Insight Media’s Chris Chinnock talks with Epsilon Photon’s CEO Ashutosh Patri about many aspects of metaoptics design, utility, fabrication, challenges and opportunities.

Q: Hi Ashutosh. Thanks for agreeing to talk with me today.

A: Hello Chris. I am glad we can have this chat.

Q: Let me begin by asking you if you see metasurfaces fully replacing conventional optics?

A: I wouldn’t characterize it that way, at least not in the near term. From my perspective, metasurfaces are more likely to be an addon to conventional optics rather than a direct replacement. There are still many challenges that need to be solved. Over time, metasurfaces may displace certain optical components, but for now they complement conventional approaches rather than supplant them.

Q: What do you see as the most critical technical challenges today?

A: The first major challenge is scale—specifically, making large‑aperture metasurfaces. If you imagine a metasurface lens for something like a DSLR camera, you immediately encounter problems. Controlling wavefront error across a large aperture is very difficult. You need extremely high uniformity, and the optical performance must be consistent across different incident angles. Achieving all of that simultaneously is challenging both in design and in fabrication.

Achromatic performance and operational bandwidth are among other major challenges. Today, most metasurface designs operate efficiently over bandwidths of about 50 to 100 nanometers. Expanding that to 300 or 400 nanometers with high efficiency, or even to 1,000 nanometers, is extremely difficult. Conventional optics can already achieve that kind of broadband performance, but metasurfaces are still far from that level of maturity.

Q: Is this primarily a design problem or a manufacturing problem?

A: It’s both, but manufacturing becomes the dominant constraint as you scale. Design challenges may be addressed over time with better simulation tools, inverse design methods, and high‑performance computing. In fact, design at the centimeter scale is already feasible computationally.

Manufacturing is a different story. If you move from a one‑centimeter‑diameter device to a ten‑centimeter‑diameter device, maintaining uniformity across the entire device is extremely difficult. Critical dimension uniformity across a full wafer requires many iterations, even with very advanced processing tools.

Q: How are metasurfaces typically fabricated today?

A: Electron beam lithography with a Gaussian spot is commonly used for prototyping. If you’re working with small structures or you want to test a new concept quickly, e‑beam lithography is very effective. But it is not scalable. For production, it simply takes too long—it could take days or even months to pattern a full wafer.

Q: What fabrication approaches are used for volume production?

A: There are two main approaches. One is UV lithography using stepper‑based tools, and the other is nanoimprint lithography. Nanoimprint lithography is becoming  a prominent manufacturing method in the metasurface industry. It can achieve very small features—down to around 60 nanometers—and it scales well to large volumes.

In nanoimprint lithography, you create a master mold, then use it to produce replica molds. Those molds are used repeatedly to imprint structures into a resist, which is then cured with UV light and the underlying high-index thin-film is plasma-etched. There is usually some trial and error at first, but once the process is optimized, it becomes very reliable.

Q: What kind of aspect ratios are achievable with these processes?

A: Aspect ratios of 10:1 are quite common, and in some cases, you can reach 20:1. The required aspect ratio also depends upon the refractive index of the structure material. The higher the refractive index the smaller size and aspect ratio of the structures needed to reach a certain phase change are. For many metasurface designs, ratios around 10:1 are sufficient. These structures aren’t trivial to fabricate though, but they are definitely achievable with current NIL processes.

Q: There’s often concern about master molds wearing out over time. Is that a real issue?

A: Not really, because the master is rarely used directly. You create a master once—often using electron beam lithography—and then you generate replica molds from it. The master is stored safely and used as little as possible, since it’s an expensive component.

Each replica mold can typically be used around tens to hundreds of times before it’s replaced. In some NIL workflows, you can even make replica molds from replica molds. This branching approach allows you to preserve the master almost indefinitely while still supporting high‑volume production.

Q: Why not use the master mold directly?

A: Masters are made from very hard materials like silicon or quartz. Pressing a hard master directly against another hard wafer is risky. Any wafer bowing or non‑planarity can lead to poor imprint uniformity or damage. Replica molds are made from more flexible materials, which allows them to conform to slight variations in the wafer surface and produce more uniform results.

Q: There has been interest in new deep‑UV tools for making NIL masters. Do you see a need for that?

A: I’m not entirely convinced. Today, NIL masters are typically fabricated using established electron beam techniques—multi‑beam systems, variable‑shaped beams, and character projection. These methods are well understood and widely available through specialized suppliers.

Unless a company wants to bring master fabrication fully in‑house, I don’t see a strong need for entirely new equipment. The real question is not just feature size, but critical dimension uniformity and mean-to-target error across the entire master. Those metrics are what ultimately matter for manufacturing.

Q: What questions would you ask a supplier offering new fabrication equipment or services?

A: The primary concerns are feature size, Critical Dimension (CD) uniformity, and mean-to-target error. Material choice also matters—silicon versus quartz—and that depends on the etch depth and structural requirements. If the goal is mass manufacturing, then repeatability becomes critical. A slightly imperfect result that is consistent is much better than a process that produces excellent results occasionally but varies run to run.

Cycle time is also important, particularly for rapid iteration during development.

Q: Your team focuses on low‑aspect‑ratio metasurface designs. Why?

A: High‑aspect‑ratio structures are difficult and expensive to manufacture, and improving yield requires multiple iterations. Traditional metasurface designs often rely on tall, single‑mode waveguides to achieve full phase coverage, especially when working with low‑index materials like glass.

We take a different approach. We design metasurfaces using low‑aspect‑ratio structures while preserving optical performance by implementing different physics within the nanostructures themselves.

Q: How does that work from a physical standpoint?

A: Conventional designs treat each nanostructure as a single‑mode waveguide. Phase control comes from varying the height and width of that structure. In our case, we deliberately design nanostructures that support multiple guided modes.

These modes propagate within a single, wider nanostructure and interfere with one another in a controlled way. That internal modal interference allows us to engineer the phase response without relying on extremely tall structures.

Q: Does that make the metasurface multimode at the system level?

A: No. Although each nanostructure supports multiple guided modes internally, these modes are only used to synthesize the desired phase response at sub-wavelength or near-wavelength scale. The metasurface as a whole depends on the designed macroscopic phase profile across the full device aperture. The modal interference occurs entirely within the nanostructure. Once the light exits, the resulting wavefront is clean, coherent, and well defined.

Q: How is this different from macroscopic interference‑based (super critical or super-oscillatory) optics?

A: The key difference is scale. In interference-based optics, interference occurs between freely propagating beams or modes in space and the resulting interference pattern depends on the full phase profile of the device aperture. In our case, the interference happens inside the nanostructure itself. The complexity is confined to the nanoscale, while the overall optical behavior of the metasurface is still dictated by the macroscopic phase profile you design across the device aperture.

Q: Is this approach more manufacturing‑friendly?

A: Yes. By reducing aspect-ratio requirements, we improve yield and reduce process complexity. That makes the designs more practical for real manufacturing environments, where repeatability and cost matter just as much as raw optical performance.

Thank you Ashutosh for this informative discussion.