-
For an open neighborhood U ⊂ M U\subset M U⊂M, show that there exists a local orthonormal basis { ω i } ⊂ Γ ( T ∗ U ) \{\omega_i\}\subset \Gamma(T^*U) {ωi}⊂Γ(T∗U), such that g ∣ U = ∑ ( ω i ) 2 g|_U=\sum(\omega_i)^2 g∣U=∑(ωi)2.
Proof:
We can use Gram-Schmidt orthogonalization to do this.
Suppose { c i ∣ i = 1 , … , n } \{c_i|i=1,\dots,n\} {ci∣i=1,…,n} is a basis which is not necessarily orthogonal in U U U, then set ω 1 = c 1 ∣ c 1 ∣ \omega_1=\frac{c_1}{|c_1|} ω1=∣c1∣c1, if ω 1 , … , ω j − 1 \omega_1,\dots,\omega_{j-1} ω1,…,ωj−1 is defined, then:
ω ~ j = c j − ∑ k = 1 j − 1 g ( c j , ω k ) ω k , \tilde{\omega}_j=c_j-\sum_{k=1}^{j-1}g(c_j,\omega_k)\omega_k, ω~j=cj−k=1∑j−1g(cj,ωk)ωk,
and set ω j = ω ~ j ∣ ω ~ j ∣ \omega_j=\frac{\tilde{\omega}_j}{|\tilde{\omega}_j|} ωj=∣ω~j∣ω~j, process in this fashion, we have a local orthonormal basis { ω i } \{\omega_i\} {ωi}. -
For a ( 0 , 2 ) (0,2) (0,2)-tensor h ∈ Γ ( T ( 0 , 2 ) M ) h\in \Gamma(T^{(0,2)}M) h∈Γ(T(0,2)M), identify it with a ( 1 , 1 ) (1,1) (1,1)-tensor and a ( 2 , 0 ) (2,0) (2,0)-tensor by raising the indices. In particular, what are the corresponding tensors for the Riemannian metric g g g?
Proof:
For any ( s , r ) (s,r) (s,r)-tensor, ∀ ω = a i δ i ∈ T p ∗ M \forall\omega=a_i\delta^i\in T_p^*M ∀ω=aiδi∈Tp∗M, we set a i = g i j a i a^i=g^{ij}a_i ai=gijai, thus ω ∗ = a i e i ∈ T p M \omega^*=a^ie_i\in T_pM ω∗=aiei∈TpM is the dual of ω \omega ω. We choose { ω } \{\omega\} {ω} as a basis of T p ∗ M T^*_pM Tp∗M, and then raise it to T p M T_pM TpM, thus we have a ( 1 , 1 ) (1,1) (1,1)-tensor, do this again, we have a ( 2 , 0 ) (2,0) (2,0)-tensor.
We can write it in detail, T 0 i 1 i 2 , i 1 , i 2 ∈ { 1 , 2 } T^{i_1i_2}_0,i_1,i_2\in\{1,2\} T0i1i2,i1,i2∈{1,2}, and i 1 ≠ i 2 i_1\neq i_2 i1=i2 is the ( 0 , 2 ) (0,2) (0,2)-tensor, then
T ~ k i 1 = g i 2 k T 0 i 1 i 2 , \tilde{T}_k^{i_1}=g_{i_2k}T^{i_1i_2}_0, T~ki1=gi2kT0i1i2,
is a ( 1 , 1 ) (1,1) (1,1)-tensor, do this again, we have a ( 2 , 0 ) (2,0) (2,0)-tensor, T ~ i 2 l 0 = g i 1 l T i 2 i 1 . \tilde{T}_{i_2l}^0=g_{i_1l}T^{i_1}_{i_2}. T~i2l0=gi1lTi2i1. -
For α i , β j ∈ Λ 1 ( M ) , 1 ≤ i , j ≤ p \alpha_i,\beta_j\in\Lambda^1(M),1\leq i,j\leq p αi,βj∈Λ1(M),1≤i,j≤p, show that:
g ( α 1 ∧ ⋯ ∧ α p , β 1 ∧ ⋯ ∧ β p ) = det ( g ( α i , β j ) ) . g(\alpha_1\wedge\dots\wedge\alpha_p,\beta_1\wedge\dots\wedge\beta_p)=\text{det }(g(\alpha_i,\beta_j)). g(α1∧⋯∧αp,β1∧⋯∧βp)=det (g(αi,βj)).
Proof:
Write it in the definition of wedge product and use
g ( α 1 ⊗ β 1 , α 2 ⊗ β 2 ) = g ( α 1 , α 2 ) ⋅ g ( β 1 , β 2 ) g(\alpha_1\otimes\beta_1,\alpha_2\otimes\beta_2)=g(\alpha_1,\alpha_2)\cdot g(\beta_1,\beta_2) g(α1⊗β1,α2⊗β2)=g(α1,α2)⋅g(β1,β2) and by the expansion of determinant we know the equality holds.
