跳转至

1 图与范畴

1.1 范畴图

  1. \(O, M\) 均为类且存在函数 \(\Phi: M \rightarrow O^{2}\) 使得 \(G = (O, \Phi[O])\) 是类上的有向图,则称 \(\mathcal G = \left<O, M, \Phi\right>\) 为范畴图,简称为图

    1. \(O\)\(\mathcal{G}\) 的结点类,其元素称为 \(\mathcal{G}\) 的结点;称 \(M\)\(\mathcal{G}\) 的箭头类,其元素称为 \(\mathcal{G}\) 的箭头
    2. \(f \in M\),定义 \(\operatorname{source}: M \rightarrow O\)\(\operatorname{target}: M \rightarrow O\) 是使得 \(\Phi(f) = (\operatorname{source}(f), \operatorname{target}(f))\) 的映射
      1. \(\operatorname{source}(f)\)\(f\) 的源,\(\operatorname{target}(f)\)\(f\) 的靶,也可将范畴图记作 \(\mathcal G = \left<O, M, \operatorname{source}, \operatorname{target}\right>\)
      2. \(f \in M\) 使得 \(A=\operatorname{source}(f)\)\(B=\operatorname{target}(f)\),则称 \(f\) 为从 \(A\)\(B\) 的箭头,记为 \(f: A \rightarrow B\)\(A \xrightarrow{f} B\)
      3. \(A, B \in O\),令 \(\mathcal{G}[A, B]=\{f \in M \mid \operatorname{source}(f)=A, \operatorname{target}(f)=B\}\),通常简写为 \([A, B]\)
  2. 特殊的范畴图

    1. \(O=\varnothing\),则 \(\mathcal{G}=\langle\varnothing, \varnothing, \varnothing\rangle\),则称 \(\mathcal{G}\) 为空图
    2. \(M=\varnothing\),则 \(\mathcal{G}=\langle O, \varnothing, \varnothing\rangle\),则称 \(\mathcal{G}\) 为离散图
    3. \(O\)\(M\) 均为集合,则称 \(\mathcal{G}\) 为小图
      1. \(O\)\(M\) 均为有穷集合,则称 \(\mathcal{G}\) 为有穷图
      2. 若当 \(A, B \in O\) 时,\([A, B]\) 均为集合,则称 \(\mathcal{G}\) 为局部小图
    4. \(f \in O\),使 \(\operatorname{source}(f)=\operatorname{target}(f)\),则称 \(f\)\(\mathcal{G}\) 的自圈
    5. \(\mathcal{G}\) 无自圈,且当 \(A, B \in O\) 时皆有 \(|[A, B]| \leqslant 1\),则称 \(\mathcal{G}\) 为简单图
  3. 图同态:设图 \(\mathcal{G}=\langle O, M, \operatorname{source}, \operatorname{target}\rangle\)\(\mathcal{G}^{\prime}=\langle O^{\prime}, M^{\prime}, \operatorname{source}', \operatorname{target}'\rangle\),设 \(\Phi_{0}: O \rightarrow O^{\prime}\)\(\Phi_{1}: M \rightarrow M^{\prime}\)

    1. \(\Phi_{0} \circ \mathrm{source}=\mathrm{source}' \circ \Phi_{1}\)\(\Phi_{0} \circ \mathrm{target}=\mathrm{target}^{\prime} \circ \Phi_{1}\),则称 \(\left\langle\Phi_{0}, \Phi_{1}\right\rangle\) 为从 \(\mathcal{G}\)\(\mathcal{G}^{\prime}\) 的图同态,记为 \(\Phi: \mathcal{G} \rightarrow \mathcal{G}^{\prime}\)\(\mathcal{G} \simeq \mathcal{G}^{\prime}\)
    2. \(\Phi_{1}: \mathcal{G} \rightarrow \mathcal{G}^{\prime}\),若 \(\Phi_{0}\)\(\Phi_{1}\) 均为双射,则称 \(\Phi\) 为图同构,记为 \(\mathcal{G} \cong \mathcal{G}^{\prime}\)

    \(\Phi=\left\langle\Phi_{0}, \Phi_{1}\right\rangle\),对任意 \(A\in O, f \in M\),规定 \(\Phi(A) =\Phi_{0}(A), \Phi(f) =\Phi_{1}(f)\)

    1. 对图 \(\mathcal{G}=\langle O, M, \operatorname{source}, \operatorname{target}\rangle\),取 \(I_{\mathcal{G}}=\left\langle I_{O}, I_{M}\right\rangle\),则 \(I_{\mathcal{G}}: \mathcal{G} \rightarrow \mathcal{G}\),称 \(I_{\mathcal{G}}\)\(\mathcal{G}\) 的恒等图同态(同构)
    2. \(\Phi: \mathcal{G} \rightarrow \mathcal{G}^{\prime}, \Psi: \mathcal{G}^{\prime} \rightarrow \mathcal{G}^{\prime \prime}\)\(\mathcal{G}=\langle O, M, \operatorname{source}, \operatorname{target}\rangle\),若令

      \[ \begin{aligned} (\Psi \circ \Phi)(A) & =\Psi(\Phi(A)), A \in O \\ (\Psi \circ \Phi)(f) & =\Psi(\Phi(f)), f \in M \end{aligned} \]

      \(\Psi \circ \Phi: \mathcal{G} \rightarrow \mathcal{G}^{\prime \prime}\),并称 \(\Psi \circ \Phi\)\(\Psi\)\(\Phi\) 的复合

    3. \(\Phi: \mathcal{G} \rightarrow \mathcal{G}^{\prime}\),若 \(\Psi: \mathcal{G}^{\prime} \rightarrow \mathcal{G}\) 使 \(\Phi \circ \Psi=\Pi_{\mathcal{G}}\)\(\Psi \circ \Phi=I_{\mathcal{G}}\),则称 \(\Psi\)\(\Phi\) 的逆.若 \(\Psi: \mathcal{G}^{\prime} \rightarrow \mathcal{G}\)\(\Phi\) 的逆,则称 \(\Phi\) 有逆

