일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | ||
6 | 7 | 8 | 9 | 10 | 11 | 12 |
13 | 14 | 15 | 16 | 17 | 18 | 19 |
20 | 21 | 22 | 23 | 24 | 25 | 26 |
27 | 28 | 29 | 30 | 31 |
- RG flow
- induced emf
- scalar QED
- Tensor Calculus
- Quantum Mechanics
- Taylor expansion
- Topological Tensor
- Paulo Freire
- spinor QED
- Quantum electodynamics
- Theories of Education
- Pen Art
- Hamiltonian Mechanics
- Classical Mechanics
- Complex analysis
- bayesian
- Orbital Dynamics
- Dirichlet problem
- Conformal Mapping
- Quantum Field Theory
- pedagogy
- 프레이리
- General Relativity
- Noether's thm
- Covariant derivatives
- higher derivatives
- Lagrangian Mechanics
- Electromagnetics
- electromotive force
- Rigid Body Dynamics
- Today
- Total
Physvillain
Basic Tensor Calculus (2) - Definition of a tensor 본문
Basic Tensor Calculus (2) - Definition of a tensor
Physvillain 2020. 10. 30. 10:37이 글은 2020.01.14에 작성됨.
원본 : physvillain.blogspot.com/2020/01/basic-tensor-calculus.html
두 개의 coordinate system $x_i$와 $\tilde{x}_i$를 생각하자. 두 coordinate간의 변환은 다음과 같을 것이다. (예를 들면 spherical과 Cartesian의 관계를 생각해보자.)
$$\tilde{x}_i = \tilde{x}_i (x_1, x_2,\cdots, x_n)$$
$$x_i = x_i ( \tilde{x}_1, \tilde{x}_2, \cdots, \tilde{x}_n)$$
infinitesimal position $d\vec{r}$의 성분 $dx_i$의 좌표변환 관계는 chain rule에 의해 다음과 같다.
$$d\tilde{x}_i = \sum_j \frac{\partial \tilde{x}_i}{\partial x_j} dx_j$$
다음으로 $\frac{\partial f}{\partial x_i}$의 좌표변환 관계도 생각해 볼 수 있을 것이다. (참고로 이것은 Cartesian에서만 gradient이고, 일반적으로 gradient가 아니다. spherical coordinate를 생각해보라.) 이것의 좌표변환 관계 역시 chain rule에 의해 다음과 같다.
$$\frac{\partial f}{\partial \tilde{x}_i} = \sum_j \frac{\partial x_j}{\partial \tilde{x}_i} \frac{\partial f}{\partial x_j}$$
두 경우를 가만히 살펴보면 좌표변환에 대해서 변화하는 방식이 다르다. 좌표 변환에 대해서 $dx_i$와 같이 변하는 벡터를 contravariant vector라 하고, $\frac{\partial f}{\partial x_i}$와 같이 변하는 벡터를 covariant vector라 한다.
DEF 2.1. The components of a covariant vector transform like a cartesian gradient and obey the transformation rule :
$$ \tilde{A}_i = \sum_{j} \frac{\partial x^j}{\partial \tilde{x}^i} A_j$$
DEF 2.2. The components of a contravariant vector transform like a coordinate differential and obey the transformation rule :
$$ \tilde{A}^i = \sum_{j} \frac{\partial \tilde{x}^i}{\partial x^j} A^j$$
DEF를 잘 살펴보면 contravariant vector의 index를 위로 올려 쓴 것을 볼 수 있을 것이다. 일반적인 관례로 contravariant 성분은 위로 올려쓰고, covariant 성분은 아래로 내려쓰는데, 분모에 covariant 성분이 오면 그 index는 contravariant한 것으로 취급한다. 또, notation의 편의성을 위해 covariant index와 contravariant index가 서로 summation이 되어 있으면 사실상 그 index는 dummy이고, summation기호를 써주지 않는다. 이것을 Einstein summation convention이라 한다. 예를들어 $A_i B^i$는 index i에 대해서 summation이 취해진 것으로 생각하지만, $A_i B_i$나 $A_{(i)} B_{(i)}$는 그렇지 않다. 이 convention을 이용하여 DEF 2.1~2를 다시 써 보면 다음과 같다.
$$\tilde{A}_i=\frac{\partial x^j}{\partial \tilde{x}^i} A_j, \:\:\:\:\: \tilde{A}^i = \frac{\partial \tilde{x}^i}{\partial x^j} A^j$$
흔히들 헷갈리는 것으로 $dx^i$는 전형적인 rank 1 contravariant tensor이지만, $x^i$는 DEF 2.1의 변환규칙을 따르지 않고, 이것은 tensor가 아니다. 변환식에 들어간 편미분 항에서 $x^i$와 같이 표기한 것은 그저 편의성을 위한 것이지, $x^i$가 contravariant vector임을 의미하는 것은 이니다. 또 kronecker delta나 levi-civita symbol도 Einstein summation convention을 위해 위 아래 index를 섞어 쓰는데, 이 index역시 contra/co-variant를 의미하지는 않는다. 마찬가지로 tensor가 아니다.
