Sea spray aerosol (SSA), also known as sea salt aerosol, refers to a kind of inorganic salt aerosol with sodium chloride (NaCl) as the main component produced on the ocean surface driven by wind^[1,2]. SSA has the characteristic of hygroscopic growth, which makes it possible to indirectly impact the climate system by acting as cloud condensation nuclei (CCN) and ice nuclei (IN), thereby altering the microphysics and radiative properties of clouds (indirect effects)^[3]. Additionally, aerosols can affect cloudiness by heating the atmosphere where clouds reside (semi-direct effect), and reduce the reflectivity of snow, land, and sea ice through deposition and melting. Therefore, in predicting future climate change, aerosols remain one of the largest sources of uncertainty^[4]. More notably, when these micrometer-sized particles come into contact with living organisms, they can easily cause harm to the respiratory and cardiovascular systems. Additionally, the atmospheric turbidity and visibility deterioration caused by the mixing of aerosol particles are easy to cause traffic jams and accidents^[5,6]. In summary, the quantitative and qualitative detection of SSA is crucial for predicting climate change and for preventing and assessing atmospheric environmental pollution^[7].

Currently, climatologists primarily simulate SSA based on wind speed and sea surface temperature^[8,9]. Given the differences in simulation results across different models, there is an urgent need for compositional analysis tools that offer real-time monitoring and remote-sensing capabilities. In addition, the concentration of SSA in the atmosphere is usually less than 10ng/m3, which requires methods with high sensitivity^[10]. Although detection methods such as X-ray fluorescence (XRF)^[11] offer high sensitivity, ranging from μg/m3 to ng/m3, they lack remote-sensing capabilities.

Laser-induced breakdown spectroscopy (LIBS) provides a versatile method for trace element measurements^[12]. The principle is that an ultrashort pulse laser is focused on the surface of the sample, vaporizing the material in the laser ablation zone to form plasma, the electrons in an excited state leap back to lower energy levels and emit characteristic fluorescence^[13,14]. Unfortunately, the current long-range detection accuracy of LIBS does not meet the requirements for atmospheric aerosol detection. The issue arises as the power density of the laser attenuates with increasing penetration depth, particularly after traversing obstacles such as clouds and raindrops^[15]. One of the important factors affecting the sensitivity of spectral detection is the power density.

Femtosecond laser pulses can overcome natural diffraction effects when transmitted through transparent media such as air and glass. Due to the dynamic balance between self-focusing, caused by the optical Kerr effect, and defocusing from the generated plasma in the self-focal region, the laser pulses converge into a plasma channel (filament) with a diameter of about 100 µm, thus enabling long-range transmission (0–10 km)^[16]. The light intensity in the “filament” reaches up to 1013–1014W/cm2, which allows direct excitation of most substances to form plasma^[17].

The current critical issue is how to improve the detection sensitivity of the filament-induced plasma spectroscopy technique at long distances. For example, a distinctive feature of the femtosecond laser filamentation process is spectrum broadening, which results from the combination of various nonlinear effects, such as self-phase modulation, self-steepening, and dielectric ionization^[18]. During filamentation, the supercontinuum spectrum generated by nonlinear effects can easily mask the fingerprint fluorescence. To obtain clean fluorescence, polarization-resolved and time-resolved technologies are commonly used to filter out the early supercontinuum spectra^[19,20], specifically by selecting the appropriate gate delay and gate width. Another issue worth analyzing is the optimal placement of the collection device for spectroscopy. When femtosecond laser filament interacts with the aerosol, amplified spontaneous emission (ASE) enhances the backscattered fluorescence^[21]. This interesting phenomenon fits well with the application requirements of atmospheric remote sensing, where both transmitting and receiving are on the same side.

