4.7.1 Injective Functions

Let $A$ and $B$ be sets.

A function $f\colon A\to B$ is injective if it satisfies the following condition:

  • (★)
  • For each $a,a'\in A$, if $f(a)=f(a')$, then $a=a'$.

Let $f\colon A\to B$ be a function.

  1. 1.

    Characterisations. The following conditions are equivalent:1

    1. (a)

      The function $f$ is injective.

    2. (b)

      The function $f$ is a monomorphism in $\mathsf{Sets}$.

    3. (c)

      The function $f$ is a retraction in $\mathsf{Sets}$.

    4. (d)

      The direct image function

      \[ f_{!}\colon \mathcal{P}(A)\to \mathcal{P}(B) \]

      associated to $f$ is injective.

    5. (e)

      The inverse image function

      \[ f^{-1}\colon \mathcal{P}(B)\to \mathcal{P}(A) \]

      associated to $f$ is surjective.

    6. (f)

      The codirect image function

      \[ f_{*}\colon \mathcal{P}(A)\to \mathcal{P}(B) \]

      associated to $f$ is injective.

    7. (g)

      The direct image functor

      \[ f_{!}\colon (\mathcal{P}(A),\subset )\to (\mathcal{P}(B),\subset ) \]

      associated to $f$ is full.

    8. (h)

      The codirect image function

      \[ f_{*}\colon \mathcal{P}(A)\to \mathcal{P}(B) \]

      associated to $f$ is full.

    9. (i)

      The direct image functor

      \[ f_{!}\colon (\mathcal{P}(A),\subset )\to (\mathcal{P}(B),\subset ) \]

      associated to $f$ is fully faithful.

    10. (j)

      The inverse image functor

      \[ f^{-1}\colon (\mathcal{P}(B),\subset )\to (\mathcal{P}(A),\subset ) \]

      associated to $f$ is essentially surjective.

    11. (k)

      The codirect image functor

      \[ f_{*}\colon (\mathcal{P}(A),\subset )\to (\mathcal{P}(B),\subset ) \]

      associated to $f$ is fully faithful.

    12. (l)

      The diagram

      commutes. That is, we have

      \[ f^{-1}(f(a))=\left\{ a\right\} \]

      for each $a\in A$.

    13. (m)

      We have

      In other words, we have

      \[ \left\{ a\in A\ \middle |\ f(a)\in f(U)\right\} =U \]

      for each $U\in \mathcal{P}(A)$.

    14. (n)

      We have

      In other words, we have

      \[ \left\{ a\in A\ \middle |\ f^{-1}(f(a))\subset U\right\} =U \]

      for each $U\in \mathcal{P}(A)$.

  2. 2.

    Interaction With Composition. Let $f\colon A\to B$ and $g\colon B\to C$ be functions.

    1. (a)

      If $f$ and $g$ are injective, then so is $g\circ f$.

    2. (b)

      If $g\circ f$ is injective, then so is $f$.

    3. (c)

      If $f$ and $g\circ f$ are injective, then $f$ may fail to be injective.

    4. (d)

      The class of injective functions does not satisfy the two-out-of-three property of Unresolved reference.

  3. 3.

    Interaction With Identities. For any set $A$, the identity function $\operatorname {\mathrm{id}}_{A}$ is injective.


  1. 1Item 1d, Item 1f, Item 1g, and Item 1h unwind respectively to the following statements:
    • For each $U,V\in \mathcal{P}(A)$, if $f_{!}(U)=f_{!}(V)$, then $U=V$.

    • For each $U,V\in \mathcal{P}(A)$, if $f_{*}(U)=f_{*}(V)$, then $U=V$.

    • For each $U,V\in \mathcal{P}(A)$, if $f_{!}(U)\subset f_{!}(V)$, then $U\subset V$.

    • For each $U,V\in \mathcal{P}(A)$, if $f_{*}(U)\subset f_{*}(V)$, then $U\subset V$.

Item 1: Characterisations
We will proceed by showing:

Step 1: Item 1a$\iff $Item 1b
We claim that Item 1a and Item 1b are equivalent:

  • Item 1a$\implies $Item 1b: We proceed in a few steps:

    • Proceeding by contrapositive, we claim that given a pair of maps $g,h\colon C\rightrightarrows A$ such that $g\neq h$, we have $f\circ g\neq f\circ h$.

    • Indeed, as $g$ and $h$ are different maps, there must exist at least one element $x\in C$ such that $g(x)\neq h(x)$.

    • But then we have $f(g(x))\neq f(h(x))$, as $f$ is injective.

    • Thus $f\circ g\neq f\circ h$, and we are done.

  • Item 1b$\implies $Item 1a: We proceed in a few steps:

    • Consider the diagram

      where $[x]$ and $[y]$ are the morphisms picking the elements $x$ and $y$ of $A$.

    • Note that we have $f(x)=f(y)$ iff $f\circ [x]=f\circ [y]$.

