（九）二阶张量的不变量

Albert M

已于 2022-10-10 20:35:09 修改

阅读量2.9k

点赞数 1

分类专栏：张量分析

于 2022-08-17 17:43:32 首次发布

本文链接：https://blog.csdn.net/qq_51453181/article/details/126375256

版权

张量分析专栏收录该内容

27 篇文章 23 订阅

订阅专栏

本文主要内容如下：

1. 张量的标量不变量
2. 二阶张量的主不变量
3. 二阶张量的矩

1. 张量的标量不变量

张量 $\bold T$ 自身不随坐标的改变而改变，张量 $\bold T$ 通过与自身、度量张量 $\bold G$ 或置换张量 $\epsilon$ 进行张量的代数运算可以得到一系列标量，这些标量也不随坐标的改变而改变，将其称之为张量的（标量）不变量。如，
$\begin{aligned} &\bold{G}:\bold{T}=\bold{G}\cdot\cdot\bold{T}=T_i^{\bullet i}=T^i_{\bullet i}=tr{\bold{T}}\\\ \\ &\bold{T}:\bold{T}=T_i^{\bullet j}T^i_{\bullet j}\\\ \\ &\bold{T}\cdot\cdot\bold{T}=T_i^{\bullet j}T_j^{\bullet i}=T^i_{\bullet j}T^j_{\bullet i}=tr(\bold{T}\cdot\bold{T})\\\ \\ &\epsilon_{ijk}\epsilon^{pql}T^{i}_{\bullet p}T^{j}_{\bullet q}T^{k}_{\bullet l}\\\ \\ &\cdots \end{aligned}$

