deepseek 给将军们讲微分几何

所谓的菌版


回复
头像
Amorphous
童生
帖子: 27
注册时间: 周三 4月 08, 2026 5:32 am
昵称: 阿毛

deepseek 给将军们讲微分几何

帖子 Amorphous »

I'll explain this mathematical construction step by step. Let me first read the content carefully.

This is a beautiful piece of differential geometry! Let me explain this in simpler terms.

What's the Big Picture?

Imagine you have a manifold (a curved space like the surface of a sphere or Earth's surface). On this manifold, you have a flow - think of it like wind blowing across the Earth's surface, moving points along.

The question is: How does this flow affect not just points, but also vectors attached to those points?

For example, if you have wind blowing across the Earth, it doesn't just move cities - it also pushes weather systems (which have velocity/direction) around!

The Main Characters

  1. M: Your manifold (like Earth's surface)
  2. TM: The tangent bundle - this is the "collection of all tangent vectors at all points"
    • Each point in M has its own "tangent space" (imagine a flat plane touching Earth at that point)
    • TM = all points × all possible tangent vectors at those points
  3. Φₜ: The flow on M (like wind moving points)
  4. V: The vector field generating the flow (describes the wind at each point)

The Key Idea: Lifting the Flow

1. From Points to Vectors

When you have a flow moving points on M, you can naturally extend it to move vectors:

If a point x moves to Φₜ(x) along the flow, then a tangent vector v at x gets:

  • Moved to the new location Φₜ(x)
  • Transformed by the Jacobian matrix dΦₜ (which tells you how nearby points are moving relative to each other)

2. Mathematically:

  • On M: Φₜ(x) = position after time t
  • On TM: Φ̃ₜ(v) = (Φₜ(x), dΦₜ(v))
    The vector gets pushed forward by the differential (Jacobian)

3. Why This Is Natural:

The Jacobian dΦₜ is the best linear approximation of how Φₜ transforms things near x. So:

  • If Φₜ is like "wind blowing across Earth"
  • Then dΦₜ tells you how the wind stretches/squeezes/rotates nearby points
  • And Φ̃ₜ applies this stretching/rotating to vectors attached to points

The Generator: Complete Lift

Every flow comes from a vector field (V on M). The lifted flow comes from a vector field on TM called Vᶜ (complete lift):

In coordinates (x, v):

  • Horizontal part: Vⁱ(x)∂/∂xⁱ - moves the base point x
  • Vertical part: (∂Vʲ/∂xⁱ)vⁱ∂/∂vʲ - changes the vector components based on how V changes

Analogy:

Imagine wind blowing over a field of grass:

  • Grass location (x): moves with the wind
  • Grass blades (v): not only move location, but also bend/twist because nearby wind speeds differ

The vertical term captures this "bending" - how the vector components change as you move.

Geometric Interpretation

Think of it as dragging vectors along with the flow:

  1. Base point moves: x → Φₜ(x)
  2. Vector transforms: v → dΦₜ(v)

This is canonical (natural) because:

  • It uses only the flow itself
  • No extra structure (like a connection/parallel transport) needed
  • The diffeomorphism Φₜ naturally identifies TₓM with T_{Φₜ(x)}M

Comparison with Other Lifts

  • Complete lift (this one): Uses the flow's Jacobian
  • Vertical lift: Only moves within the same fiber (no base point movement)
  • Horizontal lift: Requires a connection (like Levi-Civita) to define "parallel"

Why This Matters

  1. Physics: When you have a fluid flow, you need to know how velocity vectors themselves evolve
  2. Dynamics: Understanding how phase space flows
  3. Geometry: Natural way to extend diffeomorphisms from base to bundle
  4. Symmetries: If something is preserved by the flow on M, what's preserved on TM?

Simple Example

Let M = ℝ² (the plane), and V = x∂/∂x (points move radially outward).

  1. On M: Φₜ(x,y) = (eᵗx, y) (stretches horizontally)
  2. On TM: A vector v = (a,b) at (x,y) becomes:
    • New position: (eᵗx, y)
    • New vector: (eᵗa, b) (stretched by same factor!)

