unit relative to $A$ is called unital.

Let $B$ be a subset of $A$. We say that $(B, ._B)$ is a supstructure of $A$ if it is closed under $._B$, the restriction of $.$ to $B$. Such supstruture is unital if there exists some element $1_B$ in $B$ such that for all $x$ in $B$, $1_B . x = x. 1_B = x$.

We now use the word magna, monoid, ring, $\ldots$ with their common use. Example: Let the ordinal $\omega n$, and define a binary law $+$ such that if $x$ is in $\omega m$ but not in $\omega p$ with $m > p$, then for any $y$ in $\omega p$, $x+ y = x$. That is, $x$ absorbs the elements that are of lower infinite degree. The addition is defined as on $\mathbb{N}$ for elements of the same infinite degree. Clearly, $\omega n$ has $n$ relative units. This clearly gives non trivial example of supmonoids of some monoid (there are a lots of them).

A morphism of "struture" is any $f : (A,.) \rightarrow (B,.)$ such that for all $x, y$, $f(x.y) = f(x).f(y)$. We don't require that the global units are preserved. A morphism $f : (A,.) \rightarrow (B,.)$ can therefore map $A$ to a non trivial supstructure of $B$.

A structure $A$ is said to be self-similar if it is isomorphic to one of its non trivial supstructure $B$. Clearly, if this happens, there exists an infinity of such isomorphisms. Example: $(\mathbb{N},+)$ seen as a monoid is obviously isomorphic to the even numbers or any $k \mathbb{N}$. The field of complex numbers admits such isomorphism too.

Let $B$ be a non trivial supstructure in $A$. Such supstructure is said to repeat itself in $A$ if there exists an isomorphism from $B$ to another supstruture of $A$ disjoint from $B$. If $B$ is moreover self-similar, we say that $B$ is repeatedly self-similar. Example: For $(\omega n,+)$, $\omega$ is repeatedly self-similar. (We could also define its indice in $A$, which would be $n$)

If $+$ is commutative, we say that $x$ $+$-absorbs $y$ if $x + y = x$. We define the antonymic law $\#$ associated to $+$ as follow :

  • $x$ $\#$-absorbs $y$ iff $y$ $+$-absorbs $x$ (in other words, $x \# y = x$ iff $x + y = y$

  • $x \# y = x + y$ otherwise.

Clearly, if $+$ is commutative, $\#$ is commutative. If $1$ is a unit relative to $(A,+)$, then clearly $1$ is an absorbing element relative to $(A,\#)$. If $+$ is associative, then $\#$ is associative too (case by case check). This shows a new kind of duality neutral element $\leftrightarrow$ absorbing elements. Indeed a neutral element is an element absorbed by everyone, hence the name antonymic law. Let me add that algebraically speaking, the only ``ontology'' one can give to neutral elements or absorbing elements is that they are idempotent. Any idempotent element can be seen as a relative neutral or relative absorbing element of some (not necessarily non trivial) supstructure.

We define an infmonoid to be the dual of a monoid, that is a structure $(M,+,i)$ where

  • $+$ is associative,

  • $i$ is an absorbing element.

There are many such example of infmonoids. We could obviously always define the dual of any known algebraic stuctures, say, infgroups, $\ldots$ As a side remark, note that the concept of annulus $(M,+,.,0,i)$ defined as

  • $(M,+,0)$ is a commutative monoid,

  • $(M,.,i)$ is an infmonoid,

  • $.$ is distributive over $+$,

is in my opinion a much more elegant concept than the one of semi-ring (In particular, $i = 0$ iff $i$ is $+$-invertible). A field is a special case of annulus, and can (if I did no mistake) be formulated entirely algebraically (that is, it can be easily categorified). Some example of annulus are given by what I call the ``capped annulus'' $(C_{n+1}, \oplus, \otimes, n, 0)$ (finite counterpart of max-plus algebra) defined on ${0,1, \ldots, n}$ with $x \oplus y = min(x,y)$ and $x \otimes y = min(x \times y, n)$. One can easily define their additive counterpart.

Now, say that a structure $(A,.)$ is antonymic symmetric iff it is isomorphic to $(A,\#)$. Example: $(\mathbb{N} \cup \{\infty\},+)$ is clearly antonymic symmetric. The $2$-elements Boolean algebra seen as a lattice is antonymic symmetric ($\#$ is nothing else than the swap of operations). In particular, if $L$ is a totally ordered bounded lattice, then it is antonymic symmetric. An antonymic symmetric structure $(A,.)$ that is self-similar to one of its supstruture $(B,.)$ is called entirely self-similar. Example: $\omega^2 +1 $ is clearly entirely self-similar. Does anyone have any other example of entirely self-similar strutures ? In particular, non commutative ones ?


closed as off-topic by Eric Wofsey, Emil Jeřábek, Ryan Budney, Yemon Choi, Neil Strickland Nov 3 '14 at 11:27

This question appears to be off-topic. The users who voted to close gave this specific reason:

  • "MathOverflow is for mathematicians to ask each other questions about their research. See Math.StackExchange to ask general questions in mathematics." – Eric Wofsey, Emil Jeřábek, Ryan Budney
If this question can be reworded to fit the rules in the help center, please edit the question.

  • 1
    $\begingroup$ If you mean to ask more questions than the one question you have at the end (which is only even related to about a third of your post), you should state them explicitly. $\endgroup$ – Eric Wofsey Nov 1 '14 at 14:29

I haven't read your description (and don't know the area well enough) to make an evaluation, but as far as algebraic treatments go... have you seen Tom Leinster's Theory of Self-Similarity? This is based on categorical algebra and focuses on terminal coalgebras $X$ for suitably defined endofunctors $G = M \otimes -$ based on "equational systems" which are suitable "nondegenerate" bimodules $M: \mathbb{A}^{op} \times \mathbb{A} \to Set$ on a small category. (I'm skipping over a bunch of stufff; you'll have to read the paper, which is reasonably self-contained.) The structure of a terminal coalgebra $X \to G(X)$ is (by a famous result of Lambek) an isomorphism, but not just any old isomorphism -- it's a canonical isomorphism.

I might suggest that you try and tackle a few more examples in your setting; for example, can your theory single out Julia sets of various types? One thing that worries me vaguely (again I haven't read your description closely) is that mere existence of an isomorphism would let many examples through the door that are not of much mathematical interest. So a question is: is there something in your theory that lets it sift out self-similar structures of genuine interest from the rest?

  • $\begingroup$ Thanks for this reference, I'm going to try to understand this. Still, I think its important to consider semi-functors to really make sense of self-similarity in category theory. There is fundamentally no reason to give such an objective ontology to identities and absorbing elements : those are idempotent defined relatively. (That is, in the same spirit of what is a pointed object) $\endgroup$ – sure Nov 1 '14 at 13:49

Not the answer you're looking for? Browse other questions tagged or ask your own question.