Skip To Content

OSA Incubator on Flat Optics: Day 1 Reflections

OSA Incubator on Flat Optics: Day 1 Reflections

Soon Wei (Daniel) Lim, OSA Student Member, Harvard University


Have you swam or surfed in an artificial wave pool recently? The calm lapping waves and towering walls of water in these pools are typically generated by mechanical paddles placed at the boundaries. These paddles are programmed to push on the water periodically. When appropriately coordinated, they can even induce an awe-inspiring spike of water that blasts tens of meters above the pool’s surface[1]. In all cases, one can control the water waves indirectly by acting appropriately on the boundary conditions (the paddles); the physics of wave propagation handles the rest of the dynamics and produces the complex surface structures enjoyed by swimmers and watersport lovers alike.

The same principles are at work in flat optics. Flat optics rests on a delightfully simple and appealing premise: finely control the properties of light on a surface, and you can decide how light behaves everywhere. Just as in water waves, physics handles the business of propagating these boundary effects of light over space and time to produce structured light. The flatness of flat optics refers to how these surfaces do not rely solely on large-scale curved interfaces in order to shape light. While your eyeglasses and expensive microscope objectives draw upon precision-manufactured curved interfaces (between glass and air, for instance) to focus or defocus light, flat optics can achieve this degree of light control and much more with an entirely flat surface. The secret sauce is the tiny nanostructures distributed over the boundary surface (which does not actually need to be a planar surface in reality) which gives us pixel-by-pixel control of the behavior of light. In addition to manipulating the intensity distribution (brightness) of light, flat optics opens up the complete set of light parameters to on-demand control; the phase, polarization, wavelength, dispersion (effect of changing the color of light), shape (in space and in time), and even angular momentum of ultraviolet to visible and terahertz light can be shaped by the appropriately-designed flat optic metasurface. Indeed, as the word “meta”, or “beyond” in Greek, suggests, the opportunities afforded by metasurfaces transcend the straightforward light-bending behavior of undecorated flat surfaces. Multifunctional flat optics, or optics which exert control over different behaviors simultaneously (focusing and polarization selectivity, for instance), is a convenient platform to miniaturize bulky and heavy optical components onto a single lightweight system and serve as the interface between free-space optics (i.e. light propagation outside a chip) and integrated photonics (where light remains on-chip).

All this promise is well and good, but most demonstrations of flat optics have been limited to controlled laboratory environments, small device sizes, and low-volume, costly manufacturing technologies. Over the next few days from 26 to 28 February 2020, The Optical Society Incubator on Flat Optics is convening to explore the possibilities and roadmap how to take flat optics technology to mass market applications. Standing at the cusp of a potential explosion of flat optics applications in devices such as augmented reality/virtual reality (AR/VR) displays, ultralight cameras, and optical computers, we hope that the open and frank discussions between academia, research administration, and industry enabled by this meeting will catalyze a revolution in optics around these highly customizable and controllable metasurfaces.

Join me over the next three days as we leaf through the past achievements of flat optics, debate promising directions, and craft the roadmap for the future in this OSA Incubator!

Hosts Federico Capasso, Harvard University, and Paulo Dainese, Corning Inc., welcome attendees.


Day 1: Flat Optics from Components to Systems

“When a technology is successful, it creates its own applications.”

Professor Federico Capasso from Harvard University, a co-host of this incubator and pioneer in the field, kicked off the meeting with an overview presentation entitled “Flat Optics Based on Metasurfaces.” In his view, the key goal of flat optics is the creation of a technology platform upon which a universe of applications can be envisioned and built. He strongly believes that such a promising technology platform will enable its most valuable applications only when it is explored thoroughly; we cannot fully predict what the eventual applications of the platform will be before that happens. For near term research and development, Professor Capasso posits that a hybrid system is currently a more practical approach that will be accelerated into high volume applications first. These systems comprise metasurfaces coupled to conventional refractive optics (i.e. curved material interfaces) to greatly expand the bandwidth of applications and shrink the footprint of optical devices. They consolidate the strengths of both refractive and flat optics platforms: the large propagation path length of refractive optics as well as the customizability and flexibility of flat optics, as demonstrated in a diffraction-limited broadband achromatic lens. Single optic systems are also promising; two other key devices presented were a compact depth-sensing camera and a camera sensitive to the full polarization variation of an image, both implemented with a single metasurface and a conventional camera sensor.

Alexandra Boltasseva, Purdue University, presents on machine learning in empowering photonics.


