Access Official Technical PDF

Share Summary:LinkedInEmail

As data moves between processors and systems, optical interconnects can help meet the demand for high-bandwidth communication in a compact footprint. Silicon photonics provides a route to integrating optical components using silicon-compatible manufacturing infrastructure, but the performance of each component depends on materials and device design. Research into QCSE modulators based on Ge/SiGe quantum wells demonstrates how ultra-thin active stacks could support integration with sub-micron silicon photonics platforms.

The reported devices achieved an absorption contrast of 2.1 for a 1 V peak-to-peak swing using quantum-well stacks grown on approximately 100 nm strain-relaxed buffer layers. The result highlights a key engineering trade-off: thinner buffers can make optical coupling to silicon waveguides more practical, while defects and surface roughness can broaden the absorption spectrum and reduce the modulator figure of merit.

Why consider the quantum-confined Stark effect?

Germanium is already used in silicon photonics, including in photodetectors and electro-absorption modulators. Conventional Ge- and GeSi-based modulators can use the Franz–Keldysh effect, in which an applied electric field changes the material’s optical absorption. However, the indirect bandgap of Ge(Si) can limit the relationship between extinction ratio and insertion loss.

Ge/GeSi multiple-quantum-well structures offer another approach. In these stacks, thin Ge quantum wells are separated by SiGe barriers. Applying an electric field changes the energy of the quantized electronic transitions in the wells. This quantum-confined Stark effect (QCSE) shifts and alters excitonic absorption features, enabling a change in optical absorption with applied bias.

For a modulator, the useful quantity is not simply whether absorption changes, but how much it changes relative to the underlying absorption. The study evaluates this using absorption contrast, expressed as Δα/α, and also considers a figure of merit that relates extinction ratio to insertion loss. These measures help explain why material quality and device integration must be considered together.

Engineering a thin Ge/SiGe stack on silicon

The investigated structures used Ge quantum wells 14 nm thick, separated by 18 nm SiGe barriers. The quantum-well region was grown on strain-relaxed SiGe buffers, with five stacks compared across different buffer thicknesses and growth approaches. The thinnest buffers were approximately 100 nm, and the overall stack thickness was about 350–400 nm.

Growing the quantum wells on silicon involves managing strain. The buffer accommodates the mismatch between silicon and the SiGe layers, while the quantum-well and barrier layers are designed to be strain-balanced. The researchers used structural and materials characterization—including transmission electron microscopy, diffraction, and surface analysis—to assess layer structure, composition, strain, and defects.

Electrical access was provided by p-i-n test diodes, with the quantum-well region in the intrinsic part of the structure. Doping in the contact and buffer layers helps direct the applied electric field toward the active region. This matters because voltage applied to the device does not necessarily translate entirely into a potential drop across the quantum wells: thicker undoped material in the intrinsic region can reduce the field experienced by the active stack.

What thinner buffers change

A thinner buffer brings the active material closer to the silicon substrate and can help make coupling to passive silicon waveguides more compatible with compact integration schemes. In contrast, thick SiGe virtual substrates can create difficulties when coupling light to and from silicon-on-insulator waveguides, sometimes requiring a specialized platform or additional taper structures.

But reducing buffer thickness is not a purely geometric improvement. In the reported comparison, thinner buffers were associated with increased surface roughness and defect density. The resulting excitonic absorption features broadened, which affected the measured figure of merit. As buffer thickness decreased from 0.6 μm to 0.1 μm, the reported peak figure of merit fell from 3.1 to 2.1.

At the same time, the stack grown on the approximately 100 nm buffer achieved an absorption contrast of 2.1 for a 1 V swing. That result indicates that substantial field-dependent absorption can remain available in a much thinner structure, even as broadening creates a performance trade-off.

Reading the measurement results in context

The absorption response was measured through photocurrent in illuminated test diodes. Because the tunable laser did not cover the original room-temperature excitonic resonance, the devices were heated to 75 °C to shift the absorption spectrum into the measurement range. The researchers used the bias-dependent photocurrent response to track changes in absorption as the electric field altered the quantum-well transition.

Comparisons between stacks also required accounting for the electric field across the intrinsic region. A 1 V swing does not produce the same field in every design, so the reported figure-of-merit comparison rescaled voltage swings to match the field contrast experienced by the quantum wells. This is an important consideration when evaluating results across structures with different buffer thicknesses and diode geometries.

Integration implications for optical interconnects

The thin stack is significant because its total thickness is comparable to that of Ge-based Franz–Keldysh modulators already considered for silicon photonics. This makes future integration with sub-micron platforms more plausible and could avoid some of the coupling challenges associated with thick buffers.

However, photocurrent measurements characterize the material’s absorption response; they do not establish the full performance of a waveguide-integrated modulator. Transmission measurements are needed to account for additional system-level effects, including coupling loss between the silicon waveguide and modulator, metal-related loss, and free-carrier absorption in doped layers.

For engineers assessing the technology, the practical lesson is to evaluate the complete stack rather than absorption contrast alone. Buffer thickness, defect control, field distribution, optical coupling, and losses in the integrated device all influence whether a promising material response translates into a useful component.

Conclusion

Ultra-thin Ge/SiGe quantum-well stacks demonstrate a path toward QCSE modulators that are more compatible with compact silicon photonics integration. The reported absorption contrast of 2.1 at a 1 V swing on an approximately 100 nm buffer is promising, but the associated spectral broadening reduces the figure of merit compared with structures on thicker buffers. The next integration question is how to preserve the strong QCSE response while managing defects and confirming performance through waveguide transmission measurements. For optical-interconnect designers, the work illustrates the central trade-off between a thin, integration-friendly stack and the material quality needed for sharp, efficient modulation.

Access Official Technical PDF

About The Author