Extended Data Fig. 1: Characterizations of our phononic device.
From: Synthetic-domain computing and neural networks using lithium niobate integrated nonlinear phononics
a, Optical microscopic images of devices used for device characterization. Fundamental IDT pair is used in b, second-order IDT pair is used in c, nonlinear computing unit is used in d-f. b, Transmission (S21) and reflection (S11) spectra of the fundamental IDT pair. The gap between the IDT pair is 20 µm. The transmission S21 at 1023 MHz is -22 dB, leading to a power conversion efficiency of about 8.0%. c. Transmission (S21) and reflection (S11) spectra of the second-order IDT pair. The gap between the IDT pair is 100 µm. Due to weak reflections of IDT, oscillation pattern in S21 is observed with a free spectral range (FSR) of 15 MHz, close to a FSR of 18.5 MHz for a phononic cavity formed by the IDT pair. We estimate that the power conversion efficiency is 3.2% for the second-order IDT. d. The measured output spectrum from Port 3, showing both fundamental and second-order signals. An input signal of 0 dBm at 1,023 MHz is applied at Port 1. e. Output powers of second-harmonic generation (SHG) at different input frequencies. The input power is 0 dBm. Black dots are measured data, and the blue curve is a Lorentzian fitting to the data. The fitting shows a full-width-half-maximum (FWHM) bandwidth of 1.20 MHz. f. The input-output power relationship of SHG. Linear scale plot in Inset. The extracted cable-to-cable nonlinear conversion efficiency is 0.063%/W. Considering the fundamental (second-order) IDT conversion efficiency is about 8.0% (3.2%), the on-chip nonlinear conversion efficiency at the peak frequency is estimated as 24.6%/W.
