Bijective correspondence between congruences on universal algebras

190 Views Asked by At

I'm reading the chapter on univesal algebras in P. A. Grillet's Abstract Algebra. In Proposition 1.9 the author states:

Let $ \varphi\colon A\to B $ be a morphism of universal algebras. If $ \mathcal{E} $ is a congruence on $ B $, then $ \varphi^{-1}\mathcal{E} $ defined by $$ {x\mathrel{\varphi^{-1}\mathcal{E}}y}\iff{\varphi x\mathrel{\mathcal{E}}\varphi y} $$ is a congruence on $ A $. If $ \varphi $ is surjective, then $ A/\varphi^{-1}\mathcal{E}\cong B/\mathcal{E} $, and the above defines a one to one correspondence between the congruences on $ B $ and congruences on $ A $ that contain $ \operatorname{ker}\varphi $.

While I encounter no problem finding an isomorphism $ A/\varphi^{-1}\mathcal{E}\to B/\mathcal{E} $, the second part makes me a little bit unsure about how to proceed.

My attempt

Let $ \varphi $ be a surjective homomorphism $ A\to B $, and let's call $ \mathcal{E} $ a congruence on the universal algebra $ A $. Its image $ \varphi\mathcal{E} $ by the Cartesian product of $ \phi $ with itself $$ \varphi\mathcal{E}=\left(\phi^2\right)_{*}\mathcal{E} $$ is again a congruence on $ B $, since can easily find a homomorphism $ B\to A/\mathcal{E} $ such that $ \operatorname{ker}(\pi\upsilon)=\varphi\mathcal{E} $, by composing the canonical projection $ \pi\colon A\to A/\mathcal{E} $ and a right inverse $ \upsilon\colon B\to A $ of $ \varphi $ $$ \begin{array}{ccc} B & \overset{\upsilon}{\longrightarrow} & A & \overset{\pi}{\longrightarrow} & A/\mathcal{E} \end{array} $$ (following the proof strategy that proving the first part of the claim involves). Let $ \upsilon $ be such a right inverse, and let $ x,y $ points of $ B $ such that $ \pi\upsilon(x)=\pi\upsilon(y) $ (i.e., let $ x\mathrel{\operatorname{ker}(\pi\upsilon)}y $); then $ \upsilon x\mathrel{\mathcal{E}}\upsilon y $, and thus $ x=\varphi\upsilon x $ is in relation with $ y=\varphi\upsilon y $. The converse should be quite the same. $ \square $

I'm here concerned about the right inverse $ \upsilon $ not being a homomorphism: it is not always the case.

My questions

Now, is the function $$ \mathcal{F}\colon \begin{align} \left\{\text{congruences on $ A $ which contain $ \operatorname{ker}\phi $}\right\}&\to\left\{\text{congruences on $ B $}\right\}\\ \mathcal{E}&\mapsto \left(\varphi\times\varphi\right)_{*}\mathcal{E} \end{align} $$ by which each congruence $ \mathcal{E} $ is mapped in the set $$ \left(\varphi\times\varphi\right)_{*}\mathcal{E}=\left\{\left(x,y\right):\text{$ x=\varphi x_0 $ and $ y=\varphi y_0 $ such that $ x_0\mathrel{\mathcal{E}}y_0 $}\right\} $$ bijective? (I think so, but I have still to prove it; the inverse of $ \mathcal{F} $ must be the function that maps each congruence $ \mathcal{E} $ on $ B $ with the $ \phi^{-1}\mathcal{E} $ of the beginning of this post). Have I done the previous steps correctly? And, most important, is this kind of "correspondence" that the author had in mind?

I have another question: in what manner the isomorphism $ A/\varphi^{-1}\mathcal{E}\cong B/\mathcal{E} $ is related to this proof? (Maybe it isn't, and the claim of the existence of $ \mathcal{F} $ is only inserted in the same proposition with no direct relation with such, but...)

1

There are 1 best solutions below

0
On

Let $\varphi:\mathbf A \to \mathbf B$ be a homomorphism.
For each congruence $\beta$ of $\mathbf B$, define the binary relation $\alpha$ on $A$ by making $$a \,\alpha\, b \;\Longleftrightarrow\; \varphi(a) \,\beta\, \varphi(b).$$ Apparently you have no difficulty in recognizing this defines a congruence $\alpha$ on $\mathbf A$ (where $\alpha = \varphi^{-1}\beta$).

Now suppose $\varphi$ is surjective. Let $\beta$ and $\beta'$ be congruences on $\mathbf B$ and $\alpha, \alpha'$ be congruences on $\mathbf A$ such that $\alpha = \varphi^{-1}\beta$ and $\alpha' = \varphi^{-1}\beta'$.
To prove the correspondence is one-to-one (that is, injective) we must show that if $\alpha=\alpha'$ then $\beta=\beta'$.
So let $\alpha = \alpha'$ and take $u,v \in B$ such that $(u,v) \in \beta$. Since $\varphi$ is onto, there exist $a,b \in A$ such that $\varphi(a)=u$ and $\varphi(b)=v$. Now, \begin{align} (u,v) \in \beta &\;\Longleftrightarrow\; (\varphi(a),\varphi(b)) \in \beta\\ &\;\Longleftrightarrow\; (a,b) \in \alpha\\ &\;\Longleftrightarrow\; (a,b) \in \alpha'\\ &\;\Longleftrightarrow\; (\varphi(a),\varphi(b)) \in \beta'\\ &\;\Longleftrightarrow\; (u,v) \in \beta'. \end{align} Hence $\beta = \beta'$ and the correspondence is one-to-one.

Now let, again, $\beta$ be a congruence on $\mathbf B$ and $\alpha$ be the congruence on $\mathbf A$ defined by $\alpha = \varphi^{-1}\beta$.
Let $a,b \in A$. \begin{align} (a,b) \in \ker\varphi &\;\Longrightarrow\; \varphi(a) = \varphi(b)\\ &\;\Longrightarrow\; \varphi(a) \,\beta\, \varphi(b)\\ &\;\Longrightarrow\; a \,\alpha\, b, \end{align} where the second implication follows from the reflexivity of $\beta$.
Thus $\ker\varphi \subseteq \varphi^{-1}\beta$.