A novel laboratory methodology to evaluate the particle extinction-to-backscatter ratio or particle lidar ratio of aerosols Authors: Alain Miffre, Clément Pacoret-Berenger, Adrien P Genoud, Patrick Rairoux Published: 2026 ABSTRACT AbstractWhile atmospheric particles are ubiquitous in nature and contribute to the Earth’s radiative budget, their backscattering and extinction remain insufficiently quantified due to their complexity in size, shape and chemical composition. In this paper, a novel laboratory methodology is proposed to evaluate the particle extinction-to-backscatter ratio or lidar ratio of aerosols. This methodology, called LR-lab, is based on a robust light backscatter experiment at 180.0° angle, enhanced with a cavity ring down experiment to evaluate the particle extinction. No assumption is required on the particle size, shape, nor complex refractive index. Our methodology is validated on homogeneous spherical water droplets, which follow the analytical Mie theory, for different water droplets sizes and number concentrations. The particle extinction-to-backscatter ratio of mineral dust, a heterogeneous mixture of non-spherical particles, is then evaluated at 532 nm wavelength in laboratory. The dust lidar ratio is found to be equal to (47 ± 2) sr while the dust depolarization ratio is equal to (33.6 ± 0.4) %. By proposing a combined laboratory evaluation of the particle lidar and depolarization ratios, we believe our methodology can meet the growing need for lidar-based aerosol typing and for validation of light scattering numerical models on complex-shaped particles.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-61997-3.Keywords: Extinction-to-backscatter ratio, Lidar ratio, Backscattering, Extinction, Mineral dustSubject terms: Environmental sciences, Optics and photonics, Physics Introduction As underscored in the latest IPCC report1, atmospheric particles contribute to the Earth’s radiative budget by scattering and absorbing electromagnetic radiations, which reduce the amount of energy, a physical process called extinction2,3. Among all scattering directions, the backward direction is of prime importance as it governs cooling albedo effects, while absorption leads to a warming of the Earth’s climate. Hence, the particles extinction-to-backscatter ratio3, i.e., the ratio of the particle extinction and backscatter coefficients, relates to the particles’ ability to warm or cool the Earth’s climate. Quantifying this ratio is then essential to account for the role of aerosols on global warming1,3. This task is however challenging since aerosols vary in time and space over the globe through a complex vertical layering, which has to be revealed. In this context, lidar remote sensing instruments, based on light backscattering and extinction, are coveted and represent a major source of global data on atmospheric particles optical properties. Moreover, aerosols present a wide range of sizes and a high degree of complexity in shape, often irregular sometimes with sharp edges, as for mineral dust, which prevents from analytical solutions to the Maxwell’s equations, except for simplified geometries such as homogeneous spheres, for which the analytical Mie theory can be applied2,3, but are seldom encountered in nature. In the literature, the particle extinction-to-backscatter ratio has been addressed in field from ground-based4–9 and satellite-based10–13 lidar remote sensing instruments. Light backscattering and extinction have then been described in a review paper with a lidar perspective14. Retrieving the particle backscatter and extinction coefficient from a single lidar measurement is however an under-constrained problem. The particle extinction-to-backscatter ratio or lidar ratio has hence been introduced in lidar studies by Klett, Fernald and Sasano15–17 to overcome this constrain. Indeed, the solution introduced by the mathematician Klett requires prior knowledge of the particle lidar ratio. This robust formalism has since been broadly adopted in the analysis of elastic lidar observations4–13. The HSRL and Raman lidar methods18,19 enable independent measurement of the particle backscatter and extinction, hence allowing to retrieve the particle lidar ratio in field. Also, a precise knowledge of the particle lidar ratio is key for satellite missions, such as CALIPSO or EarthCARE, assuming an aerosol-type dependent lidar ratio. Moreover, prior knowledge of the lidar ratio for a given particle type, such as mineral dust, is required to apply lidar partitioning algorithms20–24, which allow revealing the complex vertical particle layering associated with specific particle types. To retrieve the particle extinction-to-backscatter ratio, an alternate approach is to take benefit from light scattering and extinction numerical simulations, which are now becoming more and more accurate, with major advances25–33. Their ability to reproduce the particle extinction-to-backscatter ratio of complex-shaped particles, such as mineral dust, however remains to be assessed as the morphology of these atmospheric particles is difficult to represent mathematically25–33. In this context, laboratory experiments on complex-shaped particles are looked for to provide quantitative evaluation of the particle extinction-to-backscatter ratio, while accounting for the particle complexity in terms of size, shape, and complex refractive index (and thus particle type). In laboratory, aerosols light extinction has been addressed by digital holography34, photo-acoustics35 or by using a ring down optical cavity36–45. Complementary use of polar nephelometry has allowed for simultaneous measurements of particle extinction and scattering46–52, without however addressing the specific backward scattering direction, as being truncated for the purpose of measuring light extinction46–50. However, for precise evaluation of the particle extinction-to-backscatter ratio in laboratory, addressing the 180.0° backward scattering direction is mandatory. Indeed, for complex-shaped particles such as mineral dust, no light scattering model2,3 can predict the deviation of the scattering matrix elements from their value at backscattering angle (θ = π) even at near backscattering angles (θ < π)53–56. The 180.0° backward scattering direction is indeed particularly sensitive to the particle inhomogeneity and non-sphericity57,58. Moreover, as published by our group59, addressing aerosol light backscattering at 180.0° angle in laboratory is difficult as it requires overcoming two challenges, that of the backward geometry at 180.0°, which has to be set without blocking the incident light with the detector, and that of the weakness of the particle backscattered light intensity, which intrinsically relates to the particle light backscatter efficiency. This is particularly the case for non-spherical particles, for which particle backscattering is weaker57. Also, to deal with aerosols (airborne particles), particles should be probed in laboratory ambient air, as in lidar applications, rather than deposited on a substrate. We overcame these difficulties by measuring the particle depolarization ratio of various aerosols in laboratory at 180.0°58–63, without however evaluating their particle extinction-to-backscatter ratio. This latter evaluation is even more demanding. To figure out this challenge, for 1 μm water droplets over a 1 cm path length, the Mie theory gives an extinction efficiency 2,3 of around 4, so single-pass Beer–Lambert transmission changes stay below 0.3 %, even at 105 particles.cm− 3, making single-path extinction measurements challenging. Longer paths or higher concentrations could improve sensitivity, but long paths are limited by sample inhomogeneity, while higher concentrations may cause multiple scattering, thus biasing the evaluation of the particle extinction-to-backscatter ratio. Conversely, lowering the particle number concentration may render the particle backscatter measurement inaccessible as the backscatter efficiency 2,3 may already be low (only about 0.4 in our example), especially for non-spherical particles for which backscatter is weaker57. Therefore, setting the aerosols sample path length and the particle concentration is one of the challenges to be addressed. Moreover, the temporal and spatial fluctuations encountered in the aerosol generation have to be minimized to allow an accurate evaluation of the particle extinction and backscatter at 180.0°, by keeping the particle size distribution stable over the whole duration of the acquisition, as the particle extinction-to-backscatter ratio is size dependent. In this paper, we report a novel laboratory methodology allowing to accurately evaluate the particle extinction-to-backscatter ratio or particle lidar ratio of aerosols in laboratory. This methodology, hereafter called LR-lab for Lidar Ratio-laboratory methodology, is based on a robust laboratory light backscatter experiment at 180.0° angle, also measuring the particle depolarization ratio, enhanced with a cavity ring down experiment to evaluate the particle extinction. To our knowledge, no published attempt exists to evaluate the particle extinction-to-backscatter ratio of aerosols in laboratory. The following will present the performances of the LR-lab methodology and propose the first laboratory evaluation of the lidar ratio of mineral dust, a heterogeneous mixture of non-spherical particles. Results The LR-lab methodology is able to evaluate the dust particle extinction-to-backscatter ratio of particles embedded in laboratory ambient air, with an accuracy of . The corresponding dust particle depolarization ratio is also evaluated. Principle of the LR-lab methodology The principle of the LR-lab methodology is illustrated in the Fig. 1 experimental set-up, where particle backscatter and extinction are addressed at 532 nm wavelength with a ns-pulsed laser source, as in most lidar remote sensing experiments. Scheme of the LR-lab experimental set-up for evaluating the extinction-to-backscatter ratio or lidar ratio of aerosol particles in laboratory ambient air. The backscatter geometry, achieved with a retroreflecting polarizing beam-splitter cube ( PBC), allows operating at 180.0° angle. To gain in accuracy, the particle backscattered light intensity is measured for a set of incident polarization states, obtained by rotating the quarter-wave plate ( QWP). Conversely, the extinction experiment is achieved with a ring down optical cavity, composed of two highly reflective mirrors () and (). A mode-matching optics, composed of lens () and (), is used to promote the propagation of the fundamental transverse mode through the optical cavity. As the laser in the optical cavity induces side scattering in the backscatter detector, the particle backscatter and extinction