Hamiltonian Optics

From the simple observation that light travels in straight lines to the complex design of modern imaging systems, the path of light has always been a subject of deep inquiry. While geometrical optics offers a practical starting point, it raises a more fundamental question: what universal principle governs the journey of a light ray through varying media? The answer lies in a remarkable analogy, a bridge between the worlds of optics and classical mechanics that allows us to treat light rays not merely as lines, but as particles following a path of least action. This powerful perspective is the essence of Hamiltonian optics.

This article delves into this elegant formalism, revealing the hidden mechanical laws that guide the propagation of light. In the first chapter, ​​Principles and Mechanisms​​, we will journey from Fermat's principle of least time to the formulation of the optical Hamiltonian. We will uncover how Hamilton's equations not only reproduce fundamental laws like Snell's law but also reveal profound truths about symmetry and conservation, introducing concepts like optical momentum, phase space, and the conservation of etendue. Following this, the chapter on ​​Applications and Interdisciplinary Connections​​ will showcase the incredible reach of this framework. We will see how Hamiltonian optics provides a systematic language for understanding and correcting aberrations in lenses, how it becomes the basis for designing particle accelerators, and how it can even describe the path of starlight bending in the curved spacetime around a black hole. Through this exploration, we will discover that Hamiltonian optics is more than a tool; it is a unifying principle that reveals the deep structural harmony woven throughout the physical world.

Imagine watching a sunbeam slice through a dusty room. We see a straight line of light. We are taught from a young age that "light travels in straight lines." This is the starting point of what we call ​​geometrical optics​​. But a deeper scientific inquiry cannot be satisfied with such a simple statement. Why does it travel in straight lines? And what happens when it doesn't? What happens when light passes through the shimmering air above a hot road, or through a sophisticated camera lens?

The path of light, it turns out, is governed by a beautifully profound principle, first guessed at by Pierre de Fermat in the 17th century. He proposed that out of all possible paths light might take between two points, it follows the path that takes the ​​least time​​. This single idea, ​​Fermat's Principle​​, is the seed from which an entire forest of understanding can grow.

This "principle of least action," as variations of it are known, should sound familiar to anyone who has studied classical mechanics. The trajectory of a ball, the orbit of a planet—these are also governed by similar principles. This hints at a deep and surprising connection: perhaps we can describe the journey of a light ray using the same powerful machinery that was developed to describe the motion of particles. This is the central idea of ​​Hamiltonian optics​​. We are going to treat a ray of light as if it were a particle on a journey, and we are going to ask what rules, what "physics," this strange particle obeys.

To build this analogy, we need to translate Fermat's principle into the language of mechanics. In mechanics, we have the Lagrangian, which is then used to construct the Hamiltonian. Let's do the same for our light "particle." The "action" for a light ray is its optical path length, $S = \int n \, ds$, where $ds$ is an infinitesimal step along the path and $n$ is the local refractive index, the property of the medium that slows light down.

If we imagine our ray traveling roughly along a $z$-axis, we can describe its path by its transverse coordinates, say $y(z)$. The path length element is $ds = \sqrt{1 + (dy/dz)^2} \, dz$. The "time" in our analogy is the coordinate $z$, and the "velocity" is the slope $y' = dy/dz$. The optical Lagrangian, the quantity whose integral we want to minimize, becomes $L = n(y,z) \sqrt{1 + (y')^2}$.

From this Lagrangian, we can perform a standard procedure in classical mechanics called a ​​Legendre transformation​​ to arrive at the Hamiltonian, the true "master equation" of our system. Just as in mechanics, we first define a momentum conjugate to our coordinate. Here, the "transverse momentum" $p_y$ is defined as $p_y = \partial L / \partial y'$. After a bit of algebra, this transformation yields the optical Hamiltonian. For a ray moving in the $xz$-plane, it takes the wonderfully compact form:

Here, $x$ is the ray's transverse position, and $p_x$ is its "optical momentum." What is this momentum, physically? It's not mass times velocity. Instead, it is the refractive index multiplied by the direction cosine of the ray's angle with the axis—essentially, it measures the tilt of the ray, scaled by the local refractive index. This Hamiltonian now contains, in a compressed and elegant form, all the rules for the ray's propagation.

