Blog Post

行列式4-拉普拉斯定理

  • note
  • Math
  • LinearAlgebra
  • 行列式

行列式4-拉普拉斯定理

作者:Eric

创建时间:2026-07-29 14:21

这篇记什么

  • kk 阶子式
  • 拉普拉斯定理

主要内容

  1. kk 阶子式及其代数余子式

    nn 阶矩阵 A=(aij)A=(a_{ij}) ,任意取定 kk 行和 kk 列

    第 i1,i2,...,iki_1,i_2,...,i_k 行(其中 i1<i2<...<iki_1<i_2<...<i_k )

    第 j1,j2,...,jkj_1,j_2,...,j_k 列(其中 j1<j2<...<jkj_1<j_2<...<j_k)

    这 kk 行和 kk 列的元素按照 原来的排列 形成一个 kk 阶的行列式,称它为矩阵 AA 的一个 kk 阶子式,记作

    A(i1,i2,...,ikj1,j2,...,jk)(1)A\begin{pmatrix}i_1,i_2,...,i_k \\ j_1,j_2,...,j_k\end{pmatrix} \tag{1}

    划去上述 kk 行和 kk 列,剩下的元素按 原来的排列 行程的 n−kn-k 阶行列式,称为子式 11 的余子式。

    令 {i1′,i2′,...,in−k′}={1,2,...,n}\{i1,i2,...,ik}\{i_1',i_2',...,i_{n-k}'\}=\{1,2,...,n\} \backslash \{i_1,i_2,...,i_k\} ,i1′<i2′<...<in−k′i_1'<i_2'<...<i_{n-k}'

    令 {j1′,j2′,...,jn−k′}={1,2,...,n}\{j1,j2,...,jk}\{j_1',j_2',...,j_{n-k}'\}=\{1,2,...,n\} \backslash \{j_1,j_2,...,j_k\} ,j1′<j2′<...<jn−k′j_1'<j_2'<...<j_{n-k}'

    子式 11 的余子式也是 AA 的一个 n−kn-k 阶余子式,记作

    A(i1′,i2′,...,in−k′j1′,j2′,...,jn−k′)A\begin{pmatrix}i_1',i_2',...,i_{n-k}' \\ j_1',j_2',...,j_{n-k}'\end{pmatrix}

    此时,

    (−1)(i1+i2+...+ik)+(j1+j2+...+jk)A(i1′,i2′,...,in−k′j1′,j2′...,jn−k′)(-1)^{(i_1+i_2+...+i_k)+(j_1+j_2+...+j_k)}A\begin{pmatrix}i_1',i_2',...,i_{n-k}' \\ j_1',j_2'...,j_{n-k}'\end{pmatrix}

    称为子式 11 的代数余子式

  2. 行列式按 kk 行 kk 列展开(拉普拉斯定理)

    nn 阶矩阵 A=(aij)A=(a_{ij}) ,取定第 i1,i2,...,iki_1,i_2,...,i_k 行(其中 i1<i2<...<iki_1<i_2<...<i_k ),则 detAdetA 等于这 kk 行形成的所有 kk 阶子式(列从 nn 列中任取 kk 列,共 CnkC_n^k 种)与它自己的代数余子式的乘积之和。

    证明如下,

    根据行列式的定义

    detA=∑u1...ukv1...vn−k(−1)τ(x)ai1u1ai2u2...aikukai1′v1ai2′v2...ain−k′vn−kdetA=\sum_{u_1...u_kv_1...v_{n-k}}(-1)^{\tau(x)}a_{{\color{red}i_1}u_1}a_{{\color{red}i_2}u_2}...a_{{\color{red}i_k}u_k}a_{{\color{green}i_1'}v_1}a_{{\color{green}i_2'}v_2}...a_{{\color{green}i_{n-k}'}v_{n-k}}

    (1)行指标的逆序数:

    第 i1\color{red}i_1 个位置由原来的 i1\color{red}i_1 位换到了 11 位,逆序数是 i1−1i_1-1

    第 im\color{red}i_m 个位置, m∈{1,2,...,k}m \in \{1,2,...,k\},由原来的 的 im\color{red}i_m 位换到了 mm 位,逆序数是 im−mi_m-m (其比前面的数都要大)