2. 二阶张量的主不变量

在二阶张量 $\bold T$ 众多的不变量中，将如下三个不变量称作二阶张量 $\bold T$ 的主不变量：
$\begin{aligned} & \mathscr{C}_1^{T}= \delta^i_pT^p_{\bullet i} =T^i_{\bullet i} =T^1_{\bullet 1}+T^2_{\bullet 2}+T^3_{\bullet 3} =tr(\bold T) =\bold{G}:\bold{T} =\bold{G}\cdot\cdot\bold{T} \\\ \\& \mathscr{C}_2^{T} =\frac{1}{2}\delta^{ij}_{pq}T^p_{\bullet i}T^q_{\bullet j} =\frac{1}{2}(T^i_{\bullet i}T^j_{\bullet j}-T^j_{\bullet i}T^i_{\bullet j}) =\begin{vmatrix} T^1_{\bullet 1} & T^1_{\bullet 2} \\\ \\ T^2_{\bullet 1} & T^2_{\bullet 2} \end{vmatrix} +\begin{vmatrix} T^1_{\bullet 1} & T^1_{\bullet 3} \\\ \\ T^3_{\bullet 1} & T^3_{\bullet 3} \end{vmatrix} +\begin{vmatrix} T^2_{\bullet 2} & T^2_{\bullet 3} \\\ \\ T^3_{\bullet 2} & T^3_{\bullet 3} \end{vmatrix} \\\ \\& \mathscr{C}_3^{T} =\frac{1}{3!}\delta^{ijk}_{pql}T^p_{\bullet i}T^q_{\bullet j}T^l_{\bullet k} =\frac{1}{6}T^i_{\bullet i}T^j_{\bullet j}T^k_{\bullet k}-\frac{1}{2}T^k_{\bullet k}T^j_{\bullet i}T^i_{\bullet j}+\frac{1}{3}T^k_{\bullet i}T^i_{\bullet j}T^j_{\bullet k} =\frac{1}{3!}(T^p_{\bullet i}T^q_{\bullet j}T^l_{\bullet k}e_{pql})e^{ijk} =\frac{1}{3!}(T^p_{\bullet 1}T^q_{\bullet 2}T^l_{\bullet 3}e_{pql}e_{ijk})e^{ijk} =\frac{1}{3!}(T^p_{\bullet 1}T^q_{\bullet 2}T^l_{\bullet 3}e_{pql})(e_{ijk}e^{ijk}) =det(\bold T) =\begin{vmatrix} T^1_{\bullet 1} & T^1_{\bullet 2} & T^1_{\bullet 3}\\\ \\ T^2_{\bullet 1} & T^2_{\bullet 2} & T^2_{\bullet 3}\\\ \\ T^3_{\bullet 1} & T^3_{\bullet 2} & T^3_{\bullet 3} \end{vmatrix} \end{aligned}$ 可知，二阶张量 $\bold T$ 的主不变量 $\mathscr{C}_1、\mathscr{C}_2、\mathscr{C}_3$ 分别为张量的矩阵 $\tau_{3}$ 的一、二、三阶主子式之和。可进一步证明存在如下关系式：
$\begin{aligned} &(1) \begin{bmatrix}\bold{T}\bullet\vec{u} & \vec{v} & \vec{w}\end{bmatrix}+ \begin{bmatrix}\vec{u} & \bold{T}\bullet\vec{v} & \vec{w}\end{bmatrix}+ \begin{bmatrix}\vec{u} & \vec{v} & \bold{T}\bullet\vec{w}\end{bmatrix}= \mathscr{C}_1^T\begin{bmatrix}\vec{u} & \vec{v} & \vec{w}\end{bmatrix}\\\ \\ &(2) \begin{bmatrix}\bold{T}\bullet\vec{u} & \bold{T}\bullet\vec{v} & \vec{w}\end{bmatrix}+ \begin{bmatrix}\vec{u} & \bold{T}\bullet\vec{v} & \bold{T}\bullet\vec{w}\end{bmatrix}+ \begin{bmatrix}\bold{T}\bullet\vec{u} & \vec{v} & \bold{T}\bullet\vec{w}\end{bmatrix}= \mathscr{C}_2^T\begin{bmatrix}\vec{u} & \vec{v} & \vec{w}\end{bmatrix}\\\ \\ &(3) \begin{bmatrix}\bold{T}\bullet\vec{u} & \bold{T}\bullet\vec{v} & \bold{T}\bullet\vec{w}\end{bmatrix}= \mathscr{C}_3^T\begin{bmatrix}\vec{u} & \vec{v} & \vec{w}\end{bmatrix} \end{aligned}$ 第三式的证明参见张量行列式的几何意义说明，前两式证明如下：
$\begin{aligned} &\quad \begin{bmatrix}\bold{T}\bullet\vec{u} & \vec{v} & \vec{w}\end{bmatrix}+ \begin{bmatrix}\vec{u} & \bold{T}\bullet\vec{v} & \vec{w}\end{bmatrix}+ \begin{bmatrix}\vec{u} & \vec{v} & \bold{T}\bullet\vec{w}\end{bmatrix} \\\ \\ &=(T^i_{\bullet m}u^m)v^jw^k\epsilon_{ijk}+u^i(T^j_{\bullet m}v^m)w^k\epsilon_{ijk}+u^iv^j(T^k_{\bullet m}w^m)\epsilon_{ijk} \\\ \\ &=u^iv^jw^k(T^m_{\bullet i}\epsilon_{mjk}+T^m_{\bullet j}\epsilon_{imk}+T^m_{\bullet k}\epsilon_{ijm}) \\\ \\ &=\frac{1}{2}u^iv^jw^k[(T^m_{\bullet i}\epsilon_{mjk}+T^m_{\bullet j}\epsilon_{imk}+T^m_{\bullet k}\epsilon_{ijm})+(T^m_{\bullet i}\epsilon_{kmj}+T^m_{\bullet