Gradient, divergence, and Laplace-Beltrami operator

[6/100%]
Show
LaTeX

Let XX be a smooth vector field on a Riemannian manifold MM. The divergence of XX, denoted by div⁡(X)\operatorname {div}(X), is defined by the function trace⁡(∇X)\operatorname {trace}(\nabla X).

?
(a)

If MM is closed (i.e., compact without boundary), show that

∫Mdiv⁡(X) dv=0. \int _M \operatorname {div}(X) \, dv = 0.
(b)

If MM is compact with boundary ∂M\partial M, show that

∫Mdiv⁡(X) dv=∫∂M⟨X,N⟩ ds, \int _M \operatorname {div}(X) \, dv = \int _{\partial M} \left\langle X, N \right\rangle \, ds,

where NN is the outer normal vector field of ∂M\partial M.

?
(a)

Let ff be a smooth function on a Riemannian manifold MM. Let grad⁡f\operatorname {grad} f be the vector field defined by the equation

⟨grad⁡f,v⟩p=dpf v,v∈TpM. \left\langle \operatorname {grad} f, v \right\rangle _p = d_pf \, v, \qquad v \in T_pM.

Let (x1,…,xn)(x^1,\ldots ,x^n) be local coordinates around pp. Find the expression for grad⁡f\operatorname {grad} f in terms of x1,…,xnx^1,\ldots ,x^n.

(b)

For a vector field XX define the divergence of XX, div⁡(X)\operatorname {div}(X) as the trace of the operator Y→DYXY \to D_Y X where DD is the Levi-Civita connection. Find the expression for the divergence of XX in a local coordinate system (x1,…,xn)(x^1,\ldots ,x^n).

(c)

Use (a) and (b) to find the expression for the Laplacian Δ\Delta in local coordinates, where Δ\Delta acting on a smooth function ff is defined by Δf=div⁡(∇f)\Delta f = \operatorname {div}(\nabla f).

Let MM be a compact connected Riemannian manifold without boundary. Let ff be a smooth function satisfying Δf=0\Delta f = 0. Show that f=constf = \text{const}.

?

Let MM be a compact Riemannian manifold without boundary. For any f∈C∞(M)f \in C^\infty (M), define ∇f∈X(M)\nabla f \in \mathcal{X}(M) and Δf∈C∞(M)\Delta f \in C^\infty (M) as follows:

At any p∈Mp \in M, choose an orthonormal frame field {e1,…,en}\{ e_1,\ldots ,e_n\} around pp and then define

(∇f)(p)=∑i(eif)(ei) (\nabla f)(p) = \sum _i (e_if)(e_i)

and

(Δf)(p)=−∑i[ei(eif)−(∇eiei)f]. (\Delta f)(p) = -\sum _i [e_i(e_if) - (\nabla _{e_i}e_i)f].

Verify first that ∇f\nabla f and Δf\Delta f are well defined (i.e., they do not depend on the choice of orthonormal frame) and then show that

∫fΔf=∫∥∇f∥2. \int f\Delta f = \int \left\| \nabla f\right\| ^2.
?

Let MnM^n be a Riemannian manifold. Show that whenever f:M→Rf : M \to \mathbb {R} is a smooth function, there is a unique vector field ∇f\nabla f (called the gradient of ff on MM) such that

⟨∇f(p),γ˙(t)⟩=ddtf(γ(t)) \left\langle \nabla f(p), \dot\gamma (t) \right\rangle = \frac{d}{dt} f(\gamma (t))

whenever γ\gamma is a smooth curve in MM with γ(t)=p\gamma (t) = p.

?
?
(a)

Consider Rn\mathbb {R}^{n} with the standard Euclidean metric and let p∈Rnp \in \mathbb {R}^{n} be an arbitrary point. For any x∈Rnx \in \mathbb {R}^{n} let ρp(x)\rho_p(x) be the distance from pp to xx. Viewing ρp(x)\rho_p(x) as a smooth function of xx away from pp, verify that ∣grad⁡(ρp(x))∣2=1\left|\operatorname {grad}(\rho_p(x))\right|^2 = 1 and that the integral curves of grad⁡(ρp(x))\operatorname {grad}(\rho_p(x)) are straight lines. (Here grad⁡(ρp(x))\operatorname {grad}(\rho_p(x)) refers to the usual gradient vector field of the function ρp(x)\rho_p(x).)

(b)

More generally, given a smooth function ff on a Riemannian manifold (M,gij)(M,g_{ij}), define grad⁡(f)\operatorname {grad}(f) to be the vector field given locally by

∑i,j(gij∂f∂xj)∂∂xi \sum _{i,j} \left(g^{ij}\frac{\partial f}{\partial x^j}\right)\frac{\partial }{\partial x^i}

show that if ∣grad⁡(f)∣2=1\left|\operatorname {grad}(f)\right|^2=1, then the integral curves of the vector field grad⁡(f)\operatorname {grad}(f) are geodesics.

References

[CJP+11] Chen J., Jiang G., Pan Y., Qin T., Tong Y., Wu Q., Xu S., Li T. Problems and Solutions in Mathematics. 2nd ed., World Scientific, 2011.