Next, Professor Alexandra Boltasseva from Purdue University discussed the role of machine learning in empowering photonics. By drawing on the case study of using Titanium Nitride metasurfaces to product selective emitters for thermophotovoltaic cells (i.e. metasurface devices that absorb radiation and re-emit radiation in a wavelength range that can be absorbed efficiently by a cell to generate electricity), she demonstrated how generative adversarial networks (GANs) and autoencoders can produce a set of highly optimal metasurface structure shapes while exploring a large parameter space. She also emphasized how physics-driven and problem-specific machine learning techniques can greatly speed up experimentation, demonstrating a two-order-of-magnitude speed up in identifying good single-photon emitters using a neural network classification model.

To wrap up the first session, Professor Xiangang Luo from the Chinese Academy of Sciences presented remotely on the topic of Flat Catenary Optics. In brief, an optical catenary is a specially shaped structure that manipulates the geometric phase of light, a property which is produced by the geometrical orientation of a structure. Professor Luo demonstrated how appropriately-designed optical catenaries can separate right and left-handed circularly polarized light, and generate structured light containing uncommon properties such as orbital angular momentum (OAM) and self-acceleration. Catenary optics can also be applied to achieve wide-angle performance, which Professor Luo exhibited by constructing a wide-angle thermal imaging camera with a one-inch diameter metasurface. Finally, Professor Luo demonstrated that the catenary shape of the electric field itself can be manipulated through surface plasmon polaritons so as to achieve high resolution photolithography.

During the discussion session following the presentations, the audience raised the issue of quality control and metrology. Professor Boltasseva described how manufacturing robustness could be built into the optimization process as a constraint, while Professor Capasso noted how the discrete few-level design of flat optics inherently provides performance stability with respect to stochastic variations in topology, unlike that of continuous shapes like aspheric lens profiles.

Another issue raised was about the level of background scatter present in the use of flat optics. In short, light sent to unwanted diffraction orders can contribute to a background signal and decrease the efficiency of the optical device. Professor Capasso emphasized that the influence of background scatter and performance efficiency is a matter that is strongly application-dependent. While flat optics will not fully replace refractive optics, since the efficiency of refractive and reflective optics (with appropriate coating layers) in traditional optics will likely remain superior for the foreseeable future. However, for specific applications that require multiple functions or smaller footprints, flat optics may prove invaluable.

Bernard Kress, Microsoft, presents on flat optical systems.


In opening up the second session, Dr. Bernard Kress from Microsoft emphasized that the community should take a system-level perspective to flat optics: focus on developing integrated planar optical systems instead of creating flat optics that directly replace traditional optical elements. In VR devices, for instance, replacing conventional lenses with flat optics does not significantly reduce the footprint of the device because the propagation distance in air remains unchanged. The direction of flat optics should be in the direction of completely redesigning optical systems to take on planar form factors, instead of incrementally updating existing topologies with newer optical components. This approach of planar optical devices is not new; miniaturized optical devices that fold the optical path into a planar system have been around since the 1990s. In fact, the cost of aligning multiple optical devices or re-aligning incorrect positions can far exceed the cost of the optical components themselves (remember the corrective optics required in the orbiting Hubble Space Telescope[2]?). Therefore, the key impact of flat optics will lie in being able to lithographically align device layers at the point of fabrication, eliminating any expensive alignment issues down the pipeline.

Stanford Professor Mark Brongersma then presented on flat optics for wavefront manipulation and AR/VR, demonstrating with multiple examples how bulky refractive and reflective optics can be replaced by optimized metasurfaces. The curved mirror in a VR headset design, for instance, can be entirely replaced by a special nanostructured mirror, substantially reducing the device volume while preserving performance. Professor Brongersma also exhibited an eye-tracking device that selectively diffracted near-infrared light while letting visible light pass through with minimal stray artifacts. This device drew upon a patterned ultrathin layer of Silicon as a high quality resonator to guide light in an orthogonal direction.

Finally, to wrap up day one of the Incubator, Rahul Trivedi from the Vuckovic group at Stanford University provided a concise but detailed introduction to the inverse design techniques used in developing photonic and metasurface devices. By treating such devices as optimization problems and developing computational tools that can leverage the explosion in computing power (especially GPUs), one can design a wide range of devices from grating couplers to dispersion-engineered resonators and scattering metasurfaces.

Day one of the Flat Optics Incubator concluded with a welcome dinner at a local restaurant. Thus far, the promise of flat optics has been very much reiterated in multiple contexts. Tomorrow promises a new deluge of novel ideas, applications, and frameworks over a full day of talks and discussions. I eagerly await what this precious intersection of visionary ideas, diverse backgrounds, and professional experience of the various distinguished incubator participants will bring. Check back tomorrow for more updates!

[1] https://youtu.be/iWKFPTgkpXo (90 ft. Vertical Spike Wave in Slow Mo, The Slow Mo Guys)

[2] https://hubblesite.org/mission-and-telescope/servicing-missions

Image for keeping the session alive