The complete lift Vᶜ = x∂/∂x + a∂/∂a (stretches both position AND vector components).

Key insight: The tangent functor T is like saying "apply the flow to both the point AND its infinitesimal neighborhood" - which is exactly what the differential does!

头像
Amorphous
童生
帖子: 27
注册时间: 周三 4月 08, 2026 5:32 am
昵称: 阿毛

Re: deepseek 给将军们讲微分几何

帖子 Amorphous »

Rigorous Treatment of Flow Lifting to Tangent and Cotangent Bundles

Preliminaries

Let $M$ be a smooth manifold of dimension $n$. We denote by:

  • $TM$ the tangent bundle of $M$, with projection $\pi: TM \to M$
  • $T^M$ the cotangent bundle of $M$, with projection $\pi^: T^*M \to M$
  • $\mathfrak{X}(M)$ the space of smooth vector fields on $M$
  • $\Omega1(M)$ the space of smooth 1-forms on $M$
  • $\text{Diff}(M)$ the group of diffeomorphisms of $M$

Section 1: Lifting to the Tangent Bundle: Pushforward Construction

1.1 The Tangent Functor

Definition 1.1 (Tangent Functor).
The tangent functor $T$ assigns to each smooth manifold $M$ its tangent bundle $TM$, and to each smooth map $f: M \to N$ the pushforward (or differential)
$$
Tf \equiv f: TM \to TN, \quad v \mapsto f(v)
$$
defined by $f(v)(g) = v(g \circ f)$ for $g \in C\infty(N)$. For $v \in T_pM$, we have $f(v) \in T_{f(p)}N$.

Lemma 1.2 (Properties of the Pushforward).
For smooth maps $f: M \to N$ and $g: N \to P$, we have:

  1. $(g \circ f)* = g* \circ f_*$ (functoriality)
  2. $(\text{id}M)* = \text{id}_{TM}$
  3. If $f$ is a diffeomorphism, then $f: TM \to TN$ is a diffeomorphism with $(f)^{-1} = (f^{-1})_*$

1.2 Canonical Lifting to $TM$

Definition 1.3 (Lifted Flow on $TM$).
Given a flow $\Phi_t: M \to M$ generated by $V \in \mathfrak{X}(M)$, the lifted flow on $TM$ is defined by applying the tangent functor:
\[
\tilde{\Phi}t := T\Phi_t = (\Phi_t): TM \to TM
\]
Explicitly, for $v \in T_pM$:
\[
\tilde{\Phi}t(v) = (\Phi_t)
(v) = d\Phi_t|p(v) \in T{\Phi_t(p)}M
\]

Theorem 1.4 (Lifted Flow Properties).
The family $\{\tilde{\Phi}t\}{t\in\mathbb{R}}$ satisfies:

  1. $\tilde{\Phi}0 = \text{id}{TM}$
  2. $\tilde{\Phi}{t+s} = \tilde{\Phi}t \circ \tilde{\Phi}_s$ for all $t,s \in \mathbb{R}$
  3. $\pi \circ \tilde{\Phi}_t = \Phi_t \circ \pi$ (bundle morphism property)
  4. $\tilde{\Phi}_t$ is a diffeomorphism of $TM$ for each $t \in \mathbb{R}$

Thus $\tilde{\Phi}_t$ is a genuine flow on the manifold $TM$.

Proof.
(1) $\tilde{\Phi}0 = T\Phi_0 = T(\text{id}M) = \text{id}_{TM}$.

