ホモトピー型理論

[0065] 自然変換

自然変換は関手の間の射の概念である。実際、関手を対象、自然変換を射とする前圏を構成できる([0069])。さらに、終域が圏であるような関手のなす前圏は圏であることを示す([006C])。

[0066] 定義

\(i\)を階数、\(C,D:\mathord {\textnormal {\textsf {PreCat}}}(i)\)を前圏、\(F,G:\mathord {\textnormal {\textsf {Fun}}}(C,D)\)を関手とする。型\(\mathord {\textnormal {\textsf {NatTrans}}}(F,G):\mathcal {U}(i)\)

\(\lbrace t:\prod _{x:C}\mathord {\textnormal {\textsf {Map}}}(F(x),G(x))\mid \forall _{x_{1},x_{2}:C}\forall _{f:\mathord {\textnormal {\textsf {Map}}}(x_{1},x_{2})}G(f)\circ t(x_{1})=t(x_{2})\circ F(f)\rbrace \)
と定義する。\(\mathord {\textnormal {\textsf {NatTrans}}}(F,G)\)の要素を\(F\)から\(G\)への自然変換(natural transformation)と呼ぶ。自然変換\(t:\mathord {\textnormal {\textsf {NatTrans}}}(F,G)\)が満たすべき性質(\(\forall _{x_{1},x_{2}:C}\forall _{f:\mathord {\textnormal {\textsf {Map}}}(x_{1},x_{2})}G(f)\circ t(x_{1})=t(x_{2})\circ F(f)\))を\(t\)自然性(naturality)と言う。

[0067] 演習

\(i\)を階数、\(C,D:\mathord {\textnormal {\textsf {PreCat}}}(i)\)を前圏とする。

  1. 1関手\(F:\mathord {\textnormal {\textsf {Fun}}}(C,D)\)に対して、恒等自然変換\(\mathord {\textnormal {\textsf {id}}}\lbrace F\rbrace :\mathord {\textnormal {\textsf {NatTrans}}}(F,F)\)\(\lambda x.\mathord {\textnormal {\textsf {id}}}\lbrace F(x)\rbrace \)と定義する。\(\mathord {\textnormal {\textsf {id}}}\lbrace F\rbrace \)の自然性を確認せよ。
  2. 2関手\(F_{1},F_{2},F_{3}:\mathord {\textnormal {\textsf {Fun}}}(C,D)\)と自然変換\(t_{1}:\mathord {\textnormal {\textsf {NatTrans}}}(F_{1},F_{2})\)\(t_{2}:\mathord {\textnormal {\textsf {NatTrans}}}(F_{2},F_{3})\)に対して、合成\(t_{2}\circ t_{1}:\mathord {\textnormal {\textsf {NatTrans}}}(F_{1},F_{3})\)\(\lambda x.t_{2}(x)\circ t_{1}(x)\)と定義する。\(t_{2}\circ t_{1}\)の自然性を確認せよ。

[0068] 命題

関数外延性を仮定する。\(i\)を階数、\(C,D:\mathord {\textnormal {\textsf {PreCat}}}(i)\)を前圏、\(F,G:\mathord {\textnormal {\textsf {Fun}}}(C,D)\)を関手、\(s,t:\mathord {\textnormal {\textsf {NatTrans}}}(F,G)\)を自然変換とすると、同値

\((s=t)\simeq (\forall _{x:C}s(x)=t(x))\)
を構成できる。

[0069] 定義

関数外延性を仮定する。\(i\)を階数、\(C,D:\mathord {\textnormal {\textsf {PreCat}}}(i)\)を前圏とする。前圏\(\mathord {\textnormal {\textsf {Fun}}}^{(\mathord {\textnormal {\textsf {Cat}}})}(C,D):\mathord {\textnormal {\textsf {PreCat}}}(i)\)を次のように定義する。

  • -\(\mathord {\textnormal {\textsf {Obj}}}\equiv \mathord {\textnormal {\textsf {Fun}}}(C,D)\)
  • -\(\mathord {\textnormal {\textsf {Map}}}\equiv \lambda (F,G).\mathord {\textnormal {\textsf {NatTrans}}}(F,G)\)
  • -恒等射と合成は[0067]の通り。
  • -残りは[0068]から容易に示せる。

[006A] 命題

関数外延性を仮定する。\(i\)を階数、\(C,D:\mathord {\textnormal {\textsf {PreCat}}}(i)\)を前圏、\(F,G:\mathord {\textnormal {\textsf {Fun}}}(C,D)\)を関手、\(t:\mathord {\textnormal {\textsf {NatTrans}}}(F,G)\)を自然変換とする。次は論理的に同値である。

  1. 1任意の\(x:C\)に対して、\(t(x):\mathord {\textnormal {\textsf {Map}}}(F(x),G(x))\)は同型である。
  2. 2\(t\)は前圏\(\mathord {\textnormal {\textsf {Fun}}}^{(\mathord {\textnormal {\textsf {Cat}}})}(C,D)\)の射として同型である。

