Nucleous-algebra-and-geometry

From Quotients to Fixed Points: A Three-Layer View of Algebra, Geometry, and Nuclei

One of the oldest questions in algebra is: when can an algebraic structure be transferred to a quotient? Groups require normal subgroups, rings require ideals, modules require submodules, and in universal algebra the fundamental notion is a congruence.

On the other hand, many geometric theories are governed not by quotients but by closure operators. A closure partitions objects into equivalence classes and selects one canonical representative from each class: its fixed point.

This naturally raises a fundamental question: when does a closure operator produce exactly the same algebra as an algebraic quotient? This article proposes a three-layer viewpoint: (1) Algebra — quotients induced by congruences; (2) Geometry — fixed points induced by closures; (3) Algebraic geometry — nuclei as the bridge between the two worlds.

---

1. Algebra: Descending a Monoidal Structure to a Quotient

Definition 1 (Commutative Monoid). A commutative monoid is a triple $(M,+,0)$ satisfying associativity, commutativity, and identity.

Definition 2 (Submonoid). A subset $N\subseteq M$ is called a submonoid if $0\in N$, and whenever $a,b\in N$, then $a+b\in N$. Notice that a submonoid alone does not generally produce a quotient monoid. To construct quotients, one needs an equivalence relation.

Definition 3 (Equivalence Relation). A relation $\sim\;\subseteq M\times M$ is an equivalence relation if it is reflexive, symmetric, and transitive. The quotient set is denoted by $M/\!\sim$.

The natural addition on equivalence classes would be $[a]+[b]=[a+b]$. However, this operation is not automatically well-defined.

Definition 4 (Monoidal Congruence). An equivalence relation $\sim$ is called a monoidal congruence if $a\sim a',\qquad b\sim b'\Longrightarrow a+b\sim a'+b'$.

Theorem 1 (Algebraic Descent). The addition on a commutative monoid descends to the quotient $M/\!\sim$ if and only if $\sim$ is a monoidal congruence. Thus, congruence is precisely the condition that allows an algebraic structure to survive quotienting.

---

2. Geometry: Closure Operators and Fixed Points

We now forget the algebra and consider only geometry.

Definition 5 (Closure Operator). Let $(M,\le)$ be a poset. A map $c:M\to M$ is called a closure operator if it is extensive, i.e. $x\le c(x)$; monotone, i.e. $x\le y\Longrightarrow c(x)\le c(y)$; and idempotent, i.e. $c(c(x))=c(x)$.

Definition 6 (Fixed Point). The set of fixed points is $\operatorname{Fix}(c)=\{x\in M:\;c(x)=x\}$.

Every closure naturally defines an equivalence relation.

Definition 7 (Closure Equivalence). Define $x\sim_c y\iff c(x)=c(y)$. The quotient set is $M/\!\sim_c$.

The remarkable fact is that the quotient and the fixed-point set are canonically identical.

Theorem 2 (Canonical Representatives). The map $[x]\longmapsto c(x)$ defines a canonical bijection $M/\!\sim_c\cong\operatorname{Fix}(c)$. Thus every equivalence class has a unique canonical representative: its closure. Geometrically, quotienting merges equivalent objects, while closure chooses one representative from each class.

---

3. Nuclei: When Quotients Become Algebras Again

We now combine algebra and geometry. Suppose $(M,+,0,\le,c)$ is simultaneously a commutative monoid, a partially ordered set, and equipped with a closure operator.

The closure already produces an equivalence relation $\sim_c$, a quotient $M/\!\sim_c$, and a fixed-point space $\operatorname{Fix}(c)$. The remaining question is purely algebraic: when does the monoidal structure descend through the closure quotient?

Definition 8 (Subadditive Closure / Nucleus). A closure operator is called a nucleus if $c(x)+c(y)\le c(x+y)$ for all $x,y\in M$. This is the compatibility condition between addition and closure.

The closure induces a natural operation on fixed points: $a\star b=c(a+b),\qquad a,b\in\operatorname{Fix}(c)$.

Theorem 3 (Nucleus Generates a Quotient Algebra). If $c$ is a nucleus, then: (1) the closure equivalence $\sim_c$ is a monoidal congruence; (2) the quotient $M/\!\sim_c$ inherits a monoid structure; (3) the quotient monoid is canonically isomorphic to the fixed-point monoid $(M/\!\sim_c,+)\cong(\operatorname{Fix}(c),\star)$, where $a\star b=c(a+b)$.

4. Proofs of the Main Theorems

Proof of Theorem 1 (Algebraic Descent)

($\Rightarrow$) Suppose the addition $[a]+[b]=[a+b]$ is well-defined on $M/\!\sim$. Let $a\sim a'$ and $b\sim b'$. Then $[a]=[a']$ and $[b]=[b']$, so by well-definedness $[a+b]=[a]+[b]=[a']+[b']=[a'+b']$, hence $a+b\sim a'+b'$. This is exactly the definition of a monoidal congruence.

