Gigahertz Phononic Integrated Circuits on Thin-Film Lithium Niobate on Sapphire

Acoustic devices play an important role in classical information processing. The slower speed and lower losses of mechanical waves enable compact and efficient elements for delaying, filtering, and storing of electric signals at radio and microwave frequencies. Discovering ways of better controlling the propagation of phonons on a chip is an important step towards enabling larger-scale phononic circuits and systems. We present a platform, inspired by decades of advances in integrated photonics, that utilizes the strong piezoelectric effect in a thin film of lithium niobate on sapphire to excite guided acoustic waves immune from leakage into the bulk due to the phononic analogue of index guiding. We demonstrate an efficient transducer matched to $50\mathrm{\Omega }$ and guiding within a $1\mu \mathrm{m}$ wide mechanical waveguide as key building blocks of this platform. Putting these components together, we realize acoustic delay lines, racetrack resonators, and meander line waveguides for sensing applications. To evaluate the promise of this platform for emerging quantum technologies, we characterize losses at low temperature and measure quality factors on the order of 50 000 at $4\mathrm{K}$. Finally, we demonstrate phononic four-wave mixing in these circuits and measure the nonlinear coefficients to provide estimates of the power needed for relevant parametric processes.

Figure 1

Design and simulation of the IDT and the mechanical waveguide. (a) Finite-element model and simulated driven mode profile of the IDT-taper-waveguide system, at the simulated IDT resonant frequency of $3.36\mathrm{GHz}$. The displacement is shown as black arrows, and the color highlights the transverse horizontal component of the displacement field. (b) First four bands of the IDT unit cell with open (solid blue line) and shorted (dashed red line) electrode boundary conditions. The strongly coupled mode is emphasized. (c) Mechanical bands of the LN waveguide (solid line) and the LN slab (dashed line). Black cones represent bulk acoustic and surface acoustic waves in sapphire. (d) Normalized displacement field of the quasi-Love guided mode in the rib waveguide at $3.5\mathrm{GHz}$. (e) Displacement field on a two-dimensional cross section of the IDT-taper-waveguide system. Black arrows indicate the direction of mechanical energy flux.

Figure 2

(a) Fabrication process of the LISA IDT. (b) Scanning electron micrograph of one fabricated transducer.

Figure 3

(a) Microwave reflection $|S_{11}|$ of a transducer with $N=9$ unit cells. (b) Transmission $|S_{21}|$ of an IDT-waveguide-IDT system where the waveguide length is $L=200\mu \mathrm{m}$. The blue curve is the unfiltered response and the black curve has been filtered to keep only the single transit response. (c) Measured conductance $G$ (blue line) and susceptance $\chi$ (red line) of the transducer. The filtered response is shown in black and corresponds to the transducer response without mechanical reflections from the other transducer.

Figure 4

Acoustic racetrack resonator. (a) Optical microscope image of an acoustic racetrack resonator. The input as well as the through and drop ports are labeled. The inset on the right shows the coupling region. (b),(c) Relative power transmission at room temperature (b) and $T\approx 4$ K (c) of the through (red line) and drop (blue line) channels. The insets show typical acoustic resonances with quality factor $Q=5000$ at room temperature (b) and $Q=47000$ at 4 K (c).

Figure 5

(a) Optical microscope image of a meander waveguide used for two-dimensional reflectometry experiments. The input and output ports are labeled as 1 and 2. The aluminum scatterer on the waveguide is labeled as $S$. (b) Optical microscope image showing a closeup of the scatterer. (c) Impulse response $h(\tau )$ in reflection for both transducers. The peaks are labeled with the corresponding sequence of scattering events. The highlighted curves have been smoothed with a $50$-point (5 ns) moving average. (d) Impulse response $h(\tau )$ in transmission.

Figure 6

(a) Four-wave mixing measurement setup. (b) Measured power for each comb line with $\mathrm{\Delta }f=10$ MHz. The two central black lines are the pumps and the generated lines are labeled by the corresponding indices. The black arrows illustrate the degenerate (dashed) and nondegenerate (solid) four-wave mixing processes. (c) The power conversion efficiency ${\eta }_{0\pm }$ scales quadratically with initial mechanical pump power in the waveguide $P_{0\pm }^{\prime }$. The slope of the black line is the fitted $\mathrm{\Gamma }$. The shaded regions are the convex hulls of the uncertainty on the measurement.

Figure 7

Simulated IDT admittance. Solid lines are fits to the simulated data.

Figure 8

Microwave transmission $|S_{21}|$ at temperature $T=4\mathrm{K}$ of an IDT-waveguide-IDT device revealing bulk acoustic wave resonances. The inset shows a typical resonance with quality factor $Q=34000$.

Figure 9

(a) Time domain impulse response $|h|$ for five delay lines with waveguide lengths $L=0.4$, $0.6$, $0.8$, $1.1$, and $1.5$ mm (from bottom to top). The black circles are at the maximum of the peaks representing the single transit and first two echoes. The responses are manually shifted by $80\mathrm{dB}\mathrm{Hz}$ for clarity. (b) The maximum impulse response is fit with a plane against the number of echoes and propagation distance.

Figure 10

Simulated bending-loss-limited quality factor $Q$ for the first (blue, quasi-Love) and second (red, quasi-Rayleigh) guided mechanical modes as a function of (a) central angle $\theta$ of the simulated sector, (b) bending radius $R$, and (c) crystal orientation $\varphi$.

Figure 11

Four-wave-mixing nonlinearity of the measurement setup, measured using two pump signals with (a) 0 dBm and (b) $-10$ dBm output powers from the signal generator.

Figure 12

Transmission spectrum of an optical ring cavity fabricated on the same chip after the full fabrication process. Inset shows a typical mode from the ring cavity and the corresponding Lorentzian fit, giving a loaded quality factor $Q=4.3\times {10}^5$ and an intrinsic quality factor $Q_i=5.4\times {10}^5$.