Randomly Rotated Photonic Molecules Keep Topological Light on Track

Topological photonics has long promised a way to guide light that shrugs off imperfections, but a stubborn problem has shadowed the field: the mathematical machinery that certifies topological protection assumes a perfectly ordered crystal. Now a team at Tsinghua University, led by Yidong Huang with Xue Feng and Yongzhuo Li, has built a silicon photonic platform that deliberately breaks that order. Their valley photonic molecular crystals, described in a peer-reviewed study published on 1 October 2026, keep topologically protected edge transport intact even when every unit cell in the lattice is rotated by a random angle, with robustness demonstrated for disorder ranges up to 60 degrees. The result offers one of the clearest experimental demonstrations yet that topology can survive in systems where the usual momentum-space description no longer applies.

The inspiration for the work came from an unexpected corner of chemistry. In molecular topological insulators built on metal–organic frameworks, the individual molecular building blocks can rotate within the lattice, and those rotations measurably alter optical and mechanical behavior. What remained largely unexplored, however, was how such rotational freedom affects topology itself. The Tsinghua group saw an opportunity: if molecular rotation could be transplanted into a photonic crystal, it would provide a clean, controllable knob for studying how disorder of a very specific geometric kind interacts with topological states. Rather than treating disorder as a nuisance to be minimized, the researchers made it a central variable of the design.

The resulting architecture is elegantly simple to describe. In a valley photonic molecular crystal, optical molecules, each a small cluster of silicon structures engineered to support valley-hall-like physics, sit on fixed and periodic lattice sites, exactly as atoms do in a conventional molecular crystal. The twist is that each molecule retains a randomly distributed orientation. The lattice positions stay ordered while the rotational degrees of freedom are scrambled. This separation is what makes the platform so useful. It isolates rotational disorder from positional disorder, allowing the team to ask a precise question: how much random rotation can a valley-topological system absorb before its protected edge states fail?

Answering that question required abandoning the standard toolkit. Band theory and topological invariants such as Chern numbers are defined in momentum space, which only exists when the structure is periodic. Random rotations destroy that periodicity, so the conventional characterization of valley topology becomes inapplicable. The researchers responded by constructing an effective Hamiltonian approach that captures the disordered system through an averaged Dirac mass and an effective valley Chern number. In parallel, they employed a spectral localizer analysis, a real-space method that certifies topology without any appeal to translational symmetry. The agreement between these two frameworks gave the team confidence that the edge states they observed were genuinely topological rather than accidental.

Before fabricating anything, the group worked through the periodic case first. They showed that molecular orientation itself determines the topological properties of an ordered valley photonic molecular crystal, establishing a baseline from which disorder could be introduced. This step matters because it reframes orientation as an additional design dimension. In ordinary photonic crystals, the topology is fixed by the geometry of the unit cell; here, the orientation of each molecule becomes a degree of freedom that can, in principle, be programmed. The effective Hamiltonian then extends this picture to the disordered regime, where orientations vary from site to site.

The experimental validation took place on silicon chips, the workhorse platform of integrated photonics. Simulations showed light propagating along a Z-shaped interface even under substantial rotational disorder, tracing a path that a conventional waveguide would struggle to maintain in the presence of such geometric scrambling. Measured transmission spectra backed this up, retaining the high-transmission peaks that signal the presence of topological edge states. Crucially, these signatures persisted as the disorder range was pushed to 60 degrees, a level of global rotational randomness that would catastrophically degrade transport in non-topological structures of comparable design.

The study did not stop at confirming robustness. The researchers also examined the competition between topological edge states and Anderson localization, the phenomenon by which strong disorder traps waves and halts transport altogether. This regime, where protection and localization fight for dominance, is among the most actively debated areas in disordered topological matter. By tuning the rotational disorder range, the platform provides a way to watch that competition unfold in a controlled laboratory setting, with the added advantage that photonic systems allow direct imaging of the propagating fields.

One of the most consequential aspects of the work is its generality. The framework developed by the Tsinghua team is not tied to the specific molecular geometry used in their devices; it applies to all molecules with threefold rotational symmetry, the C3 symmetry class that underlies valley physics. That means the effective Hamiltonian and real-space characterization methods can be carried over to any C3-symmetric building block, whether photonic, electronic, or mechanical. The authors suggest the approach could inspire analogous studies in condensed matter systems, acoustic lattices, and cold-atom platforms, where controlled rotational disorder is equally within experimental reach.

The practical implications reach toward programmable topological photonics. Because molecular orientation acts as an additional handle on topological properties, future devices could conceivably encode synthetic gauge fields, including pseudo-electric and pseudo-magnetic fields, by patterning the rotational landscape rather than the physical geometry. Such programmability has been a long-sought goal in the field, since it would allow a single fabricated chip to realize many different topological phases. The Tsinghua demonstration, published under DOI 10.1007/s44519-026-00017-3 by the Chinese Society for Optical Engineering, lays the conceptual and experimental groundwork for that ambition on a silicon platform compatible with existing fabrication infrastructure.

Beyond the immediate results, the study signals a broader shift in how disordered topological systems can be characterized. The combination of an averaged Dirac mass, an effective valley Chern number, and spectral localizer analysis forms a framework that works entirely without structural periodicity, addressing a long-standing gap between the theory of topological phases and the messy reality of fabricated devices. As integrated photonic circuits grow more complex and fabrication tolerances become a genuine constraint, tools that certify topological protection in the presence of real-world disorder are likely to become indispensable. The valley photonic molecular crystal, born from a chemical analogy and validated on a silicon chip, may prove to be one of the templates that the next generation of robust and programmable optical platforms is built upon.

Subject of Research: Topological robustness of valley photonic molecular crystals under rotational disorder on a silicon platform

Article Title: Valley photonic molecular crystals

Article References: Valley photonic molecular crystals. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: topological photonics, valley Hall effect, photonic crystals, rotational disorder, silicon photonics, edge states, Chern number, Anderson localization, molecular crystals, spectral localizer, integrated photonics, Tsinghua University