Co-optimization of spin coherence and valley splitting in Si/SiGe heterostructures

This study uses density functional theory to demonstrate that Si/SiGe heterostructures with 3–4 nm quantum wells, low $^{73}$Ge and $^{29}$Si concentrations (50 ppm), and sharp interfaces can simultaneously achieve valley splittings exceeding 500 $\mu$eV and spin dephasing times over 15 $\mu$s, thereby co-optimizing these critical parameters for semiconductor quantum devices.

The Gist

Imagine you are trying to build a tiny, super-fast computer using a single electron as a bit of information. In the world of quantum computing, this electron acts like a spinning top. To make this computer work, the spinning top needs to stay stable (coherent) for a long time, and it needs to be very distinct from other similar tops nearby.

This paper tackles two major problems that stop these "electron tops" from working well in silicon chips: Valley Splitting and Spin Decoherence.

Here is the breakdown of the research using simple analogies:

1. The Two Enemies: The "Valley" and The "Noise"

The Valley Problem (The Foggy Landscape)
Imagine the electron is a hiker walking on a mountain range. In pure silicon, there are six identical valleys where the hiker could hide. This is bad because the hiker might accidentally slip from one valley to another, losing the information it was carrying.

• The Fix: The researchers use a "strained" silicon layer (like stretching a rubber sheet) to flatten five of the valleys and leave only one deep, safe valley. The difference in height between the safe valley and the others is called Valley Splitting.
• The Goal: You want this height difference to be huge so the hiker never slips. The paper finds that making the silicon "room" (the quantum well) narrower makes this height difference bigger, keeping the hiker safer.

The Noise Problem (The Chatty Crowd)
Now, imagine the hiker is trying to think quietly, but the ground is made of rocks that are constantly chattering. These "rocks" are atomic nuclei with their own tiny magnetic spins (like tiny magnets).

• The Issue: In natural silicon, about 5% of the atoms are "chatty" (isotope 29Si). In the surrounding material (SiGe), there are even more chatty atoms (isotope 73Ge). When the electron gets too close to these chatty rocks, it gets distracted and loses its spin stability (decoherence).
• The Goal: You want the hiker to stay far away from the chatty rocks so they can focus.

2. The Dilemma: The "Goldilocks" Trap

The researchers discovered a tricky trade-off, like trying to find a chair that is both too small and too big at the same time:

• If the room is too wide: The valley splitting is small. The hiker might slip into the wrong valley (bad for stability).
• If the room is too narrow: The hiker is forced to stand very close to the walls. The walls are made of the SiGe material, which is full of the "chatty" 73Ge rocks. Even though the valley is safe, the hiker is now so close to the noise that they get distracted immediately (bad for coherence).

The Paper's Solution:
You can't just make the room narrower; you have to clean up the walls, too.

3. The Recipe for Success

The team used powerful computer simulations (Density Functional Theory) to test millions of different atomic arrangements. They found a "sweet spot" recipe:

1. Make the room narrow: Specifically, a silicon layer about 3 to 4 nanometers wide. This maximizes the valley splitting (keeps the hiker in the right valley).
2. Purify the walls: Since the narrow room forces the electron to touch the walls, you must remove the "chatty" atoms from those walls.
   • They recommend reducing the "chatty" Germanium (73Ge) in the walls to almost nothing (50 parts per million).
   • They also recommend purifying the Silicon (29Si) in the room to very low levels (50 parts per million).

The Result:
If you follow this recipe, the electron can stay in its safe valley with a huge energy gap (over 500 micro-electron volts) and stay stable for a long time (over 15 microseconds).

4. The Importance of Smooth Walls

Finally, the paper looked at the quality of the walls.

• Sharp Interface: Imagine a wall where the silicon ends and the germanium begins with a perfectly sharp, clean cut. This is ideal.
• Blurry Interface: In real life, the transition is often a bit "fuzzy" or mixed (like a gradient). The paper found that fuzzy walls are bad. They reduce the safety of the valley and increase the noise, making the electron spin unstable faster.

Summary

To build a better silicon quantum computer, you need to build a very narrow room (3–4 nm) but you must also scrub the walls clean of magnetic impurities. If you do both, the electron stays safe from slipping and quiet enough to think. If you only do one, the system fails.

Technical Summary

Technical Summary: Co-optimization of Spin Coherence and Valley Splitting in Si/SiGe Heterostructures

Problem Statement
Semiconductor spin qubits, particularly those based on silicon, offer a promising path for scalable quantum information processing due to their small size and compatibility with mature fabrication technologies. However, their performance is currently limited by two primary material challenges: small energy splittings between low-lying valley states and spin decoherence caused by hyperfine coupling to nuclear spins. In natural abundance silicon, the 5% concentration of spinful $^{29}$Si ($I=1/2$) nuclei remains a dominant source of dephasing. While isotopic enrichment of $^{28}$Si has improved coherence times, the valley splitting ($E_v$)—the energy separation between the two lowest conduction band valleys—remains a critical bottleneck. Small $E_v$ values create leakage channels for qubit operations, such as spin shuttling and exchange gates.