have to be alternately measured on a particle ensemble of determined refractive index and size, as controlled by the particle sizer over the eight minutes duration of the acquisition. The ns-pulsed laser source, emitting at 532 nm wavelength as in most lidar instruments, is then split between the backscatter and extinction parts of the experiment using halfwave plate and polarizing beamsplitter cube . Particle backscatter experiment at 180.0° Particle backscatter is addressed as previously published by our group58–63 and here briefly recalled for the sake of clarity. To overcome the intricacy imposed by the backward scattering geometry, we exploited the sensitivity of light backscattering to the light polarization property, by precisely aligning a retroreflecting polarizing beam splitter cube on the optical pathway from the laser source to the particle backscattering medium, to cover the exact lidar backward scattering angle with accuracy: , with an uncertainty determined from the field of view of our detector59. The intricacy set by the weakness of the particle backscattered light intensity is overcome by conducting a time-of-flight experiment, allowing to discriminate the particle backscattered light from the background stray light. The particle backscattered light intensity is then measured from the change in backscattered light intensity in the absence then in the presence of the particles. To gain in accuracy, is measured for a set of incident polarization states, obtained by rotating a quarter-wave plate ( QWP). From , one may retrieve accurate determinations of the particle depolarization ratio , as previously published by our group58–63. To derive the particle extinction-to-backscatter ratio, an extinction experiment is carried out in addition, as described below. Particle extinction experiment using a ring down optical cavity To access to the low particle extinction coefficient, a multiple-path extinction experiment is considered, based on a ring down optical cavity, composed of two highly reflective concave mirrors , arranged as in Fig. 1. A mode-matching optics, composed of two lenses , is used to promote the propagation of the fundamental transverse mode through the optical cavity. Due to losses within the optical cavity and through the end mirrors36–44, the light intensity transmitted through the end mirror decreases exponentially in time, with a characteristic time known as the cavity ring down time. The particle extinction coefficient αp [in m− 1 is then retrieved from the change in ring down times rates 1/τp – 1/τ0, from 1/τ0 (in the absence of particles) to 1/τp (in their presence), regardless of the laser intensity or detector gain settings. LR-lab methodology for particle extinction-to-backscatter ratio retrieval Unlike αp, the particle backscattering coefficient βp [in m− 1.sr− 1], being derived from , directly depends on the laser intensity and on the detector gain settings. Hence, from , one can determine only the product βp, where the factor , which accounts for the laser intensity and the detector gain settings, is difficult to know accurately. Calibrating βp with particle counters is also inaccurate, as systematic instrument errors40–42 would propagate to βp and consequently to . However, when using relative measurements of the product βp, the factor cancels, enabling access to ratios of βp from . In LR-lab, the particle backscattering coefficient βp is therefore expressed relative to a reference backscattering coefficient βref, thus considering the relative backscatter coefficient Rβ = βp)βref) which is independent of (subscripts ref and p denote the reference and considered particles, respectively). βref is hence the reference backscatter coefficient used to normalize βp. By likewise introducing corresponding extinction coefficients αref and αp, the targeted particle extinction-to-backscatter ratio can be determined as follows: Therefore, within LR-lab methodology, the particle extinction-to-backscatter ratio is evaluated from the change in particle backscatter from βref to βp, and from the corresponding change in ring down times rates, from 1/τref – 1/τ0 to 1/τp – 1/τ0. Moreover, to accurately evaluate the reference lidar ratio αref/βref appearing in Eq. (1), the reference particles are chosen to be homogeneous and spherical, to allow applying the analytical Mie theory2,3. The particle extinction-to-backscatter ratio is then finally retrieved by applying Eq. (1), which is then the LR-lab core equation. Validation of the LR-lab methodology The validation of the LR-lab methodology consists in evaluating the particle extinction-to-backscatter ratio of homogeneous spherical particles then comparing it with the analytical Mie theory2. Spherical water droplets were then used as targeted particles, thus following the approach published in37 for cavity ring-down experiments, with corresponding particle backscattering and extinction acquisitions given in supplementary information S4, 5 and data available statement. Within experimental error bars (see Discussion section), the laboratory-retrieved water droplets lidar ratio accurately agrees with the Mie-predicted water droplets lidar ratio. Moreover, this validation experiment has been carried out for several water droplets size distributions and number concentrations, thus proving the ability and the precision of our LR-lab methodology to evaluate