Millisecond phonon lifetimes in piezoelectric nanomechanical resonators

The gist: Lithium niobate phononic crystal resonators fabricated with this process from 150 nm thick LN, operating in the 3–4 GHz range, are measured at millikelvin temperatures with various electrode spacings that set the external coupling rate.

Fabrication and Damage Mitigation

The study addresses the limitation of phonon lifetimes in piezoelectric nanomechanical resonators by showing that fabrication-induced surface damage is a primary source of loss. Ion milling, for example, amorphizes exposed surfaces and creates a ∼10 nm layer hosting a dense population of two-level system (TLS) defects that couple to mechanical modes via strain. Conventional resist-based metallization introduces additional organic contamination that cannot be easily removed. To resolve the incompatibility between wet chemistry for surface restoration and metal contacts, a shadow-mask metallization process is introduced in which aluminum is deposited through a freestanding silicon nitride stencil as the final fabrication step after all surface cleaning is complete. This approach decouples surface preparation from metallization, enabling aggressive chemical treatment of the fully released structures before any metal touches the dearXiv:2610.10700>.

Measurement and Lifetime Determination

The phonon lifetime T1 is measured by observing the energy ringdown of coherent states using microwave reflectometry, where T1 = 1/κ by observing the energy ringdown of coherent states. The external coupling rate κe is extracted from the depth of the reflection dip with vec-. By systematically varying electrode-to-resonator spacing, researchers demonstrated that the measured lifetime is set by external coupling between the device and the measurement circuit rather than material loss. The results show that as electrode spacing increases from 360 nm to 4760 nm, the measured phonon lifetime rises from 5.0(1) µs to 4.9(1) ms, spanning three orders of magnitude.

Intrinsic Quality Factor and Material Performance

The analysis reveals that for the 4760 nm device, T1 = 4.9(1) ms compared to Te = 6.4(5) ms yields an internal lifetime of T1,in = 21(6) ms and a corresponding Qi = 5(1) × 108. This establishes that lithium niobate, once fabrication-induced surface damage is removed, achieves an intrinsic quality comparable to the best mechanical materials while remaining an intrinsically piezoelectric platform with strong electromechanical coupling. The work demonstrates that the internal phonon lifetime T1,in of our LN resonators exceeds 10 ms at gigahertz frequencies and millikelvin temperatures, corresponding to Qi > 4 × 108 and a frequency–quality factor product f·Qi > 10 18.

Performance Landscape

The researchers plot the internal quality factor Qi at single-phonon occupations against the estimated effective electromechanical coupling coefficient k 2eff (Fig. 3). This reveals a striking pattern where previous devices fall along an approximate tradeoff, but our new devices break this pattern by occupying a previously inaccessible region of parameter space; one that simultaneously has Qi > 10 8 and k 2eff = 5.2 × 10-4 to 3.5 × 10-3. The devices achieve a figure of merit between 1.3 × 10 5 and 8.6 × 10 5, exceeding all previous platforms by more than an order of magnitude.

Conclusion and Future Directions

The results demonstrate that the quality of LN as a material for nanomechanics is comparable to that of the best non-piezoelectric platforms, once fabrication-induced surface damage is removed. The historical coupling–coherence tradeoff was not a fundamental constraint but a consequence of how these devices were made. The immediate next step is to achieve strong coupling to superconducting qubits and use the ultra-long T1 times demonstrated here to enter new regimes for quantum acoustic systems. Furthermore, because the shadow-mask metallization process is material-agnostic, the same fabrication paradigm can be applied to other piezoelectric platforms such as quartz, gallium arsenide, and aluminum nitride. The dramatic reduction in defect density in our devices may make this noise manageable for the first time.

How it works

The shadow-mask process involves several critical steps to ensure surface cleanliness and proper metallization. The process begins by patterning the top side of a silicon wafer with electron beam lithography, which includes shadow-mask apertures defining the nanodevice electrode patterns. After etching through the silicon and releasing the LN film in hydrofluoric acid, it undergoes a full course of NH4OH − H2O2 − H2O (SC1) base cleaning to remove the amorphous damage layer. A shadow-mask is then aligned to the device chip in a custom jig fixture using a commercial flip-chip bonder, and aluminum is deposited by electron beam evaporation through the stencil apertures as the final fabrication step.

Device Architecture and Parameters

The devices are phononic crystal resonators fabricated from a 150 nm-thick film of X-cut MgO-doped LN on a 3 µm buried oxide layer with a 700 µm high resistivity silicon handle (NanoLN). A quasi-one-dimensional array of periodic cells supports a complete acoustic bandgap, with a larger defect cell hosting a shear-mode mechanical resonance in the 3–4 GHz range. The critical design parameter is the electrode-to-defect spacing, which controls the external coupling rate κe while leaving intrinsic material properties unchanged.

Effective Electromechanical Coupling Estimation

The effective electromechanical coupling coefficient k 2eff is defined as the ratio between converted mechanical (electrical) energy Uconv, and the input electrical (mechanical) energy Uinput. For high-overtone resonators, this can be estimated using Eq. E4, where k 2eff ≈ k squared ωFSR / ω0. For the widest electrode gap spacing device (d = 4760 nm), an estimate of k 2eff = 3.5 × 10-3 was obtained using COMSOL Multiphysics simulations and experimental parameters. This value represents a lower bound on k 2eff loaded by parasitics, such as additional stray capacitances stemming from the large device array.

Comparison to Literature

The work compares its devices against diverse mechanical resonator platforms, showing that our new devices improve upon this figure of merit by over two orders of magnitude relative to previous results. The device presented in this work achieves a figure of merit M = 1.3 × 10 5 to 8.6 × 10 5, a more than 50x improvement over all previous platforms. This demonstrates that the tradeoff between electromagnetic coupling and phonon lifetime was not fundamental.

Shadow-Mask Distortion

The study characterizes the distortion of shadow-mask metallization using a least-squares fit to determine global translation and rotation misalignments, where the residual of this fit for each grid position is called the ‘distortion’. The experiment is generally limited by uncertainties but bounds distortion at ≤ 500 nm across the chip, well below the alignment error. The lack of strong evidence for distortion in this experiment is unsurprising, as deposition at normal incidence is to first order insensitive to height variation in the mask.

1. We can improve quantum memory systems by achieving coherence times exceeding 10 ms in piezoelectric nanomechanical resonators, as demonstrated by "T1,in = 8.0(3) µs >10 ms for the 360 nm device and T1,in = 21(6) ms" for the 4760 nm device. This capability allows for significantly longer storage times in quantum information processing platforms than previously possible with lithium niobate.

2. We can design novel precision sensors by leveraging the intrinsic material quality of lithium niobate to achieve a frequency-quality factor product "f·Qi > 1018, which is comparable to the best silicon nanomechanical resonators." This enables higher sensitivity in mass sensing applications by reducing noise sources associated with intrinsic material loss.

3. We can develop more robust and scalable fabrication protocols for quantum acoustic systems by utilizing the shadow-mask metallization process, which decouples surface preparation from metallization, thereby enabling aggressive chemical treatment of the fully released structures before any metal touches the dearXiv:2610.10700v1. This material-agnostic paradigm can be applied to other platforms like quartz and aluminum nitride, improving device reproducibility.

4. We can optimize the design of electromechanical devices by utilizing a new figure of merit, M = k2eff(1 - k2eff)Qi ≈ k2effQi, which for this work achieves values between 1.3 × 105 and 8.6 × 105, exceeding all previous platforms by more than an order of magnitude. This metric allows for a more comprehensive comparison of device utility beyond the historical tradeoff between coupling and lifetime.