    • Since $f$ is assumed to be a monomorphism, if $f(x)=f(y)$, then $f\circ [x]=f\circ [y]$ and therefore $[x]=[y]$.

    • This shows that if $f(x)=f(y)$, then $x=y$, so $f$ is injective.

Step 2: Item 1a$\iff $Item 1c
Omitted.

Step 3: Item 1a$\iff $Item 1d
We claim that Item 1a and Item 1d are indeed equivalent:

  • Item 1a$\implies $Item 1d: We proceed in a few steps:

    • Assume that $f$ is injective and let $U,V\in \mathcal{P}(A)$ such that $f_{!}(U)=f_{!}(V)$. We wish to show that $U=V$.

    • To show that $U\subset V$, let $u\in U$.

    • By the definition of the direct image, we have $f(u)\in f_{!}(U)$.

    • Since $f_{!}(U)=f_{!}(V)$, it follows that $f(u)\in f_{!}(V)$.

    • Thus, there exists some $v\in V$ such that $f(v)=f(u)$.

    • Since $f$ is injective, the equality $f(v)=f(u)$ implies that $v=u$.

    • Thus $u\in V$ and $U\subset V$.

    • A symmetric argument shows that $V\subset U$.

    • Therefore $U=V$, showing $f_{!}$ to be injective.

  • Item 1d$\implies $Item 1a: We proceed in a few steps:

    • Assume that the direct image function $f_{!}$ is injective and let $a,a'\in A$ such that $f(a)=f(a')$. We wish to show that $a=a'$.

    • Since

      \begin{align*} f_{!}(\left\{ a\right\} ) & = \left\{ f(a)\right\} \\ & = \left\{ f(a')\right\} \\ & = f_{!}(\left\{ a’\right\} ), \end{align*}

      we must have $\left\{ a\right\} =\left\{ a'\right\} $, as $f_{!}$ is injective, so $a=a'$, showing $f$ to be injective.

Step 4: Item 1d$\iff $Item 1f
This follows from Item 17 of Proposition 4.6.1.1.5.

Step 5: Item 1d$\iff $Item 1g
We claim that Item 1d and Item 1g are equivalent:

  • Item 1d$\implies $Item 1g: We proceed in a few steps:

    • Let $U,V\in \mathcal{P}(A)$ such that $f_{!}(U)\subset f_{!}(V)$, assume $f_{!}$ to be injective, and consider the set $U\cup V$.

    • Since $f_{!}(U)\subset f_{!}(V)$, we have

      \begin{align*} f_{!}(U\cup V) & = f_{!}(U)\cup f_{!}(V)\\ & = f_{!}(V), \end{align*}

      where we have used Item 5 of Proposition 4.6.1.1.5 for the first equality.

    • Since $f_{!}$ is injective, this implies $U\cup V=V$.

    • Thus $U\subset V$, as we wished to show.

  • Item 1d$\implies $Item 1g: We proceed in a few steps:

    • Suppose Item 1g holds, and let $U,V\in \mathcal{P}(A)$ such that $f_{!}(U)=f_{!}(V)$.

    • Since $f_{!}(U)=f_{!}(V)$, we have $f_{!}(U)\subset f_{!}(V)$ and $f_{!}(V)\subset f_{!}(U)$.

    • By assumption, this implies $U\subset V$ and $V\subset U$.

    • Thus $U=V$, showing $f_{!}$ to be injective.

Step 6: Item 1g$\iff $Item 1h
This follows from Item 17 of Proposition 4.6.1.1.5.

Step 7: Item 1a$\iff $Item 1l
We have

\[ f^{-1}(f(a))=\left\{ a'\in A\ \middle |\ f(a')=f(a)\right\} \]

so the condition $f^{-1}(f(a))=\left\{ a\right\} $ states precisely that if $f(a')=f(a)$, then $a'=a$.

Step 8: Item 1l$\iff $Item 1m
We claim that Item 1l and Item 1m are indeed equivalent:

  • Item 1l$\implies $Item 1m: We have

    \begin{align*} [f^{-1}\circ f_{!}](U) & \mathrel {\smash {\overset {\mathclap {\scriptscriptstyle \text{def}}}=}}f^{-1}(f_{!}(U))\\ & = f^{-1}\left(f_{!}\left(\bigcup _{u\in U}\left\{ u\right\} \right)\right)\\ & = f^{-1}\left(\bigcup _{u\in U}f_{!}(\left\{ u\right\} )\right)\\ & = \bigcup _{u\in U}f^{-1}(f_{!}(\left\{ u\right\} ))\\ & = \bigcup _{u\in U}f^{-1}(f_{!}(u))\\ & = \bigcup _{u\in U}\left\{ u\right\} \\ & = U \end{align*}

    for each $U\in \mathcal{P}(A)$, where we have used Item 5 of Proposition 4.6.1.1.5 for the third equality and Item 5 of Proposition 4.6.2.1.3 for the fourth equality.

