The entropy production is not always monotone in the space-homogeneous Boltzmann equation

This paper disproves McKean's 1966 conjecture by demonstrating that entropy production in the space-homogeneous Boltzmann equation is not always monotone, specifically by constructing a counterexample with a non-physical collision kernel where entropy production increases over time.

The Gist

The Big Picture: A Broken Rule of the Universe?

Imagine a crowded dance floor where people are constantly bumping into each other, changing partners, and swapping energy. In physics, this is modeled by the Boltzmann equation. It describes how a gas (or a crowd) moves from a messy, chaotic state toward a calm, orderly state (equilibrium).

For over 150 years, physicists have relied on a rule called the H-theorem. Think of this as a "messiness meter" (Entropy).

• The Rule: As the particles collide, the system gets more disordered, and the "messiness meter" goes up.
• The Consequence: Because the system is settling down, the speed at which it gets messy (called Entropy Production) was thought to always slow down. It was like a car braking: it starts fast, then slows down smoothly until it stops.

In 1966, a famous mathematician named McKean made a bold guess (a conjecture): "Not only does the messiness increase, but the speed at which it increases should always go down." He thought the "braking" was always smooth.

This paper says: "Not so fast."

Luis Silvestre has built a mathematical "trap" where the braking system fails. He found a specific, weird scenario where the speed of the messiness speeds up instead of slowing down. It's like a car that, when you hit the brakes, suddenly revs its engine and accelerates.

---

The Ingredients: A Very Weird Collision

To prove this, Silvestre didn't use a standard gas (like air or helium). He used a very specific, artificial setup.

1. The Collision Kernel (The Rules of the Game)
In a real gas, particles can bounce off each other at any angle and with any speed.

• The Analogy: Imagine a game of pool where the balls can only hit each other if they are moving at exactly a specific speed and must bounce off at exactly a 90-degree angle.
• The Reality: Silvestre used a "kernel" (a set of rules) that forces particles to behave this way. It's not a realistic gas found in nature; it's a mathematical construct designed to break the rules.

2. The Initial Setup (The Crowd)
He arranged the particles in a very specific pattern:

• A tiny, super-dense cluster in the center (like a packed mosh pit).
• A ring of particles further out moving at a specific speed.
• Empty space everywhere else.

---

The Experiment: What Happens Next?

Silvestre let this system evolve according to his weird rules and watched the "braking" (Entropy Production).

The Surprise:
Instead of the braking force getting weaker and weaker, it suddenly got stronger.

• The Mechanism: Because of the weird rules (the 90-degree bounce and specific speed), the particles in the center and the ring interact in a way that creates a massive, sudden spike in disorder.
• The Result: The rate at which the system becomes messy actually increases for a while. The "brakes" didn't just fail; the car floored the gas pedal.

Why Does This Matter?

You might ask, "So what? This isn't a real gas."

1. It Breaks a 60-Year-Old Myth: For decades, mathematicians and physicists assumed McKean's rule was true for all space-homogeneous systems (systems where the gas looks the same everywhere). Silvestre proved that this assumption is mathematically false, even if the example is weird.
2. It Sets a Boundary: It tells us that the "smooth braking" we see in nature isn't a universal law of mathematics. It only happens because real-world collisions are "nice" and follow specific physical laws (like hard spheres or gravity).
3. The "What If" Factor: It's like finding a bridge that collapses under a specific, impossible weight. It doesn't mean the bridge is unsafe for normal cars, but it proves the bridge isn't indestructible. It forces scientists to be more careful about why things work the way they do.

The Takeaway

Think of the universe as a giant, complex machine. We thought we understood how one of its gears (Entropy Production) worked: it always slows down.

Luis Silvestre built a prototype of a different machine using strange, custom-made gears. In this prototype, the gear speeds up.

The lesson: Nature is usually well-behaved, but mathematics allows for "monsters" that break our intuition. While real gases in our atmosphere still follow the smooth braking rule, we now know that the mathematical foundation for that rule is more fragile than we thought.

In short: The paper proves that the "speed of disorder" doesn't always slow down. Sometimes, in a very weird mathematical world, it can suddenly speed up, disproving a famous guess made in 1966.

Technical Summary

1. Problem Statement

The paper addresses a long-standing open problem in kinetic theory regarding the monotonicity of entropy production in the space-homogeneous Boltzmann equation.

• Context: For a solution $f(t, v)$ to the space-homogeneous Boltzmann equation, the Boltzmann entropy $H(f) = -\int f \log f \, dv$ is known to be non-increasing in time (the H-theorem). The entropy production is defined as $D(f) = -\frac{d}{dt}H(f) = -\int Q(f,f) \log f \, dv$, which is non-negative.
• McKean's Conjecture (1966): Henry McKean conjectured that the entropy production $D(f)$ itself is monotone decreasing in time (i.e., $\frac{d}{dt}D(f) \leq 0$).
• Current Status: While numerical simulations with physically motivated kernels (e.g., hard spheres, inverse power laws) have never observed violations of this monotonicity, no rigorous proof exists for general kernels. Conversely, the monotonicity of the Fisher information has been proven for specific operators under reasonable assumptions.
• Objective: The author aims to determine if McKean's conjecture holds universally for the space-homogeneous Boltzmann equation, regardless of the collision kernel.