α 1 ∧ ⋯ ∧ α p = p ! A ( α 1 ⊗ ⋯ ⊗ α p ) = ∑ σ ∈ P p sgn σ ⋅ ( α σ ( 1 ) ⊗ ⋯ ⊗ α σ ( p ) ) . \begin{aligned} \alpha_1\wedge\dots\wedge\alpha_p=&p!\mathscr{A}(\alpha_1\otimes\dots\otimes\alpha_p)\\ =&\sum_{\sigma\in P_p}\text{sgn}\sigma\cdot(\alpha_{\sigma(1)}\otimes\dots\otimes\alpha_{\sigma(p)}). \end{aligned} α1∧⋯∧αp==p!A(α1⊗⋯⊗αp)σ∈Pp∑sgnσ⋅(ασ(1)⊗⋯⊗ασ(p)).
A similar equality holds for β 1 ∧ ⋯ ∧ β p \beta_1\wedge\dots\wedge\beta_p β1∧⋯∧βp, thus we can move the coefficient sgn σ \text{sgn}\sigma sgnσ out from the Riemann product, thus:
g ( α 1 ∧ ⋯ ∧ α p , β 1 ∧ ⋯ ∧ β p ) = ∑ σ ∈ P p sgn σ g ( α σ ( 1 ) , β σ ( 1 ) ) ⋅ ⋯ ⋅ g ( α σ ( p ) , β σ ( p ) ) , g(\alpha_1\wedge\dots\wedge\alpha_p,\beta_1\wedge\dots\wedge\beta_p)=\sum_{\sigma\in P_p}\text{sgn}\sigma g(\alpha_{\sigma(1)},\beta_{\sigma(1)})\cdot\dots\cdot g(\alpha_{\sigma(p)},\beta_{\sigma(p)}), g(α1∧⋯∧αp,β1∧⋯∧βp)=σ∈Pp∑sgnσg(ασ(1),βσ(1))⋅⋯⋅g(ασ(p),βσ(p)),
By definition, the right hand side is equal to det ( g ( α i , β j ) ) \text{det }(g(\alpha_i,\beta_j)) det (g(αi,βj)). -
Try to find the local expression of the Hodge star operator on a p p p-form. Then prove that:
∗ ∘ ∗ = ( − 1 ) p ( n − p ) id p . *\circ*=(-1)^{p(n-p)}\text{id}_p. ∗∘∗=(−1)p(n−p)idp.
Proof:
The Hodge star operator is defined:
∗ : Λ p V → Λ n − p V , *:\Lambda^pV\rightarrow \Lambda^{n-p}V, ∗:ΛpV→Λn−pV,
where V V V is an n n n-dimensional vector space.
By linearity, we only need to consider the operator acts on a oriented orthonormal basis of Λ p V \Lambda^pV ΛpV: e 1 ∧ ⋯ ∧ e p e_1\wedge\dots\wedge e_p e1∧⋯∧ep, we know acts on Hodge star operator we will have:
∗ ( e 1 ∧ ⋯ ∧ e p ) = e p + 1 ∧ ⋯ ∧ e n , *(e_1\wedge\dots\wedge e_p)=e_{p+1}\wedge\dots\wedge e_{n}, ∗(e1∧⋯∧ep)=ep+1∧⋯∧en,
do this again ,we know ∗ ( e p + 1 ∧ ⋯ ∧ e n ) = ( − 1 ) k e 1 ∧ ⋯ ∧ e p *(e_{p+1}\wedge\dots\wedge e_{n})=(-1)^ke_1\wedge\dots\wedge e_p ∗(ep+1∧⋯∧en)=(−1)ke1∧⋯∧ep, where k k k is an integer. Since
ω ∧ ( ∗ η ) = < ω , η > d v , \omega\wedge(*\eta)=<\omega,\eta>dv, ω∧(∗η)=<ω,η>dv,
take ω = e p + 1 ∧ ⋯ ∧ e n \omega=e_{p+1}\wedge\dots\wedge e_{n} ω=ep+1∧⋯∧en, thus < ω , η > = < e p + 1 ∧ ⋯ ∧ e n , e p + 1 ∧ ⋯ ∧ e n > = det g ( e i , e j ) = 1 <\omega,\eta>=<e_{p+1}\wedge\dots\wedge e_{n},e_{p+1}\wedge\dots\wedge e_{n}>=\text{det }g(e_i,e_j)=1 <ω,η>=<ep+1∧⋯∧en,ep+1∧⋯∧en>=det g(ei,ej)=1, thus ( − 1 ) k e p + 1 ∧ ⋯ ∧ e n ∧ e 1 ∧ ⋯ ∧ e p = d v = e 1 ∧ ⋯ ∧ e n , (-1)^ke_{p+1}\wedge\dots\wedge e_{n}\wedge e_1\wedge\dots\wedge e_p=dv=e_1\wedge\dots\wedge e_n, (−1)kep+1∧⋯∧en∧e1∧⋯∧ep=dv=e1∧⋯∧en,
adjust the position by the skew symmetric rule, we have ( − 1 ) k ⋅ ( − 1 ) p ( n − 1 ) = 1 (-1)^k\cdot (-1)^{p(n-1)}=1 (−1)k⋅(−1)p(n−1)=1, take k = p ( n − p ) k=p(n-p) k=p(n−p), thus ∗ ∘ ∗ = ( − 1 ) p ( n − p ) id p *\circ*=(-1)^{p(n-p)}\text{id}_p ∗∘∗=(−1)p(n−p)idp.
Riemann度量习题
最新推荐文章于 2023-05-05 19:03:24 发布