

A hitchhiker's guide to dielectric cavities*
Light's growing weightYou don't need a great deal of imagination to foresee an increasing significance of lightwave technology in data processing and telecommunications. Here are some arguments in favor of light:Miniaturization of electronic circuits leads to increased resistances and hence larger dissipation. Photons don't suffer from losses in the same degree because their interaction is much weaker than that of electrons. The speed of light in a dielectric is much larger than the propagation speed of electronic signals in metals. The bandwidths available for signal transmission are a few hundred kHz on copper cables, versus roughly a THz in a typical glass fiber  even now it is feasible to carry half a million telephone conversations over a single glass fiber. Photons are the method of choice for massively parallel data processing and storage. More could be said, but this has already been done by others: See this online article on microphotonics in general, and an article describing my own field of work in the photonics industry since May 2000. At the heart of these developments is the availability of small but efficient lasers which deliver the required intense and coherent light.
Microlaser designAll of us (physicists) have probably been "exposed" to the HeNe laser in some graduate student lab. But of course the most ubiquitous lasers are by now the semiconductor diode lasers. Both of these incarnations rely on the parallelmirror configuration to provide the feedback that makes laser action possible. This type of resonator is also known from the FabryPerot interferometer.Trapping light with interferenceOne common way of making especially good parallel mirrors is to use Bragg reflection at multiple layers of dielectric films. The current state of the art in low dimensional photonic systems is shown here at Uni Würzburg, where "VerticalCavity Surface Emitting Lasers" are made using Bragg mirror stacks. The Bragg principle is based on the destructive interference between waves in successive layers of a stack of dielectric layers.As a rather logical continuation of the same principle, one has progressed to photonic crystals which employ the Bragg principle in more than one spatial direction and can in principle be used to make extremely small photonic cavities. The price one pays is that one needs many periods of the artificial crystal lattice in order to obtain high reflectivities, so that the total size of the structure ends up being much larger than the cavity itself. Higher and higher reflectivities are required, on the other hand, if one wants to make a laser out of such a microcavity. The simple reason is that a small cavity can host only a small amount of amplifying material, and therefore it becomes more difficult for amplification to win over the losses in a microcavity laser. Whisperinggallery resonators  trapping without interferenceIn solidstate laser materials, it is often possible to realize the mirrors simply by exploiting total internal reflection at the interface between the highindex solid and the surrounding medium (e.g., air). In contrast to the Bragg principle, this confinement mechanism for light is to lowest order frequencyindependent and can therefore be called a classical effect  it can be described without explicit use of the wave nature of light, by using Fermat's principle.This is good because it means that a device based on this confinement mechanism will in principle be able to work over a very broad range of wavelengths  in stark contrast to photonic crystals. Nevertheless, one can use total internal reflection to make threedimensionally confined resonators with high frequency selectivity (or "finesse"), provided one can force wavefronts inside the cavity to interfere with themselves. This is achieved with the "whisperinggallery" resonator which is at the heart of the lowestthreshold lasers made so far. This low threshold becomes possible as a consequence of the small size that can be achieved with these resonators. They are essentially circular disks in which the light circulates around close to the dielectric interface. Such modes are especially low in losses. Semiconductors are far from being the only application of the whisperinggallery mechanism. The first laser resonators in the submillimeter size regime were made of liquid droplets containing a lasing organic dye. The highestquality optical microresonators have been achieved using fusedsilica spheres (i.e., glass). Although these materials have a refractive index closer to unity than a semiconductor, they still support whisperinggallery modes. In that context, they are often called morphologydependent resonances (MDRs). Both the semiconductor and the droplet realizations of the whispering gallery are illustrated on the cover of
Stable and unstable resonatorsOther mirror arrangements provide different advantages. In particular, there has been a considerable body of work employing concave or convex mirrors. E.g., concave mirrors separated by less than their radii of curvature added together, make a stable resonator in which light rays undergo focussing while being multiply reflected between the mirrors. Light can then be coupled out by making one of the mirrors slightly transparent.When the output couplig is small, the theoretical treatment of such a laser can often be performed by neglecting the leakage and hence assuming the existence of some orthogonal set of modal eigenfunctions.
