Weighted least action principle for Maxwell equations

Imagine you are trying to figure out the shape of a hidden object by looking at its shadow. In the world of light and physics, scientists often face a similar puzzle: How do we figure out the invisible "phase" (the timing and shape) of a light wave just by measuring how bright it is (its intensity) at two different points?

This paper by Jacob Rubinstein and Gershon Wolansky offers a new, upgraded way to solve this puzzle, specifically for light traveling through complex, "directional" materials (like certain crystals) where light doesn't behave simply.

Here is the breakdown of their idea using everyday analogies:

1. The Old Way: Following a Single Ray

Traditionally, scientists used Fermat's Principle, which is like saying, "Light takes the fastest path." Imagine a single hiker trying to get from Point A to Point B across a mountain. If you know the terrain, you can predict exactly which path that one hiker will take.

However, the authors point out a problem: A single ray of light isn't a real, measurable thing. In the real world, we can't measure a single, infinitely thin line of light. We can only measure a "bundle" of light—a patch of brightness on a wall or a sensor.

2. The New Idea: Moving a Crowd of Light

Instead of tracking one hiker, the authors treat light like a crowd of people moving from one room (Plane 1) to another room (Plane 2).

The Input: You know how crowded the first room is (the intensity of light, $I_1$ ).
The Output: You know how crowded the second room is (the intensity of light, $I_2$ ).
The Goal: You need to figure out the most efficient way to move every single person from the first room to the second so that the final crowd matches the pattern you see.

This is based on a mathematical concept called Optimal Transport (or the Monge problem). Think of it as a logistics company trying to move boxes from a warehouse to a store with the least amount of fuel. The "cost" of moving a box depends on the terrain.

3. The Twist: Light Has Two "Personalities"

In simple materials (like air or water), light's direction of travel and its wave direction are the same. But in anisotropic materials (like certain crystals), light splits its personality:

The Wave Normal: Imagine the "wavefront" as a ripple in a pond. The "normal" is a stick sticking straight up out of the water.
The Ray: This is the actual direction the energy is flowing. In these special crystals, the energy might flow diagonally while the wave ripple moves straight up.

The authors realized that to solve the "crowd movement" problem in these crystals, you have to account for both directions. They created a "Weighted Least Action Principle." Think of this as a new rulebook for the logistics company that says: "Don't just move the boxes; move them in a way that respects the weird, diagonal nature of the crystal."

4. The Solution: From Brightness to Shape

Here is the magic trick the paper describes:

Measure the Light: Take a picture of the light's brightness on a starting wall and an ending wall.
Run the Math: Use their new "Weighted Least Action" formula to calculate the most efficient path for the entire "crowd" of light to get from the first wall to the second.
Reconstruct the Wave: Once you know exactly how the light moved (the path of the rays), you can mathematically reverse-engineer the phase (the hidden shape/timing) of the wave.

It's like looking at the footprints of a crowd on a beach (the intensity) and being able to perfectly reconstruct the exact shape of the ocean waves that pushed them there, even if the sand was weird and slippery.

5. Why This Matters (According to the Paper)

The authors show that this method works for Maxwell's equations (the fundamental laws of electromagnetism) in complex materials. They provide specific mathematical formulas (cost functions) for common materials, such as:

Isotropic materials: Where light behaves normally (like glass).
Uniaxial materials: Crystals where light splits into two different behaviors.

In summary: The paper upgrades an old physics rule. Instead of guessing the path of a single, invisible light ray, it uses the measurable "brightness" of light at two points to calculate the most efficient path for the whole beam. By solving this "crowd movement" puzzle, we can finally reveal the invisible shape of the light wave itself, even when it's traveling through tricky, directional materials.

Technical Summary: Weighted Least Action Principle for Maxwell Equations

Problem Statement
The paper addresses the "phase from intensity" problem in the context of the geometric optics limit of Maxwell's equations within arbitrary (specifically anisotropic) media. While the Fermat principle of least time successfully determines the path of a single light ray given initial and final positions, a single ray is not a measurable physical object. In practical scenarios, one typically measures the intensity of a wave on two distinct planes ( $z=0$ and $z=h$ ). The challenge is to reconstruct the complete wave phase and the associated ray bundle connecting these planes using only these two intensity measurements.

