本节从适合火星的声学的提醒开始。它提供了不同数据集的详细信息：来自SuperCAM和EDL麦克风的声学数据；除了记录Libs Shock Wave和Rover噪声外，火星氧气中的人造声音（Moxie）和创造力除外；从MCD提取的MEDA和温度的风速，温度和压力数据。提出了处理方法：PSD的计算；LIBS冲击波时间序列的分析；从创造力中提取多普勒效应。最后，给出了有关距离衰减的支持解释。

　　我们证明文本简介中的三个主要断言是合理的。首先，声学阻抗描述了维持声波的介质的强度。它由z =ρc在源的远场中给出，其中ρ是密度，c是声速度。通常，使用ρ= 0.02 kg m -3和c = 238 m s -1（见下文），我们在火星表面获得z = 4.76 kg m -2 s -1，而ρ= 1.217 kg m -3和c = 340 m s -1产量z = 413 kg m -2 s -2 s = 413 kg m -2 s -1。这两个数量级的差异转化为火星的信号，大约比地球上的同一源弱20 dB。其次，在火星压力下，大约95％的二氧化碳大气可以有效地建模为理想气体。在麦克风频率范围内，鉴于声压的幅度较小，声波被认为是绝热干扰。因此，声音衍生的速度由C2 =γRT/m给出，R摩尔气体常数（8.314 J mol-1 K-1），M大气的摩尔质量（43.34 g mol-1），kelvins中的温度和γ和γ。使用γ= 9/7，二氧化碳的标准值 - 在主要文本中讨论了该值 - 我们发现C = 238 m s -1在230 K处。第三，在稀有气氛中，吸收本质上更大，因为经典（热和粘性），并且旋转吸收与压力成正比成正比。此外，在低频率下，振动吸收在经典的吸收和旋转类型上占主导地位。事实证明，与N2（参考文献3）相比，二氧化碳的振动比热比N2大20倍。因此，二氧化碳振动的双重变性模式减弱了低频的声音，而粘度强烈减弱的频率高于几个kHz。就每米衰减系数而言，这两种效应都是在相同频率下强的一个或两个数量级。

　　本研究中使用的SuperCAM麦克风数据集从SOL 1到SOL 216，当时是任务的第一个太阳连接。在此日期，总共记录了4 h和40分钟的火星声音，包括大气湍流（占总持续时间的46％），随附的液体火花的压力波（12％）和机械噪声（例如，Moxie43，Ingenition33，Ingenity33，Masterance Rotation of Perseverance of Perseverance of Perseverance of Perseverance of Perseverance of Perseverance of Perseverance of Perseverance of Perseverance of Perseverance，Mastcam-Z机械机制，42％）。在同一时期，EDL15麦克风总共记录了56分钟的火星声音，主要是在Rover操作期间（例如，Rover Drive，Arm Motion）。扩展数据表1列出了本研究中使用或提到的所有声学文件，除了图1和图1所示的声音文件。3a，4a，太多了，无法单独引用（见下文）。

　　SuperCAM的麦克风记录了25 kHz采样频率的气压波动从20 Hz到12.5 kHz，当使用100 kHz采样模式时，高达50 kHz的气压波动。使用四个电子增益之一将麦克风的模拟信号（范围从0到5 V）进行数字化（12位深度），以将灵敏度从0.6 v PA-1提高到21 V PA-1，并从2个MPa提高了敏感性。增益0用于记录校准目标上的Libs Spark的随附声音，增益1、2和3，以记录不同距离的Libs Spark的随附声音，而增益3用于大气记录。大气和机械噪声的记录通常长167 s。EDL麦克风可以记录10 MN和更长的时间序列，并具有固定增益放大器，然后是24位/44-kHz数字化器（在扩展数据表2中总结了关键特征）。

　　典型的LIBS序列由30个激光射击以3 Hz射击的相同位置（出于技术原因，仅记录29张照片）。对于给定的目标，这种序列通常在由几毫米分隔的新鲜采样点上重复5-10次。在每个激光脉冲周围以100-kHz采样频率监测激光诱导的声信号。录制窗口的开始是在激光触发器上精确定时的，因此可以通过不确定性来测量声波的传播时间 <10 μs. Up to Sol 216, SuperCam’s microphone has recorded sound sequences for 123 Martian targets located at distances ranging from 2.05 m (target Garde) to 8.01 m (target Pepin) from the microphone. On seven occasions, it has also recorded the acoustic signal related to LIBS measurements of the titanium (Ti) calibration target44 located on the rover deck 1.51 m from the microphone.