If one wants to avoid the use of partially transparent mirrors (which need
to have very low losses for highpower applications), one alternative
design is the unstable resonator containing defocussing
elements [see the
exhaustive textbook by A.E.Siegman,
Lasers
(University Science Books, Mill Valley, CA (1986)]. Such unstable lasers differ from stable resonators in their mode structure: A set of welldefined bound modes is not available for the expansion of the laser field, because they all couple to the outside. Therefore, it has been necessary to use quasibound states in the calculations. Lasers are fundamentally open systems, so a description in terms of quasibound states seems only natural. These states are, however, not as familiar a tool as the usual squareintegrable eigenfunctions one knows from bound systems. Their properties are still a topic of current research.
Chaotic resonatorsAs an extention of the unstableresonator idea, one can think of two concave mirrors in a defocussing setup combined with some lateral (sideways) guiding of the light between the mirrors. A naive reasoning could be this:We want lasing from light spilling out near one of the mirrors, but we don't want the escape angle with the optical axis to be too large, hoping thereby to improve the spatial mode pattern (focussing). So we put additional mirrors along the open sides joining the mirrors. Now combine this idea with the use of dielectric interfaces as (partially transparent) mirrors, and one is lead quite directly to consider the socalled stadium resonator (or a generalization thereof).
Here is an illustration of the stadium shape and of how it scatters
an incident ray: It is taken from J.H.Jensen, J.Opt.Soc.Am.A 10 (1993).
To arrive at the idea of using a chaotic resonator cavity, one can either start from the unstableresonator concept as described above, or from the whisperinggallery design. We came from the latter direction. The argument leading to an oval dielectric resonator is simply that a circular whisperinggallery cavity does not have a preferred emission direction, owing to its rotational symmetry. In addition, one wishes to have a parameter with which the resonance lifetimes of the cavity can be controlled. This is achieved by deforming its shape. Confocal resonatorsInbetween stable and unstable resonators, there is another useful mirror configuration, called confocal. It has the advantage of creating a focussing effect inside the resonator, which in turn amounts to producing a smaller effective mode volume for the laser. Instead of the whole volume between the mirrors, it is possible to utilize only a smaller volume around the coinciding focal points of the mirrors. The ray pattern that forms in a confocal arrangement of two concave mirrors can sometimes take on the shape of a bowtie (depending on the shape of the mirrors). This wellknown configuration is found in etalons but also in lasers. The simplest confocal cavity would consist of two circle segments with a common focus. A less trivial example is the case of two confocal paraboloids, i.e., surfaces of revolution generated by opposing parabolas that share their focal point:The righthand picture shows two bowtie rays going through the focus. There are many other ray paths that never go through the focus, but they form caustics which are reminiscent of this basic shape. For a study if this type of (threedimensional) mirror configuration, see my work with Izo Abram's group at CNET, "Mode structure and ray dynamics of a parabolic dome microcavity". This is the manuscript: gzipped Postscript, PDF. The Cavity QED group of T. Mossberg at the University of Oregon provides some info on why the ensuing internal focussing is desirable: Cavity quantum electrodynamic effects allow to modify the rate of spontaneous emission of the laser active material. To that end, one has to go to small mode volumes. But as is pointed out in this web seminar, the cavity volume isn't necessarily what counts. With a focused ray pattern as in the confocal resonator, the light field is especially strong in only certain portions of the resonator, notably the focal point in the center. And that is where the desired strong coupling between the light and the active medium occurs. Bowtie laserNow we put all of the above together, but for the price of one...
The microcylinder laser shown here is not circular, but not a stadium shape, either. The stadium has fully chaotic ray dynamics, the circle has no chaos at all. This oval shape has a mixed phase space. As a byproduct of the transition to chaos which takes place with increasing deformation, a bowtieshaped ray path is born that does not exist below a certain eccentricity. This pattern combines internal and external focussing, and its lifetime is long enough for lasing because the rays hit the surface close to the critical angle for total internal reflection.
This is the world's most powerful microlaser to
date.