Having the Hamiltonian is like having the rulebook for a game. The rules for how the position $x$ and momentum $p_x$ change as the ray propagates along $z$ are given by ​​Hamilton's equations​​:

Let's pause to appreciate what these are telling us. The first equation says that the slope of the ray ($dx/dz$) is determined by how the Hamiltonian changes with momentum. The second equation says that the change in momentum (the "bending" of the ray) is determined by how the Hamiltonian changes with position. Since the Hamiltonian depends on the refractive index $n(x,z)$, this second equation is really saying that a ray bends when it encounters a gradient in the refractive index. So, in a uniform medium where $n$ is constant, $\partial H / \partial x = 0$, which means $dp_x/dz = 0$. The momentum $p_x$ is constant, and so is the slope $dx/dz$. The ray travels in a straight line! Our grand new formalism successfully reproduces the most basic rule of optics.

But it can do so much more. Let's test it on a classic problem: refraction at an interface between two different media, say with refractive indices $n_1$ and $n_2$. The interface is a plane at $z=0$. The refractive index $n$ changes at $z=0$, but for any given $z$, it does not depend on the transverse position $x$. According to our second Hamilton's equation, this means $dp_x/dz = 0$. The transverse momentum $p_x$ is conserved! It has the same value before and after the interface. By applying the first Hamilton's equation, this conservation of $p_x$ can be shown to be mathematically identical to Snell's Law. The Hamiltonian formalism hasn't just re-derived an old law; it has revealed a deeper truth. Snell's law is a consequence of the conservation of transverse optical momentum.

Now for a more interesting case. Imagine a special medium where the refractive index decreases linearly with height, say $n(x) = n_0(1 - \alpha x)$. This is a simplified model for phenomena like atmospheric refraction or mirages. What path does a light ray follow if it propagates mainly along the $z$-axis? Using Hamilton's equations in the paraxial limit, we find that the ray bends with a constant curvature. The resulting trajectory $x(z)$ is a perfect parabola: $x(z) = x_0 + z \tan\theta_0 - \frac{\alpha}{2} z^2$ (where $x_0$ and $\theta_0$ are the initial position and angle). This is exactly the same equation that describes a projectile thrown in a uniform gravitational field! The analogy is not just a loose metaphor; it is mathematically precise. The light ray "falls" in a GRIN (graded-index) medium just as a ball falls in a gravitational field.

One of the most profound ideas in physics, expressed in ​​Noether's theorem​​, is that every continuous symmetry of a system corresponds to a conserved quantity. If the laws of physics are the same here as they are there (spatial translation symmetry), then linear momentum is conserved. If the laws are the same now as they were then (time translation symmetry), energy is conserved.

This deep principle holds true in our optical world. We already saw a hint of it: when the medium was uniform in the $x$-direction (translation symmetry), the corresponding momentum $p_x$ was conserved. What if we have a different symmetry, like rotation?

Consider an optical fiber where the refractive index doesn't depend on the angle $\theta$ around the central axis, but only on the radial distance $r$. This is a system with rotational symmetry. Our Hamiltonian framework predicts that there must be a conserved quantity associated with this symmetry. Indeed, by analyzing the Hamiltonian in cylindrical coordinates, we find an "optical invariant" that remains constant along the ray's entire twisting path through the fiber. This quantity, the optical analogue of ​​angular momentum​​, is given by:

This is a beautiful and powerful result. It means that no matter how complex the path, this specific combination of the local refractive index, the ray's distance from the axis, and its angular velocity will not change. This is the guiding principle behind how light is confined in optical fibers. The beauty of the Hamiltonian approach is that we didn't have to painstakingly trace the ray; the conservation law fell right out of the symmetry.

So far, we have been following the lonely journey of a single ray. But a beam of light, like from a laser or a light bulb, is a vast collection—a swarm—of individual rays. How does the entire swarm evolve? To answer this, we need to think about the ​​phase space​​ of our system. This is an abstract space whose coordinates are not just the positions $(x, y)$, but also the momenta $(p_x, p_y)$. Every possible ray at a given plane $z$ is represented by a single point in this 4D phase space. An entire beam of light is a cloud of points, a volume in this space.

Here, we encounter another deep theorem from classical mechanics: ​​Liouville's theorem​​. It states that for a Hamiltonian system, the volume occupied by a cloud of points in phase space is conserved as the system evolves. The cloud may stretch, twist, and deform, but its fundamental volume does not change. The flow of points in phase space is "incompressible." For Hamiltonian optics, we can prove this explicitly by showing that the divergence of the flow in phase space is exactly zero.