such a particle lidar ratio. Laboratory evaluation of the dust particle lidar ratio using LR-lab methodology As a case study, the LR-lab methodology has been applied to mineral dust particles, a heterogeneous mixture of complex-shaped particles. Arizona Test Dust was chosen as a natural mixture of several oxides present in mineral dust, with corresponding backscattering and extinction acquisitions given in the Methodology section and data available statement. The main finding of this paper is then the evaluation of the lidar ratio of mineral dust in laboratory, , by applying the LR-lab methodology. The achieved precision on is hence 4.3% and related to the achieved precision in the experimental evaluation of the dust particle backscatter and extinction. To the best of our knowledge, this is the first time that a dust particle lidar ratio is evaluated in laboratory. The corresponding dust size distribution, with its modal parameters, is given in Fig. S2. The laboratory-retrieved dust lidar ratio substantially differs from the Mie-predicted dust lidar ratio, , attributed to the non-sphericity and heterogeneity of mineral dust (see Discussion section). Moreover, the intensive character of the laboratory-retrieved dust lidar ratio has been carefully verified for different particle concentrations, as presented in the Discussion section. In addition, the corresponding dust depolarization ratio has been found equal to . Discussion This section discusses on the advantages, limitations and underlying assumptions behind the LR-lab methodology. The laboratory-retrieved particle lidar ratios are compared with field-retrieved and numerically-retrieved particle lidar ratios. Advantages and limitations of the LR-lab methodology The LR-lab methodology presents several distinct advantages. Firstly, following Eq. (1), the accuracy on primarily depends on the precision in the experimental evaluation of the backscatter ratio Rβ = βp /βref and in the corresponding extinction ratio Rα = αp /αref. Hence, in LR-lab, the ratio of the optical cavity length to the sample length does not need to be accurately known as Rα is independent on , in contrary to αp and αref taken separately. Determining the latter would require precise knowledge of , with potential errors documented in the literature on cavity-ring-down experiments41,42. Likewise, the scattering volume does not need to be quantified as long as it is identical for the reference and the targeted particles. This is a strength of our novel methodology. Secondly, the particle extinction-to-backscatter ratio is determined without demanding prior knowledge of any particle concentration. While αp, αref, βp and βref depend on the particle number concentration, Rα, Rβ, and can be determined without requiring any particle concentration measurement. Systematic errors inherent to particle counters40–42 are thus avoided, which brings accuracy on . Thirdly, the LR-lab methodology does not either require assumptions on the particle size, shape, nor complex refractive index of the considered particles. The particle size is however measured over the eight minutes duration of the acquisition as a control measurement, as particle extinction-to-backscatter ratios are size dependent. Fourthly, the fluctuations in the aerosol generation, which inevitably occur during the duration of the acquisition, are accounted for in the LR-lab methodology by considering the resulting fluctuations in the ring down time rates as ultimate evaluations of 1/τp – 1/τ0 (see Methodology section). Nevertheless, the accuracy on depends on the accuracy achieved in the numerical evaluation of and to apply the Mie theory, the reference particles must be homogeneous perfect spheres, of known size and complex refractive index. A key advantage of our methodology is that the particle sphericity can be checked over the whole duration of the backscatter experiment (see Methodology section). Monodisperse particles were left out as reference particles since, as we experienced, the particle depolarization ratio of micron-sized silica spheres was found to differ from the expected zero-value for perfect spheres, thus highlighting discrepancies with Mie theory. In the literature, such discrepancies are attributed to the formation of dimers or/and higher aggregates during the particle aerosolization, as reported in cavity ring-down40 and in light backscattering experiments62. The main limitation in the evaluation of then stems from the precision achieved in evaluating the size distribution of reference particles with the particle sizer. Light backscatter and extinction from reference particles Following Fig. 1, the reference particle backscattering (βref) and the reference extinction (αref) coefficients are independently measured on a particle ensemble, using DEHS-particles as reference particles (see Methodology section). The linearity between the reference backscattering (βref) and the reference extinction (αref) coefficients has then been carefully verified for different reference particle concentrations, as shown in Fig. 2, where, in line with the Results section, βref is plotted through the values of βref while αref is plotted through the change in ring down time rate 1/τref – 1/τ0 (the underlying assumption of linearity of the cavity ring down experiment is discussed and validated in Fig. 3). Table 1 gathers the retrieved values of βref (for