🔬 mesoscale physics

Depth-Resolved Lattice Distortions in a Silicon-Germanium Qubit Host

This paper utilizes high-resolution X-ray nano-structural mapping to demonstrate how growth-induced lattice defects and crosshatch patterns in Intel Si/SiGe heterostructures propagate through the device, creating permanent strain that directly impacts the energy spectra and performance of silicon-germanium qubits.

Original authors: Jonathan C. Marcks, E. S. Joseph, J. Reily, Talise Oh, Abigail Postlewaite, Tao Zhou, M. A. Eriksson, Mark Friesen, Martin V. Holt

Published 2026-07-30

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Jonathan C. Marcks, E. S. Joseph, J. Reily, Talise Oh, Abigail Postlewaite, Tao Zhou, M. A. Eriksson, Mark Friesen, Martin V. Holt

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to build a super-fast, super-quiet computer, but instead of using silicon chips like your phone, you are trying to trap tiny, spinning electrons to act as the brain's memory. This is the world of quantum computing, a field where scientists are trying to harness the weird rules of the very small to solve problems that would take today's supercomputers thousands of years. The star player in this game is the "qubit," a quantum bit that can be in two states at once. To make these qubits work, scientists often use silicon, the same material in your computer chips, because it's clean, quiet, and we know how to build with it. But to trap the electrons, they have to sandwich a thin layer of silicon between layers of a mix called silicon-germanium (SiGe). Think of it like making a perfect, flat sandwich where the bread is slightly different from the filling. The problem is that when you grow this sandwich in a lab, the layers don't always fit together perfectly. They get wobbly, they stretch, and they develop tiny wrinkles and bumps, kind of like a rug that was rolled up too tight and then unrolled. These imperfections can mess up the delicate spin of the electron, causing the computer to make mistakes. So, the big question is: how bad are these wrinkles, and do they actually ruin the sandwich?

This paper is like a high-tech X-ray vision session that peeks inside a silicon-germanium sandwich made by Intel to see exactly how those wrinkles form and where they go. The researchers used a super-powerful microscope that uses X-rays to look at the crystal structure of the material with incredible precision—about 30 nanometers wide (that's roughly 3,000 times thinner than a human hair) and able to see changes in depth down to 200 nanometers. They found that the "wrinkles," which scientists call a "crosshatch pattern," start deep in the bottom layers of the material and travel all the way up to the top where the qubit lives. It's as if a ripple in the bottom of a swimming pool travels all the way to the surface, distorting the water right where you are trying to float. The paper shows that these ripples create a bumpy landscape for the electrons, changing the energy levels they need to function.

The team discovered that these distortions aren't random; they are linked to how the material was grown and the slight angle of the starting crystal. They mapped out the "tilt" of the crystal planes and found that where the crystal tilts more, the material is more strained, creating a bumpy energy hill for the electrons. This is a big deal because if the energy landscape is too bumpy, the electrons might get stuck or jump around when they are supposed to stay put, which could cause the quantum computer to fail. The researchers also found that these bumps create tiny "steps" on the surface of the quantum well, like a staircase where there should be a smooth ramp. These steps can mess up the electron's behavior, potentially making the computer less reliable.

Interestingly, the paper also looked at whether these bumps were the main reason for another problem called "valley splitting" (a specific energy gap that helps the qubit work). After comparing their X-ray maps with previous data, they found that the bumps didn't seem to be the main culprit for the variations in valley splitting; instead, it was likely caused by the random mixing of atoms in the material itself. So, while the wrinkles are real and do distort the landscape, they aren't the only thing to worry about. The authors suggest that as we try to build bigger quantum computers with more qubits, we can't just hope to find a perfectly flat spot in the material. We will need to either design the devices to avoid these bumpy areas or figure out how to grow the material so the wrinkles don't form in the first place. This study gives us a clear, 3D map of the problem, helping engineers know exactly what they are up against as they try to build the quantum computers of the future.

Technical Summary: Depth-Resolved Lattice Distortions in a Silicon-Germanium Qubit Host

Problem Statement
Semiconductor spin qubits, particularly those based on silicon quantum dots (QDs), rely on the host material's crystal lattice for their operation. While silicon-based systems offer advantages such as low spin-orbit coupling and industrial manufacturing compatibility, the growth of Silicon-Germanium (SiGe) heterostructures necessary to confine electrons introduces significant structural challenges. The lattice mismatch between Si and SiGe necessitates the introduction of dislocations to relieve strain. These dislocations often bundle along specific slip planes, creating a "crosshatch pattern" (CHP) characterized by surface height variations and lattice plane tilts. This structural disorder leads to inhomogeneous strain and lattice tilt, which distort the local band structure. Since qubit energy levels are determined by this local band structure, such distortions are expected to cause fluctuations in qubit energy spectra, potentially leading to device-to-device variability, qubit failure, and increased operational overhead. Previous characterization methods lacked the simultaneous lateral and depth resolution required to map these defects specifically within the active quantum well of an industrially manufactured device.

