Introduction In order to overcome the long term stability issues caused by functionalized films in gas sensing [1], uncoated sensors have become increasingly attractive for applications where the selectivity is not a major concern such as industrial gas monitoring. Despite their poor selectivity, discrimination can be achieved by measuring different properties of the gas mixture [2]. The sound attenuation of a gas depends on several of its physical properties such as mass density, viscosity and sound velocity among several others [3]. Its value depends on the frequency in a non-linear manner which makes measuring large parts of its spectrum interesting for gas discrimination.In this abstract, an uncoated sensor capable of measuring the attenuation spectrum continuously over a frequency range is presented. Measurements on binary mixtures such as nitrogen (N2) with either hydrogen (H2), carbon dioxide (CO2) or methane (CH4) are presented. Then, a simple method based on the construction of a preliminary mixture signature allowing to distinguish each type of mixture demonstrates the potential of this sensor to be used in smart gas sensors as a perspective to future work. Although attenuation sensors can be found in the literature [3], to the best of our knowledge, it is the first of this kind with easy integration thanks to the use of capacitive micromachined ultrasonic transducers (CMUTs) and allowing discrimination of binary mixtures. Setup The manufacturing process of the CMUT arrays is similar to the one used in reference [4] and their characteristics are reported in Table 1. Schematics of the experimental setup are shown in Figure 1. An electrical signal (1) is sent to an emitter CMUT array (2) which generates a continuous ultrasonic wave at a given frequency f. The wave travels a distance d through the gas and is attenuated by an amount that depends on the gas composition, through the attenuation coefficient α, before reaching the receiver (3) which is connected to a charge amplifier (4). Both the emitter and receiver signals are fed to a network analyzer (5) in order to measure the total transfer function of the setup |H| as a function of frequency. Far from the resonant frequency of the CMUT array (fr = 9.6MHz), |H| is given by Equation 1, where |He | is the transfer function of the setup which is independent of the gas. Thus, by measuring first |H| under pure N2, |HN2 |, it is possible to know the shift in attenuation Δαaccording to Equation 2. Results Tests under H2, CO2 and CH4 in N2 at different concentrations were performed at 20°C and 1atm for f ranging from 2MHz to 4.5MHz. The normalized measurements are shown in Figures 2, 3 and 4, respectively. The normalization consists in multiplying Δα by the wavelength in N2, λN2 , at the optimal frequency, fopt , for each mixture. Theses frequencies correspond to the best theoretical limit of detection LOD (Equation 3). The noise standard deviation σ increases with the frequency from 1.7x10-5 at 2MHz to 6.1x10-5 at 4.5MHz. For each mixture, the optimal frequency, fopt , sensitivity, S, and theoretical limit of detection, LOD, are reported in Table 2. The calculated values of the LOD were then verified for the three binary mixtures at lower concentrations. The results are shown in Figures 5, 6 and 7 and are consistent with the calculations and even better in the case of CH4 where the step at 0.25% is still visible. These values of LOD correspond approximately to an attenuation of 1m-1,which corresponds to the state of the art on attenuation sensors [3]. Finally the normalized attenuation spectra for the three mixtures, shown in Figure 8, makes the discrimination of the three mixtures possible. In this study, it is done by simply considering both the sign of λN2 Δα and the one of the mean slope Σ in the region between 2MHz and 3.6MHz defined by Equation 4 (both signs are reported in Table 2 for each mixture). Conclusion A CMUTs based device is used to measure the attenuation spectrum of a gas. Its characterization shows performances comparable to the state of the art of attenuation sensors. Such measurements allow to determine the concentration of binary mixtures of N2 with either H2, CO2 or CH4. Finally, a signature of each of the three binary mixture is introduced to show the potential of this sensor as part of a smart sensor network. Possible perspectives include increasing the frequency range and testing in other gases.