This has a monumental consequence for optics. This conserved phase space volume, often called ​​etendue​​ (or étendue), is given by $\text{d}x\,\text{d}y\,\text{d}p_x\,\text{d}p_y$. By relating the optical momenta back to the physical angles of the rays, we find that the quantity $n^2 \, dA \, d\Omega$ is conserved, where $dA$ is the cross-sectional area of the beam and $d\Omega$ is the solid angle it occupies. This is the famous ​​law of conservation of etendue​​. It tells you that you can't focus a beam of light from a large, diffuse source (like the sun) onto an arbitrarily small point. If you decrease the area $dA$, the solid angle $d\Omega$ must increase proportionally. This law is the mortal enemy of anyone trying to build a death ray, but it is the foundation of all optical instrument design.

By combining this with the conservation of energy (the power in a beamlet is constant), we get an even more practical conserved quantity. The power is proportional to the radiance $L$ times the etendue. If both power and etendue are conserved (up to the $n^2$ factor), it must be that the ​​basic radiance​​, $L/n^2$, is conserved along any ray. This simple and elegant law governs the brightness of images formed by any optical system, no matter how complex.

The real power of a formalism reveals itself when building complex things from simple parts. In optics, we build telescopes, microscopes, and cameras from a series of lenses, mirrors, and sections of free space. The Hamiltonian framework provides a beautifully systematic way to do this.

Each optical element can be viewed as performing a ​​canonical transformation​​ on the ray's phase-space coordinates. It takes an input ray $(y, p_y)$ and maps it to an output ray $(Y, P_Y)$. For a simple thin lens of focal length $f$, this transformation is a sudden "kick" to the ray's momentum, while its position remains unchanged: $Y = y$, $P_Y = p_y - y/f$. Such transformations can be elegantly described by a single ​​generating function​​. For the thin lens, the transformation is generated by the function $F_2(y, P_Y) = y P_Y + \frac{y^2}{2f}$. This function contains all the information about the lens's action.

For many simple systems, particularly in the ​​paraxial approximation​​ (where rays stay close to the axis and at small angles), these transformations are linear. This means they can be represented by ​​matrices​​. Free space propagation over a distance $d$ has a matrix. A GRIN lens has a matrix. A thin lens has a matrix. To find the effect of a complex system, we simply multiply the matrices of its components in order. This reduces the art of optical design to the systematic science of linear algebra. The fact that these transformations are Hamiltonian imposes a strict constraint on the matrices: their determinant must be 1. This is the matrix manifestation of Liouville's theorem.

Perhaps the most breathtaking aspect of the Hamiltonian formalism is its universality. The same mathematical structure describes wildly different physical phenomena. Let's step outside of optics for a moment and consider the motion of charged particles, like electrons in an electron microscope or a particle accelerator. There, the particles are guided not by varying refractive indices, but by magnetic fields.

Can we apply our optical framework? Absolutely. In a uniform magnetic field, the paraxial motion of a charged particle can also be described by a Hamiltonian where the propagation distance $z$ acts as time. For a system with two non-interacting particles, there exists a conserved quantity, a bilinear "invariant" that links their positions and momenta. This is the direct analogue of the ​​Lagrange invariant​​ in optics, now modified by the presence of the magnetic field. The fact that the same mathematical skeleton underpins both light optics and charged particle optics is a stunning testament to the unity of physical law.

Finally, we must ask: where does this road lead? Is this ray-based picture the ultimate truth? No. It is a powerful and accurate approximation, but it is an approximation nonetheless. The ultimate reality is that of ​​wave optics​​. Light is fundamentally a wave. The beautiful, classical world of Hamiltonian optics emerges from the wavy quantum world in the limit of very short wavelengths, in a process analogous to how classical mechanics emerges from quantum mechanics.

The connection is made through Hamilton's own ​​characteristic functions​​, like the generating functions we saw earlier. The "point characteristic function" $V$, which gives the optical path length between two points, turns out to be nothing more than the phase of the wave propagator in the Fresnel diffraction theory of wave optics. The other characteristic functions, like the mixed function $W_2$, can be found from $V$ via a Legendre transformation, perfectly mirroring the relationship between different propagators in wave theory. The "particles" of light, our rays, are merely the paths of constant phase, the normals to the wavefronts. The entire magnificent edifice of Hamiltonian optics is, in the end, a sophisticated way of tracing the shadows cast by the underlying waves. And in that, it is not a lesser theory, but a bridge connecting two worlds, revealing the deep and harmonious structure of physical law.