The basic ideas of our work are illustrated on picture pages &
What is chaos ? And what in the world is quantum chaos ?Chaos is not just chaosWe are talking here about deterministic chaos. The term refers to the fact that even simple classical systems governed by simple equations such as Newton's laws can exhibit highly irregular motion that defies longterm predictions. One example for such a simple physical system is the driven pendulum, which you can explore with this Java applet, including additional information.The driven pendulum is not autonomous. A standard autonomous system showing chaos is the double pendulum, which can be admired in this JAVA simulation. In Optics, there is a slight confusion of terminology about the concept of chaos, because it is traditionally found when people want to describe the statistical properties of a photon source. "Chaotic light" in that context has a much shallower meaning  it just means "random" thermal distribution of photons as it is found in blackbody radiation. Chaos in the deterministic sense already has a place in optics as well, but again we have to make a distinction to our work. In multimode lasing one can look at the temporal and/or spatial evolution of the laser emission and finds that the signal can become very irregular. By mapping this behavior onto an artificial (usually manydimensional) space, e.g. by a socalled timedelay embedding, one then sometimes finds that the system follows a trajectory on a "chaotic attractor". That's a type of structure one finds in dissipative nonlinear classical systems. This is what people have studied in nonlinear optics for a long time now  an example of a chaotic attractor for a laser can be seen here. There are many lists of chaosscience links, one example being the one at NIST. Chaos in billiardsIn the classical ray picture for our microresonators, the fact that boundaries are penetrable does not (to lowest order in the wavelength) affect the shape of the trajectories, and hence our internal ray dynamics is that of a nondissipative, closed system.The optical resonator in the ray picture is a realization of what mathematicians call a billiard. See this short article for an entertaining introduction to billiards. Only nonchaotic billiards are shown there: the circle and the ellipse (note that this math definition of a billiard doesn't conform with what we know from the local pub). But generic oval billiards display chaotic dynamics. To take the step into the world of chaotic billiards, follow this link to the polygonal and stadium billiard (among others). If you have any further questions about chaos, you may well find an answer at this informative FAQ site maintained by Jim Meiss. Further information, including a host of graphics and animations, is also available from the chaos group at the University of Maryland. Quantum ChaosQuantum chaos sounds like a contradiction in terms because linear wave equations such as the Schrödinger equation do not exhibit the sensitivity to initial conditions that gives rise to chaos. Nonetheless, classical mechanics is just a limiting case of quantum mechanics, just as ray optics is the limit of wave optics for short wavelengths. So one should expect "signatures of chaos" in the wave solutions. To find and understand these, semiclassical methods are indispensable.One of the pioneers of quantum chaos, Martin C. Gutzwiller, has written a beautiful introduction to this field in Scientific American. See in particular the third figure describing the central place of quantum chaos in our our understanding of quantum mechanics. An important lesson here is: Playing around with the simple standard systems, such as harmonic oscillators, we barely scratch the surface of what the classicalquantum transition really entails. If we want to go beyond pedestrian descriptions of this transition, classically chaotic systems are where the action is! This also holds for muchdiscussed fundamental topics such as "decoherence", see the example of periodically "kicked" Cesium atom. As a byproduct, quantum chaos has brought together an arsenal of powerful techniques. My first chance to study these was a graduate course at Yale taught by Prof. Gutzwiller in 1993/94; he also accompanied my thesis work on chaotic optical cavities through discussions and as a reader at dissertation time. As it turns out, many of the intrinsic emission properties of dielectric optical resonators have a classical origin. The significance of this for quantum chaos is that comparison between ray model and numerical solutions of the wave equations uncover corrections to the ray model. Alternatively, one can also discover such wave corrections by comparing the ray predictions to an actual experiment. We follow both approaches.
Such wave corrections become especially interesting when the underlying
classical dynamics is partially chaotic, as is the case in the asymmetric
dielectric resonators. In that setting, two major new effects arise: In dielectric cavities, the effect of such phenomena on resonance lifetimes and emission directionality, and of course on resonance frequencies, can be studied. Emission directionality is in itself a completely new question to investigate from the viewpoint of quantum chaos: when decay occurs, e.g.,in nuclear physics or chemistry, any anisotropy of the individual process is averaged out in the observation of an ensemble  but microlasers can be looked at individually, and from various directions. If they are bounded only by a dielectric interface, the emission pattern is determined by the phasespace structure. This is an important focus of my work: the shortwavelength asymptotics of systems that are chaotic and open. What this means is illustrated in a slightly different example on picture page . There, we studied the relation between resonance lifetimes and dynamical tunneling (since it involves tunneling into a chaotic portion of phase space, it is also called "chaosassisted tunneling"). Eric Heller's group provides some information on dynamical tunneling in a different system. A list of Quantum Chaos groups is found here.


Related information is found on the following
web pages:

© Jens Uwe Nöckel 2001Aug21