  4. \(\mathcal{G}=\langle O, M, \operatorname{source}, \operatorname{target}\rangle\)\(\mathcal{G}^{\prime}=\langle O^{\prime}, M^{\prime}, \operatorname{source}', \operatorname{target}'\rangle\) 均为图,\(\Phi: \mathcal{G} \rightarrow \mathcal{G}^{\prime}\)\(\Psi: \mathcal{G} \rightarrow \mathcal{G}^{\prime}\)

    1. 定义 \(M^{\prime}\) 上的局部二元运算满足当 \(f, g \in M^{\prime}\)\(\operatorname{target}^{\prime}(f)=\operatorname{source}^{\prime}(g)\) 时,恒有

      1. \(g \circ f \downarrow\) 当且仅当 \(\operatorname{source}^{\prime}(g)=\operatorname{target}^{\prime}(f)\)
      2. \(g \circ f \downarrow\)\(\operatorname{source}(g \circ f)=\operatorname{source}^{\prime}(f)\)\(\operatorname{target}^{\prime}(g \circ f)=\operatorname{target}^{\prime}(g)\)

      称具有该局部二元运算的图为演算系统

    2. \(\alpha: O \rightarrow M^{\prime}\) 对任意 \(f \in M\)\(\alpha(\operatorname{target}(f)) \circ \Phi(f)=\Psi(f) \circ \alpha(\operatorname{source}(f))\),则称 \(\alpha\)\(\Phi\)\(\Psi\) 的自然变换,记为 \(\alpha: \Phi \rightarrow \Psi\)

1.2 范畴

1.2.1 范畴的概念

  1. 若图 \(\langle O, M, \mathrm{dom}, \mathrm{cod}\rangle\)\(\mathrm{comp}: M \times M \rightharpoonup M\)(记作 \(g \circ f\)\(gf\))满足如下公理

    1. \(C_{0}\):若 \(f, g \in M\),则 \(g \circ f \downarrow\) 当且仅当 \(\operatorname{dom}(g)=\operatorname{cod}(f)\)
    2. \(C_{1}\):若 \(f, g \in M\)\(g \circ f \downarrow\),则 \(\operatorname{dom}(g \circ f)=\operatorname{dom}(f)\)\(\operatorname{cod}(g \circ f)= \operatorname{cod}(g)\)
    3. \(C_{2}\):若 \(A \in O\),则有 \(I_{A} \in M\) 使得
      1. \(\operatorname{dom}\left(I_{A}\right)=A=\operatorname{cod}\left(I_{A}\right)\)
      2. \(g \in M\) 使 \(\operatorname{dom}(g)=A\),则 \(g \circ I_{A}=g\)
      3. \(f \in M\) 使 \(\operatorname{cod}(f)=A\),则 \(I_{A} \circ f=f\)
    4. \(C_{3}\):若 \(f, g, h \in M\),则有
      1. \(h \circ (g \circ f) \downarrow\) 当且仅当 \((h \circ g) \circ f \downarrow\)
      2. \(h \circ (g \circ f) \downarrow\)\(h \circ (g \circ f)=(h \circ g) \circ f\)

    称五元组 \(\mathscr{C}=\langle O, M, \mathrm{dom}, \mathrm{cod}, \mathrm{comp}\rangle\) 为范畴,称图 \(\langle O, M, \operatorname{dom}, \operatorname{cod}\rangle\)\(\mathscr{C}\) 的基图,记为 \(\mathbf{G}(\mathscr{C})\)

    1. \(O\)\(\mathscr{C}\) 的对象类,记为 \(\operatorname{Ob} \mathscr{C}\),其元素称为 \(\mathscr{C}\) 的对象
      1. \(A, B \in O\),则记 \(\mathbf{G}(\mathscr{C})[A, B]=\mathscr{C}[A, B]\),通常简写为 \([A, B]\)
      2. \(A \in O\),则称由公理 \(C_{2}\) 给出的态射 \(\mathbf{I}_{A}\) 为关于 \(A\) 的幺态射,简称幺态射
    2. \(M\)\(\mathscr{C}\) 的态射类,记为 \(\operatorname{Mor} \mathscr{C}\),其元素称为 \(\mathscr{C}\) 的态射
      1. \(f \in M\),则称 \(\operatorname{dom}(f)\)\(f\) 的论域,\(\operatorname{cod}(f)\)\(f\) 的余论域
      2. \(f \in M\),若 \(A=\operatorname{dom}(f)\)\(B=\operatorname{cod}(f)\),则称 \(f\) 为一个从 \(A\)\(B\) 的态射,记为 \(f: A \rightarrow B\)\(A \xrightarrow{f} B\)
      3. \(f, g \in M\),使 \(\operatorname{comp}(g, f) \downarrow\),则称 \(\operatorname{comp}(g, f)\)\(g\)\(f\) 的合成
    3. \(\mathbf{I}\)\(\mathbf{J}\) 为指标类
      1. \(\eta: \mathbf{I} \rightarrow \operatorname{Ob} \mathscr{C}\),当记 \(A_{i}=\eta(i)(i \in \mathbf{I})\) 时,则称 \(\left(A_{i}\right)_{i \in I}\) 为由 \(\mathbf{I}\) 标记的对象族
      2. \(\xi: \mathbf{J} \rightarrow \operatorname{Mor} \mathscr{C}\),当记 \(f_{j}=\xi(j)(j \in \mathbf{J})\) 时,则称 \(\left(f_{j}\right)_{j \in J}\) 为由 \(\mathbf{J}\) 标记的态射族
    4. \(\mathscr{C}\) 为范畴且 \(A, B, C, D \in \operatorname{Ob} \mathscr{C}\)
      1. \(A \neq C\)\(B \neq D\),则 \([A, B] \cap [C, D]=\varnothing\)
      2. \(f \in[A, B]\)\(g \in[B, C]\),则 \(g \circ f \in[A, C]\)
    5. \(\mathscr{C}\) 为范畴,若 \(A \in \operatorname{Ob} \mathscr{C}\),则关于 \(A\) 的幺态射存在且唯一
  2. \(\mathscr{C}\) 为范畴且 \(\mathbf{I}\)\(\mathbf{J}\) 均为类,若 \(\mathcal{D}=\langle\left(A_{i}\right)_{i \in \mathbf{I}}, \left(f_{j}\right)_{j \in \mathbf{J}}, \Psi\rangle\) 满足