j}\epsilon_{kim}-T^m_{\bullet k}\epsilon_{mji})] \\\ \\ &=\frac{1}{2}u^iv^jw^k (T^m_{\bullet p}\epsilon_{mql}\delta^{p}_{i}\delta^{q}_{j}\delta^{l}_{k}+ T^m_{\bullet p}\epsilon_{qml}\delta^{p}_{j}\delta^{q}_{i}\delta^{l}_{k}+ T^m_{\bullet p}\epsilon_{qlm}\delta^{p}_{k}\delta^{q}_{i}\delta^{l}_{j}+ T^m_{\bullet p}\epsilon_{qml}\delta^{p}_{i}\delta^{q}_{k}\delta^{l}_{j}+ T^m_{\bullet p}\epsilon_{qlm}\delta^{p}_{j}\delta^{q}_{k}\delta^{l}_{i}- T^m_{\bullet p}\epsilon_{mql}\delta^{p}_{k}\delta^{q}_{j}\delta^{l}_{i})\\\ \\ &=\frac{1}{2}u^iv^jw^k(T^m_{\bullet p}\epsilon_{mql}) (\delta^{p}_{i}\delta^{q}_{j}\delta^{l}_{k}- \delta^{p}_{j}\delta^{q}_{i}\delta^{l}_{k}+ \delta^{p}_{k}\delta^{q}_{i}\delta^{l}_{j}- \delta^{p}_{i}\delta^{q}_{k}\delta^{l}_{j}+ \delta^{p}_{j}\delta^{q}_{k}\delta^{l}_{i}- \delta^{p}_{k}\delta^{q}_{j}\delta^{l}_{i})\\\ \\ &=\frac{1}{2}u^iv^jw^kT^m_{\bullet p}(\epsilon_{mql}\epsilon^{pql})\epsilon_{ijk}\\\ \\ &=u^iv^jw^kT^m_{\bullet p}\epsilon_{ijk}\delta^p_m\\\ \\ &=T^m_{\bullet m}(u^iv^jw^k\epsilon_{ijk})\\\ \\ &=\mathscr{C}_1^T\begin{bmatrix}\vec{u} & \vec{v} & \vec{w}\end{bmatrix} \\\ \\ %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% &\quad \begin{bmatrix}\bold{T}\bullet\vec{u} & \bold{T}\bullet\vec{v} & \vec{w}\end{bmatrix}+ \begin{bmatrix}\vec{u} & \bold{T}\bullet\vec{v} & \bold{T}\bullet\vec{w}\end{bmatrix}+ \begin{bmatrix}\bold{T}\bullet\vec{u} & \vec{v} & \bold{T}\bullet\vec{w}\end{bmatrix}\\\ \\ &=(T^i_{\bullet m}u^m)(T^j_{\bullet n}v^n)w^k\epsilon_{ijk}+ u^i(T^j_{\bullet m}v^m)(T^k_{\bullet n}w^n)\epsilon_{ijk}+ (T^i_{\bullet m}u^m)v^j(T^k_{\bullet n}w^n)\epsilon_{ijk}\\\ \\ &=u^iv^jw^k(T^m_{\bullet i}T^n_{\bullet j}\epsilon_{mnk}+ T^m_{\bullet j}T^n_{\bullet k}\epsilon_{imn}+ T^m_{\bullet i}T^n_{\bullet k}\epsilon_{mjn})\\\ \\ &=\frac{1}{2}u^iv^jw^k(T^m_{\bullet i}T^n_{\bullet j}\epsilon_{mnk}+ T^m_{\bullet j}T^n_{\bullet k}\epsilon_{imn}+ T^m_{\bullet i}T^n_{\bullet k}\epsilon_{mjn})+ (T^n_{\bullet i}T^m_{\bullet j}\epsilon_{nmk}+ T^n_{\bullet j}T^m_{\bullet k}\epsilon_{inm}+ T^n_{\bullet i}T^m_{\bullet k}\epsilon_{njm})\\\ \\ &=\frac{1}{2}u^iv^jw^k (T^m_{\bullet p}T^n_{\bullet q}\epsilon_{mnl}\delta^p_i\delta^q_j\delta^l_k+ T^m_{\bullet p}T^n_{\bullet q}\epsilon_{lmn}\delta^p_j\delta^q_k\delta^l_i+ T^m_{\bullet p}T^n_{\bullet q}\epsilon_{mln}\delta^p_i\delta^q_k\delta^l_j+ T^n_{\bullet q}T^m_{\bullet p}\epsilon_{nml}\delta^p_j\delta^q_i\delta^l_k+ % T^n_{\bullet q}T^m_{\bullet p}\epsilon_{lnm}\delta^p_k\delta^q_j\delta^l_i+ T^n_{\bullet q}T^m_{\bullet p}\epsilon_{nlm}\delta^p_k\delta^q_i\delta^l_j)\\\ \\ &=\frac{1}{2}u^iv^jw^kT^m_{\bullet p}T^n_{\bullet q}\epsilon_{mnl}( \delta^p_i\delta^q_j\delta^l_k+ \delta^p_j\delta^q_k\delta^l_i- \delta^p_i\delta^q_k\delta^l_j- \delta^p_j\delta^q_i\delta^l_k- \delta^p_k\delta^q_j\delta^l_i+ \delta^p_k\delta^q_i\delta^l_j)\\\ \\ &=\frac{1}{2}u^iv^jw^kT^m_{\bullet p}T^n_{\bullet q}(\epsilon_{mnl}\epsilon^{pql})\epsilon_{ijk}\\\ \\ &=(u^iv^jw^k\epsilon_{ijk})(\frac{1}{2}T^m_{\bullet p}T^n_{\bullet q}\delta^{pq}_{mn})\\\ \\ &=\mathscr{C}_2^T\begin{bmatrix}\vec{u} & \vec{v} & \vec{w}\end{bmatrix} \end{aligned}$