    第 ic′\color{green}i_c' 个位置, c∈{1,2,...,n−k}c \in \{1,2,...,n-k\} ,均属顺序数内,因此 ic′\color{green}i_c' 的逆序数都为 00

    (2)列指标的逆序数:τ(u1u2...ukv1v2...vn−k)\tau(u_1u_2...u_kv_1v_2...v_{n-k})

    由此,

    detA=∑u1...ukv1...vn−k(−1)(i1−1)+(i2−2)+...+(ik−k)⏞行+τ(u1...ukv1...vn−k)⏞列ai1u1ai2u2...aikukai1′v1ai2′v2...ain−k′vn−kdetA=\sum_{u_1...u_kv_1...v_{n-k}}(-1)^{\overbrace{(i_1-1)+(i_2-2)+...+(i_k-k)}^{\text{行}}+\overbrace{\tau(u_1...u_kv_1...v_{n-k})}^{\text{列}}}a_{{\color{red}i_1}u_1}a_{{\color{red}i_2}u_2}...a_{{\color{red}i_k}u_k}a_{{\color{green}i_1'}v_1}a_{{\color{green}i_2'}v_2}...a_{{\color{green}i_{n-k}'}v_{n-k}}

    上式中较难处理的有 ∑u1...ukv1...vn−k\sum_{u_1...u_kv_1...v_{n-k}} 和 τ(u1...ukv1...vn−k)\tau(u_1...u_kv_1...v_{n-k}) 两部分

    (a) 先试着拆解 ∑u1...ukv1...vn−k\sum_{u_1...u_kv_1...v_{n-k}} ,

    由 kk 阶子式的定义,取定第 j1,j2,...,jkj_1,j_2,...,j_k 列(j1<j2<...<jkj_1<j_2<...<j_k)

    u1...uku_1...u_k 取遍 j1...jkj_1...j_k 的 kk 元排列,而 v1...vn−kv_1...v_{n-k} 取遍 j1′...jn−k′j_1'...j_{n-k}' 的 n−kn-k 元排列

    因此,

    ∑u1...ukv1...vn−k=∑j1<j2<...<jn∑u1...uk∑v1...vn−k\sum_{u_1...u_kv_1...v_{n-k}}=\sum_{j_1<j_2<...<j_n}\sum_{u_1...u_k}\sum_{v_1...v_{n-k}}

    (b) 下计算 τ(u1...ukv1...vn−k)\tau(u_1...u_kv_1...v_{n-k})

    仅关注 u1...uku_1...u_k 部分,若 u1...uk→s次对换j1...jku_1...u_k \xrightarrow[]{s\text{次对换}} j_1...j_k ,

    则 τ(u1...uk)=s+τ(j1...jk)\tau(u_1...u_k)=s + \tau(j_1...j_k)

    由于 j1...jkj_1...j_k 为偶排列(顺序排列,逆序数为 00 ),因此 τ(u1...uk)=s\tau(u_1...u_k)=s

    (−1)τ(u1...ukv1...vn−k)=(−1)s(−1)τ(j1...jkv1...vn−k)=(−1)τ(u1...uk)(−1)τ(j1...jkv1...vn−k)(-1)^{\tau(u_1...u_kv_1...v_{n-k})}=(-1)^{\color{red}s}(-1)^{\tau(j_1...j_kv_1...v_{n-k})}=(-1)^{{\color{red}\tau(u_1...u_k)}}(-1)^{\tau(j_1...j_kv_1...v_{n-k})}

    接着,计算 τ(j1...jkv1...vn−k)\tau({\color{red}j_1...j_k}{\color{green}v_1...v_{n-k}})

    同前文行指标的逆序数中 im\color{red}i_m 的部分,不同的是前文处理中的 ic′\color{green}i_c' 逆序数为 00,此处后半部分的逆序数为 τ(v1...vn−k)\tau(v_1...v_{n-k}) ,

    (−1)τ(j1...jkv1...vn−k)=(−1)(j1−1)+(j2−2)+...+(jk−k)+τ(v1...vn−k)(-1)^{\tau({\color{red}j_1...j_k}{\color{green}v_1...v_{n-k}})}=(-1)^{{\color{red}(j_1-1)+(j_2-2)+...+(j_k-k)}+{\color{green}\tau(v_1...v_{n-k})}}