    1. \(\left\{A_{i} \mid i \in \mathbf{I}\right\} \subseteq \operatorname{Ob} \mathscr{C}\)\(\left\{f_{j} \mid j \in \mathbf{J}\right\} \subseteq \operatorname{Mor} \mathscr{C}\)
    2. \(j \in \mathbf{J}\) 使得 \(\Psi\left(f_{j}\right)=\left\langle A_{l}, A_{k}\right\rangle\),且 \(l, k \in \mathbf{I}\),则 \(\operatorname{dom}\left(f_{j}\right)=A_{l}\)\(\operatorname{cod}\left(f_{j}\right)=A_{k}\)

    则称 \(\mathcal{D}\)\(\mathscr{C}\) 中的一个图,并令 \(\mathcal{D}_{V}=\left(A_{i}\right)_{i \in \mathbf{I}}\)\(\mathcal{D}_{E}=\left(f_{j}\right)_{j \in \mathbf{J}}\)

    1. \(\mathscr{C}\) 为范畴,\(\mathcal{D}=\langle V, E, \Psi\rangle\) 为图且 \(\Gamma: \mathcal{D} \rightarrow G(\mathscr{C})\) 为图同态,若有

      1. 对任意 \(v \in V, e \in E\),定义 \(A_{v}=\Gamma(v), f_{e}=\Gamma(e), A_{e}=\operatorname{dom}\left(f_{e}\right), B_{e}=\operatorname{cod}\left(f_{e}\right)\)
      2. 定义 \(\mathcal{D}_{V}=\left(A_{v}\right)_{v \in V}, \mathcal{D}_{S}=\left(A_{e}\right)_{e \in E}, \mathcal{D}_{E}=\left(B_{e}\right)_{e \in E}\)
      3. 函数 \(\Psi^{\prime}\left(f_{e}\right)_{e \in E} \rightarrow\left(A_{v}\right)_{v \in V} \times\left(A_{v}\right)_{v \in V}\)\(\Psi^{\prime}\left(f_{e}\right)=\left\langle\operatorname{dom}\left(f_{e}\right), \operatorname{cod}\left(f_{e}\right)\right\rangle\) 对任意 \(e \in E\) 成立

      则称 \(\left\langle\left(A_{v}\right)_{v \in V},\left(f_{e}\right)_{e \in E}, \Psi^{\prime}\right\rangle\)\(\mathscr{C}\) 中一个图,记为 \(\langle\mathcal{D}, \Gamma\rangle\)\(\Gamma(\mathcal{D})\)

    2. \(\mathscr{C}\) 为范畴,\(\mathcal{D}\)\(\mathscr{C}\) 中的图

      1. \(\mathcal{D}\) 的如上图形式的三角形子图为交换的当且仅当 \(h=g \circ f\)
      2. \(\mathcal{D}\) 的如上图形式的四边形子图为交换的当且仅当 \(g \circ f=k \circ h\)
      3. \(\mathcal{D}\) 的所有上述形式的子图都是交换的,则称 \(\mathcal{D}\) 为交换的
  3. 特殊的范畴:设 \(\mathscr C\) 为范畴

    1. \(\mathbf{0} = \left<\varnothing, \varnothing, \varnothing, \varnothing, \varnothing\right>\) 为空范畴
    2. 对任意 \(A, B \in \operatorname{Ob} \mathscr{C}\),若每个 \([A, B]\) 均为集合,则称 \(\mathscr{C}\) 为局部小范畴
    3. \(\operatorname{Ob} \mathscr{C}\)\(\operatorname{Mor} \mathscr{C}\) 均为集合,则称 \(\mathscr{C}\) 为小范畴
    4. 若每个 \(f \in \operatorname{Mor} \mathscr{C}\) 均为幺态射,则称 \(\mathscr{C}\) 为离散范畴
    5. 若当 \(A, B \in \operatorname{Ob} \mathscr{C}\) 时皆有 \(|[A, B]| \leqslant 1\),则称 \(\mathscr{C}\) 为简单范畴
  4. 典型范畴

    1. 集合相关典型范畴

      范畴 对象 态射 复合
      \(\operatorname{ANMFn}(X)\) \(X\) \(f: X \Rightarrow X\) \(\circledast\)
      \(\operatorname{MFn}(X)\) \(X\) \(f: X \Rightarrow X\) \(\circ\)
      \(\operatorname{PFn}(X)\) \(X\) \(f: X \rightharpoonup X\) \(\circ\)
      \(\operatorname{Set}(X)\) \(X\) \(f: X \rightarrow X\) \(\circ\)
    2. 经典结构的典型范畴