Now that we have tinkered with the intricate machinery of Hamiltonian optics, let’s take it for a joyride. And what a ride it is! We have seen that this way of thinking, recasting the laws of light propagation in the language of Hamilton’s mechanics, provides a deep and elegant foundation. But the real fun begins when we use it. This is the part of the story where we discover that this framework is not just a neater way to organize old optical problems. It is a kind of skeleton key, a master principle that unlocks doors to rooms of the physical world we scarcely imagined were connected. Our journey will take us from the mundane to the magnificent—from the design of a better camera lens to the path of light around a spinning black hole.

The history of optics is, in many ways, an endless quest for the perfect image. Nature, however, seems to have a mischievous streak. Every time we try to focus light with a simple lens or mirror, we run into a zoo of imperfections called aberrations. Rays of different colors bend by different amounts, rays from the edge of a lens focus at a different spot from those at the center, and the image of a flat object becomes annoyingly curved. For centuries, lens design was a dark art, a craft of trial, error, and hard-won intuition.

Hamiltonian optics transforms this art into a science. It gives us a systematic calculus of imperfections. The elegant Hamiltonian function, which contains the entire physics of the system, can be expanded in a series. The first, simple term gives us perfect, grade-school optics. But the higher-order terms—those are the aberrations! Each term corresponds to a specific type of flaw: spherical aberration, coma, astigmatism, and so on.

The beauty of this is that we can now manipulate these terms mathematically. The formalism elegantly handles not just simple mirrors and lenses, but also exotic modern components like graded-index (GRIN) lenses, where the refractive index varies continuously within the glass. By sculpting this internal index profile, we can play one aberration off against another. The Hamiltonian framework allows us to calculate precisely how a specific index gradient, say $n(\rho)^2 = n_0^2 - \gamma^2 \rho^2$, gives rise to a specific amount of spherical aberration, allowing designers to correct for it with unprecedented control.

This perspective also reveals deep truths. It shows that some aberrations are fundamental. For any system of spherical lenses, there is an unavoidable, intrinsic curvature of the focal plane, described by the famous Petzval sum. In the Hamiltonian picture, this emerges not as a coincidence, but as a fundamental invariant of the system, a property as intrinsic as the materials themselves, which can be derived from the basic interaction of a ray with a single surface.

In modern optical design, this approach is taken to its logical conclusion using the language of Lie algebra. A complex lens system is seen as a sequence of mathematical maps—a 'kick' from a surface, followed by a 'drift' through space, another kick, and so on. Each map can be represented by a Lie operator. To understand chromatic aberration, for example, one can calculate how the final map changes as the refractive index shifts with wavelength. This powerful technique allows engineers to compute the chromatic properties of a complex thick lens system almost automatically, isolating the contributions from each surface and drift space.

Perhaps most elegantly, this algebraic viewpoint reveals that aberrations can interact. What happens when a system has both spherical aberration and a slight tilt? The two effects don't just add up; they "talk" to each other. Their interaction, described by a mathematical operation called the Poisson bracket, can generate entirely new aberrations, like coma. This tells us that the order of components in an optical system is critically important, as their combined effect is more than the sum of their parts. The Hamiltonian framework gives us the rules for this conversation between imperfections.

Here is where the story takes a dramatic turn. What if I told you that the giant, kilometer-long tunnels of a particle accelerator are, from a certain point of view, nothing more than enormous camera lenses? The "light" they focus is not photons, but beams of protons or electrons. And the "lenses" are not glass, but powerful magnets.

The profound connection is, once again, the Hamiltonian. The equations of motion for a charged particle zipping through a magnetic field can be cast in a form that is mathematically identical to the Hamiltonian for a light ray in a medium with a peculiar, direction-dependent refractive index. This means we can take our entire toolbox of optical concepts—focal length, aberrations, image planes—and apply it directly to the world of particle accelerators.

The magnets that steer and focus particle beams are never perfect. Just as a spherical glass surface creates spherical aberration, a simple quadrupole magnet has its own characteristic aberrations. More complex magnets, like octupoles, are often introduced deliberately to correct these flaws. Using the Hamiltonian formalism, an accelerator physicist can calculate the precise 'kick' a particle's momentum receives as it passes through, say, an octupole field. This analysis reveals nonlinear aberrations in the beam's trajectory that are a direct analogue to the third-order aberrations in a lens system. By understanding the 'magnetic aberrations', we can design systems of magnets to cancel them out, keeping the beam tightly focused over vast distances.

