1: \begin{abstract} We establish { small energy} H\"{o}lder bounds for  minimizers $u_\e$ of 

2: \[E_\e (u):=\int_\Om  W(\nabla u)+ \frac{1}{\e^2} \int_\Om f(u),\]

3:  where $W$ is a positive definite quadratic form { and the potential $f$ constrains $u$ to be close to a given manifold $\mathcal N$}. This implies that, up to subsequence, $u_\e$ converges locally uniformly to { an $\mathcal N$-valued} $W$-harmonic map, away from its singular set.

4: We treat general energies, covering in particular the 3D Landau-de Gennes model  for liquid crystals, with three distinct elastic constants.  Similar results are known in the isotropic case $W(\nabla u)=\vert \nabla u\vert^2$ and rely on three ingredients: a monotonicity formula for the scale-invariant energy on small balls, a uniform pointwise bound, and a Bochner equation for the energy density. In the level of generality we consider, all of these ingredients are absent. 

5: { In particular, the lack of monotonicity formula is an important reason why optimal estimates on the singular set of $W$-harmonic maps constitute an open problem.}

6: Our novel argument relies on showing appropriate decay for the energy on small balls, separately at scales smaller and larger than $\e$: the former is obtained from the regularity of solutions to elliptic systems while the latter is inherited from the regularity of $W$-harmonic maps.  This also allows us to handle physically relevant boundary conditions for which, even in the isotropic case, uniform convergence up to the boundary was open.

7: \end{abstract}

8: