🔬 physics

Optimizing bounds for energy-constrained optimal cooling problems in two dimensions

This paper develops a Lagrange duality framework and a semidefinite programming hierarchy to derive new analytical upper bounds on the cooling efficiency of two-dimensional incompressible flows, proving that efficiency scales as $\mathrm{Pe}^2$ for general domains and improves to $\mathrm{Pe}^2/\ln^2\mathrm{Pe}$ for specific geometries like disks and annuli.

Original authors: Pedro Blöss Braga, Giovanni Fantuzzi

Published 2026-08-17

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Pedro Blöss Braga, Giovanni Fantuzzi

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to cool down a hot room, but you have a strict limit on how much energy your air conditioner can use. You can't just blast cold air everywhere; you have to be clever. You need to figure out the perfect way to swirl the air around so that it picks up heat from the hot spots and carries it to the cold walls as efficiently as possible. This is the kind of puzzle scientists face when designing heat exchangers for engines, cooling systems for computer chips, or even understanding how heat moves in the atmosphere.

In the world of physics, this "swirling" is called fluid flow, and the "heat" is often just a passive tag along, like a dye in water, that gets carried by the current. The challenge is that the air (or water) has to follow strict rules: it can't disappear or appear out of nowhere (it must be incompressible), and it can't push through the walls of the room. Scientists measure how hard they are working with a number called the "Péclet number" (Pe), which basically tells you how much kinetic energy the fluid has. The goal is simple: use a fixed amount of energy to make the temperature as uniform as possible. If the temperature is uneven, the fluid is doing a bad job. If it's perfectly smooth, the cooling is perfect. But finding the absolute best way to swirl the fluid is incredibly hard because the math gets messy and twisty when the fluid moves fast.

This is where a team of mathematicians from Friedrich–Alexander University stepped in. They tackled the question of: "What is the absolute worst-case scenario for cooling efficiency?" Instead of trying to find the perfect swirl (which is like finding a needle in a haystack), they asked, "No matter how clever you are, how bad can the cooling get if you only have this much energy?" They used a clever mathematical trick called "Lagrange duality," which is like flipping a problem upside down to see it from a different angle. By turning the problem into a "dual" version, they could use powerful computer tools called semidefinite programs (SDPs) to calculate strict upper limits on how efficient the cooling could possibly be.

Their main discovery is a new set of rules that act as a speed limit for cooling efficiency. They proved that for any shape of room and any pattern of heat, the cooling efficiency ratio $E(Pe)$ cannot exceed a value proportional to the square of the energy you put in (specifically, it is bounded by $Pe^2$ ). This confirms and improves upon previous guesses. But they didn't stop there. When they looked at circular rooms (like disks or rings) where the heat is spread out evenly, they found an even stricter rule. They showed that the efficiency ratio is at most a value proportional to $Pe^2$ divided by the square of the natural logarithm of $Pe $($ Pe^2 / \ln^2 Pe$). Think of it like this: if you double your energy, the efficiency doesn't just scale up linearly with the square of the energy; it is held back by a "logarithmic penalty" that gets bigger as you try to push harder. It is important to note that this is a proven ceiling on performance, but the authors explicitly state that the true optimal cooling might actually be even better (lower) than this limit, as the mathematical gap between the bound and the true optimum is not yet closed.

The authors didn't just write equations; they built a digital playground. They simulated these cooling problems in a square room and a ring-shaped room (an annulus). They used a method that breaks the continuous fluid into tiny, manageable chunks to run the calculations. They found that as they increased the energy (the Péclet number), the "critical" swirls needed to achieve the best cooling became incredibly complex, with more and more tiny ripples appearing near the walls. Their computer simulations matched their new mathematical rules perfectly. They also showed how to use the symmetry of the room (like how a square looks the same if you flip it) to make the computer calculations faster and less demanding.

One of the most exciting parts of their work is how they used the computer results to inspire a new, better mathematical proof. The simulations showed them what the "ideal" cooling pattern looked like in a ring-shaped room. They took that visual clue and built a specific mathematical function that proved their new, tighter upper bound ( $Pe^2 / \ln^2 Pe$ ). This is a rare and powerful moment in math: the computer didn't just check the answer; it helped write the proof.

However, the authors are careful to point out what they haven't solved. They proved that you cannot do better than their new limits, but they did not prove that you can actually reach those limits. The gap between the "best possible" limit they found and the "best possible" limit that actually exists might still be there. It's like knowing you can't run faster than 100 mph, but not knowing if the fastest human can actually run 99 mph or only 50 mph. They also note that their methods work best for steady, unchanging flows. If the fluid starts churning and changing over time, the math gets even trickier, though they believe their tools could eventually be adapted for that too.

In the end, this paper is a masterclass in using computers and advanced math to set boundaries on what is physically possible. It tells engineers and scientists that no matter how smart their cooling designs get, there is a fundamental ceiling on performance dictated by the laws of physics and the amount of energy available. For circular rooms with even heating, that ceiling is slightly lower than we previously thought, thanks to a new logarithmic factor. While the perfect cooling flow might still be a mystery, we now have a much clearer map of the territory, knowing exactly where the walls are.