証明

2から1は自明である。

1から2を示す。各\(x:C\)に対して、射\(s(x),u(x):\mathord {\textnormal {\textsf {Map}}}(G(x),F(x))\)と同一視\(p(x):s(x)\circ t(x)=\mathord {\textnormal {\textsf {id}}}\)\(q(x):t(x)\circ u(x)=\mathord {\textnormal {\textsf {id}}}\)を得る。[0068]より、\(s\)\(u\)の自然性を確認すれば十分である。\(x_{1},x_{2}:C\)を対象、\(f:\mathord {\textnormal {\textsf {Map}}}(x_{1},x_{2})\)を射とする。\(F(f)\circ s(x_{1})=s(x_{2})\circ G(f)\)を言うには、[005G]より\(F(f)\circ s(x_{1})\circ t(x_{1})=s(x_{2})\circ G(f)\circ t(x_{1})\)を示せば十分で、これは次のように分かる。

\(F(f)\circ s(x_{1})\circ t(x_{1})\)\(=\) {\(p(x_{1})\)}\(F(f)\)\(=\) {\({(p(x_{2}))}^{-1}\)}\(s(x_{2})\circ t(x_{2})\circ F(f)\)\(=\) {\(t\)の自然性}\(s(x_{2})\circ G(f)\circ t(x_{1})\)
\(u\)の自然性も同様である。

[006C] 命題

関数外延性を仮定する。\(i\)を階数、\(C,D:\mathord {\textnormal {\textsf {PreCat}}}(i)\)を前圏とする。\(D\)が圏ならば、\(\mathord {\textnormal {\textsf {Fun}}}^{(\mathord {\textnormal {\textsf {Cat}}})}(C,D)\)は圏である。

証明

\(F:\mathord {\textnormal {\textsf {Fun}}}^{(\mathord {\textnormal {\textsf {Cat}}})}(C,D)\)を対象とする。[006A]よりレトラクト\((\sum _{G:\mathord {\textnormal {\textsf {Fun}}}^{(\mathord {\textnormal {\textsf {Cat}}})}(C,D)}F\cong G)\triangleleft (\sum _{G:\mathord {\textnormal {\textsf {Fun}}}(C,D)}\lbrace t:\mathord {\textnormal {\textsf {NatTrans}}}(F,G)\mid \forall _{x:C}\mathord {\textnormal {\textsf {IsIso}}}(t(x))\rbrace )\)を得る。後者が可縮であることを示すには、[006B]より

\(\sum _{G_{0}:C.\mathord {\textnormal {\textsf {Obj}}}\to D.\mathord {\textnormal {\textsf {Obj}}}}\sum _{G_{1}:\prod _{\lbrace x_{1},x_{2}:C\rbrace }\mathord {\textnormal {\textsf {Map}}}(x_{1},x_{2})\to \mathord {\textnormal {\textsf {Map}}}(G_{0}(x_{1}),G_{0}(x_{2}))}\lbrace t:\prod _{x:C}F(x)\cong G_{0}(x)\mid \forall _{x_{1},x_{2}:C}\forall _{f:\mathord {\textnormal {\textsf {Map}}}(x_{1},x_{2})}G_{1}(f)\circ t(x_{1})=t(x_{2})\circ F(f)\rbrace \)
が可縮であることを示せばよい。\(\sum _{G_{0}:C.\mathord {\textnormal {\textsf {Obj}}}\to D.\mathord {\textnormal {\textsf {Obj}}}}\prod _{x:C}F(x)\cong G_{0}(x)\)の部分は\(\prod _{x:C}\sum _{y:D}F(x)\cong y\)のレトラクトなので、[0029]\(D\)が圏であるという仮定から可縮である。よって、件の型は\(\lbrace G_{1}:\prod _{\lbrace x_{1},x_{2}:C\rbrace }\mathord {\textnormal {\textsf {Map}}}(x_{1},x_{2})\to \mathord {\textnormal {\textsf {Map}}}(F(x_{1}),F(x_{2}))\mid \forall _{x_{1},x_{2}:C}\forall _{f:\mathord {\textnormal {\textsf {Map}}}(x_{1},x_{2})}G_{1}(f)\circ \mathord {\textnormal {\textsf {id}}}=\mathord {\textnormal {\textsf {id}}}\circ F(f)\rbrace \)のレトラクトである。この型はさらに\(\prod _{x_{1},x_{2}:C}\prod _{f:\mathord {\textnormal {\textsf {Map}}}(x_{1},x_{2})}\lbrace g:\mathord {\textnormal {\textsf {Map}}}(F(x_{1}),F(x_{2}))\mid g=F(f)\rbrace \)のレトラクトであるが、これは[0029][0026]より可縮である。

\(\mathord {\textnormal {\textsf {Fun}}}^{(\mathord {\textnormal {\textsf {Cat}}})}(C,D)\)普遍性でも特徴づけられる([006R])。

[006Q] 定義

\(i\)を階数、\(C_{1},C_{2},D:\mathord {\textnormal {\textsf {PreCat}}}(i)\)を前圏とする。型\(\mathord {\textnormal {\textsf {BiFun}}}(C_{1},C_{2};D):\mathcal {U}(i)\)を次のレコード型と定義する。

