Mean curvature flows with prescribed singular sets

This paper constructs mean-convex ancient solutions to mean curvature flow in a smooth Riemannian metric arbitrarily close to Euclidean space, demonstrating that any closed set $K \subset \mathbb{R}^n$ can be realized as the first-time singular set of such a flow in $\mathbb{R}^{m+n}$.

The Gist

Imagine you have a giant, invisible sheet of rubber floating in space. If you let it shrink naturally, it will try to minimize its surface area, kind of like a soap bubble popping. In mathematics, this process is called Mean Curvature Flow.

Usually, when these sheets shrink, they get weird and crumpled at specific points called "singularities." Think of these singularities as the moment the rubber sheet tears or pinches off. For decades, mathematicians believed that in a perfectly flat, empty universe (Euclidean space), these tears could only happen in very simple, predictable ways—like isolated dots or smooth lines. It was thought that the "shape" of the tear was strictly limited by the laws of physics in that flat universe.

The Big Breakthrough

Raphael Tsiamis, the author of this paper, asks a bold question: What if the universe isn't perfectly flat?

He proves that if you take that flat universe and wiggle the fabric of space just a tiny, tiny bit (so small that you'd never notice it with a ruler), you can force the rubber sheet to tear in any shape you want.

Here is the simple breakdown of how he did it, using some creative analogies:

1. The "Magic Trampoline" (The Metric)

Imagine the universe is a giant trampoline. In a normal trampoline, the fabric is flat. Tsiamis says, "What if we secretly sewed a few microscopic wrinkles into the trampoline fabric?"

These wrinkles are what mathematicians call a Riemannian metric. They are so tiny that the trampoline looks perfectly flat to the naked eye, but they create invisible "gravity wells" or "speed bumps" that guide the rubber sheet as it shrinks.

2. The "Ghostly Blueprint" (The Singular Set)

Tsiamis starts with a "Ghostly Blueprint." This is any shape you can imagine—a fractal snowflake, a messy scribble, a perfect circle, or a jagged mountain range. Let's call this shape K.

His goal is to make the rubber sheet tear exactly along the lines of this blueprint K at the very last moment before it disappears.

3. The "Staircase" Construction (The Solution)

Building a solution for such a complex shape is like trying to build a staircase that leads to a cloud. You can't just jump to the top; you have to build step by step.

• The Cylinder: He starts with a simple, boring shape: a shrinking cylinder (like a soda can getting smaller and smaller). This is the "easy" part.
• The Barrier: He builds invisible walls (barriers) around this cylinder. These walls act like a cage, forcing the shrinking sheet to stay close to the cylinder but allowing it to wiggle slightly.
• The Staircase: He breaks the problem down into tiny steps. He solves the math for one small "step" of the staircase, then uses that answer to solve the next step, and so on. This is called an inductive argument.
• The Glue: Once he has all the steps, he uses a special mathematical "glue" (a transport equation) to stitch them all together smoothly. This glue is the function f mentioned in the paper. It's the secret sauce that adjusts the "trampoline fabric" just enough to guide the sheet into the exact shape of the blueprint K.

4. The Result: Controlled Chaos

The final result is a "Mean-Convex Ancient Solution." Let's translate that:

• Ancient: It has been shrinking forever, from the distant past up until the present moment.
• Mean-Convex: It's shrinking in a "nice" way (it doesn't have weird inward dents that would make it collapse prematurely).
• Prescribed Singular Set: At time $t=0$, the sheet vanishes, but the pattern of its disappearance is exactly the shape K you chose.

Why This Matters

Before this paper, mathematicians thought the universe was a strict teacher that only allowed simple, clean tears in space-time. Tsiamis showed that the universe is actually a very flexible teacher.

If you are willing to tweak the "rules of the game" (the geometry of space) by an amount so small it's practically zero, you can make the universe behave in incredibly complex and chaotic ways. You can make a singularity look like a fractal, a maze, or a random scribble.

In a Nutshell:
Tsiamis proved that by adding invisible, microscopic wrinkles to the fabric of space, you can force a shrinking object to break apart in any pattern you can draw, no matter how messy or complicated that pattern is. It's like having a magic eraser that, instead of just wiping a line, can be programmed to erase a specific, intricate drawing on a piece of paper.