($\Leftarrow$) Suppose $\sim$ is a monoidal congruence, i.e. $a\sim a',\ b\sim b'\Longrightarrow a+b\sim a'+b'$. Define $[a]+[b]:=[a+b]$. To check this is well-defined, let $a,a'$ be two representatives of the same class ($a\sim a'$) and $b,b'$ two representatives of another class ($b\sim b'$). By the congruence property, $a+b\sim a'+b'$, so $[a+b]=[a'+b']$; the result does not depend on the choice of representatives. Associativity, commutativity, and the identity $[0]$ are inherited directly from $M$ since they hold at the level of representatives and pass to classes unchanged. $\blacksquare$

---

Proof of Theorem 2 (Canonical Representatives)

Define $\varphi:M/\!\sim_c\to\operatorname{Fix}(c)$ by $\varphi([x])=c(x)$.

Well-defined. If $x\sim_c y$ then by definition $c(x)=c(y)$, so $\varphi([x])=\varphi([y])$.

Lands in $\operatorname{Fix}(c)$. By idempotence, $c(c(x))=c(x)$, so $c(x)\in\operatorname{Fix}(c)$.

Injective. Suppose $\varphi([x])=\varphi([y])$, i.e. $c(x)=c(y)$. By definition of $\sim_c$, this means $x\sim_c y$, hence $[x]=[y]$.

Surjective. Let $f\in\operatorname{Fix}(c)$, so $c(f)=f$. Then $\varphi([f])=c(f)=f$, so every fixed point is hit.

Since $\varphi$ is well-defined, injective, and surjective, it is a bijection $M/\!\sim_c\cong\operatorname{Fix}(c)$. $\blacksquare$

---

Proof of Theorem 3 (Nucleus Generates a Quotient Algebra)

Assume $c$ is a nucleus: $c(x)+c(y)\le c(x+y)$ for all $x,y\in M$.

(1) $\sim_c$ is a monoidal congruence.

Let $x\sim_c x'$ and $y\sim_c y'$, i.e. $c(x)=c(x')$ and $c(y)=c(y')$. We must show $c(x+y)=c(x'+y')$.

First, since $x\le c(x)$ and $y\le c(y)$ (extensivity), monotonicity gives
$$x+y\le c(x)+c(y).$$
Applying $c$ (monotone) and then the nucleus inequality $c(x)+c(y)\le c(x+y)$ together with idempotence,
$$c(x+y)\le c\big(c(x)+c(y)\big)\le c\big(c(x+y)\big)=c(x+y),$$
so all these are equal; in particular $c(x+y)=c\big(c(x)+c(y)\big)$.

Since $c(x)=c(x')$ and $c(y)=c(y')$, the right-hand side is symmetric in the primed and unprimed data:
$$c(x+y)=c\big(c(x)+c(y)\big)=c\big(c(x')+c(y')\big)=c(x'+y').$$

Hence $x+y\sim_c x'+y'$, proving $\sim_c$ is a monoidal congruence.

(2) The quotient inherits a monoid structure.

This follows directly from part (1) together with Theorem 1: since $\sim_c$ is a monoidal congruence, the operation $[x]+[y]:=[x+y]$ is well-defined on $M/\!\sim_c$, and associativity, commutativity, and the identity class $[0]$ are inherited from $M$.

(3) $(M/\!\sim_c,+)\cong(\operatorname{Fix}(c),\star)$ with $a\star b=c(a+b)$.

Take the bijection $\varphi$ from Theorem 2, $\varphi([x])=c(x)$. We check it is a monoid homomorphism with respect to $\star$:
$$\varphi([x]+[y])=\varphi([x+y])=c(x+y).$$
On the other hand,
$$\varphi([x])\star\varphi([y])=c(x)\star c(y)=c\big(c(x)+c(y)\big).$$
From the computation in part (1), $c(x+y)=c\big(c(x)+c(y)\big)$, so
$$\varphi([x]+[y])=\varphi([x])\star\varphi([y]).$$
Thus $\varphi$ is a monoid isomorphism, and identities match since $\varphi([0])=c(0)$, which acts as the unit for $\star$ because $c(0)\star a=c(c(0)+a)$, and extensivity plus the nucleus/idempotence machinery force this to equal $a$ for $a\in\operatorname{Fix}(c)$. Therefore
$$(M/\!\sim_c,+)\cong(\operatorname{Fix}(c),\star). \qquad\blacksquare$$

---

The Big Picture

The theory naturally separates into three layers.

Layer 1 — Algebra. Congruence determines when algebra descends to a quotient: $(M,+)\longrightarrow M/\!\sim$.

Layer 2 — Geometry. Closure determines canonical representatives: $M\longrightarrow M/\!\sim_c\cong\operatorname{Fix}(c)$.

Layer 3 — Algebra + Geometry. A nucleus is precisely the compatibility condition that makes the two constructions coincide.

$$\text{Congruence}\Longleftrightarrow\text{Quotient Algebra}$$

$$\text{Closure}\Longleftrightarrow\text{Fixed Points}$$

$$\text{Nucleus}\Longrightarrow\text{Quotient Algebra}\cong\text{Fixed-Point Algebra}$$

This viewpoint suggests that nuclei play a role analogous to ideals or normal subgroups: rather than merely producing canonical representatives, they guarantee that these representatives inherit the full algebraic structure of the quotient.