根据上述推导过程可知：
$\begin{aligned} &(1) \ T^m_{\bullet i}\epsilon_{mjk}+ T^m_{\bullet j}\epsilon_{imk}+ T^m_{\bullet k}\epsilon_{ijm} =\mathscr{C}_1^T\epsilon_{ijk}\\\ \\ &(2) \ T^m_{\bullet i}T^n_{\bullet j}\epsilon_{mnk}+ T^m_{\bullet j}T^n_{\bullet k}\epsilon_{imn}+ T^m_{\bullet i}T^n_{\bullet k}\epsilon_{mjn} =\mathscr{C}_2^T\epsilon_{ijk}\\\ \\ &(3) \ T^m_{\bullet i}T^n_{\bullet j}T^l_{\bullet k}\epsilon_{mnl} =\mathscr{C}_3^T\epsilon_{ijk} \end{aligned}$

此外，对于正则张量的线性变换还有如下的Nanson公式 —— 描述了映射前后面元的变化情况
$(\bold{T}\bullet\vec{u})\times(\bold{T}\bullet\vec{v})=\mathscr{C}_3^T(\bold{T}^T)^{-1}(\vec{u}\times\vec{v})$ 证明如下：
(法一)
$\begin{aligned} & \quad \bold{T}^T\bullet[(\bold{T}\bullet\vec{u})\times(\bold{T}\bullet\vec{v})]\\\ \\ &=[(T^T)_p^{\bullet q}\vec{g}^p\vec{g}_q]\bullet[(T^i_{\bullet m}u^m)(T^j_{\bullet n}v^n)\epsilon_{ijk}\vec{g}^k]\\\ \\ &=(T^q_{\bullet p}\vec{g}^p\vec{g}_q)\bullet[(T^i_{\bullet m}u^m)(T^j_{\bullet n}v^n)\epsilon_{ijk}\vec{g}^k]\\\ \\ &=(T^i_{\bullet m}T^j_{\bullet n}T^k_{\bullet p}\epsilon_{ijk})u^mv^n\vec{g}^p\\\ \\ &=\mathscr{C}_3^T(u^mv^n\epsilon_{mnp}\vec{g}^p)\\\ \\ &=\mathscr{C}_3^T(\vec{u}\times\vec{v}) \end{aligned}$ 则：
$(\bold{T}\bullet\vec{u})\times(\bold{T}\bullet\vec{v}) =(\bold{T}^T)^{-1}\bullet[\mathscr{C}_3^T(\vec{u}\times\vec{v})] =\mathscr{C}_3^T(\bold{T}^T)^{-1}\bullet(\vec{u}\times\vec{v})$ (法二)

对任意矢量 $\vec{w}$ 有： $(\bold{T}\bullet\vec{w})\bullet[(\bold{T}\bullet\vec{u}) \times (\bold{T}\bullet\vec{v} )] = \mathscr{C}_3^T \vec{w}\bullet(\vec{u} \times \vec{v})= \vec{w}\bullet\bold{T}^T\bullet[(\bold{T}\bullet\vec{u}) \times (\bold{T}\bullet\vec{v} )] \\\ \\ \Longrightarrow \vec{w}\bullet\{\bold{T}^T\bullet[(\bold{T}\bullet\vec{u}) \times (\bold{T}\bullet\vec{v} )]-\mathscr{C}_3^T (\vec{u} \times \vec{v})\}=0$ 由 $\vec{w}$ 任意性知： $\bold{T}^T\bullet[(\bold{T}\bullet\vec{u}) \times (\bold{T}\bullet\vec{v} )]-\mathscr{C}_3^T (\vec{u} \times \vec{v})=0$ 由于 $\bold T$ 正则，故：
$(\bold{T}\bullet\vec{u}) \times (\bold{T}\bullet\vec{v} )=\mathscr{C}_3^T(\bold{T}^T)^{-1} (\vec{u} \times \vec{v})$