      范畴 对象 态射 范畴 对象 态射
      \(\mathrm{Graph}\) 小图 图同态 \(\mathrm{Preoset}\) 拟序集 保序映射
      \(\mathrm{Mon}\) 幺半群 幺半群同态 \(\mathrm{Poset}\) 偏序集 保序映射
      \(\mathrm{Grp}\) 群同态 \(\mathrm{Metre}\) 度量空间 连续函数
      \(\mathrm{Rng}\) 环同态 \(\mathrm{Top}\) 拓扑空间 连续函数
      \(\mathrm{Field}\) 域同态 \(\mathrm{BanSP}\) \(\text{Banach}\) 空间 有界线性变换
      \(\mathrm{Lat}\) 格同态 \(\mathrm{HilbSP}\) \(\text{Hilbert}\) 空间 有界线性变换
      \(\mathrm{Mod}\) 模同态 \(\mathrm{LinSP}\) 连续线性空间 线性变换

1.2.2 范畴运算

  1. 子范畴:设范畴 \(\mathscr{A}\)\(\mathscr{B}\) 满足

    1. \(\operatorname{Ob} \mathscr{B} \subseteq \operatorname{Ob} \mathscr{A}\)\(\operatorname{Mor} \mathscr{B} \subseteq \operatorname{Mor} \mathscr{A}\)
    2. \(A \in \operatorname{Ob} \mathscr{B}\),则 \(\mathscr{B}\)\(\mathscr{A}\) 关于 \(A\) 的幺态射相同
    3. \(\operatorname{dom}_{\mathscr{B}} \subseteq \operatorname{dom}_{\mathscr{A}}, \operatorname{cod}_{\mathscr{B}} \subseteq \operatorname{cod}_{\mathscr{A}}\)\(\operatorname{comp}_{\mathscr{B}} \subseteq \operatorname{comp}_{\mathscr{A}}\)

    则称 \(\mathscr{B}\)\(\mathscr{A}\) 的子范畴,记为 \(\mathscr{B} \subseteq \mathscr{A}\).易知子范畴关系 \(\subseteq\) 是自反的、传递的和反对称的

    1. \(\mathscr{B} \subseteq \mathscr{A}\)\(\mathscr{B} \neq \mathscr{A}\),则称 \(\mathscr{B}\)\(\mathscr{A}\) 的真子范畴,记为 \(\mathscr{B} \subseteq \mathscr{A}\).易知真子范畴关系 \(\subseteq\) 是反自反的和传递的
    2. \(\mathscr{B} \subseteq \mathscr{A}\)\(\operatorname{Ob} \mathscr{B}=\operatorname{Ob} \mathscr{A}\),则称 \(\mathscr{B}\)\(\mathscr{A}\) 的宽子范畴
    3. \(\mathscr{B} \subseteq \mathscr{A}\),且当 \(A, B \in \operatorname{Ob} \mathscr{B}\) 时皆有 \(\mathscr{B}[A, B]=\mathscr{A}[A, B]\),则称 \(\mathscr{B}\)\(\mathscr{A}\) 的完全子范畴
  2. 商范畴:设 \(\mathscr{A}\) 为范畴,若 \(\operatorname{Mor} \mathscr{A}\) 上的等价关系 \(\approx\) 满足

    1. \(A, B \in \operatorname{Ob} \mathscr{A}\)\(f \in \mathscr{A}[A, B]\)\([f]_{\approx} \subseteq \mathscr{A}[A, B]\)
    2. \(f, f^{\prime}, g, g^{\prime} \in \operatorname{Mor} \mathscr{A}\) 使 \(f \approx f^{\prime}\)\(g \approx g^{\prime}\),则 \(g \circ f \approx g^{\prime} \circ f^{\prime}\)

    则称 \(\approx\)\(\mathscr{A}\) 上的同余关系.若 \(f, g \in \operatorname{Mor} \mathscr{A}\),则令

    \[ \begin{aligned} \operatorname{dom}^{\prime}\left([f]_{\approx}\right)&=\operatorname{dom}(f) \\ \operatorname{cod}^{\prime}\left([f]_{\approx}\right)&=\operatorname{cod}(f) \\ \operatorname{comp}^{\prime}\left([g]_{\approx},[f]_{\approx}\right)&= \left\{\begin{aligned} & [\operatorname{comp}(g, f)]_{\approx}, & \operatorname{comp}(g, f) \downarrow \\ & \uparrow, & \textsf{否则} \end{aligned}\right. \end{aligned} \]

    则称 \(\mathscr{A} / \approx=\left\langle\mathrm{Ob} \mathrm{\mathscr{A}}, \mathrm{Mor} \mathscr{A} / \approx, \mathrm{dom}^{\prime}, \mathrm{cod}^{\prime}, \mathrm{comp}^{\prime}\right\rangle\)\(\mathscr{A}\) 关于 \(\approx\) 的商,则 \(\mathscr{A} / \approx\) 是一个范畴,称为 \(\mathscr{A}\) 关于 \(\approx\) 的商范畴

  3. 积范畴:设 \(\mathscr{A}_{i}=\left\langle O_{i}, M_{i} ; \operatorname{dom}_{i}, \operatorname{cod}_{i}, \operatorname{comp}_{i}\right\rangle \ (i=1,2)\) 均为范畴,令 \(O=O_{1} \times O_{2}, M=M_{1} \times M_{2}\),若对任意 \(\langle f, g\rangle,\left\langle f^{\prime}, g^{\prime}\right\rangle \in M\)

    \[ \begin{aligned} \operatorname{dom}(\langle f, g\rangle)&=\left\langle\operatorname{dom}_{1}(f), \operatorname{dom}_{2}(g)\right\rangle \\ \operatorname{cod}(\langle f, g\rangle)&=\left\langle\operatorname{cod}_{1}(f), \operatorname{cod}_{2}(g)\right\rangle \\ \operatorname{comp}\left(\left\langle f^{\prime}, g^{\prime}\right\rangle,\langle f, g\rangle\right)&=\left\{\begin{aligned} & \left\langle\operatorname{comp}_{1}\left(f^{\prime}, f\right), \operatorname{comp}_{2}\left(g^{\prime}, g\right)\right\rangle, & \operatorname{comp}_{1}\left(f^{\prime}, f\right) \downarrow \wedge \operatorname{comp}_{2}\left(g^{\prime}, g\right) \downarrow \\ & \uparrow, & \textsf{否则} \end{aligned}\right. \end{aligned} \]

