Modern integrated photonics is built almost entirely on the back of the silicon-on-insulator (SOI) platform. By leveraging the mature manufacturing infrastructure of the semiconductor industry, researchers can mass-produce complex optical circuits on microchips. However, this reliance on standard semiconductor materials introduces a fundamental physical bottleneck: spectral transparency constraints.
Every material possesses an intrinsic spectral window dictated by its atomic structure. At short wavelengths (the visible and ultraviolet spectrum), photons carry enough energy to kick electrons across the material's electronic bandgap, causing intense interband absorption. At long wavelengths (the infrared spectrum), photon energies match the natural vibrational frequencies of the chemical bonds within the crystal lattice, triggering strong phonon absorption.
Standard silicon photonics operates in the near-infrared telecommunications window (around 1550 nm) precisely because it sits in the quiet valley between these two absorption regimes. But move outside this narrow window, and the platform fails:
- In the visible and near-visible spectrum (<1000 nm): Silicon becomes highly absorptive. As David Sharp notes in Near-Visible Topological Edge States in a Silicon Nitride Platform, this absorption edge creates a severe spectral mismatch that prevents standard silicon circuits from interfacing with high-performance quantum emitters like quantum dots, which primarily operate in the visible range.
- In the mid-infrared spectrum (>2.4–4 µm): Standard silica ($\text{SiO}_2$) fiber claddings and silicon cores begin to absorb light due to oxide and semiconductor phonon bands. As highlighted by Kai Xia et al. in The generation of high-power, broadband supercontinuum spanning from visible to mid-infrared via incoherent beam combination in a GeO 2 -core fiber bundle, standard silica-based fibers exhibit crippling absorption beyond 2.4 µm.
This mid-infrared (MIR) blindness is particularly costly because the 3–12 µm window is the "molecular fingerprint region"—the spectral zone where vital chemical and biological molecules exhibit unique vibrational absorption signatures. To build compact, on-chip sensors capable of identifying environmental gases or medical biomarkers, researchers must bypass these material absorption limits.
Shifting the Phonon Edge with Chalcogenide Glasses
To extend the transparency window deeper into the infrared, we must alter the physical properties of the lattice itself. The resonant frequency $\omega$ of a molecular bond can be modeled conceptually as a simple spring-mass system:
$$\omega \propto \sqrt{\frac{k}{\mu}}$$
where $k$ is the bond stiffness and $\mu$ is the reduced mass of the vibrating atoms. To lower the resonant frequency—and thus push the phonon absorption edge to longer, less energetic infrared wavelengths—we must either weaken the bonds or increase the mass of the constituent atoms.
This is the core intuition behind chalcogenide glasses (ChGs). By replacing light oxygen atoms with heavier chalcogen elements from group 16 of the periodic table—such as sulfur (S), selenium (Se), and tellurium (Te)—the material's fundamental lattice vibrations are shifted deep into the far-infrared.
[Silica Glass (SiO2)] --> Light atoms (O) --> High vibrational frequency --> Absorbs past 2.4 µm
[Chalcogenide Glass (ChG)] --> Heavy atoms (Se) --> Low vibrational frequency --> Transparent up to 11 µm
As a result, compounds like Ge-Sb-S offer a massive transparency window spanning from 1 µm up to 11 µm. In Towards on-chip, integrated chalcogenide glass based biochemical sensors, Anu Agarwal et al. demonstrated that these glasses can be fabricated into high-Q microdisk resonators. Because these resonators do not suffer from the infrared absorption that plagues silicon, they enable cavity-enhanced infrared spectroscopy directly on a chip. Furthermore, the material's flexibility allows for advanced manufacturing techniques; Fan Zhang et al. showed in Fabrication of a Chalcogenide Glass Microlens Array for Infrared Laser Beam Homogenization that ChGs can be shaped into microlens arrays to manipulate high-power infrared laser beams without thermal degradation.
Dispersion Engineering for Mid-IR Broadband Light
Simply having a transparent material is not enough to build a practical sensor; you also need a broadband light source to probe multiple molecular absorption lines simultaneously. This is typically achieved via soliton microcomb generation, where a continuous-wave pump laser generates a coherent cascade of equidistant spectral lines inside a microresonator.
However, microcomb generation requires anomalous group velocity dispersion (GVD), a regime where higher-frequency (bluer) light travels faster than lower-frequency (redder) light. This anomalous dispersion is required to balance the nonlinear self-phase modulation that holds the soliton pulse together. Unfortunately, almost all optical materials naturally exhibit normal dispersion in the mid-infrared region.
To overcome this, researchers use geometric dispersion engineering. By tightly confining light within a waveguide or microresonator, the physical boundaries of the structure force the light to interact with the geometry in a way that counteracts the natural material dispersion.
Total Dispersion = Material Dispersion (Naturally Normal) + Waveguide Dispersion (Engineered Anomalous)
By carefully tailoring the cross-sectional dimensions of chalcogenide microresonators, the waveguide dispersion can be tuned to flip the total dispersion into the anomalous regime.
- Siqi Lu et al. utilized a specialized "sandwich" microresonator design in Broadband Mid-Infrared Frequency Comb in Integrated Chalcogenide Microresonator to precisely control this dispersion balance, achieving broadband mid-IR comb generation.
- To expand the bandwidth even further, Zifu Wang et al. developed a concentric dual-ring microresonator architecture in Engineered octave frequency comb in integrated chalcogenide dual-ring microresonators. By inducing mode hybridization between the two coupled rings, the authors created a highly customized dispersion profile that yielded an octave-spanning Kerr frequency comb stretching from the telecommunications band (1224 nm) all the way into the mid-infrared (2913 nm).