Previous researchers have conducted a series of studies on aerosols using filament-induced plasma technology. The first issue discussed was the long-range interaction between the filament and the aerosol, specifically whether the filament can penetrate obstacles such as clouds or raindrops in a complex atmospheric environment. Analysis has shown that, due to the intensity clamping effect, only 10% of the energy is concentrated inside the filament, while the remaining energy forms a background energy reservoir around the filament^[22]. The energy stored in the background energy reservoir can be replenished to the filament, facilitating long-range transmission.

In our group’s previous work, we innovatively proposed using an asymmetric incidence method to optimize the astigmatism issue during the long-distance filamentation process of a femtosecond laser. Compared to femtosecond laser incidence at the center of the concave mirror, the intensity of the acoustic signal emitted by the filament increased by 69.5 times, and the detection limit of the Na element in aerosol was reduced by 86%, reaching 320ng/m3^[23]. In this work, we continue to use the asymmetric incidence method to design a long-distance filamentation system. Unlike our previous work, we have thoroughly explored the effects of different interaction regions between the filament and aerosol on the fluorescence efficiency. Additionally, we have employed a high-power industrial-grade ultrafast laser platform for filamentation. As the input power increases, the starting point of the filament is advanced, forming a longer filament.

In this paper, we design a remote trace element spectroscopy detection system. Through theoretical and experimental analyses, considering both the Mie scattering and ASE effects of aerosol on the filament, we find that the optimal interaction region for the filament and aerosol occurs when the aerosol completely engulfs the filament. Using this system, we achieved a detection limit of 0.015ng/m3 for the Na element in SSA at a detection distance of 30 m. This result demonstrates the potential value of the filament in the field of atmospheric remote sensing.

We have established a remote-sensing system for the SSA fluorescence spectrum, as shown in Fig. 1(a). The Magma200 laser, manufactured by Amplitude, operates at a repetition rate of 100 Hz with a pulse energy output ranging from 0 to 200 mJ and a peak power exceeding 400 GW. Figure 1(d) shows the pulse duration of the incident laser, measured using the frequency-resolved optical gating (FROG) technique from Femtoeasy, with a full width at half-maximum (FWHM) of 496.9 fs. The incident laser is first reflected by mirror 1 and then focused by a combined lens system (concave lens, concave mirror). The starting point of femtosecond laser filamentation was controlled at a distance of approximately 30 m from the concave mirror by adjusting the relative distance between the concave lens (f=−15cm; R=2.54cm) and the concave mirror (f=200cm; R=20.7cm). SSAs are generated using an aerosol generator (HRH-WAG3, Beijing HuiRongHe Technology Co., Ltd, China), which is capable of uniformly and stably producing particles with a median mass particle size between 1 and 3 µm. The mass concentration of the Na element ranging from 1.5×10−8 to 0.3mg/m3 in the aerosol particle system was achieved using a NaCl solution with a mass fraction ranging from 5×10−8 to 1% in a 1 m long glass tube (For detailed information, see lines 24–81 of the Supplementary Material).

According to the principle of reversible optical paths, the filament interacts with the aerosol, generating a backward fluorescent signal from plasma radiation. This signal is then collected and focused by a concave mirror. We placed a fiber-optic probe behind mirror 2 (dichroic mirror: @1030 nm reflectance >99%, @589 nm transmittance >94.9%). The fiber transmitted the received fluorescence signal to a grating spectrometer (Omni-λ300, Zolix) with a slit, and the spectral information was output by an Istar-sCMOS camera and recorded to a computer. The gate delay, gate width, and exposure time of Istar-sCMOS were set at 220 ns, 1000 ns, and 30 s, respectively, to obtain spectral images with a high signal-to-noise ratio.