    \(\mathscr{A}_{1} \times \mathscr{A}_{2}=\langle O, M, \mathrm{dom}, \mathrm{cod}, \mathrm{comp}\rangle\) 是一个范畴,称为 \(\mathscr{A}_{1}\)\(\mathscr{A}_{2}\) 的积范畴

    1. 一般地,\(n\) 个范畴 \(\mathscr A_1, \mathscr A_2, \cdots, \mathscr A_n\) 的积范畴 \(\left(\cdots\left(\mathscr{A}_{1} \times \mathscr{A}_{2}\right) \times \cdots\right) \times \mathscr{A}_{n}\) 记作 \(\mathscr{A}_{1} \times \mathscr{A}_{2} \times \cdots \times \mathscr{A}_{n}\)
    2. \(\mathscr{A}^{0}=\mathbf{0}\),同时定义 \(\mathscr{A}^{1}=\mathscr{A}\)\(\mathscr{A}^{n+1}=\mathscr{A}^{n} \times \mathscr{A}\)
  4. 和范畴:设 \(\mathscr{A}_{i}=\left\langle O_{i}, M_{i} ; \operatorname{dom}_{i}, \operatorname{cod}_{i}, \operatorname{comp}_{i}\right\rangle \ (i=1,2)\) 均为范畴,令 \(O=O_{1} \sqcup O_{2}, M=M_{1} \sqcup M_{2}\),若对任意 \(f_{1}, g_{1} \in M_{1}, g_{2}, g_{2} \in M_{2}\)

    \[ \begin{aligned} \operatorname{dom}\left(\left\langle f_{i}, i\right\rangle\right)&=\left\langle\operatorname{dom}_{i}\left(f_{i}\right), i\right\rangle \\ \operatorname{cod}\left(\left\langle f_{i}, i\right\rangle\right)&=\left\langle\operatorname{cod}_{i}\left(f_{i}\right), i\right\rangle \\ \operatorname{comp}\left(\left\langle f_i, i\right\rangle,\langle g_j, j\rangle\right)&=\left\{\begin{aligned} & \left\langle\operatorname{comp}_{i}\left(f_{i}, g_{i}\right), i\right\rangle, & i = j \wedge \operatorname{comp}_{i}\left(f_{i}, g_{i}\right) \downarrow \\ & \uparrow, & \textsf{否则} \end{aligned}\right. \end{aligned} \]

    \(\mathscr{A}_{1} \sqcup \mathscr{A}_{2}=\langle O, M, \mathrm{dom}, \mathrm{cod}, \mathrm{comp}\rangle\) 是一个范畴,称为 \(\mathscr{A}_{1}\)\(\mathscr{A}_{2}\) 的和范畴,并可推广到 \(n\) 个范畴的情形

  5. 对偶范畴:设 \(\mathscr{C}=\langle O, M\), dom, cod, comp \(\rangle\) 为范畴,若令 \(O^{\prime}=O\)\(M^{\prime}=\left\{f^{\mathrm{op}} \mid f \in M\right\}\),并对任意的 \(f, g \in M\)

    \[ \begin{aligned} \operatorname{dom}^{\prime}\left(f^{\mathrm{op}}\right) & =\operatorname{cod}(f) \\ \operatorname{cod}^{\prime}\left(f^{\mathrm{op}}\right) & =\operatorname{dom}(f) \\ \operatorname{comp}^{\prime}\left(g^{\mathrm{op}}, f^{\mathrm{op}}\right) & = \left\{\begin{aligned} & (\operatorname{comp}(f, g))^{\mathrm{op}}, & \operatorname{comp}(f, g) \downarrow \\ & \uparrow, & \textsf{否则} \end{aligned}\right. \end{aligned} \]

    \(\left\langle O^{\prime}, M^{\prime}, \operatorname{dom}^{\prime}, \mathrm{cod}^{\prime}, \mathrm{comp}^{\prime}\right\rangle\) 为范畴,称为 \(\mathscr{A}\) 的对偶范畴,记为 \(\mathscr{A}^{\mathrm{op}}\)

    1. \(\left(\mathscr{A}^{\mathrm{op}}\right)^{\mathrm{op}}=\mathscr{A}\)
    2. 对每个 \(A \in O\),令 \(A^{\mathrm{op}}=A\),称 \(A^{\mathrm{op}}\)\(A\) 的对偶对象;对每个 \(f \in M\),称 \(f^{\mathrm{op}}\)\(f\) 的对偶态射
    3. 对偶原理:若 \(S\) 是一个对所有范畴皆真的命题,则 \(S\) 的对偶命题 \(S^{\mathrm{op}}\) 亦然

1.2.3 特殊态射与对象

  1. \(\mathscr{C}\) 为范畴,\(A, B \in \operatorname{Ob} \mathscr{C}\)\(f \in[A, B]\)

    1. 若有 \(g \in[B, A]\) 使 \(f \circ g=\mathbf{I}_{B}\),则称 \(f\) 有右逆或右可逆,\(g\)\(f\) 的一个右逆,此时称 \(f\)\(\text{retraction}\)
    2. 若有 \(g \in[B, A]\) 使 \(g \circ f=\mathbf{I}_{A}\),则称 \(f\) 有左逆或左可逆,\(g\)\(f\) 的一个左逆,此时称 \(f\)\(\text{section}\)