Substrate Engineering via Transfer Printing
While alternative materials like chalcogenides solve the transparency problem, abandoning silicon entirely means losing access to standard CMOS fabrication lines. Is there a way to keep the high refractive index and mature processing of silicon while eliminating its infrared limitations?
The bottleneck in a standard silicon-on-insulator (SOI) wafer is not actually the silicon core itself—silicon remains transparent up to roughly 8 µm. The true culprit is the underlying silicon dioxide ($\text{SiO}_2$) substrate cladding, which begins absorbing light heavily at wavelengths beyond 4 µm. When light guides through a silicon waveguide, its evanescent field extends into this oxide substrate, causing severe attenuation.
To solve this, Yu Chen et al. developed a physical transplantation technique detailed in Heterogeneously integrated silicon photonics for the mid-infrared and spectroscopic sensing. Instead of leaving the silicon devices on their native oxide substrates, the authors used a "transfer-printing" method to chemically etch away the underlying $\text{SiO}_2$ layer, lift the pre-fabricated silicon structures off the wafer, and stamp them onto an arbitrary, infrared-transparent substrate like calcium fluoride ($\text{CaF}_2$).
[Standard SOI Wafer] [Transfer-Printed Platform]
+------------------------+ +------------------------+
| Silicon Waveguide Core | | Silicon Waveguide Core |
+------------------------+ +------------------------+
| SiO2 Cladding (LOSSY!) | | CaF2 Substrate (CLEAR) |
+------------------------+ +------------------------+
By replacing the lossy oxide cladding with $\text{CaF}_2$, the authors completely bypassed the substrate absorption bottleneck, unlocking the 4–8 µm mid-infrared spectrum for silicon devices and enabling on-chip spectroscopic sensing with sub-nanogram sensitivity.
Non-Volatile Tuning and Wide-Bandgap PCMs
Beyond guiding and generating light, practical optical circuits require active components like switches and modulators. In reconfigurable photonics, this is often achieved using phase-change materials (PCMs) like $\text{Ge}_2\text{Sb}_2\text{Te}_5$ (GST), which can be electrically or optically toggled between amorphous and crystalline states to alter their refractive index.
However, traditional PCMs suffer from narrow electronic bandgaps, meaning they absorb a massive amount of light in their crystalline state. This high insertion loss severely limits their utility across broad spectral ranges, particularly in the visible and near-infrared bands.
To address this limitation, Zhuoran Fang et al. turned to a wide-bandgap PCM: antimony trisulfide ($\text{Sb}_2\text{S}_3$). In Non-Volatile Reconfigurable Integrated Photonics Enabled by Broadband Low-Loss Phase Change Material, the authors integrated $\text{Sb}_2\text{S}_3$ into silicon and silicon nitride circuits. Because $\text{Sb}_2\text{S}_3$ has a much wider electronic bandgap than traditional GST, its absorption edge is pushed out of the visible and telecommunication windows. This provides strong phase modulation with near-zero insertion loss, enabling energy-efficient optical switches that retain their state indefinitely without requiring constant holding power.
Bypassing Detector Limitations via Frequency Upconversion
Even if an optical chip successfully guides and modulates mid-infrared light, detecting that light presents a final, steep technological hurdle. Mid-infrared photodetectors typically require narrow-bandgap semiconductors (like Mercury Cadmium Telluride) that are expensive, prone to thermal noise, and often require bulky cryogenic cooling to function. In contrast, visible and near-infrared silicon cameras are incredibly cheap, highly efficient, and operate reliably at room temperature.
To exploit this asymmetry, researchers use nonlinear optical processes to translate mid-infrared signals into the visible spectrum before detection occurs.
One approach relies on nonlinear frequency upconversion on a metasurface. In Nonlinear Mid-infrared Metasurface based on a Phase-Change Material, Fuyong Yue et al. engineered a reconfigurable metasurface using a specialized phase-change material (GSST) that amplifies third-harmonic generation (THG). By mixing an incoming mid-infrared signal within the metasurface, the frequency is multiplied, shifting the light into a shorter wavelength band that can be imaged directly by standard, low-cost near-infrared cameras.
A more exotic, quantum-mechanical alternative bypasses classical frequency conversion entirely through quantum imaging with undetected photons. As reviewed by Hugo Defienne et al. in Advances in quantum imaging, a nonlinear crystal can be pumped to create a pair of entangled photons: a mid-infrared "signal" photon and a visible "idler" photon.
+-------------------+
| Pump Photon |
+---------+---------+
|
(Spontaneous Parametric Down-Conversion)
|
+---------------+---------------+
| |
v v
+-----------------------+ +-----------------------+
| Mid-IR Signal Photon | | Visible Idler Photon |
| (Interacts with Specimen) | (Goes to Si Camera) |
+-----------------------+ +-----------------------+
The mid-infrared photon is sent through the chemical sample being analyzed, while the visible photon is sent directly to a standard silicon camera. Because the two photons are quantum-mechanically entangled, measuring the quantum correlations of the visible photon reveals the absorption profile experienced by the mid-infrared photon. The mid-infrared photon is never actually detected; its spectral information is read entirely out of its visible partner, completely sidestepping the mid-infrared detector bottleneck.
Key Foundational Work
Mid-IR super-continuum generation (2009) — contributed 4 concept elements to this direction.
Towards on-chip, integrated chalcogenide glass based biochemical sensors (2010) — contributed 1 concept elements to this direction.
Integrating optics and micro-fluidic channels using femtosecond laser irradiation (2009) — contributed 1 concept elements to this direction.