Interface Piezoelectric Loss in Superconducting Qubits

This paper reports the direct observation of interface piezoelectricity at the aluminum-silicon boundary as a distinct dissipation channel in superconducting qubits, demonstrating that it can significantly reduce qubit lifetimes and potentially dominate over two-level system losses at high frequencies.

The Gist

Imagine you are trying to keep a spinning top perfectly balanced on a table. In the world of quantum computing, this "spinning top" is a superconducting qubit, a tiny machine that holds information. The biggest problem scientists face is that these tops eventually wobble and fall over (lose their information) because of "dissipation" or energy loss.

For a long time, scientists thought the main reason these tops fell was because the table itself was bumpy or dirty. They called these bumps "Two-Level Systems" (TLS)—basically, tiny defects in the materials that steal energy. They spent years polishing the table (improving materials) to make it smoother, and it worked. The tops spun longer.

But this paper discovered a new, invisible force knocking the tops over.

Here is the story of what they found, explained simply:

1. The "Ghost" Piezoelectric Effect

The researchers built their quantum tops on silicon, a material that is supposed to be "non-piezoelectric."

• The Analogy: Think of piezoelectricity like a trampoline. If you jump on a trampoline (apply electricity), it bounces (creates sound/vibration). If you push a trampoline, it makes a sound. Materials like quartz are like trampolines; silicon is supposed to be like a solid concrete floor—it shouldn't bounce or make sound when you push it.
• The Discovery: The team found that even though the bulk silicon floor is solid concrete, the very thin interface (the boundary) where the metal qubit touches the silicon acts like a tiny, invisible trampoline. When the qubit vibrates with electricity, it accidentally pushes on this "concrete-trampoline," creating sound waves (phonons) that travel away, stealing the qubit's energy.

2. The Experiment: Tuning the Radio

To prove this, they built a special device.

• The Setup: They made a qubit that also acted as a speaker and a microphone for sound waves. They placed it inside a "sound cage" (a Surface Acoustic Wave resonator) made of mirrors that trap sound waves.
• The Trick: They tuned the qubit to sing at specific notes.
  • The Result: When the qubit sang a note that perfectly matched the "room tone" of the sound cage, the qubit's energy vanished twice as fast as normal.
  • The Proof: They applied a voltage to the qubit. If the energy loss was caused by the "bumpy table" (TLS defects), the voltage would have changed the loss pattern. But it didn't. The loss pattern stayed exactly the same, proving it wasn't the defects, but the sound waves (phonons) stealing the energy.

3. Why This Matters (The "Frequency" Problem)

The paper explains that this "ghost trampoline" effect gets much worse as the qubits get faster (higher frequency).

• The Analogy: Imagine pushing a child on a swing. If you push slowly, the swing doesn't go far. But if you push at just the right fast rhythm, the swing goes huge.
• The Finding: The researchers found that as they tried to make the qubits operate at higher speeds (like moving from a slow walk to a sprint), the energy loss from these sound waves exploded.
• The Prediction: They used computer simulations to predict that for future, super-fast qubits (operating at very high frequencies), this "sound wave theft" will become the biggest problem, potentially worse than the "bumpy table" defects they have been fighting for years.

4. The Solution? Build a Different Floor

Since this loss comes from the shape of the device and the boundary between materials, simply making the silicon "cleaner" won't fix it.

• The Idea: The paper suggests we need to change the design of the "floor."
  • Option A: Carve out the silicon underneath the edges of the metal (like an undercut) so the "trampoline" effect has nowhere to push.
  • Option B: Put the qubit on a thin, floating membrane (like a drum skin) instead of a thick block of concrete. This changes how the sound waves behave and can stop them from stealing energy.

Summary

This paper reveals that superconducting qubits on silicon are losing energy not just because of dirty materials, but because the metal-silicon boundary accidentally turns electricity into sound waves. It's like a silent alarm that steals the battery of a quantum computer. As we try to build faster quantum computers, this "sound theft" will become a major obstacle, and we will need to redesign the physical shape of the chips to stop it.