Figures 1(b) and 1(c) show the starting position and length of the filament as characterized by the ultrasound signal and the nitrogen fluorescence signal, respectively. We placed a microphone (V306, Olympus) on the side of the filament to collect the acoustic signal excited by the filament. The acoustic signals were amplified by an ultrasonic pulse receiver (5072PR, Olympus) and displayed on a digital fluorescent oscilloscope (DPO3034, Tektronix Inc), then recorded by a computer. Meanwhile, a convex lens with a focal length of 5 cm, a monochromator (WGD-100, Gang Dong Sci. & Tech. Co., Ltd.), and a PMT (H11902, Hamamatsu) were used to collect the nitrogen fluorescence signal (337 nm) excited by the filament. By moving the ultrasonic probe (megahertz) and fluorescence receiver device point by point (in 5 cm steps) along the axial direction of the filament using a manual displacement table, the spatial position distribution and axial relative intensity distribution of the femtosecond laser filament formation in this experiment can be recorded in detail. After simple data processing, the starting point of the filament was determined to be 29.30 m from the concave mirror, and the ending point was 29.85 m from the concave mirror, resulting in a total filament length of 0.55 m.

It is worth noting that the detection sensitivity of the aerosol signal is related to the fluorescence collection efficiency of the spectrometer; therefore, the following optimizations were undertaken to ensure that the fiber-optic probe was in the optimum position. A small bulb was placed at the focal point of the femtosecond laser to simulate the backward fluorescence. According to the principle of optical path reversibility, light from the small bulb is collected by the concave mirror and converges to the fiber behind the mirror 2. The fiber’s position was incrementally optimized using a precision manual 3D translation table to achieve the maximum signal value, thereby determining the optimum fiber position.

When the incident laser passes through the aerosol particle system, it inevitably undergoes absorption and scattering, attenuating the fluorescence collected at the detection end. Given that the particle size of the aerosol (1–3 µm) is similar to the wavelength of the incident laser (1030 nm), it is classified as Mie scattering. Additionally, based on previous filamentation experiences, different ions and molecules along the forward or backward propagation path of the filament can generate ASE. Neutral Na in SSA can act as a gain medium, enhancing the fluorescence collected at the detection end^[24,25]. Overall, aerosol particles both attenuate and enhance the fluorescence. Therefore, we performed numerical simulations using the nonlinear wave equation (omit the time term) and Mie scattering theory to comprehensively analyze the impact of ASE and Mie scattering on the fluorescence. (For detailed numerical simulation, see lines 83–170 of the Supplementary Material).

Initially, the propagation of forward fluorescence can be expressed by^[26], −2ik∂A∂z=Δ⊥A+2k2n0nnlA−ikαA.(1)

In Eq. (1), A represents the fluorescence distribution, k is the wavenumber corresponding to λ=589nm, nnl is the refractive index change caused by the filament during the fluorescence propagation, n0 indicates the linear refractive index of air, and α describes the air lasing gain and Mie scattering loss. The next step is to analyze the spatial distribution of the Mie scattering signal intensity, that is, to calculate the backscattered fluorescence radiation of the aerosol, which can be represented by^[27], I(θ)=I0F(θ,φ)/(k2d2),|Sj(θ)|2=|∑n2n+1n(n+1)(anπn+bnτn)|2.(2)

Assuming that the aerosol consists of symmetrically distributed spherical particles, I(θ) represents the scattering intensity measured at a distance d from the scattering sphere, and I0 represents the integral fluorescence intensity on the interaction length calculated by Eq. (1). F (θ,φ) refers to the scattering coefficient in direction (θ,φ), where θ points to the scattering angle, and φ is the azimuth. Sj(θ) is the amplitude function of the scattering intensity. αn and bn are Mie scattering coefficients. Thus, based on the geometric relationship of spatial collection angles, the spatially integrated intensity of the backward fluorescence collected by the concave mirror is calculated, as shown by the red curve in Fig. 2(c). We observed that when the laser focuses and forms a filament after entering the aerosol channel, the Mie scattering effect of aerosol particles significantly attenuates the filament’s intensity. Conversely, when the laser focuses and forms a filament before entering the aerosol channel, the losses due to Mie scattering are greatly reduced because of the energy replenishment effect from the background energy reservoir.