(2) By functoriality:
\[
\tilde{\Phi}{t+s} = T(\Phi{t+s}) = T(\Phi_t \circ \Phi_s) = T\Phi_t \circ T\Phi_s = \tilde{\Phi}t \circ \tilde{\Phi}s
\]

(3) For $v \in T_pM$:
\[
\pi(\tilde{\Phi}t(v)) = \pi(d\Phi_t|p(v)) = \Phi_t(p) = \Phi_t(\pi(v))
\]

(4) Since $\Phi_t$ is a diffeomorphism, $T\Phi_t = (\Phi_t)$ is a diffeomorphism with inverse $T\Phi{-t} = (\Phi_t)_^{-1}$. ∎

Proposition 1.5 (Commutative Diagram).
The lifting construction yields the following commutative diagram:
\[
\begin{array}{ccc}
TM & \xrightarrow{\tilde{\Phi}t = (\Phi_t)*} & TM \\
\pi\downarrow & & \downarrow\pi \\
M & \xrightarrow{\quad\Phi_t\quad} & M
\end{array}
\]
This makes $\tilde{\Phi}_t$ a bundle automorphism covering $\Phi_t$.

1.3 The Generating Vector Field: Complete Lift

Definition 1.6 (Complete Lift).
For $V \in \mathfrak{X}(M)$ generating $\Phi_t$, the complete lift $VC \in \mathfrak{X}(TM)$ is defined by:
\[
VC_v := \left.\frac{d}{dt}\right|{t=0} \tilde{\Phi}t(v) = \left.\frac{d}{dt}\right|_{t=0} d\Phi_t(v)
\]
for $v \in TM$.

Theorem 1.7 (Coordinate Expression).
In local coordinates $(xi)$ on $M$ and induced coordinates $(xi, vi)$ on $TM$, if $V = Vi(x)\frac{\partial}{\partial xi}$, then:
\[
VC = Vi(x)\frac{\partial}{\partial xi} + \frac{\partial Vj}{\partial xi}(x)vi\frac{\partial}{\partial vj}
\]

Proof.
Let $\Phi_t(x) = (\Phi_t1(x), \ldots, \Phi_tn(x))$. Then:
\[
d\Phi_t\left(\frac{\partial}{\partial xi}\right) = \frac{\partial \Phi_tj}{\partial xi}\frac{\partial}{\partial xj}
\]
For a vector $v = vi\frac{\partial}{\partial xi}$, we have:
\[
\tilde{\Phi}t(v) = vi\frac{\partial \Phi_tj}{\partial xi}\frac{\partial}{\partial xj}
\]
Differentiating at $t=0$:
\[
VC_v = \left.\frac{d}{dt}\right|
{t=0} \left(vi\frac{\partial \Phi_tj}{\partial xi}\right)\frac{\partial}{\partial xj}
= vi\frac{\partial Vj}{\partial xi}\frac{\partial}{\partial xj}
\]
In coordinates $(xi, vi)$ on $TM$, this becomes the stated expression. ∎

Definition 1.8 (Horizontal and Vertical Components).
The complete lift decomposes as:
\[
VC = VH + VV
\]
where:

  • $VH = Vi(x)\frac{\partial}{\partial xi}$ is the horizontal lift (base motion)
  • $VV = \frac{\partial Vj}{\partial xi}vi\frac{\partial}{\partial vj}$ is the vertical component (fiber linearization)

Section 2: Lifting to the Cotangent Bundle: Pullback Construction

2.1 The Cotangent Functor

Definition 2.1 (Cotangent Functor).
The cotangent functor $T^$ assigns to each smooth manifold $M$ its cotangent bundle $T^M$, and to each diffeomorphism $f: M \to N$ the pullback
\[
T^f \equiv f^: T^N \to T^M, \quad \alpha \mapsto f^\alpha
\]
defined by $(f^
\alpha)p(v) = \alpha{f(p)}(f_*(v))$ for $v \in T_pM$.

Lemma 2.2 (Properties of the Pullback).
For diffeomorphisms $f: M \to N$ and $g: N \to P$, we have:

  1. $(g \circ f)^* = f^* \circ g^*$
  2. $(\text{id}M)^* = \text{id}{T^*M}$
  3. $(f^)^{-1} = (f^{-1})^$

2.2 Canonical Lifting to $T^*M$

Definition 2.3 (Lifted Flow on $T^M$).
Given a flow $\Phi_t: M \to M$ generated by $V \in \mathfrak{X}(M)$, the lifted flow on $T^
M$ is defined by the pullback:
\[
\hat{\Phi}t := (\Phi_t)^: T^M \to T^M
\]
Explicitly, for $\alpha \in T^
pM$:
\[
\hat{\Phi}t(\alpha) = (\Phi_t)^(\alpha) \in T^{\Phi_t^{-1}(p)}M
\]
with the action: $\hat{\Phi}t(\alpha)(v) = \alpha((\Phi_t)*(v))$ for $v \in T_{\Phi_t^{-1}(p)}M$.