Technical Summary

Technical Summary: Interface Piezoelectric Loss in Superconducting Qubits

Problem Statement
Dissipation remains a primary obstacle to improving the coherence of superconducting quantum circuits. While material disorder, typically modeled as two-level systems (TLSs) in dielectrics, has been extensively studied and mitigated, the microscopic origins of remaining losses are not fully understood. Specifically, it has been unclear whether electromechanical coupling at interfaces could serve as a distinct loss channel in qubits fabricated on nominally non-piezoelectric substrates, such as silicon. Although bulk silicon possesses centrosymmetry which suppresses piezoelectricity, recent classical measurements suggest that superconductor-substrate interfaces can exhibit effective piezoelectric responses. Whether this interface piezoelectricity limits qubit coherence and how it scales with frequency and geometry has remained an open question.

Methodology
To investigate this, the authors fabricated transmon qubits on high-resistivity intrinsic silicon substrates. The devices integrate a superconducting interference device (SQUID) shunted by an interdigital capacitor ($C_{idt}$). Crucially, this $C_{idt}$ is positioned between two Bragg reflectors, functioning simultaneously as the qubit's shunt capacitor and as an interdigital transducer (IDT) for a surface acoustic wave (SAW) resonator. This geometry ensures maximal spatial overlap between the qubit's electric field and the mechanical strain field of the SAW modes.

The experimental approach involved:

1. Frequency Tuning: Tuning the qubit transition frequency ($f_q$) into resonance with discrete mechanical modes of the SAW resonator.
2. Bias Averaging: Applying a DC bias voltage ($V_{bias}$) via the XY line to distinguish between piezoelectric coupling and TLS defects. TLS defects exhibit Stark shifts (frequency changes) with bias, whereas SAW resonator modes are determined by lithographic geometry and are largely bias-independent. By averaging measurements over $V_{bias}$, TLS contributions are suppressed, isolating the mechanical resonances.
3. Coherence Measurements: Measuring the qubit excited-state population ($p_e$) and relaxation time ($T_1$) as a function of delay time and qubit frequency to quantify energy loss.
4. Multiphysics Simulations: Performing finite-element simulations of aluminum-on-silicon coplanar capacitors to model interface piezoelectricity using a thin piezoelectric layer beneath the electrodes. These simulations calculated the energy participation ratio (EPR) and predicted the frequency scaling of the loss.

Key Results

• Observation of Loss Reduction: When the qubit frequency was tuned into resonance with the SAW modes, the qubit lifetime ($T_1$) decreased by up to a factor of two (approximately 50% reduction). This reduction was observed consistently across multiple samples (A, B, and C) and thermal cycles.
• Distinction from TLS: The resonances causing $T_1$ reduction remained fixed in frequency regardless of the applied bias voltage, confirming they are mechanical in origin rather than TLS defects.
• Coupling Strength: Fitting the data to a master equation model yielded electromechanical coupling strengths ($g_{m,k}$) of approximately 100 kHz and mechanical linewidths ($\kappa_{m,k}$) of roughly 1 MHz. The system operates in the weak-coupling regime, where coherent exchange is not resolved, but dissipative relaxation is significant.
• Frequency Scaling: Simulations revealed that the interface piezoelectric decay rate follows a steep power law, $\Gamma_{piezo} \propto f_q^{3.4}$. In contrast, dielectric loss from TLSs exhibits only weak frequency dependence.
• Crossover Prediction: The simulations predict that while interface piezoelectricity is subdominant to TLS loss at current operating frequencies (sub-10 GHz), it becomes comparable to TLS loss near 10 GHz and is expected to dominate at higher frequencies (e.g., millimeter-wave regimes).

Significance and Claims
The paper provides direct experimental evidence that interface piezoelectricity is a distinct and previously overlooked source of dissipation in superconducting qubits, even on substrates like silicon that are nominally non-piezoelectric. The authors claim that this loss mechanism arises from symmetry breaking at the metal-substrate interface, making it a structurally rooted decoherence channel that cannot be eliminated solely by improving bulk material quality.

The findings suggest that as TLS losses are further reduced and as qubit frequencies increase (e.g., for millimeter-wave qubits), interface piezoelectric loss will become a limiting factor. The paper posits that mitigating this loss will require device-level engineering, such as modifying the phonon density of states (e.g., using membrane geometries) or engineering spatial profiles to minimize field overlap at the interface, rather than just material optimization. Conversely, the authors note that this mechanism could be harnessed for coherent qubit-phonon interactions in hybrid quantum acoustic systems.