Next, we tested the optimal interaction length between the filament and the aerosol in the experiment. The process is as follows: We moved the glass tube filled with aerosol particles (Na element mass concentration of 0.3mg/m3) horizontally to the left, gradually increasing the interaction length between the filament and the NaCl aerosol, as shown in Fig. 2(a). The fluorescence intensity reaches its maximum when the aerosol nearly “engulfs” the filament (interaction length of 45 cm, total filament length of 55 cm), as shown in Fig. 2(b). The experimental results, represented by the blue spheres in Fig. 2(c), closely match the theoretical calculations depicted by the red curve. Our analysis indicates that as the glass tube moves toward the filament, the gain length increases, allowing fluorescence photons to continuously generate and amplify within the gain medium. However, femtosecond laser pulses lose energy due to Mie scattering while transmitting through aerosols before filamentation occurs. Thus, the gain of Na on the filament is limited by Mie scattering.

The Na element detected in our work is in the neutral state. The fluorescence observed at 589 nm corresponds to the well-known D-line transitions of neutral Na (Na I)^[28,29] Specifically, these transitions are: D1 line: transition from the 3pP21/2 state to the 3sS21/2 ground state at 589.6 nm.D2 line: transition from the 3pP23/2 state to the 3sS21/2 ground state at 589.0 nm.

Considering that the resolution of the spectrometer used in our experiment cannot distinguish between the D1 and D2 lines, the spectral lines we measured are collectively referred to as the D-line of Na I.

Finally, we conducted a mass concentration detection limit analysis of SSA and tested the fluorescence spectra of the Na element at various concentrations; the test results are displayed in Fig. 3(a). The emission spectrum of the Na element at 589 nm is clearly visible for each concentration of SSA excited by the filament, and the concentration is positively correlated with the peak intensity of the spectral line. To more clearly illustrate the detection limit of SSA, the integrated intensity corresponding to the spectral line of the Na element at 589 nm was used to generate Fig. 3(b).

The optimal detection limit of Na element in SSA was directly measured at 0.015ng/m3 in the experiment. Here, σ represents the standard deviation, and 3σ indicates that there is a greater than 99.7% probability of the blank sample signal being distributed within this interval. The Na element concentration higher than that corresponding to the signal intensity multiplied by 3σ of the blank sample was selected as the detection limit. After a literature survey, it was found that the detection limit of the Na element in sea salt aerosols, as measured by inductively coupled plasma mass spectrometry (ICP-MS), was recorded at 0.6ng/m3^[30]. By comparison, the detection limits obtained in this study are superior.

The intensity of the fluorescence signal generated by the filament-induced aerosol can be described using the lidar equation as follows^[31]: E(λ,R)=ELK0(λ)T(λ,λLR)ξ(R)A0R2N(R)σF(λL,λ)4πcτd2,(3)where EL represents the energy density in the filament as a function of the propagation distance R, A₀ is the detector area on the telescope focal plan, σ^F(λ_L, λ) is the cross section of the fluorescence excited by λ_L and emitted at λ, which is independent to the range, c is the light speed, and τ_d is the detector response time. K0(λ) denotes the spectral collection efficiency of the acquisition system. T(λ,λL,R) is the atmosphere transmission factor, and ξ(R) is the geometrical overlapping factor, which depends on the geometrical dimensions of the telescope and the divergence of the laser beam (filament). N(R) represents the number density of the excited molecules. Clearly, the relationship between the aerosol particle concentration and the integrated intensity of the fluorescence is proportional, which is consistent with our experimental measurements.

In this paper, we have developed a sea salt aerosol fluorescence detection system that enables long-range and real-time measurements using a filament as the pump source. Through theoretical analysis of the Mie scattering and ASE effects during the interaction of the filament with aerosol, we discover that increasing the filament length enhances the fluorescence intensity. Experimental measurements confirm that the optimal interaction region between the filament and aerosols aligns with our theoretical calculations. We also experimentally determined the optimal Na element detection limit to be 0.015ng/m3 at a detection distance of 30 m. This work serves as an effective guide to the experimental methodologies for remote trace detection of atmospheric aerosols using high-power density femtosecond lasers.