Last week our Perspective article "Limitations and possibilities of topological photonics" was published in Nature Reviews Physics. As the title suggests, we address some overblown claims of topological robustness frequently made in the literature and clarify in which areas topological protection can play a useful role for applications in photonics.
We first thought of writing such an article in July 2023, in response to several papers somehow being published in high impact journals despite their central claims being based on a misunderstanding of the nature of topological protection and robustness in the systems they studied. For example, claims of "topologically enhanced" or "topologically protected" localization are generally unfounded, given that the localization length is generally determined by the width of the band gap, a non-topological quantity.
Another common problem we wanted to address was the frequent use of comparisons between trivial and non-trivial structures to claim various forms of topological "enhancement". Sadly, such claims also frequently appear in top journals. As we discuss in the article, such a comparison ends up being meaningless because trivial and non-trivial structures host modes with differing dimensionality. For example, in 2D structures the edge modes (localized along the 1D boundary of the system) will naturally give a stronger light localization than a trivial 2D structure without any edge states. However, there are many ways to create edge states that do not require complicated topologically non-trivial designs. What matters is whether unidirectional chiral edge states (which are unique to topologically non-trivial systems) offer some advantage compared to non-chiral states, appearing either as trivial edge states or, more simply, as bulk states of a one-dimensional system. This kind of fair comparison is surprisingly rare in the literature - the most prominent example I know of is the 2014 paper "Topologically Robust Transport of Photons in a Synthetic Gauge Field".
Unfortunately, this methodology was not widely adopted, and there was little progress on the hard problem of demonstrating quantitative performance enhancements of topological designs compared to state-of-the-art non-topological designs; for example, we had to wait until 2023 to see a rigorous comparison between scattering in valley Hall and non-topological photonic crystal waveguides. In this work, the non-topological W1 photonic crystal waveguide had lower scattering losses in the slow light regime.
Promoters of topological photonics may argue that such a comparison is also unfair, given that the W1 photonic crystal waveguide design is the result of years of testing, experimentation, and optimization, whereas the valley Hall design is much newer, with the potential for further optimization. This point brings me to the "possibilities" of topological photonics we discuss in our article: a topologically non-trivial band structure should not be the end of the design process. Rather, topological bands provide a unique starting point for further optimization, for example by guaranteeing the creation of localized modes near the middle of a band gap. Before the advent of topological band theory we did not have a systematic way to do this!
In the next phase of research in topological photonics, the focus will not be on demonstrating ever more exotic topological phenomena in increasingly more complicated setups. Rather, we should be aiming to integrate this new design tool with other approaches such as fine-tuning or inverse design to move from proofs of concept to genuinely better devices. Photonic crystal waveguides and fibers, integrated lasers, and frequency combs are three areas ripe for further breakthroughs, in my opinion. Watch this space for more on these topics!