    因此,

    (−1)τ(u1...ukv1...vn−k)=(−1)τ(u1...uk)+(j1−1)+(j2−2)+...+(jk−k)+τ(v1...vn−k)(-1)^{\tau(u_1...u_kv_1...v_{n-k})}=(-1)^{{\color{red}\tau(u_1...u_k)}+{\color{green}(j_1-1)+(j_2-2)+...+(j_k-k)}+{\color{red}\tau(v_1...v_{n-k})}}

    将 ∑u1...ukv1...vn−k\sum_{u_1...u_kv_1...v_{n-k}} 和 τ(u1...ukv1...vn−k)\tau(u_1...u_kv_1...v_{n-k}) 两部分代入,整合得

    detA=∑j1<j2<...<jn∑u1...uk∑v1...vn−k(−1)(i1−1)+(i2−2)+...+(ik−k)⏞行+τ(u1...uk)+(j1−1)+(j2−2)+...+(jk−k)+τ(v1...vn−k)⏞列ai1u1ai2u2...aikukai1′v1ai2′v2...ain−k′vn−kdetA=\sum_{j_1<j_2<...<j_n}\sum_{u_1...u_k}\sum_{v_1...v_{n-k}}(-1)^{\overbrace{(i_1-1)+(i_2-2)+...+(i_k-k)}^{\text{行}}+\overbrace{{\color{red}\tau(u_1...u_k)}+{\color{green}(j_1-1)+(j_2-2)+...+(j_k-k)}+{\color{red}\tau(v_1...v_{n-k})}}^{\text{列}}}a_{{\color{red}i_1}u_1}a_{{\color{red}i_2}u_2}...a_{{\color{red}i_k}u_k}a_{{\color{green}i_1'}v_1}a_{{\color{green}i_2'}v_2}...a_{{\color{green}i_{n-k}'}v_{n-k}}

    其中,由于出现了两次 1+2+...+k1+2+...+k ,由等差数列的求和公式,这两项可消除

    提取可提项,化简得

    detA=∑j1<j2<...<jn(−1)(i1+...+ik)+(j1+...+jk)∑u1...uk(−1)τ(u1...uk)ai1u1ai2u2...aikuk∑v1...vn−k(−1)τ(v1...vn−k)ai1′v1ai2′v2...ain−k′vn−kdetA=\sum_{j_1<j_2<...<j_n}(-1)^{(i_1+...+i_k)+(j_1+...+j_k)}\sum_{u_1...u_k}(-1)^{\tau(u_1...u_k)}a_{{\color{red}i_1}u_1}a_{{\color{red}i_2}u_2}...a_{{\color{red}i_k}u_k}\sum_{v_1...v_{n-k}}(-1)^{\tau(v_1...v_{n-k})}a_{{\color{green}i_1'}v_1}a_{{\color{green}i_2'}v_2}...a_{{\color{green}i_{n-k}'}v_{n-k}}

    此式即为,

    detA=∑j1<j2<...<jn(−1)(i1+...+ik)+(j1+...+jk)A(i1,i2,...,ikj1,j2...,jk)A(i1′,i2′,...,in−k′j1′,j2′,...,jn−k′)det A =\sum_{j_1<j_2<...<j_n}(-1)^{(i_1+...+i_k)+(j_1+...+j_k)}A\begin{pmatrix}i_1,i_2,...,i_k \\ j_1,j_2...,j_k\end{pmatrix} A\begin{pmatrix}i_1',i_2',...,i_{n-k}' \\ j_1',j_2',...,j_{n-k}'\end{pmatrix} detA=∑j1<j2<...<jnA(i1,i2,...,ikj1,j2...,jk)⋅(−1)(i1+...+ik)+(j1+...+jk)A(i1′,i2′,...,in−k′j1′,j2′,...,jn−k′)det A =\sum_{j_1<j_2<...<j_n}A\begin{pmatrix}i_1,i_2,...,i_k \\ j_1,j_2...,j_k\end{pmatrix}\cdot(-1)^{(i_1+...+i_k)+(j_1+...+j_k)} A\begin{pmatrix}i_1',i_2',...,i_{n-k}' \\ j_1',j_2',...,j_{n-k}'\end{pmatrix}

    上述式子可表述为: detAdetA 等于这 kk 行形成的所有 kk 阶子式与它自己的代数余子式的乘积之和。

    原定理得证。

相关

  • 高等代数(上册) - 丘维声著. 北京:清华大学出版社,2010.6