Theorem 2.4 (Lifted Flow Properties on $T^*M$).
The family $\{\hat{\Phi}t\}{t\in\mathbb{R}}$ satisfies:

  1. $\hat{\Phi}0 = \text{id}{T^*M}$
  2. $\hat{\Phi}{t+s} = \hat{\Phi}t \circ \hat{\Phi}_s$ for all $t,s \in \mathbb{R}$
  3. $\pi^* \circ \hat{\Phi}_t = \Phi_t^{-1} \circ \pi^*$ (bundle morphism property)
  4. $\hat{\Phi}_t$ is a diffeomorphism of $T^*M$ for each $t \in \mathbb{R}$

Proof.
(1) $\hat{\Phi}0 = (\Phi_0)^* = (\text{id}M)^* = \text{id}_{T^*M}$.

(2) Using Lemma 2.2(1):
\[
\hat{\Phi}{t+s} = (\Phi{t+s})^* = (\Phi_t \circ \Phi_s)^* = \Phi_s^* \circ \Phi_t^* = \hat{\Phi}s \circ \hat{\Phi}t
\]
Note the order reversal due to contravariance.

(3) For $\alpha \in T^pM$, we have $\pi^(\hat{\Phi}t(\alpha)) = \pi^((\Phi_t)^\alpha) = \Phi_t^{-1}(p) = \Phi_t^{-1}(\pi^*(\alpha))$.

(4) Since $\Phi_t$ is a diffeomorphism, $(\Phi_t)^$ is a diffeomorphism with inverse $(\Phi_t^{-1})^ = (\Phi_{-t})^*$. ∎

Proposition 2.5 (Commutative Diagram for Cotangent Lift).
The lifting construction yields:
\[
\begin{array}{ccc}
T^M & \xrightarrow{\hat{\Phi}_t = (\Phi_t)^} & T^M \\
\pi^
\downarrow & & \downarrow\pi^* \\
M & \xrightarrow{\quad\Phi_t^{-1}\quad} & M
\end{array}
\]

2.3 The Generating Vector Field: Cotangent Lift

Definition 2.6 (Cotangent Lift).
For $V \in \mathfrak{X}(M)$ generating $\Phi_t$, the cotangent lift $V^* \in \mathfrak{X}(T^M)$ is defined by:
\[
V^
\alpha := \left.\frac{d}{dt}\right|{t=0} \hat{\Phi}t(\alpha) = \left.\frac{d}{dt}\right|{t=0} (\Phi_t)^\alpha
\]
for $\alpha \in T^
M$.

Theorem 2.7 (Coordinate Expression).
In local coordinates $(xi)$ on $M$ and induced coordinates $(xi, p_i)$ on $T^M$, if $V = Vi(x)\frac{\partial}{\partial xi}$, then:
\[
V^
= Vi(x)\frac{\partial}{\partial xi} - \frac{\partial Vj}{\partial xi}(x)p_j\frac{\partial}{\partial p_i}
\]

Proof.
Let $\alpha = p_i dxi$. Under the flow $\Phi_t$, coordinates transform as $xi \mapsto \Phi_ti(x)$. The pullback acts on forms as:
\[
(\Phi_t)^(p_i dxi) = p_i \frac{\partial \Phi_ti}{\partial xj} dxj
\]
Differentiating at $t=0$:
\[
\left.\frac{d}{dt}\right|_{t=0} (\Phi_t)^
\alpha = p_i \frac{\partial Vi}{\partial xj} dxj
\]
In Hamiltonian vector field form, this yields the stated expression. ∎

Definition 2.8 (Hamiltonian Vector Field).
For a smooth function $H: T^M \to \mathbb{R}$, define the Hamiltonian vector field $X_H$ on $T^M$ by:
\[
\iota_{X_H}\omega = dH
\]
where $\omega = dp_i \wedge dxi$ is the canonical symplectic form. Then $V^*$ can be identified with $X_H$ for $H(x,p) = p_i Vi(x)$.


