Fig. 4: Experimental results for ultrasensitive detection.
a Comb-line-resolved PA spectra measured with (blue) and without (grey) optomechanical enhancement, respectively, for 1 ppm 12C2H2 /N2. b PA measurements for the 12C2H2 v1 + v3 band P(17) line. A HITRAN model (red curve) is displayed for comparison. The shaded areas (light grey, orange, and purple) show the standard-error intervals for measurement times of 40 ms, 200 ms, and 2 s, respectively. The residuals between the 2-s data and the HITRAN simulation are plotted in grey. The systematic discrepancies around 195.258 THz are due to the additional electronic amplifier we used for amplifying the PA signal in the low-concentration measurement. c The SNR of the PA signal versus the acquisition time (t). The fitting (black) indicates that the SNR is proportional to √t within 100 s. For A, B, and C, the dual-comb power is set to 150 mW. The number of comb lines is reduced to > 40 per comb for achieving a high power per comb line. Also, the line spacings are set to ~700 MHz with Δfr = 25 Hz. d The excitation power dependences of the PA signal measured under different C2H2 concentrations (615, 500, 54 and 1 ppm, respectively). The inset shows the concentration dependence of the PA signal (with a fixed excitation power of 1 mW). The lines represent linear fitting. e Allan–Werle deviation analysis. The noise data of the MIM sensor over a period of 15 min are measured, with (red) and without (blue) switching on the excitation laser (200 mW), using a lock-in amplifier. The gas cell was filled with pure N2.
