Weighted Łojasiewicz inequalities and regularity of harmonic map flow
Alex Waldron · 2025
Read Weighted Łojasiewicz inequalities and regularity of harmonic map flow by Alex Waldron (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-07054.txt.
Introduction
Alex Waldron proves a key estimate in 2025 for a flow that smooths maps between shapes, at the moment it threatens to tear. When the flow runs out of time in the tightest setting, maps can bubble: energy flees into tiny spheres. Two things then matter: the body map stays unbroken, and no neck of stray energy links the bubbles.
Waldron shows both follow from one weighted inequality linking the terms of a known energy ledger. He then proves that inequality when the target is the round sphere, settling both points there, the abstract reports. Such results are the load-bearing walls of the theory: they certify that a breakdown is tidy, a neat tree of bubbles, not a mess.
Catalog record
| CREATOR | Alex Waldron |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 25 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-07054
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Weighted Łojasiewicz inequalities and regularity of harmonic map flow Alex Waldron Abstract . At a finite-time singularity of harmonic map flow in the critical dimension, we show that a Łojasiewicz inequality between the quantities appearing in Struwe’s monotonicity formula implies continuity of the body map and the no-neck property for bubble-tree decompositions. We prove such an inequality when the target is S 2 , S^{2}, yielding both properties in this case. 1. Introduction Łojasiewicz(-Simon) inequalities play a key role in the theory of geometric evolution equations; they rule out the notorious non-uniqueness phenomenon for subsequential limits at infinite time [ 6 ] . In recent years, most prominently in the work of Colding and Minicozzi, Łojasiewicz inequalities suitable for finite-time singularity analysis have become available. These have led to uniqueness theorems for the most basic multiplicity-one blowup limits of mean curvature flow, namely, round cylinders [ 3 ] and asymptotically conical shrinkers [ 1 ] , as well as mean-convex cylindrical shrinkers [ 14 , 15 ] . Łojasiewicz inequalities are also strongly tied to questions of regularity. Simon used his classical inequality in the elliptic setting to obtain a strong Hausdorff-dimension estimate and rectifiability of the singular set for area-minimizing currents [ 7 ] . Colding and Minicozzi have applied their results to obtain improved regularity of singular sets for mean curvature flow with generic (i.e. cylindrical) singularities at finite time [ 4 ] . They have also demonstrated that differentiability of the “arrival time” in mean-convex mean curvature flow depends on a Łojasiewicz inequality [ 2 ] . This paper establishes a link between a different regularity question and a Łojasiewicz inequality. The question concerns the so-called “body map” u(T)=\lim_{t\nearrow T}u(t) of a solution of harmonic map flow u:\Sigma\times\left[0,T\right)\to N in the critical dimension, \dim\left(\Sigma\right)=2. This map is by definition of class H^{1}, but may fail to be continuous in certain cases, as shown by Topping [ 11 ] . Meanwhile, Topping conjectured that if N is real-analytic then the body map must be continuous [ 11 , p. 286] . Our main result, Theorem 4.2 below, is that continuity of the body map follows from the validity of a Łojasiewicz inequality between the Gaussian-weighted quantities appearing in Struwe’s monotonicity formula [ 9 ] . We prove such an inequality, Theorem 2.2 , in the case that the target manifold is the round 2-sphere, allowing us to settle the fundamental case of the above question. 