    若有 \(g \in[B, A]\) 使 \(f \circ g=\mathbf{I}_{B}\)\(g \circ f=\mathbf{I}_{A}\),则称 \(f\) 有逆或可逆,也称 \(f\) 为同构态射,\(g\)\(f\) 的一个逆,记为 \(f^{-1}\)

    1. \(\mathscr{C}\) 为范畴,\(A, B \in \operatorname{Ob} \mathscr{C}\)\(f \in[A, B]\),则 \(f\) 有逆当且仅当 \(f\) 有唯一的逆
    2. \(\mathscr{C}\) 为范畴,\(A, B, C \in \operatorname{Ob} \mathscr{C}\)\(f \in[A, B], g \in[B, C]\)
      1. \(f\) 有逆,则 \(f^{-1}\) 也有逆且 \(\left(f^{-1}\right)^{-1}=f\)
      2. \(f\)\(g\) 均有逆,则 \(g \circ f\) 有逆且 \((g \circ f)^{-1}=f^{-1} \circ g^{-1}\)
    3. \(\mathscr{C}\) 为范畴,\(A, B, C \in \operatorname{Ob} \mathscr{C}\)\(f \in[A, B], g \in[B, C]\)
      1. \(f\)\(g\) 均为 \(\text{section}\),则 \(g \circ f\) 亦然
      2. \(f\)\(g\) 均为 \(\text{retraction}\),则 \(g \circ f\) 亦然
      3. \(f\)\(g\) 均为同构态射,则 \(g \circ f\) 亦然
      4. \(g \circ f\)\(\text{section}\),则 \(f\)\(\text{section}\)
      5. \(g \circ f\)\(\text{retraction}\),则 \(g\)\(\text{retraction}\)
      6. \(g \circ f\) 为同构态射,则 \(f\)\(\text{section}\)\(g\)\(\text{retraction}\)
    4. \(\mathscr{C}\) 为范畴且 \(A, B \in \operatorname{Ob} \mathscr{C}\).若有同构态射 \(f \in[A, B]\) 则称 \(A\)\(B\) 等价,记为 \(A \simeq B\)
      1. \(\simeq\)\(\operatorname{Ob} \mathscr{C}\) 上的一个等价关系
      2. \(\mathscr{C}\) 为范畴且 \(A, B \in \operatorname{Ob} \mathscr{C}\).若 \(A \simeq B\),则对任意的 \(C \in \operatorname{Ob} \mathscr{C}\),皆有双射 \(\sigma_{C}:[A, C] \rightarrow[B, C]\) 及双射 \(\widetilde{\sigma}_{C}:[C, A] \rightarrow[C, B]\)
  2. \(\mathscr{C}\) 为范畴,\(A, B \in \operatorname{Ob} \mathscr{C}\)\(f \in[A, B]\)

    1. \(f\) 可左消去,即如果 \(C \in \operatorname{Ob} \mathscr{C}\)\(h, k \in[C, A]\) 使 \(f \circ h=f \circ k\),则 \(h=k\),则称 \(f\) 为单态射
    2. \(f\) 可右消去,即如果 \(C \in \operatorname{Ob} \mathscr{C}\)\(h, k \in[B, C]\) 使 \(h \circ f=k \circ f\),则 \(h=k\),则称 \(f\) 为外态射

    \(f\) 可消去,即 \(f\) 既可左消去又可右消去,则称 \(f\) 为双态射

    1. \(\mathscr{C}\) 为范畴, \(A, B, C \in \operatorname{Ob} \mathscr{C}, f \in[A, B]\)\(g \in[B, C]\)
      1. \(f\)\(g\) 均为单态射,则 \(g \circ f\) 亦然
      2. \(f\)\(g\) 均为外态射,则 \(g \circ f\) 亦然
      3. \(f\)\(g\) 均为双态射,则 \(g \circ f\) 亦然
      4. \(g \circ f\) 为单态射,则 \(f\) 为单态射
      5. \(g \circ f\) 为外态射,则 \(g\) 为外态射
      6. \(g \circ f\) 为双态射,则 \(f\) 为单态射且 \(g\) 为外态射
    2. \(\mathscr{C}\) 为范畴,\(A, B \in \operatorname{Ob} \mathscr{C}\)\(f \in[A, B]\)
      1. \(f\)\(\text{section}\),则 \(f\) 为单态射
      2. \(f\)\(\text{retraction}\),则 \(f\) 为外态射
      3. \(f\) 为同构态射,则 \(f\) 为双态射
    3. \(\mathscr{C}\) 为范畴且 \(A, B \in \operatorname{Ob} \mathscr{E}\),若 \(f \in[A, B]\),则 \(f\) 为同构态射当且仅当以下任一成立
      1. \(f\) 既是 \(\text{section}\),又是外态射
      2. \(f\) 既是 \(\text{retraction}\),又是单态射
    4. \(f \in \operatorname{Set}[A, B]\)
      1. \(A \neq \varnothing\),则下列条件等价:① \(f\) 为单射;② \(f\)\(\text{section}\);③ \(f\) 为单态射
      2. 下列条件等价:① \(f\) 为满射;② \(f\)\(\text{retraction}\);③ \(f\) 为外态射
      3. 下列条件等价:① \(f\) 为双射;② \(f\) 为同构态射;③ \(f\) 为双态射
  3. \(\mathscr{C}\) 为范畴且 \(A \in \operatorname{Ob} \mathscr{C}\)

    1. 若对每个 \(B \in \operatorname{Ob} \mathscr{C}\) 皆有 \(|[A, B]|=1\),则称 \(A\) 为初始对象,用 \(\mathbf{I}_{A B}\) 表示 \([A, B]\) 中的唯一态射
    2. 若对每个 \(B \in \operatorname{Ob} \mathscr{C}\) 皆有 \(|[B, A]|=1\),则称 \(A\) 为终止对象,用 \(\mathbf{T}_{B A}\) 表示 \([B, A]\) 中的唯一态射

    \(A\) 同时是初始对象与终止对象,则称 \(A\) 为零对象.设 \(B \in \operatorname{Ob} \mathscr{C}\),用 \(\mathbf{O}_{A B}\)\(\mathbf{O}_{B A}\) 分别表示 \([A, B]\)\([B, A]\) 中的唯一态射

    1. \(\mathbf{I}_{AA}=\mathbf{T}_{AA}=\mathbf{O}_{AA}=\mathbf{I}_{A}\)\(\mathbf{O}_{AB}=\mathbf{I}_{A B}, \mathbf{O}_{B A}=\mathbf{T}_{BA}\)
    2. \(\mathscr{C}\) 为范畴且 \(A, B \in \operatorname{Ob} \mathscr{C}\)
      1. \(A\) 为初始对象,则 \(B\) 为初始对象当且仅当 \(B \simeq A\)
      2. \(A\) 为终止对象,则 \(B\) 为终止对象当且仅当 \(B \simeq A\)
      3. \(A\) 为零对象,则 \(B\) 为零对象当且仅当 \(B \simeq A\)
  4. \(\mathscr{C}\) 为范畴,\(A, B \in \operatorname{Ob} \mathscr{C}\)\(f \in[A, B]\)

    1. 若对每个 \(C \in \operatorname{Ob} \mathscr{C}\) 及任意 \(h, k \in[C, A]\) 皆有 \(f \circ h=f \circ k\),则称 \(f\) 为常态射
    2. 若对每个 \(C \in \operatorname{Ob} \mathscr{C}\) 及任意 \(h, k \in[B, C]\) 皆有 \(h \circ f=k \circ f\),则称 \(f\) 为余常态射

    \(f\) 既是常态射,又是余常态射,则称 \(f\) 为零态射

    1. \(\mathscr{C}\) 为范畴,若任意 \(A, B \in \operatorname{Ob} \mathscr{C}\) 都有 \([A, B]\) 中存在零态射,那么称 \(\mathscr{C}\) 为点化范畴
    2. \(\mathscr{C}\) 为范畴,\(A, B, C, D \in \operatorname{Ob} \mathscr{C}\)\(f \in[A, B], g \in[B, C], h \in[C, D]\)
      1. \(g\) 为常态射,则 \(h \circ g\)\(g \circ f\) 均为常态射
      2. \(g\) 为余常态射,则 \(h \circ g\)\(g \circ f\) 均为余常态射
      3. \(g\) 为零态射,则 \(h \circ g\)\(g \circ f\) 均为零态射
    3. \(\mathscr{C}\) 为范畴,\(A, B \in \operatorname{Ob} \mathscr{C}\)\(f, g \in[A, B]\).若 \(f\) 为常态射,\(g\) 为余常态射且 \([B, A] \neq \varnothing\),则 \(f=g\)
    4. \(\mathscr{C}\) 为范畴且 \(A, B \in \operatorname{Ob} \mathscr{C}\)
      1. \(A\) 为初始对象,则 \(\mathbf{I}_{A B}\) 为余常态射
      2. \(A\) 为终止对象,则 \(\mathbf{T}_{A B}\) 与常态射
      3. \(A\) 为零对象,则 \(\mathbf{O}_{A B}\) 为余常态射,\(\mathbf{O}_{B A}\) 为常态射
      4. \(A\)\(B\) 均为零对象,则 \(\mathbf{O}_{A B}\)\(\mathbf{O}_{B A}\) 皆为零态射
    5. 如果 \(\mathscr{C}\) 为范畴,则下列条件等价:

      1. \(A, B \in \operatorname{Ob} \mathscr{C}\),则 \([A, B]\) 中有零态射
      2. \(A, B \in \operatorname{Ob} \mathscr{C}\),则 \([A, B]\) 中有唯一的零态射
      3. \(A, B \in \operatorname{Ob} \mathscr{C}\),则 \([A, B]\) 中有唯一的常态射
      4. \(A, B \in \operatorname{Ob} \mathscr{C}\),则 \([A, B]\) 中有唯一的余常态射
      5. \(A, B \in \operatorname{Ob} \mathscr{C}\),则 \([A, B]\) 中既有常态射,又有余常态射
      6. 存在函数 \(\sigma: \operatorname{Ob} \mathscr{C} \times \operatorname{Ob} \mathscr{C} \rightarrow \operatorname{Mor} \mathscr{C}\) 满足如下条件:

        1. \(A, B \in \operatorname{Ob} \mathscr{C}\),则 \(\sigma(A, B) \in[A, B]\)
        2. \(A, B, C \in \operatorname{Ob} \mathscr{C}\)\(f \in[B, C]\),则 \(f \circ \sigma(A, B)=\sigma(A, C)\)
        3. \(A, B, C \in \operatorname{Ob} \mathscr{C}\)\(g \in[A, B]\),则 \(\sigma(B, C) \circ g=\sigma(A, C)\)

        称函数 \(\sigma\)\(\mathscr{A}\) 的零态射选择函数

    6. \(\mathscr{C}\) 为范畴,\(Z \in \operatorname{Ob} \mathscr{C}\) 为零对象且 \(A, B \in \operatorname{Ob} \mathscr{C}\)

      1. \(\mathbf{I}_{Z B} \circ \mathbf{T}_{A Z}\) 为零态射
      2. \(Z^{\prime} \in \operatorname{Ob} \mathscr{C}\) 也是零对象,则 \(\mathbf{I}_{Z^{\prime} B} \circ \mathbf{T}_{A Z^{\prime}}=\mathbf{I}_{Z B} \circ \mathbf{T}_{A Z}\)