Section 3: Functorial Properties and Naturality

Theorem 3.1 (Functoriality).
The lifting constructions define functors:

  1. Tangent functor: $T: \text{Diff}(M) \to \text{Diff}(TM)$, $\Phi \mapsto \Phi_*$
  2. Cotangent functor: $T^: \text{Diff}(M) \to \text{Diff}(T^M)$, $\Phi \mapsto \Phi^*$

satisfying:
\[
T(\Phi \circ \Psi) = T\Phi \circ T\Psi, \quad T^(\Phi \circ \Psi) = T^\Psi \circ T^*\Phi
\]

Theorem 3.2 (Naturality with Respect to Vector Fields).
Let $V, W \in \mathfrak{X}(M)$ with flows $\Phi_tV, \Phi_tW$. Then:

  1. The complete lifts satisfy: $[VC, WC] = [V,W]C$
  2. The cotangent lifts satisfy: $[V^, W^] = [V,W]^*$

Proposition 3.3 (Relation to Lie Derivative).
For $\alpha \in \Omega1(M)$ and $V \in \mathfrak{X}(M)$:
\[
\left.\frac{d}{dt}\right|{t=0} (\Phi_t)^*\alpha = \mathcal{L}V\alpha
\]
where $\mathcal{L}_V$ is the Lie derivative.

Similarly, for $X \in \mathfrak{X}(M)$:
\[
\left.\frac{d}{dt}\right|{t=0} (\Phi_t)*X = \mathcal{L}_VX
\]


Section 4: Summary and Comparison

Theorem 4.1 (Summary of Lifting Constructions).

PropertyTangent Bundle ($TM$)Cotangent Bundle ($T^M$)
FunctorTangent functor $T$Cotangent functor $T^*$
ActionPushforward $(\Phi_t)$Pullback $(\Phi_t)^*$
Flow$\tilde{\Phi}_t = T\Phi_t$$\hat{\Phi}t = T^\Phi_t$
GeneratorComplete lift $VC$Cotangent lift $V^*$
Coordinates$VC = Vi\partial_{xi} + (\partial_i Vj)vi\partial{vj}$$V^* = Vi\partial_{xi} - (\partial_i Vj)p_j\partial{p_i}$
Diagram$\pi \circ \tilde{\Phi}_t = \Phi_t \circ \pi$$\pi^* \circ \hat{\Phi}_t = \Phi_t^{-1} \circ \pi^*$
CovarianceCovariant functorContravariant functor

Corollary 4.2 (Naturality of Constructions).
Both lifting constructions are canonical (require no additional structure) and functorial, making them the natural ways to extend flows from the base manifold to its tangent and cotangent bundles.

Remark 4.3 (Physical Interpretation).

  • Tangent bundle lift: Describes how velocity vectors are transported by the flow
  • Cotangent bundle lift: Describes how momentum/force covectors are transported by the flow
  • The sign difference in the vertical components reflects the dual nature of vectors and covectors under coordinate transformations

Section 5: Applications

Definition 5.1 (Natural Lift of Tensor Fields).
For any tensor field $T$ of type $(r,s)$ on $M$, define its natural lift under $\Phi_t$ by:
\[
(\Phi_t)T := (\Phi_t)^{\otimes r} \otimes (\Phi_t)^{*\otimes s} T
\]

Theorem 5.2 (Infinitesimal Generator).
The infinitesimal generator of the lifted flow on tensor fields is the Lie derivative:
\[
\left.\frac{d}{dt}\right|{t=0} (\Phi_t)*T = \mathcal{L}_VT
\]

Application 5.3 (Symplectic Geometry).
On $T^M$ with canonical symplectic form $\omega = d\theta$ (where $\theta = p_i dxi$ is the tautological 1-form), we have:
\[
\mathcal{L}{V^
}\omega = 0 \quad \text{and} \quad \mathcal{L}{V^}\theta = 0
\]
Thus $V^
$ is a Hamiltonian vector field preserving the symplectic structure.