　　For the derivation of the sound speed (Fig. 3a), targets farther than 6 m are excluded because of a small signal-to-noise ratio that prevents a good time-of-flight measurement. Recordings of the LIBS acoustic signals from Ti are also excluded, as their sound propagates above the rover and are biased by extra heating and turbulence induced by the warm body of the rover. In total, 109 targets between Sol 1 and Sol 216 are considered for Fig. 3. For the attenuation study (Fig. 4), regolith or loose material targets, which generally lead to a lower sound amplitude, are excluded, as well as out-of-focus points for the same reason. In total, 96 targets are used. The measurements from Ti are included and provide a useful constraint on the attenuation at short distance. As the laser energy used on Ti is lower than the laser energy used on Mars targets (110 A pumping current on Ti compared with 155 A on Mars targets), the PSD amplitude from the Ti measurements are normalized by a factor of 155/110, as the pumping current is proportional to the laser energy, which is proportional to the laser irradiance, as the spot size and the pulse duration remains the same.

　　Both EDL and SuperCam microphones are also used to record sounds produced by the rover. They help to inform operators about equipment health (for example, rover driving, MOXIE) and provide sources of sound that are well localized in space and time (for example, during Ingenuity flights or LIBS sparks). All recordings also pick up some perturbations, such as intense single-frequency emissions at 195 Hz, 198.75 Hz and the following harmonics at 780 Hz and 795 Hz that result from the rover’s internal heat pump used for thermal management. Sounds of the rover’s instruments or pumps propagate through both structural vibrations (microphonics) and acoustic propagation in the atmosphere.

　　The EDL microphone15 was used to record the rover drive on Sol 16 (Extended Data Fig. 1a). Broad, quasi-continuous ‘screech’ signals in the 520–700 Hz, 1.2–1.4 kHz and 1.6–1.9 kHz bandwidths are assumed to arise directly from frictional interaction of the metal wheel tread with surface rocks. Sonorous transients or ‘clanks’ are seen at 13 s with several narrow-frequency components but with a lower total sound intensity than the aforementioned phenomenon. It is suggested that these are structural resonances of mobility system elements (for example, suspension) excited by near-impulsive changes in loading, for example, when a wheel slips off the edge of a rock.

　　The Ingenuity rotorcraft33 provides a localized but moving source of sound on Mars. On Sol 69 during the fourth flight, SuperCam’s microphone recorded the entire 116-s duration of the flight. A prominent acoustic signal, up to 2 × 10−7 Pa2 Hz−1 (1.5 mPa sound pressure level), associated with the BPF at 84 Hz and its first overtone 168 Hz was detected by SuperCam’s microphone (Extended Data Fig. 1b). All phases of the flight are visible but the take-off occurred during a gust (at the rover location) as high as 20 mPa. The BPF clearly stands out but its overtone is much fainter, owing to greater atmospheric absorption at higher frequency. After landing, the microphone captured the blades spinning down.

　　The MOXIE instrument43 operates every 1–2 months to produce a few grams of gaseous O2. The primary objective of these repeated operations is to look for possible degradation of the O2 production efficiency associated with the harsh environment of Mars. MOXIE uses SuperCam’s microphone recordings for independent diagnosis of compressor performance, including precise measurements of the motor rotation rates, as indicated by the fundamental frequency of the observed comb of harmonics. Distinct transitions in Extended Data Fig. 1c, recorded during a night-time run (Sol 81), correspond to commanded changes in motor speed from 50 to 58.3 Hz. The loudest harmonics are near 500 Hz, at which several more frequencies are also excited. This range corresponds to resonant frequencies of the MOXIE instrument, as observed during dynamics testing. Even recorded during one of the quietest times of the day, the amplitude of the signal only reaches 1.5 mPa.