1 1 1 The continuity question for N=S^{2} was raised again recently by Jendrej, Lawrie, and Schlag [ 5 ] . Our main estimate also gives the “no-neck” property of the bubble-tree decomposition along any sequence approaching the singular time. The weighted Łojasiewicz inequality for maps to the 2-sphere is obtained by localizing one of the inequalities proved recently by the author [ 12 ] . The localization is enabled by a weighted Poincaré inequality, Lemma 2.1 , similar to those used in Type-I singularity analysis; this inequality turns out to be particularly strong in the critical dimension. We use the inequality in two places: once in the localization, Theorem 2.2 , and later in the proof of main theorem, cf. ( 4.15 ). An interesting feature of the proof is that we do not obtain Hölder continuity of the body map, as in the case of a “strict” type-II blowup [ 13 ] , unless the Łojasiewicz inequality holds with the optimal exponent ( \alpha=2 ). In the case of a fractional exponent ( 1<\alpha<2 ), we obtain the oscillation bound \Osc_{B_{r}}u(T)=O\left|\log r\right|^{\frac{\alpha-1}{\alpha-2}}. See Theorem 4.2 -Corollary 4.3 for detailed statements. 2. Weighted Łojasiewicz inequality 2.1. Weighted Poincaré inequality Let (\Sigma,g) be a Riemannian surface and (N,\left\langle\cdot,\cdot\right\rangle) any compact Riemannian manifold. Given a map u:\Sigma\to N, denote its stress-energy tensor by (2.1) S_{ij}=\left\langle{\partial}_{i}u,{\partial}_{j}u\right\rangle-\frac{1}{2}g_{ij}|du|_{g}^{2}. We have (2.2) \nabla^{i}S_{ij}=\left\langle{\mathcal{T}}(u),{\partial}_{j}u\right\rangle, see e.g. [ 8 , (2.5)] . Let X^{i} be any conformal Killing field, satisfying \nabla^{i}X^{j}+\nabla^{j}X^{i}=\mu g^{ij} for some function \mu. Note that g^{ij}S_{ij}=0 in dimension two. Contracting the identity ( 2.2 ) with X, we have (2.3) \nabla^{i}\left(X^{j}S_{ij}\right)=\left\langle{\mathcal{T}}(u),X^{j}{\partial}_{j}u\right\rangle. We now specialize to the case \Sigma={\mathbb{R}}^{2} with the flat metric; the general case requires only small modifications. Let x=x^{i}{\partial}_{i} be the radial vector field and r=\sqrt{(x^{1})^{2}+(x^{2})^{2}} the radial coordinate in {\mathbb{R}}^{2}. Define the weighted L^{2} -norm (2.4) \|\alpha\|_{\tau}=\sqrt{\int_{{\mathbb{R}}^{2}}|\alpha|^{2}\,e^{-\frac{r^{2}}{4\tau}}\,dV}. Let {\widehat{{\mathcal{T}}}}_{\tau}(u):={\mathcal{T}}(u)-\frac{x}{2\tau}\intprod du. Lemma 2.1 . For any W^{2,2} map u:{\mathbb{R}}^{2}\to N, we have (2.5) \left\|rdu\right\|_{\tau}\leq 4\tau\left\|{\widehat{{\mathcal{T}}}}_{\tau}(u)\right\|_{\tau} and (2.6) \|{\mathcal{T}}(u)\|_{\tau}\leq 3\left\|{\widehat{{\mathcal{T}}}}_{\tau}(u)\right\|_{\tau}. Proof. We take X=x in ( 2.3 ) and integrate by parts against e^{-\frac{r^{2}}{4\tau}}, to obtain \frac{1}{2\tau}\int x^{i}x^{j}S_{ij}e^{-\frac{r^{2}}{4\tau}}\,dV=\int\left\langle{\mathcal{T}}(u),x^{j}{\partial}_{j}u\right\rangle e^{-\frac{r^{2}}{4\tau}}\,dV. Unpacking the definition of the stress-energy tensor ( 2.1 ), we have \begin{split}-\frac{1}{4\tau}\int r^{2}|du|^{2}e^{-\frac{r^{2}}{4\tau}}\,dV&=\int\left\langle{\mathcal{T}}(u),x^{j}{\partial}_{j}u\right\rangle e^{-\frac{r^{2}}{4\tau}}\,dV-\int\left\langle\frac{x^{i}}{2\tau}{\partial}_{i}u,x^{j}{\partial}_{j}u\right\rangle e^{-\frac{r^{2}}{4\tau}}\,dV\\ &=\int\left\langle{\widehat{{\mathcal{T}}}}_{\tau}(u),x^{j}{\partial}_{j}u\right\rangle e^{-\frac{r^{2}}{4\tau}}\,dV.