Methodology
The authors employed Nanoscale X-ray Diffraction Microscopy (nXDM) to characterize an industrially manufactured Si/SiGe heterostructure from Intel's "Tunnel Falls" spin qubit processor. The experimental setup utilized a focused 9.5 keV X-ray beam with a 30 nm lateral spot size. By scanning the beam across the region of interest and collecting diffraction patterns with a pixelated X-ray photon detector, the team mapped the local lattice vector, including lattice constants, strain, and rotation.

A key methodological innovation was the use of "functional depth resolution." Because the SiGe buffer layer is graded (increasing Ge concentration from 0% to ~31%), the lattice constant varies with depth. By defining "virtual detectors" over specific angular ranges ( $\Delta 2\theta$ ) in the diffraction signal, the authors could isolate diffraction signals corresponding to specific Ge concentrations and, consequently, specific depths within the heterostructure. This allowed for depth-resolved mapping with approximately 200 nm vertical resolution, circumventing the limitation of the large X-ray penetration depth. The study focused on correlating these lattice distortions with the physical location of a qubit device defined by gate electrodes.

Key Contributions and Results

Depth-Resolved Lattice Mapping: The study successfully mapped lattice tilt ( $\Omega$ ) and strain ( $\epsilon_{zz}$ ) throughout the graded SiGe buffer layer and the virtual substrate. The authors observed that lattice tilt features (specifically $\gamma$ tilt) emerge and propagate through the buffer layer, correlating with the formation of the final crosshatch pattern.
Correlation of Strain and Tilt: At the device scale (~1 µm), the authors found a strong spatial correlation between lattice tilt and c-axis strain. Specifically, regions of higher lattice tilt exhibited more negative (compressive) strain, with a correlation decay length of approximately 820 nm. Curvature ( $\kappa$ ) also showed a positive correlation with strain, suggesting that CHP features bundle strain, creating localized regions of low tilt and small strain.
Fine Structure Identification: Beyond the macroscopic crosshatch, the high-resolution imaging revealed "fine structure" in the lattice tilt (sub-1 µm features) that is typically obscured by the larger CHP. These fine features contribute to local strain relaxation and atomic step formation.
Impact on Qubit Energy Landscape: The measured strain was translated into shifts in the conduction band minimum ( $\Delta E_C$ ). The authors calculated energy gradients ( $|\nabla E_C|$ ) reaching up to 68 $\mu$ eV/ $\mu$ m. Given that valley splitting energies in SiGe can be around 100 $\mu$ eV, these gradients imply that energy fluctuations over the length of a moderate-sized device could be comparable to the valley splitting energy, potentially affecting electron shuttling fidelity and exchange coupling.
Atomic Step Density: The authors calculated the atomic step density at the quantum well interface based on the integrated lattice tilt. They found step densities up to 6.6 steps/ $\mu$ m. At the scale of a few quantum dots, this implies that multiple dots may be intersected by atomic steps, a condition known to significantly reduce valley splitting.
Valley Splitting Analysis: The authors compared their measured lattice distortions against previously measured valley splitting data for the same device. They found no direct correlation between the measured lattice tilt/strain and the observed valley splitting fluctuations, concluding that alloy disorder in the quantum well remains the dominant factor for valley splitting in this specific device, likely due to the device sitting in a relatively flat strain region and the broadness of the quantum well interface.

Significance and Claims
The paper claims to provide the first depth-resolved nXDM mapping of an industrially manufactured Si/SiGe qubit host with sub-micron lateral resolution. By linking the growth-induced evolution of lattice tilt and strain directly to the quantum well, the study demonstrates that structural imperfections propagate from the buffer layer to the active device region.

The authors modestly conclude that while individual few-qubit devices might be situated in low-strain regions, scaling to larger processors with components like electron shuttlers will inevitably encounter these inhomogeneities. The work suggests that future device design may need to account for crosshatch features or develop new growth strategies to guide dislocation bundles. Furthermore, the quantitative dataset provided serves as a benchmark for future theoretical modeling of heterostructure effects on spin qubit performance. The study emphasizes that fully correlating heterostructure features with qubit properties will ultimately require cryogenic qubit characterization with resolution comparable to the nXDM measurements presented.

More like this