The 2D Crystal That Bends Light Like Nothing Else on Earth

Physicists at Delft University of Technology have found that a 2D ferroelectric crystal called CuInP2S6 — barely thicker than a few dozen atoms — bends and slows light more dramatically than any known material in the blue-ultraviolet spectrum. The discovery, published in Advanced Optical Materials, also reveals a previously unknown way to tune a material's refractive index simply by changing how thick it is, potentially placing a powerful new dial in the hands of photonic engineers.

Why Light-Bending Materials Matter

Modern telecommunications, quantum computing, and precision sensing all depend on devices that can steer, split, and shape beams of light with minimal energy on ultra-small chips. The key ingredient is a material that treats light polarized in one direction differently from light polarized in another — a property called birefringence. High birefringence in the violet and ultraviolet range is especially coveted: it underpins polarization optics for UV lithography used to etch microchips, next-generation ultraviolet communications, and quantum photonic circuits. Today's go-to materials — calcite, quartz, rutile — work reasonably well, but they are bulky, difficult to integrate into chip-scale devices, and their optical response cannot be switched electronically. Researchers have been searching for a 2D alternative that can be fabricated onto a wafer like a semiconductor.

What Makes CuInP2S6 (CIPS) Special

CuInP2S6 — abbreviated CIPS — is a layered mineral in which copper, indium, and phosphorus atoms are sandwiched between planes of sulfur. It is ferrielectric: even at room temperature, each molecular layer carries a spontaneous electric polarization that can be flipped or modified by an applied field. The secret to its unusual optics lies in the copper ions. Unlike their neighbours, the Cu(I) cations are spatially restless — they can sit inside the sulfur cage, at an octahedral site, or hop out into the gap between layers. This ionic instability couples the crystal's polarization to its conductivity, and to the way it interacts with light. Crucially, this coupling grows much stronger as the flake gets thinner, because the crystal undergoes a structural phase transition below a critical thickness of roughly 90 nm, losing its in-plane polarization while retaining its out-of-plane polarization.

The Record-Breaking Measurements

The Delft team — led by Houssam El Mrabet Haje and Roald van der Kolk — exfoliated a bulk CIPS crystal onto a silicon wafer and then painstakingly thinned it in roughly 10 nm steps using a low-angle argon beam, pausing at each step to measure the material's optical constants with a variable-angle spectroscopic ellipsometer. Working across a wavelength range spanning deep ultraviolet to near-infrared (211 to 1688 nm), they mapped the refractive index and extinction coefficient in both the in-plane and out-of-plane directions at seventeen different thicknesses. At just 22 nm thick and a wavelength of 339.5 nm, the difference between the in-plane and out-of-plane refractive indices — the birefringence — reached 1.24. For comparison, calcite, the standard birefringent crystal, peaks around 0.17 in this wavelength range, and hexagonal boron nitride, the leading 2D contender, sits below 0.7. CIPS at 22 nm outpaces every known material intrinsically, without any nanostructuring or metasurface engineering.

A Thickness-Controlled Optical Switch

Perhaps more surprising than the record birefringence is a second finding: the refractive index of CIPS changes anomalously as a function of thickness across a wide range — from 22 nm up to about 170 nm — with shifts as large as 23.2% at a wavelength of 280 nm. This is not the gentle tapering expected from quantum confinement or surface effects; it is an abrupt, step-like modulation that tracks the crystal's internal polarization state. Raman spectroscopy measurements revealed that the vibrational signature of the copper ions shifts sharply at specific thicknesses, suggesting that the Cu(I) ions redistribute between lattice sites as the flake thins, altering the polarization and, in turn, the refractive index. Density functional theory calculations corroborated this picture: moving copper ions from the van der Waals gap into the interior of the sulfur cage reproduces the observed drop in refractive index. Because an applied electric field can also move those copper ions, this raises a tantalising prospect — tuning the refractive index of a chip-integrated CIPS layer electronically, on demand, without mechanical parts or temperature changes. Comparison experiments on LiNbO3, the workhorse of commercial electro-optic modulators, showed the same anomalous thickness dependence, albeit more modestly, suggesting the phenomenon may be general to ferroelectric materials.

What Comes Next for Photonics

The researchers are candid that a complete mechanistic account of the anomalous thickness regime remains to be written — the interplay of polarization, ionic conductivity, and quantum confinement in CIPS is complex, and more experimental and computational work is needed. Nonetheless, the practical implications are already clear. CIPS flakes within the anomalous thickness window (22 to 170 nm) have an electro-optic coefficient comparable to bulk LiNbO3, meaning they could one day replace today's centimetre-scale lithium niobate modulators with devices just nanometres thick. The giant UV birefringence opens a path to miniaturised polarization optics for chip-scale UV lithography and ultraviolet quantum communication links. And because CIPS is stable at room temperature and can be exfoliated and stacked like other van der Waals materials, it is inherently compatible with the fabrication techniques already used to build 2D heterostructure devices.

• Record intrinsic birefringence — at delta-n of 1.24 in the near-UV, CIPS surpasses every known material, enabling thinner and more efficient polarization-control elements without metasurface tricks.
• Electrically tunable refractive index — the copper-ion mechanism suggests that applying a voltage across a CIPS flake could modulate its refractive index in real time, a long-sought capability for integrated electro-optics.
• Generalisable to other ferroelectrics — anomalous thickness-dependent optical behaviour also observed in LiNbO3 hints at a broader principle, potentially applicable across the ferroelectric materials family.

"Our work paves the way for a paradigm shift in electro-optics and photonics based on the CIPS family of materials." — El Mrabet Haje et al., Advanced Optical Materials, 2025.