Application 5.4 (Geometric Mechanics).
If $\Phi_t$ is a symmetry of a Lagrangian $

头像
xexz
童生
帖子: 29
注册时间: 周二 4月 07, 2026 4:05 pm

Re: deepseek 给将军们讲微分几何

帖子 xexz »

deepseeek这两天强制更新客户端了,似乎有大动作。

randomforest
童生
帖子: 22
注册时间: 周二 4月 07, 2026 11:29 pm

Re: deepseek 给将军们讲微分几何

帖子 randomforest »

怎么latex 不显示了?

Similarly, for $$X \in \mathfrak{X}(M)$$:
$$
\left.\frac{d}{dt}\right|{t=0} (\Phi_t)*X = \mathcal{L}_VX
$$

嗯, 试了一下, 现在只支持把数学公式放在\$$ 和 \$$ 之间, 单个$ 和 \[都不支持

头像
zeami
秀才
帖子: 233
注册时间: 周一 4月 06, 2026 10:57 pm
昵称: 贼阿米

Re: deepseek 给将军们讲微分几何

帖子 zeami »

@Amorphous 阿毛帮我问一哈:block function in forum

巫咸上天 识者其谁

randomforest
童生
帖子: 22
注册时间: 周二 4月 07, 2026 11:29 pm

Re: deepseek 给将军们讲微分几何

帖子 randomforest »

我也让deepseek写了个基础的。

协变矢量、逆变矢量与度规张量在基变换中的角色

在微分几何与张量分析中,协变矢量(余矢)、逆变矢量以及度规张量的概念与基矢量的变换密切相关。下面从基变换出发,逐层解释它们各自扮演的角色。

1. 基变换与对偶基

设 $$V$$ 是一个 $$n$$ 维矢量空间,选取两组基:

  • 旧基:$$\{\mathbf{e}1,\dots,\mathbf{e}n\}$$
  • 新基:$$\{\mathbf{e}{1'},\dots,\mathbf{e}{n'}\}$$

基变换由矩阵 $$A$$ 及其逆 $$A^{-1}$$ 描述:

$$
\mathbf{e}{i'} = \sum_j Aj{}{i'} \,\mathbf{e}j, \qquad
\mathbf{e}
i = \sum{j'} (A^{-1})^{j'}{}i \,\mathbf{e}_{j'}.
$$

$$Aj{}_{i'}$$ 是新基矢在旧基下的分量,下指标为行索引,上指标为列索引。

对偶空间 $$V^*$$ 中存在与 $$\{\mathbf{e}i\}$$ 对偶的基 $$\{\boldsymbol{\theta}i\}$$,满足 $$\boldsymbol{\theta}i(\mathbf{e}j)=\deltai_j$$。当矢量基按 $$A$$ 变换时,对偶基的变换规则恰好相反(逆步变换):

$$
\boldsymbol{\theta}^{i'} = \sum_j (A^{-1})^{i'}{}j \,\boldsymbol{\theta}j, \qquad
\boldsymbol{\theta}i = \sum
{j'} Ai{}_{j'} \,\boldsymbol{\theta}^{j'}.
$$

这一差异直接导致了矢量分量出现两种相反的变换行为。

2. 逆变矢量(Contravariant Vector)

逆变矢量 $$\mathbf{v}$$ 本身是与基选取无关的几何对象,但其分量依赖于基。展开式唯一:

$$
\mathbf{v} = \sum_i vi \mathbf{e}i = \sum{i'} v^{i'} \mathbf{e}_{i'}.
$$

代入基变换关系,得到分量的变换规律:

$$
v^{i'} = \sum_j (A^{-1})^{i'}{}_j \,vj.
$$

  • 分量变换矩阵是基变换矩阵的,因此称“逆”变(contra-variant)。
  • 物理上,逆变分量与基矢量“反向补偿”,使几何对象不变。

3. 协变矢量 / 对偶一形式(Covariant Vector / Dual One‑Form)