The analogy holds even in fantastically complex situations, such as a charged particle beam moving through a background gas (a GRIN-like medium) while also being guided by a powerful solenoidal magnetic field. The Hamiltonian handles it all, allowing us to calculate even higher-order aberrations, like the fifth-order terms that become critical in next-generation machines. What we call "beam optics" is truly just Hamiltonian optics in a different guise.

The power of this viewpoint extends from the subatomic to the cosmic. Einstein's theory of general relativity tells us that massive objects warp the fabric of spacetime, and light follows the "straightest possible paths," or geodesics, through this curved geometry. This sounds complicated, but an amazing trick emerges: we can describe the path of a light ray in a curved spacetime as if it were a ray in flat space, but one filled with a medium of a certain effective refractive index, $n_{eff}$. Gravity becomes an optical medium!

This isn't just a metaphor. It allows us to apply the principles of Hamiltonian optics to astrophysics. For instance, the phenomenon of gravitational lensing, where a massive galaxy bends the light from a more distant object, can be analyzed as light passing through a giant, cosmic GRIN lens.

The connection even lets us explore abstract mathematical worlds. The strange, non-Euclidean world of hyperbolic geometry, for example, can be visualized with the Poincaré disk model. Within this disk, the "straight lines" are circular arcs that meet the boundary at right angles. It turns out that these paths are exactly the trajectories light would take if the disk were a flat piece of glass with a refractive index profile of $n(r) = 2/(1-r^2)$. We can, in principle, build a physical object that visually reproduces the geometry of a purely mathematical space.

The ultimate test of any physical idea is to throw it at a black hole. Let's consider a light ray skimming the edge of a spinning Kerr black hole. The black hole's rotation is so powerful that it literally drags the fabric of spacetime around with it, an effect called frame-dragging. A ray of light gets caught in this cosmic whirlpool. The physics seems as far from a simple lens as one could imagine. And yet, if we analyze a bundle of paraxial rays orbiting the black hole, we find that a quantity known as the Lagrange invariant—a cornerstone of classical optics, $L = y_2 p_{y1} - y_1 p_{y2}$—is perfectly conserved. Think about that. A rule of thumb for 18th-century telescope makers remains steadfast as light itself is tortured by the gravity of a collapsed star. If that doesn't send a shiver down your spine, you may not have a pulse. It is a stunning testament to the power and universality of the Hamiltonian structure.

Having soared to the edge of the cosmos, let us now shrink our perspective down to the nanoscale. The same principles that guide stars and galaxies are now being used to design technologies on the scale of billionths of a meter. In the field of nanophotonics, scientists seek to control and guide light on computer chips, creating circuits that use photons instead of electrons.

One of the most exciting tools in this field is the Surface Plasmon Polariton (SPP). This is an exotic quasiparticle, a hybrid of light and a wave of electrons, that can be trapped on the surface of a metal like gold or silver. These SPPs act like a two-dimensional form of light, tightly bound to the interface. And how do they propagate? You guessed it—they obey the laws of Hamiltonian optics.

By engineering the dielectric material on top of the metal surface, for instance by giving it a slight, linear gradient in its refractive index ($n_d(y) = n_0 + \alpha y$), we can create a "2D GRIN lens" for plasmons. A beam of SPPs launched along this surface will not travel in a straight line, but will follow a predictable parabolic arc. We are literally bending light on a chip, using the same principles that guide the design of a camera lens. This opens the door to creating ultra-compact lenses, waveguides, and switches for the optical computers of the future.

So, what have we learned on this journey? We have seen the same set of ideas—the same Hamiltonian—appear again and again, describing the imperfections in a photograph, the focus of a proton beam, the bending of starlight by a black hole, and the guidance of a plasmon on a chip.

This is the real magic. Hamiltonian optics is not just a bag of mathematical tricks. It is a point of view. It is the recognition that a deep unity pervades the physical world, all flowing from the simple, yet profound, principle of stationary action. This principle dictates that physical systems will always follow a path of extremal "cost," whether that cost is time, action, or optical path length. The graceful arc of a light ray in a GRIN lens and the path of a photon skirting the abyss of spacetime are, in a fundamental sense, cousins. To see this unity, to appreciate this shared mathematical skeleton beneath the surface of seemingly different phenomena—that is the true gift of the Hamiltonian perspective. It reveals the inherent beauty and interconnectedness of the universe.