\end{split} Applying Cauchy-Schwarz, we obtain \frac{1}{4\tau}\left\|rdu\right\|_{\tau}^{2}\leq\left\|{\widehat{{\mathcal{T}}}}_{\tau}(u)\right\|_{\tau}\|x\intprod du\|_{\tau}. Since |x\intprod du|\leq r|du|, this implies ( 2.5 ). Then ( 2.6 ) follows from ( 2.5 ) and the triangle inequality. ∎ 2.2. Weighted Łojasiewicz inequality for maps to S^{2} Let \Phi(u):=\frac{1}{2}\int|du|^{2}e^{\frac{-r^{2}}{4}}\,dV and {\widehat{{\mathcal{T}}}}(u):={\widehat{{\mathcal{T}}}}_{1}(u)={\mathcal{T}}(u)-\frac{x}{2}\intprod du. Theorem 2.2 . Fix k\in{\mathbb{N}} and \beta>0. Given any map u:{\mathbb{R}}^{2}\to S^{2} with E(u)\leq 4\pi k, there exists n\in\{0,\ldots,k\} such that |\Phi(u)-4\pi n|\leq C_{k,\beta}\|{\widehat{{\mathcal{T}}}}(u)\|_{1}^{2-\beta}, where \|\cdot\|_{1} is the weighted L^{2} -norm defined by ( 2.4 ). Proof. By the weighted Poincaré inequality of Lemma 2.1 , we have \int r^{2}|du|^{2}e^{-\frac{r^{2}}{4}}+\|{\mathcal{T}}(u)\|^{2}_{L^{2}(B_{2})}\leq C\|{\widehat{{\mathcal{T}}}}(u)\|_{1}^{2}. In particular, this gives E(u,B_{2}\setminus B_{1})\leq C\|{\widehat{{\mathcal{T}}}}(u)\|_{1}^{2}. We can assume without loss of generality that \|{\widehat{{\mathcal{T}}}}(u)\|^{2}_{1}\leq{\varepsilon}_{0}/C, where {\varepsilon}_{0} is the standard {\varepsilon} -regularity constant for maps to S^{2}. Let 0\leq\lambda\leq 1 be the smallest radius such that \int_{B_{2}\setminus B_{\lambda}}|du|^{2}\leq{\varepsilon}_{0}. We have \lambda^{2}\int_{B_{2}\setminus B_{\lambda}}|du|^{2}\leq\int r^{2}|du|^{2}e^{-\frac{r^{2}}{4}}\,dV\leq C\|{\widehat{{\mathcal{T}}}}(u)\|_{1}^{2}. This gives \lambda\leq C\|{\widehat{{\mathcal{T}}}}(u)\|_{1}. Now, [ 12 , Corollary 1.3] reads (2.7) \begin{split}&\left|E(u,B_{1})-4\pi n\right|\\ &\qquad\leq C_{k,\beta}\left(\|{\mathcal{T}}(u)\|_{L^{2}(B_{2})}^{2}+E(u,B_{2}\setminus B_{1})+\lambda^{1-\beta}\left(\|{\mathcal{T}}(u)\|_{L^{2}(B_{2})}+\sqrt{E(u,B_{2}\setminus B_{1})}\right)\right).\end{split} Inserting the above estimates into the RHS of ( 2.7 ), we get \left|E(u,B_{1})-4\pi n\right|\leq C\|{\widehat{{\mathcal{T}}}}(u)\|_{1}^{2-\beta}. We further have \left|E(u,B_{1})-\frac{1}{2}\int_{B_{1}}|du|^{2}e^{\frac{-r^{2}}{4}}\right|\leq\int_{B_{1}}r^{2}|du|^{2}\leq C\|{\widehat{{\mathcal{T}}}}(u)\|_{1}^{2}. Finally, we have \int_{{\mathbb{R}}^{2}\setminus B_{1}}|du|^{2}e^{\frac{-r^{2}}{4}}\leq\int_{{\mathbb{R}}^{2}\setminus B_{1}}r^{2}|du|^{2}e^{\frac{-r^{2}}{4}}\leq C\|{\widehat{{\mathcal{T}}}}(u)\|_{1}^{2}. Combining these, we obtain the desired estimate. ∎ 3. Monotonicity formulae in dimension two Recall the pointwise energy identity for harmonic map flow, (3.1) \frac{1}{2}\frac{{\partial}}{{\partial}t}|du|^{2}+|{\mathcal{T}}(u)|^{2}=\nabla^{i}\nabla^{j}S_{ij}, see [ 8 , (2.7)] . Lemma 3.1 . Let w=w(r,t)=\exp\left(-{\frac{\phi(x)}{2\tau(t)}}\right). For a solution u(t):\Sigma\to N of harmonic map flow, we have \begin{split}\frac{1}{2}\frac{{\partial}}{{\partial}t}\left(|du|^{2}w\right)+\left|{\mathcal{T}}(u)-\frac{\nabla\phi}{2\tau}\intprod du\right|^{2}w&=\nabla^{i}\nabla^{j}\left(S_{ij}w\right)\\ &\qquad+\frac{w}{2\tau}\nabla^{i}\nabla^{j}\phi S_{ij}+\frac{\left(|\nabla\phi|^{2}+2\phi\tau^{\prime}(t)\right)w}{8\tau^{2}}|du|^{2}.