1.2.4 范畴的同构

  1. \(F: \mathscr{A} \rightarrow \mathscr{B}\) 为协变函子,若有协变函子 \(G: \mathscr{B} \rightarrow \mathscr{A}\) 使得

    \[ G \circ F=\mathbf{I}_{\mathscr{A}} \wedge F \circ G=\mathbf{I}_{\mathscr{B}} \]

    则称 \(F\) 为同构函子,并称 \(G\)\(F\) 的一个逆.若存在同构函子 \(F: \mathscr{A} \rightarrow \mathscr{B}\),则称 \(\mathscr{A}\) 同构于 \(\mathscr{B}\),记为 \(\mathscr{A} \cong \mathscr{B}\)

    1. 范畴的同构 \(\cong\) 是等价关系
    2. 恒等函子 \(I_{\mathscr{A}}: \mathscr{A} \rightarrow \mathscr{A}\) 是同构函子
    3. \(F: \mathscr{A} \rightarrow \mathscr{B}\) 为同构函子
      1. \(F\) 的逆存在且唯一,常用 \(F^{-1}\) 表示
      2. \(F^{-1}: \mathscr{B} \rightarrow \mathscr{A}\) 也是同构函子,且有 \(\left(F^{-1}\right)^{-1}=F\)
      3. \(G: \mathscr{B} \rightarrow \mathscr{A}\) 为同构函子,则 \(G \circ F: \mathscr{A} \rightarrow \mathscr{C}\) 也是同构函子,且有 \((G \circ F)^{-1}=F^{-1} \circ G^{-1}\)
    4. \(F: \mathscr{A} \rightarrow \mathscr{B}\) 为协变函子,则以下条件等价
      1. \(F\) 为同构函子
      2. 函数 \(F: \operatorname{Mor} \mathscr{A} \rightarrow \mathrm{Mor} \mathscr{B}\) 为双射
      3. \(F\) 为忠信的和完全的,且函数 \(F: \operatorname{Ob} \mathscr{A} \rightarrow \operatorname{Ob} \mathscr{B}\) 为双射
  2. \(\mathscr{A}\) 为范畴,称 \(\mathscr{A}\) 为骨架范畴当且仅当 \(\mathscr{A}\) 满足 \(A, B \in \operatorname{Ob} \mathscr{A}\) 使得 \(A \simeq B\) 蕴含 \(A=B\)

    1. 称范畴 \(\mathscr{B}\)\(\mathscr{A}\) 的骨架当且仅当 \(\mathscr{B}\) 满足
      1. \(\mathscr{B}\)\(\mathscr{A}\) 的完全子范畴
      2. \(\mathscr{B}\) 为骨架范畴
      3. \(\mathscr{B}\) 为极大的,即当 \(\mathscr{A}\) 的完全子范畴 \(\mathscr{C}\) 为骨架范畴且有 \(\mathscr{B} \subseteq \mathscr{C}\) 时,皆有 \(\mathscr{C}=\mathscr{B}\)
    2. \(\mathscr{A}\) 为范畴,则有
      1. \(\mathscr{A}_{\varphi}\)\(\mathscr{A}\) 的骨架
      2. \(\widetilde{\mathscr{A}}\)\(\mathscr{A}\) 的骨架,则 \(\widetilde{\mathscr{A}} \cong \mathscr{A}_{\varphi}\)
      3. \(\mathscr{A}_{1}\)\(\mathscr{A}_{2}\) 均为 \(\mathscr{A}\) 的骨架,则 \(\mathscr{A}_{1} \cong \mathscr{A}_{2}\)
    3. 若范畴 \(\mathscr{A}\) 有骨架 \(\widetilde{\mathscr{A}}\) 且范畴 \(\mathscr{B}\) 有骨架 \(\widetilde{\mathscr{B}}\) 使得 \(\widetilde{\mathscr{A}} \cong \widetilde{\mathscr{B}}\),则称 \(\mathscr{A}\) 等价于 \(\mathscr{B}\)\(\mathscr{A}\)\(\mathscr{B}\) 等价,记作 \(\mathscr{A} \simeq \mathscr{B}\)
      1. \(\mathscr{A} \simeq \mathscr{A}\)
      2. \(\mathscr{A} \simeq \mathscr{B}\),则 \(\mathscr{B} \simeq \mathscr{A}\)
      3. \(\mathscr{A} \simeq \mathscr{B}\)\(\mathscr{B} \simeq \mathscr{C}\),则 \(\mathscr{A} \simeq \mathscr{C}\)
      4. \(\mathscr{A}\)\(\mathscr{B}\) 均为骨架范畴,则 \(\mathscr{A} \simeq \mathscr{B}\) 当且仅当 \(\mathscr{A} \cong \mathscr{B}\)
  3. 若协变函子 \(F: \mathscr{A} \rightarrow \mathscr{C}\) 为忠信的、完全的和稠密的,则称 \(F\) 为等价函子

    1. \(\widetilde{\mathscr{A}}\) 为范畴 \(\mathscr{A}\) 的骨架且 \(E: \widetilde{\mathscr{A}} \rightarrow \mathscr{A}\) 为包含函子

      1. \(A \in \operatorname{Ob} \mathscr{A}\),则存在唯一的 \(\widetilde{A} \in \operatorname{Ob} \widetilde{\mathscr{A}}\) 使得 \(\widetilde{A} \simeq A\)
      2. \(E\) 为等价函子
      3. 存在等价函子 \(P: \mathscr{A} \rightarrow \widetilde{\mathscr{A}}\) 使得 \(P \circ E=I_{\widetilde{\mathscr{A}}}\)

      通常称函子 \(P\) 为投影函子

    2. \(\mathscr{A} \simeq \mathscr{B}\) 当且仅当存在等价函子 \(F: \mathscr{A} \rightarrow \mathscr{B}\)

    3. \(\mathscr{A} \cong \mathscr{B}\),则 \(\mathscr{A} \simeq \mathscr{B}\)