  • -\(\mathord {\textnormal {\textsf {obj}}}:C_{1}.\mathord {\textnormal {\textsf {Obj}}}\to C_{2}.\mathord {\textnormal {\textsf {Obj}}}\to D.\mathord {\textnormal {\textsf {Obj}}}\)
  • -\(\mathord {\textnormal {\textsf {map}}}_{1}:\prod _{\lbrace x_{11},x_{12}:C_{1}\rbrace }\mathord {\textnormal {\textsf {Map}}}(x_{11},x_{12})\to (\prod _{x_{2}:C_{2}}\mathord {\textnormal {\textsf {Map}}}(\mathord {\textnormal {\textsf {obj}}}(x_{11},x_{2}),\mathord {\textnormal {\textsf {obj}}}(x_{12},x_{2})))\)
  • -\(\mathord {\textnormal {\textsf {map}}}_{2}:\prod _{x_{1}:C_{1}}\prod _{\lbrace x_{21},x_{22}:C_{2}\rbrace }\mathord {\textnormal {\textsf {Map}}}(x_{21},x_{22})\to \mathord {\textnormal {\textsf {Map}}}(\mathord {\textnormal {\textsf {obj}}}(x_{1},x_{21}),\mathord {\textnormal {\textsf {obj}}}(x_{1},x_{22}))\)
  • -\(\_:\forall _{x_{1}:C_{1}}\forall _{x_{2}:C_{2}}\mathord {\textnormal {\textsf {map}}}_{1}(\mathord {\textnormal {\textsf {id}}}\lbrace x_{1}\rbrace ,x_{2})=\mathord {\textnormal {\textsf {id}}}\lbrace \mathord {\textnormal {\textsf {obj}}}(x_{1},x_{2})\rbrace \)
  • -\(\_:\forall _{x_{11},x_{12},x_{13}:C_{1}}\forall _{x_{2}:C_{2}}\forall _{f_{1}:\mathord {\textnormal {\textsf {Map}}}(x_{11},x_{12})}\forall _{f_{2}:\mathord {\textnormal {\textsf {Map}}}(x_{12},x_{13})}\mathord {\textnormal {\textsf {map}}}_{1}(f_{2}\circ f_{1},x_{2})=\mathord {\textnormal {\textsf {map}}}_{1}(f_{2},x_{2})\circ \mathord {\textnormal {\textsf {map}}}_{1}(f_{1},x_{2})\)
  • -\(\_:\forall _{x_{1}:C_{1}}\forall _{x_{2}:C_{2}}\mathord {\textnormal {\textsf {map}}}_{2}(x_{1},\mathord {\textnormal {\textsf {id}}}\lbrace x_{2}\rbrace )=\mathord {\textnormal {\textsf {id}}}\lbrace \mathord {\textnormal {\textsf {obj}}}(x_{1},x_{2})\rbrace \)
  • -\(\_:\forall _{x_{1}:C_{1}}\forall _{x_{21},x_{22},x_{23}:C_{2}}\forall _{f_{1}:\mathord {\textnormal {\textsf {Map}}}(x_{21},x_{22})}\forall _{f_{2}:\mathord {\textnormal {\textsf {Map}}}(x_{22},x_{23})}\mathord {\textnormal {\textsf {map}}}_{2}(x_{1},f_{2}\circ f_{1})=\mathord {\textnormal {\textsf {map}}}_{2}(x_{1},f_{2})\circ \mathord {\textnormal {\textsf {map}}}_{2}(x_{1},f_{1})\)
  • -\(\_:\forall _{x_{11},x_{12}:C_{1}}\forall _{x_{21},x_{22}:C_{2}}\forall _{f_{1}:\mathord {\textnormal {\textsf {Map}}}(x_{11},x_{12})}\forall _{f_{2}:\mathord {\textnormal {\textsf {Map}}}(x_{21},x_{22})}\mathord {\textnormal {\textsf {map}}}_{1}(f_{1},x_{22})\circ \mathord {\textnormal {\textsf {map}}}_{2}(x_{11},f_{2})=\mathord {\textnormal {\textsf {map}}}_{2}(x_{12},f_{2})\circ \mathord {\textnormal {\textsf {map}}}_{1}(f_{1},x_{21})\)
\(\mathord {\textnormal {\textsf {BiFun}}}(C_{1},C_{2};D)\)の要素を\(C_{1},C_{2}\)から\(D\)への双関手(bifunctor)と呼ぶ。

[006R] 演習

関数外延性を仮定する。\(i\)を階数、\(C,D,X:\mathord {\textnormal {\textsf {PreCat}}}(i)\)を前圏とする。同値

\(\mathord {\textnormal {\textsf {Fun}}}(X,\mathord {\textnormal {\textsf {Fun}}}^{(\mathord {\textnormal {\textsf {Cat}}})}(C,D))\simeq \mathord {\textnormal {\textsf {BiFun}}}(C,X;D)\)
を構成せよ。