마찬가지 규칙으로 higher rank tensor에 대해 일반화 해보면, rank n tensor의 coordinate transformation rule은 다음과 같다.
$$\tilde{T}_{i_1 \cdots i_p}^{\;\;\;\;\;\;\;k_1 \cdots k_q}= \frac{\partial x^{j_1}}{\partial \tilde{x}^{i_1}} \cdots \frac{\partial x^{j_p}}{\partial \tilde{x}^{i_p}} \frac{\partial \tilde{x}^{k_1}}{\partial x^{l_1}} \cdots \frac{\partial \tilde{x}^{k_q}}{\partial x^{l_q}} T_{j_1 \cdots j_p}^{\;\;\;\;\;\;\;l_1 \cdots l_q}$$
여기서 $p+q=n$이 성립해야 할 것이다. tensor를 더하거나 양변에 등식을 쓸 때는 dummy indices를 제외하고 식을 구성하는 tensor의 알짜 indices는 같은 순서에, 같은 up/down position에 존재해야 한다. 일부 책에서는 tensor $T_{i \: kl}^{\:j\:\:m}$의 index를 순서 상관없이 $T_{ikl}^{jm}$과 같이 쓰기도 하나, 엄연히 이 물리량이 tensor라면 이렇게 쓰면 안 된다. 허나 뒤에 나올 Christoffel symbol과 같이 tensor가 아닌 경우에는 이와 같이 순서를 무시하고 섞어 쓰기도 한다.
THM 2.1. The sum of two like-tensor is a like-tensor of the same type.
한 가지 주의해야 할 점으로, 우리가 흔히 vector "identities" 라고 배운 것들은 좌표계와 상관없이 참이다. 예를 들자면, $\nabla \cdot f\vec{A} = f \nabla \cdot \vec{A} + \vec{A} \cdot \nabla f$는 어느 좌표계든지 성립한다. 그러나, (당연하겠지만) 성분으로 쓴 식들은 보통 Cartesian에서만 성립하므로, 좌표계의 선택에 따라 성분별 표기는 언제든지 달라질 수 있다. 즉 일반적인 좌표계에서는 $(A \times B)^i \neq \epsilon^{ijk} A_j B_k$이다. 이것은 오로지 Cartesian에서만 등호가 성립한다.
특별히 rank 1 tensor (vector) 두 개의 product를 dyad라 한다.
DEF 2.3. A rank 2 dyad, $\mathbf{D}$, results from taking the dyadic product of 2 vectors, $\vec{A}$ and $\vec{B}$.
$$\mathbf{D}=\vec{A}\vec{B}, \:\:\: D_i^{\:j}=A_iB^j, \:\:\: D^i_{\:j}=A^i B_j, \:\:\: D^{ij}=A^iB^j,\:\:\: D_{ij}=A_iB_j$$
THM 2.2. A rank 2 dyad is a rank 2 tensor.
이것의 증명은 간단하므로 생략한다. 다음으로 tensor끼리의 inverse relation을 살펴보자. (matrix처럼 생각하면 된다.)
DEF 2.4. If $A_{ij}$ is a rank 2 covariant tensor and $B^{kl}$ is a rank 2 contravariant tensor, then they are each other's inverse if $A_{ij} B^{jk} = \delta_i^{\:k}$.
THM 2.3. The derivatives $\partial x^i / \partial \tilde{x}^j$ and $\partial \tilde{x}^j / \partial x^k$ are each other's inverse. That is, $\frac{\partial x^i}{\partial \tilde{x}^j} \frac{\partial \tilde{x}^j}{\partial x^k} = \delta^i_{\:k}$.
이것의 증명은 chain rule을 이용하면 자명하므로 생략한다.
DEF 2.5. A tensor contraction occurs when one of a tensor's free covariant indices is set equal to one of its free contravariant indices. In this case, a sum is performed.
예를들어서 $T_{ij}^{\:\:j}$는 rank 1 tensor이고, 어느 책에서는 줄여서 $T_{ij}^{\:\:j}=T_i$로 적기도 한다. 마찬가지로 $T_i^{\:i}$는 matix T의 trace이고 coordinate system에 무관한 scalar이다.
THM 2.4. A contraction of a rank 2 tensor (its trace) is a scalar (rank 0 tensor) whose value is independent of the coordinate system chosen.
증명은 간단하다.
$$\tilde{T}_k^{\:k} = \frac{\partial x^i}{\partial \tilde{x}^k} \frac{\partial \tilde{x}^k}{\partial x^j} T_i^{\:j}=\delta^i_{\:j}T_i^{\:j}=\mathrm{tr}\:T$$
'General Relativity' 카테고리의 다른 글
Basic Tensor Calculus (5) - Scalar & Product (0) | 2020.10.30 |
---|---|
Basic Tensor Calculus (4) - Physical components & Basis vectors (0) | 2020.10.30 |
Basic Tensor Calculus (3) - Metric (0) | 2020.10.30 |
Basic Tensor Calculus (1) - Introduction (0) | 2020.10.30 |
Classical Tests of General Relativity (only using Schwartzchild metric) (0) | 2020.10.19 |