🔬 mesoscale physics

Valley Splitting Correlations Across a Silicon Quantum Well Containing Germanium

This study demonstrates that valley splitting variations in an Intel-manufactured SiGe/Si/SiGe quantum dot array exhibit spatial correlations across both sub-100nm and micrometer scales, confirming that microscopic alloy disorder is the dominant factor and providing essential insights for designing scalable silicon-based quantum computers.

Original authors: Jonathan C. Marcks, Emily Eagen, Emma C. Brann, Merritt P. Losert, Talise Oh, J. Reily, Christopher S. Wang, Daniel Keith, Fahd A. Mohiyaddin, Florian Luthi, Matthew J. Curry, Jiefei Zhang, F. Joseph

Published 2026-02-05

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jonathan C. Marcks, Emily Eagen, Emma C. Brann, Merritt P. Losert, Talise Oh, J. Reily, Christopher S. Wang, Daniel Keith, Fahd A. Mohiyaddin, Florian Luthi, Matthew J. Curry, Jiefei Zhang, F. Joseph Heremans, Mark Friesen, M. A. Eriksson

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-precise digital clock using tiny, invisible marbles (electrons) trapped in a microscopic box (a quantum dot). For this clock to work perfectly, the marble needs to stay in one specific state. However, in the material used to build these boxes (silicon), there is a sneaky problem: the marble can accidentally slip into a "ghost" state that looks almost exactly like the real one. This is called valley splitting.

If the gap between the real state and the ghost state is too small, your clock gets confused, and the information gets corrupted. The goal of this research was to figure out how big this gap is across a long, industrial chip and whether the size of the gap changes smoothly or randomly as you move from one spot to another.

Here is a breakdown of what the researchers found, using simple analogies:

1. The Terrain is Rough (The "Alloy Disorder")

Think of the silicon chip not as a smooth, flat floor, but as a bumpy hiking trail made of a mixture of two types of rocks: Silicon and a tiny bit of Germanium. Even though the mixture is supposed to be uniform, the rocks are scattered randomly, like sprinkles in a cake.

The researchers found that these random "sprinkles" (alloy disorder) create tiny hills and valleys in the energy landscape. Because of this, the "ghost state" gap (valley splitting) isn't the same everywhere. Sometimes the gap is wide (safe), and sometimes it's narrow (risky).

2. The "Flashlight" Probe

To measure these gaps, the team didn't just look at one spot and stop. They used a clever trick called DAPS (Detuning Axis Pulsed Spectroscopy).

Imagine you have a flashlight beam (the electron) that you can slide back and forth along a 1.3-micron-long track (about 1/50th the width of a human hair). As you slide this flashlight under different gates (the "fingers" of the chip), you are essentially scanning the terrain.

The Discovery: When they slid the flashlight just a few nanometers (the width of a few atoms), the gap size changed dramatically. It was like walking from a sunny spot to a shady spot in just a few steps.
The Scale: They found that the "memory" of the terrain only lasts for about 19 nanometers. If you move your electron just a tiny bit further than that, the new spot has no relationship to the old one. It's like flipping a coin: the result of the flip 19 nanometers away has nothing to do with the flip you just did.

3. The "Long Walk" Across the Chip

After scanning the tiny distances, they looked at the whole 1.3-micron track, checking 21 different spots (like checking the weather at 21 different towns along a highway).

The Pattern: They found that while the gap size varies wildly from spot to spot, the variations aren't completely chaotic over long distances. There were some subtle patterns where the gap size seemed to "anticipate" or "echo" what happened further down the line.
The Surprise: However, when they ran computer simulations to see if this was a special "hidden order" in the material, they found something interesting: It wasn't special at all.
- The Analogy: Imagine you walk down a street and notice that every 6th house is red, and every 12th house is blue. You might think there is a secret pattern. But if you generate a random list of house colors on a computer, you will often see those same "patterns" just by pure luck.
- The researchers concluded that the patterns they saw on the big chip were likely just random coincidences caused by sampling a limited number of spots from a random distribution. The material is essentially a "random scatter" of bumps, and the patterns we see are just the natural noise of looking at a small sample.

4. Why This Matters (According to the Paper)

The paper doesn't promise a new product or a fix for tomorrow's computers. Instead, it provides a crucial "map" for future engineers.

The Problem: If you build a massive quantum computer with thousands of qubits, you need every single one to have a "safe" gap size.
The Finding: Because the gap size changes so quickly and randomly (every 19 nanometers), you cannot assume that if one spot is good, the spot next to it will be good.
The Conclusion: To build reliable quantum computers, engineers need to design materials and devices that can handle this inherent "roughness." They need to understand that the material is naturally uneven, and they must design their systems to work even when the electron lands on a "bumpy" spot.

In short: The researchers mapped the "energy terrain" of a silicon chip and found it to be a bumpy, random landscape where the safety of the quantum bits changes every few atoms. The patterns they saw across the whole chip were just random noise, not a hidden code, reminding us that nature is messy and we need to build our quantum computers to be robust enough to handle that mess.