Technical Summary

1. Problem Statement and Context

The Core Question:
The paper addresses a fundamental question in geometric analysis regarding the flexibility of singularity formation in Mean Curvature Flow (MCF). Specifically, it investigates whether the singular set of a mean-convex flow at its first singular time can be prescribed arbitrarily.

Background and Constraints:

• Standard Theory: In Euclidean space $\mathbb{R}^{n+1}$, deep results by White, Huisken-Sinestrari, and Colding-Minicozzi establish that the singular set of a mean-convex flow is highly constrained. White proved the parabolic Hausdorff dimension is at most $n-1$, and Colding-Minicozzi showed the set is contained in a finite union of codimension-2 Lipschitz submanifolds plus a set of codimension $\ge 3$.
• The Conjecture: It was an open question (attributed to Colding-Minicozzi, Ilmanen, and White) whether any closed set could arise as a singular set. White conjectured that in $\mathbb{R}^3$, singular sets might be restricted to isolated points and curves.
• Known Examples: Prior to this work, known examples were limited to product structures (shrinking cylinders $R^k \times S^{n-k}$), "marriage rings," and specific non-product flows with singular sets like rays or $\mathbb{R}^n \times \{0\}$.

The Main Objective:
To demonstrate that by allowing an arbitrarily small smooth perturbation of the ambient Euclidean metric, one can construct a mean-convex ancient solution to MCF in $\mathbb{R}^{m+n}$ ($m \ge 2$) whose first-time singular set is exactly $K \times \{0\}$, where $K$ is an arbitrary non-empty closed subset of $\mathbb{R}^n$. This implies that in $\mathbb{R}^3$, the singular set can be a fractal or any other closed set, provided the metric is perturbed slightly.

---

2. Methodology and Construction Strategy

The proof relies on a sophisticated construction involving warped product metrics, barrier methods, and an inductive gluing procedure.

A. Geometric Setup

The author works in $\mathbb{R}^{m+n} = \mathbb{R}^n_x \times \mathbb{R}^m_\xi$ with a metric of the form:
$$g_f = \sum_{i=1}^n dx_i^2 + f(x, |\xi|^2) \sum_{j=1}^m d\xi_j^2$$
where $f(x, r)$ is a smooth function close to 1. The flow is restricted to $O(m)$-invariant hypersurfaces (tubes) defined by a radius function $u(x,t)$, or equivalently $w(x,t) = u^2(x,t)$.

B. The Cylindrical Mean Curvature Flow (CMCF)

The evolution equation for the radius function $w$ in this background is derived as a quasilinear parabolic PDE:
$$M_\zeta(w) := \partial_t w - \Delta w + 2(m-1) + \zeta(x,t)Q(w) + \frac{2|\nabla w|^2}{4w + |\nabla w|^2} = 0$$
where $\zeta$ is a cutoff function controlling the nonlinearity. The goal is to find a solution $w$ that stays close to the shrinking cylinder solution $\phi(t) = -2(m-1)t$ but decays to 0 exactly on the set $K$.

C. Key Analytic Ingredients

1. Cut-off Function Construction (Definition 1):
   For a closed set $K$, a smooth function $h(x)$ is constructed such that $h \equiv 0$ on $K$, $h > 0$ on $U = \mathbb{R}^n \setminus K$, and $h$ satisfies specific derivative bounds controlled by a small parameter $\tau$. This function dictates the "distance" to the singular set.

2. Barrier Functions (Section 2.1):
   The author constructs upper and lower barriers $w_\pm$ of the form $-2(m-1)t + F_\pm(x,t)$, where $F_\pm$ involve terms like $h(x)^3 \exp(-h(x)^{-9/8})$. These barriers are designed to:

   • Satisfy $M_\zeta(w_-) < 0 < M_\zeta(w_+)$.
   • Remain exponentially close to the cylinder solution.
   • Force the solution to vanish exactly where $h=0$ (i.e., on $K$).

3. Inductive Gluing on a "Staircase" Domain (Section 2.2):
   The domain is approximated by a sequence of cylindrical domains (a "staircase") defined by level sets of $h(x)^4$.

