Skip to main content

Section 3.9 Session 9: Retracts and idempotents

At last! The exercises have returned!
The reading this week was interesting in how it approached the relationship between "size of sets" and "properties of maps". The art of counting without counting seemed to be a recurring theme.

Example 3.9.1. Exercise 1:.

...unless the set \(A\) has a point and \(B\) has none...
Solution.
So \(A \mathrel{\vcenter{\huge \triangleleft} \!\!\!\! \vcenter{\tiny \rightarrow}} B\) means that there exists some map \(A \xrightarrow{f} B\text{.}\) In the category of sets, we know that the number of possible maps from \(A \longrightarrow B\) is related to the size of the sets by a formula we arrived at in Session 2: \({|B|}^{|A|}\text{.}\) The only time when this expression for number of maps evaluates to 0 is when \(|B| = 0\) and \(|A| > 0\text{.}\)
Another way of looking at this might to consider the contrapositive of this statement. If we know that there does not exist a map from \(A\) to \(B\text{,}\) that means there must be some origin point in \(A\) for the map to act on but no destination point in \(B\) to send it to. Not having a place to send the point to directly implies \(B=\emptyset\text{.}\) If we also had \(A =\emptyset\) then we’d have a "null map", so the non-existance of a map also implies \(A\) must be non-empty.

Example 3.9.2. Exercise 2:.

...Show that (R) ... [and] (T) ...
Solution.
Let’s start with the reflexive property (R): \(A \leq_R A\text{.}\) This notation indicates that there exist maps \(\set{s,r}\) such that \(A \xrightarrow{s} A \xrightarrow{r} A\) with \(rs = 1_A\text{.}\) Clearly the map \(1_A\) already mets both these conditions since \(1_A \circ 1_A = 1_A\text{.}\)
To prove the transitive property (T), we can construct a retraction for the composition of the functions using a composition of their retractions. If \(A \leq_R B\text{,}\) then there exist maps \(\set{s_1,r_1}\) such that \(A \xrightarrow{s_1} B \xrightarrow{r_1} A\) with \(r_1 \circ s_1 = 1_A\text{.}\) Likewise, \(B \leq_R C\text{,}\) then there exist maps \(\set{s_2,r_2}\) such that \(B \xrightarrow{s_2} C \xrightarrow{r_2} B\) with \(r_2 \circ s_2 = 1_B\text{.}\)
Given these \(\set{s_1,r_1,s_2,r_2}\text{,}\) we can define \(A \xrightarrow{s_3} C = s_2 \circ s_1\) and define \(C \xrightarrow{r_3} A = r_1 \circ r_2\text{.}\) Note that \(r_3 \circ s_3\) can be evaluated as follows using the associative property:
\begin{equation*} r_3 \circ s_3 = (r_1 \circ r_2) \circ (s_2 \circ s_1) \end{equation*}
\begin{equation*} = r_1 \circ (r_2 \circ s_2) \circ s_1 \end{equation*}
\begin{equation*} = r_1 \circ 1_B \circ s_1 = r_1 \circ s_1 = 1_A \end{equation*}
This completes the proof that \(A \leq_R C\text{.}\)

Example 3.9.3. Exercise 3:.

... Use these maps to construct an isomorphism \(A \xrightarrow{f} A'\text{.}\)
Solution.
So we’re given that both \(A \mathrel{\substack{s \\[-0.6ex] \textstyle\longrightarrow \\[-0.6ex] \textstyle\longleftarrow \\[-0.6ex] r}} B\) and \(A' \mathrel{\substack{ s' \\[-0.6ex] \textstyle\longrightarrow \\[-0.6ex] \textstyle\longleftarrow \\[-0.6ex] r'}} B\) "split the same idempotent" \(B \xrightarrow{e} B\text{.}\)
The definition of "splitting an idempotent" tells us that \(sr = e = s'r'\) with \(rs = 1_A\) and \(r's' = 1_{A'}\text{.}\) Given these maps, we’re trying to establish a composition that goes from \(A\) to \(A'\) using \(B\) as a step in between. To get from \(A\) to \(B\) we can use map \(s\text{,}\) and then to get from \(B\) to \(A'\) we can use map \(s'\text{.}\)
Let us define \(A \xrightarrow{f} A'\) by \(r' \circ s\text{.}\) This map should be a well-defined isomorphism with the inverse map given by \(f^{-1} = r \circ s'\text{.}\) To prove this, we must show that \(f^{-1} \circ f = 1_A\) and \(f \circ f^{-1} = 1_{A'}\text{.}\)
By the associative property, the first expression can be simplified as follows:
\begin{equation*} f^{-1} \circ f = (r \circ s') \circ (r' \circ s) \end{equation*}
\begin{equation*} = r \circ (s' \circ r') \circ s \end{equation*}
\begin{equation*} = r \circ e \circ s \end{equation*}
\begin{equation*} = r \circ (s \circ r) \circ s \end{equation*}
\begin{equation*} = (r \circ s) \circ (r \circ s) \end{equation*}
\begin{equation*} = 1_A \circ 1_A = 1_A \end{equation*}
Likewise, the second expression can be similarly simplified:
\begin{equation*} f \circ f^{-1} = (r' \circ s) \circ (r \circ s') \end{equation*}
\begin{equation*} = r' \circ (s \circ r) \circ s' \end{equation*}
\begin{equation*} = r' \circ e \circ s' \end{equation*}
\begin{equation*} = r' \circ (s' \circ r') \circ s' \end{equation*}
\begin{equation*} = (r' \circ s') \circ (r' \circ s') \end{equation*}
\begin{equation*} = 1_{A'} \circ 1_{A'} = 1_{A'} \end{equation*}
These two conditions are sufficient to demonstrate that \(f\) is an isomorphism.