Technical Summary: Valley Splitting Correlations Across a Silicon Quantum Well Containing Germanium

Problem Statement
Silicon quantum dots in SiGe/Si/SiGe heterostructures are a leading platform for electron spin qubits. However, the indirect bandgap and six-fold crystal symmetry of silicon create near-degenerate valley states. While strain lifts the in-plane valley degeneracy, the out-of-plane confinement typically results in a low-lying excited valley state with a small energy splitting ( $E_V$ ), often between 10 and 200 $\mu$ eV. This low-lying state degrades qubit control and state preparation and measurement (SPAM) fidelities. Theoretical work suggests that microscopic disorder in the SiGe alloy (random Ge atom distribution) is the dominant factor limiting $E_V$ . A critical challenge for scalable quantum computing is achieving a uniformly large valley splitting across mesoscopic distances (many-qubit devices), as local variations can create "hot-spots" of low splitting that reduce yield and impact electron shuttling schemes.

Methodology
The authors investigated valley splitting variations in an industrially fabricated Intel "Tunnel Falls" device containing a 1.3 $\mu$ m-long channel of a Si $_{0.972}$ Ge $_{0.028}$ quantum well. The device features a 1D array of 21 accessible quantum dots formed by plunger and barrier gates with a 60 nm pitch.

Device Modeling: The authors utilized the MaSQE simulator to model electron density, confirming the elliptical shape of the dots and the orientation of the first excited orbital state. They simulated 1000 valley splitting values based on Alloy Disorder-Dominated (ADD) theory to establish a theoretical Rayleigh distribution reference.
Measurement Technique: The study employed Detuning Axis Pulsed Spectroscopy (DAPS). This technique involves pulsing a double quantum dot (DQD) across a charge stability diagram to induce tunneling between source and target dots. By measuring the Single Electron Transistor (SET) current after a variable hold time, the authors extracted the valley splitting energy ( $E_V$ ) and orbital energy ( $E_O$ ) from the resulting spectra.
Spatial Probing:
- Sub-100 nm Scale: The authors performed a "continuous valley probe" by depleting an outer barrier gate in a DQD configuration. This allowed them to shift the electron wavefunction continuously across ~100 nm of material, measuring $E_V$ with ~2 nm resolution.
- Device Scale: They measured $E_V$ discretely under each of the 21 accessible finger gates across the full 1.3 $\mu$ m channel.
Statistical Analysis: The authors calculated autocorrelations of $E_V$ at various spatial separations. They utilized permutation tests to determine the statistical significance of observed correlations and compared experimental data against simulations of random sampling from the theoretical ADD distribution.

Key Results

Nanoscale Fluctuations: Under individual gate fingers, $E_V$ exhibited rapid fluctuations, ranging from a maximum of ~500 $\mu$ eV to a minimum of ~50 $\mu$ eV over distances of ~40 nm.
Correlation Length: Analysis of the nanoscale fluctuations revealed a correlation length ( $\ell_C$ ) of approximately 19.2 nm. This matches the average electron wavefunction radius (19 nm), indicating that the electron wavefunction smooths out faster variations in the underlying material disorder.
Device-Scale Correlations: Across the full 1.3 $\mu$ m array, the authors observed non-trivial correlations in $E_V$ . Specifically, they found significant anti-correlation at a gate separation of $n=6$ (360 nm) and positive correlation at $n=12$ (720 nm).
Statistical Consistency: Permutation tests and simulations demonstrated that these observed device-scale correlations are consistent with the random sampling of a single underlying ADD distribution. The apparent long-range correlations arise from finite sampling effects rather than a distinct, large-scale material inhomogeneity.
Thermal Population: In regions where $E_V < 60$ $\mu$ eV, the thermal population of the excited valley state at 150 mK is approximately 1%, which could impact Pauli spin blockade fidelity.

Key Contributions

Experimental Validation of ADD Theory: The work provides experimental evidence that valley splitting variations in state-of-the-art, industrially fabricated Si/SiGe devices are dominated by alloy disorder, consistent with theoretical predictions.
Mesoscopic Characterization: The study bridges the gap between single-gate and device-scale understanding, quantifying how valley splitting varies over both the electron wavefunction scale (19 nm) and the full processor scale (1.3 $\mu$ m).
Correlation Analysis: The authors demonstrate that fluctuations appearing as correlations between spatially separated qubits can be expected outcomes of random sampling from a disorder-dominated distribution, rather than evidence of systematic fabrication defects.

Significance
The paper claims that understanding these inherent material variations is critical for the design of scalable quantum devices. By verifying that valley splitting behaves according to ADD theory, the study supports the use of this theoretical framework for predicting qubit infidelity. The identification of the 19 nm correlation length suggests that the electron wavefunction acts as a natural low-pass filter for material disorder. Furthermore, the finding that device-scale correlations can arise from random sampling implies that larger datasets and larger devices are required to distinguish true large-scale material inhomogeneities from statistical noise. Ultimately, these insights are necessary for improving Si/SiGe material design and modeling valley-induced qubit infidelity to enable the manufacturing of uniform, low-noise quantum dot processors.

1. The Terrain is Rough (The "Alloy Disorder")

2. The "Flashlight" Probe

3. The "Long Walk" Across the Chip

4. Why This Matters (According to the Paper)

More like this