   • Induction Step: A solution is constructed on each step $Q_j$ using the terminal data from the previous step as initial data.
   • Gluing: A smooth cutoff function interpolates between the solution on the current step and the barrier function on the next step.
   • Regularity: The Parabolic Small Perturbation Theorem (Savin's theorem, adapted by Wang) is crucial here. It ensures that if the solution stays sufficiently close to the cylinder solution (which is guaranteed by the barriers), it inherits high regularity ($C^\infty$) and satisfies necessary derivative estimates.

4. Metric Perturbation via Transport Equation (Section 3 & Appendix A):
   The constructed solution $w_\tau$ is an exact solution for a specific operator $M_1$ in a specific region, but not globally for the standard Euclidean metric.

   • The author defines an "arrival time" function $T(x, r)$ based on $w_\tau$.
   • The goal is to find a metric function $f(x,r)$ such that $T$ satisfies the arrival time equation $P_f(T) = 0$ for the metric $g_f$.
   • This reduces to solving a transport equation for a variable $z = (1-f)/r$:
     $$z_r + z^2 = -\langle a, \nabla_x z \rangle + bz + c$$
   • Using the Method of Characteristics, the author proves the existence of a smooth solution $z$ (and thus $f$) that is arbitrarily close to 1 (Euclidean) and vanishes appropriately to make the flow singular exactly at $K \times \{0\}$.

---

3. Key Results

Theorem 1.1 (Main Theorem):
Let $K \subset \mathbb{R}^n$ be a non-empty closed set and $m \ge 2$. There exists a smooth Riemannian metric $g_f$ on $\mathbb{R}^{m+n}$, which is $C^\infty$-close to the Euclidean metric (specifically, $|D^k(f-1)| < C\tau$), such that there exists a mean-convex ancient solution $\{G_t\}_{t<0}$ to the MCF in $(\mathbb{R}^{m+n}, g_f)$ with the property:
$$\text{sing}(G_0) = K \times \{0\}$$
where the singularity occurs at time $t=0$.

Corollary/Implications:

• Fractal Singularities: In $\mathbb{R}^3$ (where $n=1, m=2$), $K$ can be a fractal set, meaning the singular set can have arbitrary topological complexity.
• Instability of Euclidean Structure: The structural constraints on singular sets (e.g., codimension-2 structure) established for Euclidean space are not stable under arbitrarily small smooth perturbations of the ambient metric.

---

4. Significance and Impact

1. Resolution of Flexibility Question: The paper definitively answers the question of how flexible singular sets can be. It shows that without the strict rigidity of the Euclidean metric, the singular set can be any closed set.
2. Parallel to Minimal Surfaces: The result parallels a recent breakthrough by Leon Simon (2023) regarding stable minimal hypersurfaces with prescribed singular sets. However, Tsiamis achieves this for the parabolic MCF setting using distinct methods (barriers and transport equations rather than minimal surface theory).
3. Methodological Innovation:
   • The use of an infinite staircase of cylindrical domains combined with Savin's small perturbation theorem provides a robust framework for constructing solutions to quasilinear parabolic equations with prescribed behavior near singular sets.
   • The reduction of the metric construction to a transport equation solvable by characteristics is a powerful technique for modifying the ambient geometry to fit a desired flow.
4. Theoretical Implications: It highlights that the "nice" structure of singularities in Euclidean MCF is a fragile property dependent on the specific symmetries and flatness of the Euclidean metric. In generic Riemannian manifolds (even those very close to Euclidean), the singularity structure can be arbitrarily complex.

Summary Conclusion

Raphael Tsiamis's paper constructs a counterexample to the rigidity of mean-convex MCF singularities in the context of perturbed metrics. By combining barrier methods, inductive gluing on a discretized domain, and solving a transport equation to adjust the metric, the author proves that any closed set can be realized as the first-time singular set of a mean-convex ancient flow, provided the ambient metric is allowed to be slightly perturbed from Euclidean. This fundamentally alters the understanding of singularity formation, showing that the constraints observed in Euclidean space are not universal but metric-dependent.