　　Recording LIBS sparks was the main rationale to develop SuperCam’s microphone to infer physical properties of rock targets, such as their hardness13. Typical LIBS sequences consist of 30 laser shots at the same position per observation, fired at 3 Hz (Extended Data Fig. 1d). LIBS operations are monitored by the microphone at a 100-kHz sampling rate for 60 ms around each laser pulse. The mean amplitude of the signal is 0.25 ± 0.08 Pa (1σ) for this shot sequence.

　　The wind, temperature and pressure data are recorded from the MEDA21 instrument. The wind data are acquired up to 2 Hz, pressure at 1 Hz and temperature at 1 or 2 Hz. Wind speed and direction are independently acquired from two individual booms separated by 120° (termed boom 1 and boom 2), for which one is preferred for a given wind direction (see accuracies and resolutions in Extended Data Table 2).

　　The MCD32 provides climate predictions derived from 3D simulations of Mars atmosphere performed with the Mars global climate model developed at the Laboratoire de Météorologie Dynamique (http://www-mars.lmd.jussieu.fr). The Laboratoire de Météorologie Dynamique Mars global climate model is described in ref. 32 but — since then — it has adopted more sophisticated and realistic modelling for the CO2, dust and water cycles, photochemistry, radiative transfer and so on. In this work, we use the climatology scenario45 from the MCD Version 5.3, in which: (1) the simulated spatial and vertical dust distributions are reconstructed from observations during Martian years 24 to 31 without global dust storms (thus representative of standard climate conditions) and (2) average solar extreme ultraviolet conditions are assumed. In this study, the MCD outputs (surface temperature and atmospheric temperature at 2 m above the surface) are provided for daytime local times in increments of 1 h and between Ls = 5.2° and Ls = 104.7° in increments of 10° to capture the seasonal variations in temperature.

　　The microphone data from SuperCam are converted from volts to pascals using the instrument sensitivity for each gain (0.6, 1.3, 5.3 and 21.6 V Pa−1, corresponding to amplification factors of 29 to 972). The microphone’s electronic response function for each gain (bandpass filter between 100 Hz and 10 kHz) is used to correct raw spectra below 100 Hz and above 10 kHz. EDL microphone data are not converted into physical units. PSDs represented in Fig. 1, 2b were computed from a Fourier transform, using a Welch’s estimator. Spectrograms represented in Fig. 2a and Extended Data Fig. 1b, c are computed with a Hanning window of 2 s. Extended Data Figure 1a is computed with a window of 1 s and Extended Data Fig. 1d with a window of 5 ms.

　　The creation of the laser-induced plasma is accompanied by a shock wave, which can be described as an N-wave acoustic pulse46 primarily, a short, approximately 300-μs-long compression/rarefaction acoustic signal. This signal is followed by echoes on nearby rocks and the rover structure, plus diffraction. The whole acoustic signal typically lasts less than 5 ms. A bandpass filter is applied to remove electromagnetic interferences, atmospheric signal below 2 kHz and to reduce noises above 20 kHz. There are residuals of the laser warm-ups but they do not affect the determination of the sound speed (Extended Data Fig. 2).

　　Time series data are used to calculate the local speed of sound. The distance to each target, which is returned by the instrument’s autofocus, is known to an accuracy of ±0.5% (ref. 8). The laser trigger time is known to a few microseconds and the shock wave becomes sonic after 1 μs (ref. 12), which is less than 0.1% of the propagation time. The arrival of the pressure wave is considered to be detected when the signal increases 3σ above the background.