  • Item 1m$\implies $Item 1l: Applying the condition $f^{-1}\circ f_{!}=\operatorname {\mathrm{id}}_{\mathcal{P}(A)}$ to $U=\left\{ a\right\} $ gives

    \[ f^{-1}(f_{!}(\left\{ a\right\} ))=\left\{ a\right\} . \]

Step 9: Item 1a$\iff $Item 1n
We claim that Item 1a and Item 1n are equivalent:

  • Item 1a$\implies $Item 1n: If $f$ is injective, then $f^{-1}(f(a))=\left\{ a\right\} $, so we have

    \begin{align*} f^{-1}(f_{*}(a)) & = \left\{ a\in A\ \middle |\ \left\{ a\right\} \subset U\right\} \\ & = U. \end{align*}
  • Item 1n$\implies $Item 1a: For $U=\left\{ a\right\} $, the condition $f^{-1}(f_{*}(U))=U$ becomes

    \[ \left\{ a'\in A\ \middle |\ f^{-1}(f(a'))\subset \left\{ a\right\} \right\} =\left\{ a\right\} . \]

    Since the set $f^{-1}(f(a'))$ is given by

    \[ \left\{ a\in A\ \middle |\ f(a)=f(a')\right\} , \]

    it follows that $f$ is injective.

This finishes the proof.

Step 10: Item 1m$\implies $Item 1e
Given $U\in \mathcal{P}(A)$, we must show there exists some $V\in \mathcal{P}(B)$ such that $f^{-1}(V)=U$. Choosing $V=f_{!}(U)$, it follows from Item 1m that we have $f^{-1}(f_{!}(U))=U$. Thus $f^{-1}$ is surjective.

Step 11: Item 1e$\implies $Item 1a
We proceed in a few steps:

  • Suppose that $f^{-1}$ is surjective and let $a,a'\in A$ with $f(a)=f(a')$.

  • Since $f^{-1}$ is surjective, there exists some $V\in \mathcal{P}(B)$ such that $f^{-1}(V)=\left\{ a\right\} $.

  • This means that $x\in \left\{ a\right\} $ iff $f(a)\in V$.

  • Since $f(a)=f(a')$, we have $f(a')\in V$.

  • This means that $a'\in \left\{ a\right\} $, so we must have $a=a'$.

  • Thus $f$ is injective.

This finishes the proof.

Step 12: Item 1g$\iff $Item 1i
This follows from the fact that $\mathcal{P}(B)$ is locally posetal.

Step 13: Item 1h$\iff $Item 1j
This follows from the fact that $\mathcal{P}(A)$ is locally posetal.

Step 14: Item 1e$\iff $Item 1k
This follows from the fact that $\mathcal{P}(B)$ is locally posetal.

Item 2: Interaction With Composition
We claim that Item 2a and Item 2b are indeed true:

  • Item 2a: Suppose $f$ and $g$ injective and let $a,a'\in A$ such that $[g\circ f](a)=[g\circ f](a')$. Then $g(f(a)) = g(f(a'))$, so by injectivity of $g$, we have $f(a) = f(a')$, and by injectivity of $f$, we have $a = a'$. Thus $g \circ f$ is injective.

  • Item 2b: Suppose $g\circ f$ is injective and let $a,a'\in A$ such that $f(a) = f(a')$. We have

    \begin{align*} [g\circ f](a) & \mathrel {\smash {\overset {\mathclap {\scriptscriptstyle \text{def}}}=}}g(f(a))\\ & = g(f(a'))\\ & \mathrel {\smash {\overset {\mathclap {\scriptscriptstyle \text{def}}}=}}[g\circ f](a’), \end{align*}

    so by the injectivity of $g\circ f$, we have $a = a'$. Thus $f$ must be injective.

  • Item 2c: Let:

    • $A=\left\{ 1\right\} $;

    • $B=\left\{ p,q\right\} $;

    • $C=\left\{ z\right\} $;

    • $f(1)=p$;

    • $g(p)=z$ and $g(q)=z$.

    Since $A$ is a singleton set, the functions $f$ and $g\circ f$ are trivially injective. However, $g(p)=g(q)$ while $p\neq q$, so $g$ fails to be injective.

  • Item 2d: This follows from Item 2c.

This finishes the proof.

Item 3: Interaction With Identities
Let $a,a'\in A$ such that $\operatorname {\mathrm{id}}_{A}(a)=\operatorname {\mathrm{id}}_{A'}(a')$. We have

\begin{align*} a & \mathrel {\smash {\overset {\mathclap {\scriptscriptstyle \text{def}}}=}}\operatorname {\mathrm{id}}_{A}(a)\\ & = \operatorname {\mathrm{id}}_{A'}(a')\\ & \mathrel {\smash {\overset {\mathclap {\scriptscriptstyle \text{def}}}=}}a’ \end{align*}

so $\operatorname {\mathrm{id}}_{A}$ is injective.


Noticed something off, or have any comments? Feel free to reach out!


You can also use the contact form below: