3

Harmonic vector fields are critical points of Dirichlet energy function on the set of all unit vector fields on $M$, which is defined as follows:

$$E(X):=\frac{1}{2}\int_M\|dX\|^2\mathrm{dVol_g}\qquad X: (M,g)\to (TM,g_s), \|X\|=1.$$

Where it's Euler-Lagrange equation satisfies the following: $$\bar{\Delta}X=\lambda X,\quad \lambda\in C^\infty(M)$$ where $\bar{\Delta}$ denote rough Laplacian.

$p$-Harmonic vector field: If we use $p$-energy function instead of Dirichlet energy function: $$E_p(X):=\frac{1}{p}\int_M\|\mathrm{d}X\|^p\mathrm{dVol_g},\qquad X: (M,g)\to (TM,g_s),$$ The Euler-Lagrange equation of $E_p(.)$ in all vector fields (not in unit case) is as follows (for function case see this paper): $$\|dX\|^{p-2}\left(\tau(X) + dX\left(\mathrm{grad}(\log \|dX\|^{p-2})\right)\right) =0.$$ where $\tau(X)=\mathrm{div}(dX)$ called tension field of $X$.

My questions are :

Question 1: What is the Euler-Lagrange of $E_p(.)$ on the set of all unit vector fields on $M$?

Question 2: Can one hope to get statements which is generalized and new (not same to exactly) 2-harmonic (harmonic vector field) case?

Thanks.

C.F.G
  • 4,165
  • 5
  • 30
  • 64
  • Isn't Question 1 just Lagrange multipliers? Also, what do you mean by a tension field of $X$? Is it not just $\overline\Delta X$? How do you understand the term involving $dX$? Is $dX$ supposed to be the antisymmetrized derivative for just the covariant derivative of $X$ w.r.t. $g$? – Willie Wong Jul 07 '17 at 13:14
  • Tension field is just an equivalent equation to Euler-Lagrange equation and here I don't know what is it exactly. $dX: TM\to TTM$ is derivation of vector field $X$. – C.F.G Jul 07 '17 at 17:35
  • 2
    Just to check my understanding, is it the case that $| dX|^2 = n + |\nabla X|^2$? Imposing that you are keeping variations within the sphere bundle, and denoting by the infinitesimal variation $V$, you have that $\langle X, V\rangle = 0$, which means that $| d (X + V)|^2 = n + |\nabla (X+V)|^2$ which to linear order gives $n + |\nabla X|^2 + 2 \langle \nabla X, \nabla V\rangle$. This would imply that for the $p$ case you have the ELE being – Willie Wong Jul 08 '17 at 02:36
  • 2
    $$ \mathrm{div}_g \cdot \left[ (n + |\nabla X|^2)^{p/2 - 1} \nabla X \right] = \lambda X $$ since the integral of the inner product of the left hand side against any vector field orthogonal to $X$ must be 0. The constant $\lambda$ can also be computed using that $X$ is of unit norm as $$ \lambda = - (n + |\nabla X|^2)^{p/2 - 1} |\nabla X|^2 $$ (These are all predicate in my assumption that $|dX|^2 = n + |\nabla X|^2$ is correct.) – Willie Wong Jul 08 '17 at 02:43
  • Yes, I think your assumption $|dX|^2 = n + |\nabla X|^2$ is correct. – C.F.G Jul 08 '17 at 08:16

0 Answers0