2. Methodology

The author constructs a counterexample using a singular collision kernel and a specific piecewise constant distribution function.

A. The Collision Kernel

The paper utilizes a highly singular collision kernel $B(|v-v_*|, \cos \theta) = \Phi(|v-v_*|)b(\cos \theta)$, which is not physically motivated but mathematically tractable:

• Angular part ($b$): A Dirac mass concentrated at $\theta = \pm \pi/2$. This restricts collisions to 90-degree scattering angles.
• Radial part ($\Phi$): A Dirac mass concentrated at the relative velocity magnitude $|v-v_*| = \sqrt{2}$.
• Geometric Implication: This kernel restricts the collision integral to configurations where the pre-collision velocities ($v, v_*$) and post-collision velocities ($v', v'_*$) form a square of side length 1. This simplifies the Boltzmann collision operator $Q(f,f)$ to an integral over a circle $S^1$.

B. The Distribution Function ($f$)

The author constructs a specific non-negative, compactly supported function $f: \mathbb{R}^2 \to [0, \infty)$ defined by three regions:

1. Core: A small ball $B_\rho$ around the origin where $f(v) = c a^2$ (with $c$ small, $a$ large).
2. Ring: An annulus at radius $\sqrt{5}$ where $f(v) = a$.
3. Background: The region between the core and the ring (and outside the support) where $f(v) = 1$ (or 0 outside a larger ball).

The parameters are chosen such that $a \gg 1$ and $c$ is sufficiently small. The specific radii ($\sqrt{2}$ and $\sqrt{5}$) are chosen to align with the geometric constraints of the singular kernel.

C. Analytical Approach

The proof relies on analyzing the time derivative of the entropy production, $\partial_t D(f)$.

1. Derivation: The author derives an explicit formula for $\partial_t D(f)$:
   $$\partial_t D(f) = -\int \frac{Q^2}{f} \, dv + \int Q L \, dv$$
   where $L$ is a specific integral term involving the logarithm of the ratio of pre- and post-collision densities.
2. Scaling Analysis: The author estimates the magnitude of the terms $Q$ (collision operator) and $L$ for the constructed function $f$:
   • Gain/Loss Terms ($Q$): $Q$ is of order $a^2$ only in specific regions where the geometry of the square configuration aligns with the support of $f$.
   • The Logarithmic Term ($L$): Crucially, $L$ scales as $a^2 \log a$ in the specific ring where $|v| \approx \sqrt{2}$. This is because $v_*$ can fall into the high-density core ($f \approx ca^2$) while $v$ is in the ring, creating a large logarithmic ratio $\log(f/f')$.
3. Dominance Argument:
   • The negative term $-\int \frac{Q^2}{f}$ scales as $O(a^4)$.
   • The positive term $\int Q L$ scales as $O(a^4 \log a)$ in the critical ring.
   • By choosing $a$ sufficiently large, the positive term dominates the negative term, resulting in $\partial_t D(f) > 0$.

3. Key Contributions

• Disproof of McKean's Conjecture: The paper provides a rigorous counterexample showing that entropy production is not universally monotone decreasing in time for the space-homogeneous Boltzmann equation.
• Counterexample Construction: It demonstrates that the sign of the second derivative of entropy ($H''(t)$) can be positive, disproving the conjecture at the lowest possible order of derivatives.
• Kernel Specificity: The result highlights that the monotonicity property is not a universal feature of the Boltzmann equation but depends heavily on the collision kernel. The counterexample relies on a kernel that is singular and physically unrealistic.
• Clarification of Numerical Observations: The paper explains why numerical tests (which use physical kernels like hard spheres) have never observed this behavior: the property likely holds for physically relevant kernels but fails for pathological ones.

4. Results

• Theorem 1: There exists a non-negative, compactly supported function $f$ and a collision kernel of the form (2) such that the entropy production $D(f(t))$ increases in time.
• Quantitative Estimate: For the constructed example, as the parameter $a \to \infty$, the time derivative of entropy production behaves as:
  $$\partial_t D(f) \approx a^4 \log a > 0$$
• Approximation: While the proof uses singular Dirac masses for the kernel and piecewise constant functions for $f$, the author notes that these can be approximated by smooth functions and smooth kernels to obtain the same result, provided the kernel remains highly concentrated near specific relative velocities and angles.

5. Significance

• Theoretical Impact: This work resolves a 50-year-old conjecture by McKean, establishing that the "super H-theorem" (monotonicity of higher-order entropy derivatives) does not hold in the general setting of the Boltzmann equation.
• Physical vs. Mathematical Models: It draws a sharp distinction between mathematical generality and physical reality. While the result disproves the conjecture mathematically, the author suggests that McKean's conjecture might still hold for physically relevant kernels (e.g., Maxwell molecules, hard spheres) or under additional assumptions similar to those required for Fisher information monotonicity.
• Future Directions: The paper opens new questions regarding the specific conditions on collision kernels required to ensure entropy production monotonicity. It suggests that the geometry of the kernel (localization in relative velocity and scattering angle) is a critical factor in the dynamics of entropy production.

In summary, Silvestre demonstrates that the intuitive expectation of "entropy production always decreasing" is false in the general mathematical framework of the Boltzmann equation, relying on a carefully engineered singular kernel to expose the non-monotonic behavior.