Previous work established a "weighted least action principle" for scalar wave equations, linking the geometric optics limit to the Monge optimal transport problem. However, in scalar isotropic media, the wave normals (gradients of the phase) and the energy propagation rays coincide. In anisotropic media governed by Maxwell's equations, these two families of curves are distinct: the Fresnel wave normals ( $p = \nabla s$ ) are orthogonal to wavefronts, while the Fresnel rays ( $q$ ) are tangent to the Poynting vector and carry energy. The existing scalar theory does not directly apply to this anisotropic case where the transport vectors and phase gradients diverge.

Methodology
The authors extend the optimal transport framework to the vector-valued Maxwell equations by leveraging the Hamiltonian structure of the geometric optics limit.

Hamiltonian Formulation: The limit is characterized by two surfaces: the Fresnel surface of wave normals $H(p, x) = 0$ and the dual Fresnel surface of rays $H^*(q, x) = 0$ . The authors utilize the geometric connection between the wave normal $p$ and the ray direction $q$ , specifically the relation $q = \nabla_p H / (p \cdot \nabla_p H)$ , which implies $p \cdot q = 1$ .
Transport Equation: The energy transport is governed by $\nabla \cdot (qG) = 0$ . By projecting this onto the propagation axis $z$ , the authors define a wave action $a = Gq_3$ and a velocity vector $v = (q_1/q_3, q_2/q_3)$ . This allows the transport equation to be rewritten in a form suitable for optimal transport between the intensity distributions $I_1$ and $I_2$ on the two planes.
Legendre Transform and Action Definition: The authors define a cost function based on the Legendre transform of the Hamiltonian. Specifically, they define $D(p, x)$ such that $p_3 = D(p, x)$ , and its dual $D^*(v, x)$ . The action $Q(x_1, x_2)$ for a ray connecting two points is defined as the minimum integral of $D^*$ along the orbit.
Variational Principle: A weighted action functional $M(T)$ is constructed by integrating the product of the initial intensity $I_1(x)$ and the action $Q(x, T(x))$ over the domain, subject to the constraint that the mapping $T$ transports $I_1$ to $I_2$ .

Key Contributions and Results

Derivation of the Principle: The paper derives a weighted least action principle specifically for the geometric optics limit of Maxwell equations in anisotropic media. It proves that the ray mapping $T^*$ , generated by integrating the characteristic equations of the Hamiltonian system, is the minimizer of the Monge functional $M(T)$ .
Solution to Phase from Intensity: The authors demonstrate how to solve for the phase of an electromagnetic wave using two intensity measurements.
- In the homogeneous case (where $H=H(p)$ ), rays are straight lines. The optimal mapping $\bar{T}$ is found by minimizing the functional $M(T) = \int I_1(x) D^*(\bar{T}(x)-x) dx$ .
- Once the optimal mapping $\bar{T}$ is determined, the ray direction vector $q$ is recovered (since $v = (\bar{T}(x)-x)/h$ ).
- Using the Fresnel ray surface equation $H^*(q)=0$ , the full vector $q$ is determined.
- Finally, the wave normal vector $p$ (and thus the initial phase gradient $\nabla s$ ) is recovered via the dual relation $p = \nabla_q H^* / (q \cdot \nabla_q H^*)$ .
Explicit Cost Functions: The paper provides explicit expressions for the action $Q$ $Q$ and the cost function $D^*$ $D^{*}$ for specific media configurations, including:
- General anisotropic media with a diagonal dielectric matrix $\varepsilon$ .
- Isotropic media (recovering the standard optical path length).
- Uniaxial materials, where distinct expressions are derived for the two sheets of the Fresnel surface.

Significance
The paper claims that its primary significance lies in generalizing the weighted least action principle from scalar wave equations to the more complex vector-valued Maxwell equations in anisotropic media. This is crucial because, unlike in isotropic media, the energy transport rays and phase normals are distinct in anisotropic settings. By establishing the equivalence between the geometric optics limit and an optimal transport problem involving Fresnel rays, the authors provide a rigorous variational framework to reconstruct the electromagnetic phase from intensity data.

The authors note that while the solution to the Monge problem (finding the optimal map) is unique under the convexity of the dual function, the physical realization of the illumination problem may admit multiple solutions corresponding to different sheets of the Fresnel surface. The theory is presented as applicable not only to electromagnetism but also to other anisotropic dispersive vector-valued wave problems, such as linear elasticity. The work clarifies the historical connection between Monge's transport problem and optics, noting that while early suggestions linked them via a simple Euclidean cost function, the correct cost function is determined by the specific Hamiltonian structure of the wave equation.