\end{split} Proof. Multiplying the pointwise energy identity ( 3.1 ) by w, we obtain \begin{split}\frac{1}{2}\frac{{\partial}}{{\partial}t}\left(|du|^{2}w\right)-\frac{\tau^{\prime}(t)}{4\tau^{2}}\phi|du|^{2}w+|{\mathcal{T}}(u)|^{2}w&=\nabla^{i}\nabla^{j}\left(S_{ij}w\right)-2\nabla^{i}w\nabla^{j}S_{ij}-\nabla^{i}\nabla^{j}wS_{ij}\\ &=\nabla^{i}\nabla^{j}\left(S_{ij}w\right)+2\left\langle{\mathcal{T}}(u),\frac{\nabla\phi\intprod du}{2\tau}\right\rangle w\\ &\quad-\frac{\left(2\tau\nabla^{i}\nabla^{j}\phi+\nabla^{i}\phi\nabla^{j}\phi\right)}{4\tau^{2}}S_{ij}w.\end{split} Recalling the definition of the stress-energy tensor ( 2.1 ), we can rearrange to obtain the desired formula. ∎ We have the following versions of Struwe’s monotonicity formula in the critical dimension. Theorem 3.2 . Let \Sigma={\mathbb{R}}^{2} and take \Phi_{\tau}(u)=\frac{1}{2}\int|du|^{2}e^{-\frac{r^{2}}{4\tau}}\,dV,\quad\Psi_{\tau}(u)=\frac{1}{2}\int r^{2}|du|^{2}e^{-\frac{r^{2}}{4\tau}}\,dV. We have (3.2) \frac{d}{dt}\Phi_{\tau}(u(t))+\|{\widehat{{\mathcal{T}}}}_{\tau}(u(t))\|_{\tau}^{2}=\frac{1+\tau^{\prime}(t)}{4\tau^{2}}\Psi_{\tau}(u(t)) and (3.3) \frac{d}{dt}\Psi_{\tau}(u(t))+\|r{\widehat{{\mathcal{T}}}}_{\tau}(u(t))\|_{\tau}^{2}=\frac{1+\tau^{\prime}(t)}{8\tau^{2}}\int r^{4}|du|^{2}e^{-\frac{r^{2}}{4\tau}}\,dV. Proof. Take \phi=\frac{r^{2}}{2} in the previous Lemma, so that w=e^{\frac{-r^{2}}{4\tau}}. We have (3.4) \begin{split}\frac{1}{2}\frac{{\partial}}{{\partial}t}\left(|du|^{2}w\right)+\left|{\widehat{{\mathcal{T}}}}_{\tau}(u)\right|^{2}w&=\nabla^{i}\nabla^{j}\left(S_{ij}w\right)+\frac{\left(1+\tau^{\prime}(t)\right)r^{2}}{8\tau^{2}}|du|^{2}w.\end{split} The identity ( 3.2 ) follows by integrating ( 3.4 ) in space. The identity ( 3.3 ) follows by integrating by parts against r^{2} and using \nabla^{i}\nabla^{j}r^{2}=2\delta^{ij} together with tracelessness of S_{ij}. ∎ 4. Estimates on energy density and oscillation Lemma 4.1 . Fix 0<R\leq\sqrt{T}, 0<\lambda\leq\frac{1}{2}, and \kappa>0. Let u:{\mathbb{R}}^{2}\times\left[0,T\right)\to N be a classical solution of harmonic map flow. Suppose that R^{2}\leq T_{1}\leq T is such that (4.1) \sup_{T_{1}-R^{2}\leq t\leq T_{1}-(R\lambda)^{2}}\frac{\Psi_{T_{1}-t}\left(u(t)\right)}{T_{1}-t}\leq{\varepsilon}_{0}, where {\varepsilon}_{0}>0 depends on the geometry of N. Letting (4.2) \eta:=\frac{\sqrt{\Psi_{4R^{2}}\left(u(T_{1}-R^{2})\right)}}{R} and (4.3) \psi\left(\log\frac{R}{\sqrt{T_{1}-t}}\right):=\sqrt{\frac{\Psi_{T_{1}-t}\left(u(t)\right)}{T_{1}-t}}, we have r|du(x,t)|\leq C_{\kappa}\left(\int_{0}^{\log\frac{R}{r}}\psi\left(\sigma\right)e^{\kappa\left(\sigma-\log\frac{R}{r}\right)}\,d\sigma+\eta\left(\frac{r}{R}\right)^{\kappa}\right) for \lambda R\leq r\leq\min\left\{\left(C_{N,\kappa}\sqrt[\kappa]{\eta}\right)^{-1},1\right\}R and T_{1}-\frac{1}{4}r^{2}\leq t<T_{1}. Proof. We may assume without loss of generality that R=T_{1}=1. Let s=-\frac{1}{2}\log\left(1-t\right), and define \bar{\psi}\left(s\right):=\sqrt{\frac{\Psi_{4(1-t)}(u(t))}{1-t}}. We