A fundamental trade-off exists in optimizing these parameters. Strategies to increase valley splitting, such as narrowing the silicon quantum well (QW) width ($t_{QW}$), force a larger fraction of the electronic wavefunction to penetrate into the surrounding SiGe barrier layers. While this enhances valley splitting, it simultaneously increases the wavefunction's overlap with spinful $^{73}$Ge ($I=9/2$) nuclei in the SiGe barriers, potentially degrading spin coherence times ($T_2^*$). Previous studies have not fully addressed the simultaneous optimization of these competing factors in realistic, disordered heterostructures.

Methodology
The authors employ large-scale Density Functional Theory (DFT) calculations to model realistic Si/Si$_{0.7}$Ge$_{0.3}$ heterostructures. Key methodological features include:

• Structural Modeling: The study utilizes Special Quasirandom Structures (SQS) to accurately simulate alloy disorder in the SiGe barriers, generating random configurations containing up to 5,800 atoms.
• System Variation: The authors model Si QWs with widths ($t_{QW}$) ranging from 2.15 nm to 9.1 nm, sandwiched between relaxed SiGe barriers. They investigate both idealized atomically abrupt interfaces and structures with finite-width transition layers (modeled as 1.1 nm thick Si$_{0.85}$Ge$_{0.15}$ layers) to account for growth imperfections.
• Computational Techniques: Structural optimizations use the PBEsol exchange-correlation functional. Valley splittings are calculated with high convergence precision ($\lesssim 1$ $\mu$eV) to resolve small energy scales.
• Hyperfine Coupling: To calculate contact hyperfine interactions, the authors use the Projector Augmented Wave (PAW) method to reconstruct all-electron spin densities, which is necessary as standard pseudopotential methods cannot access wavefunction density at nuclear sites.
• Scaling: Calculated dephasing times are rescaled from the simulation cell area ($\sim 7.6$ nm$^2$) to an effective quantum dot area of 700 nm$^2$ to match experimental conditions.

Key Results

1. Valley Splitting vs. Well Width: The calculations confirm a general trend where reducing the QW width increases the valley splitting ($E_v$). For ideal abrupt interfaces, $E_v$ can exceed 1 meV for $t_{QW} < 4$ nm. However, for wider wells ($t_{QW} \ge 8$ nm), $E_v$ drops to 0.1–0.25 meV. The study highlights significant fluctuations in $E_v$ due to alloy disorder and interface abruptness, with values varying by over 1 meV during structural relaxation.
2. Wavefunction Penetration: As $t_{QW}$ decreases below $\sim 5$ nm, the percentage of the electronic wavefunction residing in the SiGe barriers ($p_B$) increases rapidly. This increased penetration significantly enhances the hyperfine coupling to $^{73}$Ge nuclei.
3. Co-optimization of Isotopes: The study maps the inhomogeneous spin dephasing time ($T_2^*$) against $^{29}$Si and $^{73}$Ge concentrations.
   • In wider wells or with high $^{29}$Si concentrations ($>1000$ ppm), dephasing is dominated by the QW nuclei.
   • In narrow wells ($t_{QW} \approx 3.2$ nm) with high $^{28}$Si enrichment (50 ppm $^{29}$Si), the contribution from $^{73}$Ge in the barriers becomes significant.
   • To achieve $T_2^* > 15$ $\mu$s in narrow wells, the authors find that isotopic depletion of $^{73}$Ge in the SiGe barriers is essential. For instance, reducing $^{73}$Ge from natural abundance (7.7%) to 0.1% can triple $T_2^*$ in a 3.22 nm well.
4. Impact of Interface Abruptness: The inclusion of transition layers (simulating non-abrupt interfaces) results in a nearly 50% reduction in valley splitting compared to ideal abrupt interfaces. Furthermore, broadened interfaces lead to a more rapid decrease in $T_2^*$ as the well width decreases, as the softened confinement potential increases wavefunction overlap with barrier spins.

Significance and Claims
The paper claims to provide a quantitative guide for the co-optimization of valley splitting and spin coherence in Si/SiGe heterostructures. The authors conclude that:

• Si/SiGe heterostructures with 3–4 nm wide quantum wells and isotopic concentrations of 50 ppm for both $^{73}$Ge and $^{29}$Si can support average valley splittings $E_v > 500$ $\mu$eV and spin dephasing times $T_2^* > 15$ $\mu$s (assuming a 700 nm$^2$ effective area).
• Achieving fault-tolerant spin qubits with large valley splittings requires a dual approach: narrowing the quantum well to maximize $E_v$ while simultaneously pursuing high levels of $^{28}$Si enrichment and $^{73}$Ge depletion to mitigate the resulting increase in hyperfine coupling.
• Improvements in growth processes to achieve sharper Si/SiGe interfaces are critical, as interface broadening detrimentally affects both valley splitting and coherence times.

The work distinguishes itself from prior studies by performing large-scale first-principles calculations on realistic disordered structures, including narrow well models and transition layers, rather than relying solely on simplified models or idealized interfaces.