协变矢量(余矢)$$\boldsymbol{\omega} \in V^*$$ 是用对偶基展开的一形式:

$$
\boldsymbol{\omega} = \sum_i \omega_i \boldsymbol{\theta}i = \sum{i'} \omega{i'} \boldsymbol{\theta}^{i'}.
$$

利用对偶基的变换 $$\boldsymbol{\theta}^{i'} = \sum_j (A^{-1})^{i'}{}_j \boldsymbol{\theta}j$$,可得分量变换:

$$
\omega{i'} = \sum_j Aj{}{i'} \,\omega_j.
$$

  • 分量变换矩阵与基矢量变换矩阵相同,因此称“协”变(co-variant)。
  • 协变分量直接跟随基矢量的方向一起变,从而保证余矢作用在矢量上的值不变:$$\boldsymbol{\omega}(\mathbf{v}) = \sum_i \omega_i vi$$ 是标量。

4. 度规张量(Metric Tensor)在基变换中的作用

度规张量 $$g$$ 是一个 $$(0,2)$$ 型张量,赋予矢量空间内积结构:

$$
g(\mathbf{u},\mathbf{v}) = \sum{i,j} g{ij} ui vj.
$$

在基变换下,其分量按双重协变方式变换:

$$
g{i'j'} = \sum{k,l} Ak{}{i'} Al{}{j'} \, g_{kl}.
$$

度规张量的核心作用是建立逆变与协变之间的同构映射

4.1 指标升降

利用度规可将逆变分量转为协变分量(降指标),反之亦然(升指标):

$$
v_i = \sum_j g_{ij} vj, \qquad
vi = \sum_j g^{ij} v_j,
$$

其中 $$g^{ij}$$ 是度规的逆矩阵,满足 $$\sum_k g^{ik} g_{kj} = \deltai_j$$。

关键:基变换下,由逆变分量通过度规降指标得到的协变分量,其变换规则自动与协变矢量一致。这正是因为度规分量的双重协变变换抵消了逆变分量的逆变换。

4.2 定义对偶基

度规还给出一种自然方式将矢量基与对偶一形式基对应起来:

$$
\mathbf{e}i := \sum_j g^{ij} \mathbf{e}j \quad (\text{更常见是将 } \boldsymbol{\theta}i \text{ 与 } g^{ij}\mathbf{e}j \text{ 等同}),
$$

但实际上严格的对偶基 $$\boldsymbol{\theta}i$$ 由双线性配对定义,与度规无关。一旦存在度规,我们就可以用度规把逆变基矢量变成协变余矢,并将逆变与协变分量联系起来。

5. 总结:各自在基变换中的角色

对象基变换时的行为几何意义
基矢量 $$\mathbf{e}i$$按矩阵 $$A$$ 变换标架本身
对偶基 $$\boldsymbol{\theta}i$$按矩阵 $$A^{-1}$$ 变换与基矢量配对,给出分量
逆变矢量分量 $$vi$$按 $$A^{-1}$$ 变换(与基矢反向)保证矢量对象 $$\mathbf{v} = vi \mathbf{e}_i$$ 不变
协变矢量分量 $$\omega_i$$按 $$A$$ 变换(与基矢同向)保证余矢对象 $$\boldsymbol{\omega} = \omega_i \boldsymbol{\theta}i$$ 不变
度规张量 $$g_{ij}$$双重协变变换(按 $$A$$ 两次)定义内积、升降指标、连接逆变与协变世界

度规张量的出现使得我们可以自由地在两种表示间切换,并在坐标变换下保持内积、线元等物理量不变。理解这些变换规则,是掌握广义相对论、微分几何以及现代物理中张量语言的基础。

回复