Technical Summary: Optimizing Bounds for Energy-Constrained Optimal Cooling Problems in Two Dimensions

Problem Statement
This work addresses optimal control problems for incompressible fluids in two-dimensional domains ( $\Omega \subset \mathbb{R}^2$ ) containing internal heat sources and sinks ( $f$ ) and a boundary held at a constant temperature (zero). The objective is to determine the maximum cooling efficiency achievable by a flow field $u$ subject to a kinetic energy budget. The flow is governed by the steady advection-diffusion equation:
$-\Delta T + u \cdot \nabla T = f \quad \text{in } \Omega, \quad T = 0 \quad \text{on } \partial\Omega$
where $T$ is the temperature. The cooling efficiency is defined as $E(\text{Pe}) = \|f\|_{H^{-1}}^2 / J(\text{Pe})$ , where $J(\text{Pe})$ is the minimum mean square temperature gradient $\int_\Omega |\nabla T|^2 dx$ achievable by any divergence-free velocity field $u$ with $\|u\|_{L^2} \leq \text{Pe}$ . Here, $\text{Pe}$ is a nondimensional Péclet number representing the kinetic energy constraint.

While lower bounds on $J(\text{Pe})$ (and thus upper bounds on $E(\text{Pe})$ ) can be found via numerical optimization of specific flows, finding rigorous upper bounds on $E(\text{Pe})$ —which must hold for all possible flows—is significantly more difficult. Previous analytical bounds for energy-constrained flows in disks, such as $E(\text{Pe}) \lesssim \text{Pe}^2$ , were found to be loose compared to numerical lower bounds suggesting $E(\text{Pe}) \gtrsim \text{Pe}$ .

Methodology
The authors employ a combination of Lagrange duality and Semidefinite Programming (SDP) to derive rigorous upper bounds on cooling efficiency.

Streamfunction Reformulation: The velocity field is parameterized via a streamfunction $\psi$ such that $u = \text{Pe} \nabla^\perp \psi$ . This transforms the nonconvex optimal control problem (OCP) into a minimization over $\psi$ and $T$ .
Lagrange Duality: The authors formulate the Lagrange dual of the OCP. By introducing Lagrange multipliers for the advection-diffusion equation and the energy constraint, they derive an infinite-dimensional convex maximization problem (LDP). The optimal value of this dual problem, $L(\text{Pe})$ $L (Pe)$ , provides a lower bound on the primal cost $J(\text{Pe})$ $J (Pe)$ , and consequently an upper bound on the efficiency $E(\text{Pe})$ $E (Pe)$ .
- The dual problem is shown to be an infinite-dimensional SDP involving a positive semidefinite constraint on a quadratic form $Q(a, \xi)$ defined over the streamfunction and temperature spaces.
SDP Discretization: To compute these bounds numerically, the infinite-dimensional dual is discretized using conforming Galerkin schemes (e.g., spectral methods or finite elements). This yields a hierarchy of finite-dimensional SDPs.
- The authors prove that under standard approximation assumptions (existence of recovery sequences in Sobolev spaces), the discrete solutions converge to the true dual optimum as the discretization resolution increases.
Symmetry Exploitation: For domains with geometric symmetries (squares, annuli, disks), the authors demonstrate how to decompose the function spaces into invariant subspaces. This reduces the computational complexity of the resulting SDPs by exploiting the symmetry of the heat distribution and the domain.

Key Contributions and Results

Numerical Framework for Rigorous Bounds: The paper establishes a computationally tractable method to compute the best possible upper bound on cooling efficiency provable via Lagrange duality. Numerical experiments on a square and an annulus demonstrate rapid convergence of the SDP discretizations.
General Analytical Bound ( $E(\text{Pe}) \lesssim \text{Pe}^2$ ): By constructing a specific admissible point for the dual problem, the authors provide an elementary proof that for arbitrary domains and heat distributions, $E(\text{Pe}) \leq 1 + C\text{Pe}^2$ . This generalizes previous results known for disks to arbitrary 2D domains.
Improved Logarithmic Bounds for Symmetric Domains: Inspired by numerical results showing that the optimal dual variable $\xi$ $ξ$ concentrates near boundaries and exhibits specific radial structures, the authors construct refined analytical admissible points for disks and annuli with heat distributions having a strictly positive azimuthal average.
- They prove that for these cases, the bound improves to $E(\text{Pe}) \lesssim \text{Pe}^2 / \ln^2 \text{Pe}$ .
- This result significantly tightens the known upper bounds for uniformly heated disks and annuli, bringing them closer to the observed lower bounds.

Significance and Claims
The paper claims to bridge the gap between numerical optimization (which provides lower bounds) and analytical estimates (which often provide loose upper bounds) by utilizing the Lagrange dual formulation.

Generalization: The work extends the understanding of energy-constrained cooling beyond specific geometries, providing a unified framework for arbitrary 2D domains.
Refinement of Asymptotics: The derivation of the $1/\ln^2 \text{Pe}$ factor represents a significant improvement over the standard $\text{Pe}^2$ scaling for symmetric heating scenarios, suggesting that the efficiency of cooling does not degrade as rapidly as previously thought for large Péclet numbers in these geometries.
Computational Guidance: The authors highlight that their numerical SDP results were instrumental in guiding the construction of the improved analytical bounds, demonstrating a feedback loop between computational optimization and rigorous analysis.
Limitations: The authors note that while the dual bounds converge, the duality gap (the difference between the primal nonconvex problem and the convex dual) may prevent the bounds from being sharp. They acknowledge that proving convergence rates for the discretization requires higher regularity of the dual variables, which remains an open technical challenge. Furthermore, they state that if the duality gap affects the large- $\text{Pe}$ asymptotics, methods beyond standard Lagrange duality (such as moment-sum-of-squares hierarchies) would be required to obtain quantitatively sharp bounds.

The work does not propose new physical experiments or specific engineering designs but rather advances the mathematical theory of bounding transport properties in fluid mechanics, offering a robust tool for evaluating the theoretical limits of cooling efficiency.

More like this