　　The fourth (Sol 69), fifth (Sol 76), sixth (Sol 91) and eighth (Sol 120) flights of Ingenuity were recorded by SuperCam at a 25-kHz sampling rate. We use data from the fourth flight, as this flight came closer to Perseverance than any other flight SuperCam could record. During this flight, Ingenuity climbed to an altitude of 5 m, accelerated to 3.5 m s−1, travelled 130 m at constant height, decelerated, turned around and returned to its base by the same route. Taking off at a distance of 76 m from Perseverance, it came as close as 69 m and moved as far as 123 m from Perseverance (Extended Data Fig. 3b, bottom).

　　On the PSD obtained during the whole recording, the BPF (two times the rotation rate for a two-bladed rotor) at 84 Hz and its first harmonic at 168 Hz are clearly visible above the background, which itself is higher than that of Sol 38a, a very quiet recording on Mars (Extended Data Fig. 3a). There is no other tone above the background. There is a period of atmospheric turbulence up to 56 s into the recording that explains why the spectrum at low frequency is above that of Sol 38a. Discontinuities in the amplitude of the tones at 84 Hz and 168 Hz are visible during the cruise phase of the flight. Such a modulation beat results from the interferences of two signals with slightly different frequencies (about 50 mHz apart), each originating from the two blades that are frequency shifted. The study of this phase shift is outside the scope of this paper.

　　Each tone is fitted by a Gaussian function every 0.5 s. In the main text, we report on the study of the BPF at 84 Hz. The received frequency varies along Ingenuity’s flight (Extended Data Fig. 3b, top) as a function of the variation of the distance range between the rover and the helicopter. The received frequency, the classical Doppler effect, varies by ±1.5%. The fit of this tone, when the atmosphere is quiet (t >60 s），作为范围速率的函数，在源bpf中得出F = 84.44 Hz，C = 237.7±3 m s -1。与第一个谐波产量相似的拟合f = 168.90 Hz，c = 236.9±4 m s -1，与源自BPF得出的值相干。

　　当声波通过大气传播时，一部分声能通过称为大气（或固有）衰减的吸收机制将声能作为热量转移到传播培养基中。该过程已在34,47,48的大气中进行了很大的描述和验证。与分子运动有关的大气衰减取决于波的频率。它可以归因于两个现象。首先，经典的衰减包括由粘性摩擦引起的热量损失以及压缩区域和稀疏区域之间非绝热热扩散的损失。当建立平衡的时间较少时，这种现象对于短周期波就更为重要。经典的衰减与频率平方成正比。第二种现象是分子衰减，这是由于多原子分子的内部自由度的激发（旋转和振动模式），每个分子的自由度（旋转和振动模式）花了一些时间，称为“松弛时间”，以返回平衡。与松弛时间相比，波浪的时间越短，分子必须放松其能量，因此声学能的吸收越大。

　　该理论已应用于火星大气中，以计算经验衰减模型3,4（图4）。尽管火星模型在10 kHz以上非常吻合（由于对CO2的动态粘度和热导率的了解，经典的衰减受到了良好的限制），但根据分子弛豫的方式，它们在不传动和一部分可听见的范围内差异很大。Williams Model4推断了在1 BAR和273 K以上获取的CO2的实验数据。该模型认为分子衰减与频率成比例地增加，在松弛频率（240 Hz）下达到最大值，然后降低为1/f。另一方面，低音和钱伯斯Model3将分子松弛区分为旋转和振动弛豫。旋转放松是通过F平方建模的，就像经典的衰减一样。对于振动放松，该模型认为，在FR下方，振动衰减如F平方生长。在FR上方，振动模式并没有激发，并且振动衰减保持恒定水平。

　　作为补充说明，CO2在1,341 cm -1（ν1对称拉伸），667 cm -1（ν2退化弯曲）和2,349 cm -1（ν3非对称拉伸）时具有三种振动模式。相关的振动温度分别为1,890 K，960 K和3,360 K。在240 K时，每种模式对振动比热的贡献分别为3.7％（ν1），96.3％（ν2）和<0.1％（ν3）在240 K时。这证明了为什么一阶模型只能考虑一阶模型仅考虑ν2弯曲模式对振动特异性热的贡献。