3. 二阶张量的矩

$n$ 个二阶张量 $\bold T$ 依次点积得到的二阶张量的迹称作二阶张量 $\bold T$ 的 $n$ 阶矩 $\mathscr{C}_n^{*T}$ 。如：
$\begin{aligned} &\mathscr{C}_1^{*T} =tr(\bold{T})=T^i_{\bullet i} \\\ \\ &\mathscr{C}_2^{*T}=tr(\bold{T \bullet T})=T^i_{\bullet j}T^j_{\bullet i} \\\ \\ &\mathscr{C}_3^{*T}=tr(\bold{T \bullet T \bullet T})=T^i_{\bullet j}T^j_{\bullet k}T^k_{\bullet i} \end{aligned}$ 可以证明，高于三阶的矩可由一、二、三阶矩得到。另外，二阶张量的一、二、三阶矩与其主不变量之间并不独立，可以互相转化:
$\begin{aligned} & \begin{cases} \mathscr{C}_1^{*T}= \mathscr{C}_1^{T}\\\ \\ \mathscr{C}_2^{*T}= (\mathscr{C}_1^{T})^2-2\mathscr{C}_2^{T}\\\ \\ \mathscr{C}_3^{*T}= (\mathscr{C}_1^{T})^3-3\mathscr{C}_1^{T}\mathscr{C}_2^{T}+3\mathscr{C}_3^{T} \end{cases} \Longleftarrow \begin{cases} \mathscr{C}_1^{T}= T^i_{\bullet i}=\mathscr{C}_1^{*T}\\\ \\ (\mathscr{C}_1^{T})^2-2\mathscr{C}_2^{T} =T^i_{\bullet i}T^j_{\bullet j}-2\cdot\frac{1}{2}(T^i_{\bullet i}T^j_{\bullet j}-T^j_{\bullet i}T^i_{\bullet j}) =T^j_{\bullet i}T^i_{\bullet j} =\mathscr{C}_2^{*T}\\\ \\ (\mathscr{C}_1^{T})^3-3\mathscr{C}_1^{T}\mathscr{C}_2^{T}+3\mathscr{C}_3^{T} =T^i_{\bullet i}T^j_{\bullet j}T^k_{\bullet k}-\frac{3}{2}T^k_{\bullet k}(T^i_{\bullet i}T^j_{\bullet j}-T^j_{\bullet i}T^i_{\bullet j})+3(\frac{1}{6}T^i_{\bullet i}T^j_{\bullet j}T^k_{\bullet k}-\frac{1}{2}T^k_{\bullet k}T^j_{\bullet i}T^i_{\bullet j}+\frac{1}{3}T^k_{\bullet i}T^i_{\bullet j}T^j_{\bullet k}) =T^k_{\bullet i}T^i_{\bullet j}T^j_{\bullet k} =\mathscr{C}_3^{*T} \end{cases} \\\ \\ & \begin{cases} \mathscr{C}_1^{T}= \mathscr{C}_1^{*T}\\\ \\ \mathscr{C}_2^{T}= \frac{1}{2}[(\mathscr{C}_1^{*T})^2-\mathscr{C}_2^{*T}]\\\ \\ \mathscr{C}_3^{T}= \frac{1}{6}(\mathscr{C}_1^{*T})^3-\frac{1}{2}\mathscr{C}_1^{*T}\mathscr{C}_2^{*T}+\frac{1}{3}\mathscr{C}_3^{*T} \end{cases} \Longleftarrow \begin{cases} \mathscr{C}_1^{*T}= T^i_{\bullet i}=\mathscr{C}_1^{T}\\\ \\ \frac{1}{2}[(\mathscr{C}_1^{*T})^2-\mathscr{C}_2^{*T}] =\frac{1}{2}(T^i_{\bullet i}T^j_{\bullet j}-2T^i_{\bullet j}T^j_{\bullet i}) =\mathscr{C}_2^{T}\\\ \\ \frac{1}{6}(\mathscr{C}_1^{*T})^3-\frac{1}{2}\mathscr{C}_1^{*T}\mathscr{C}_2^{*T}+\frac{1}{3}\mathscr{C}_3^{*T} =\frac{1}{6}T^i_{\bullet i}T^j_{\bullet j}T^k_{\bullet k}-\frac{1}{2}T^k_{\bullet k}T^j_{\bullet i}T^i_{\bullet j}+\frac{1}{3}T^k_{\bullet i}T^i_{\bullet j}T^j_{\bullet k} =\mathscr{C}_3^{T} \end{cases} \end{aligned}$