backscattering) and 1/τref – 1/τ0 (for extinction) for four distinct particle concentrations. Interestingly, this experimental dataset can be fitted with a straight line passing through the origin (proving the linearity between αref and βref at different particle concentrations. Therefore, the measured particle backscatter and extinction consistently relate within LR-lab, with no observed systematic bias, despite being obtained from independently performed measurements of particle backscatter and extinction. Moreover, by applying the Mie theory to the reference spherical particles with the size distribution presented in Fig. 5, the error associated with approximating the scattering phase function at 180.0°, instead of integrating it over the range from 179.8° to 180.0°, was quantified. The resulting error was below 0.02%, therefore negligible compared with the uncertainty in the backscattering measurements on βref reported in Table 1. Reference particle backscattering and extinction for four distinct particle number concentration of reference particles. βref [a.u.] evaluated at lidar 180.0° angle, is proportional to the reference backscattering coefficient βref. Conversely, the corresponding change in ring down times rates 1/τref – 1/τ0 [µs− 1], evaluated from the cavity ring down experiment, is proportional to the reference extinction coefficient αref. A linear fit that passes through the origin is obtained , with no observed systematic bias, highlighting that, in our experiment, reference particle backscattering consistently relates with reference particle extinction. (a) Observed linearity between backscattering and particle number concentration. βp is plotted as a function of the particle number concentration for reference particles (left graph) and for dust particles (right graph). The particle number concentration is plotted in arbitrary units as being proportional to the particle scattered light intensity , measured at a scattering angle q0, different from 180.0° angle, from a polarization-insensitive detector (see supplementary information S.I.−3). (b) Same as (a), except that βp is replaced with 1/τp – 1/τ0. Table 1Reference particle backscattering and extinction. Laboratory evaluation of βref, 1/τref – 1/τ0 µs− 1], with corresponding retrieved reference particle depolarization ratio Data point #3 corresponds to Fig. 4a, with other curves for other data points given in data availability statement). In between the first and the fourth data point, the particle concentration is multiplied by a factor of about four, as obtained from the independent scattering measurement explained in Supplementary Information.Data point #1234 Reference particle backscattering and extinction. Laboratory evaluation of βref, 1/τref – 1/τ0 µs− 1], with corresponding retrieved reference particle depolarization ratio Data point #3 corresponds to Fig. 4a, with other curves for other data points given in data availability statement). In between the first and the fourth data point, the particle concentration is multiplied by a factor of about four, as obtained from the independent scattering measurement explained in Supplementary Information. Measured particle size distribution of reference particles (example of DEHS droplets), as measured with SMPS (triangles) and APS (full circles) instruments. The size distribution is fitted with a triple-mode log-normal distribution to evaluate corresponding modal parameters (). In addition, the observed linearity between βref and 1/τref – 1/τ0 can be exploited to improve the accuracy on the particle lidar ratio retrieved from Eq. (1), by utilizing the slope illustrated in Fig. 2, rather than relying on any one of its four individual data points. Indeed, Eq. (1) providing the particle lidar ratio can be rewritten by highlighting the contribution of reference particles to : where βref/(1/τref – 1/τ0) accounts for all the dependence on reference particles. The reference lidar ratio appearing in is determined by applying the Mie theory, as allowed by the zero-value of the laboratory-measured (see Table 1), and was found equal to (see Methodology section). Consequently, is a constant quantity in our experiment. Moreover, an accurate determination of can be derived from the slope of the straight line observed in Fig. 2 and we found W. m− 2.sr. µs− 1, when expressing cavity ring down times in microseconds. differs from unity since and are respectively expressed in units of βref and 1/τref – 1/τ0. Therefore, in LR-lab, the particle lidar ratio can be evaluated by applying Eq. (2), by taking care to express cavity ring down times in microseconds and backscatter measurements in the same unit as βref, thus maintaining constant for all backscatter acquisitions. Following Eq. (2), three uncertainty sources affect the particle lidar ratio evaluation: that in the constant of our experimental set-up, that in the backscatter measurement βp, that in the extinction measurement 1/τref – 1/τ0. The latter uncertainties are given in Table 2 (for water droplets) and Table 3 (for mineral dust), with details given in the Methodology section. Table 2Water droplets particle backscattering and extinction. is retrieved by applying Eq. (2) to targeted water droplets particles. Data point #2 corresponds to Fig. S.I.1 (see supplementary information S.I.−4), with other curves for other data points given in data availability statement. is obtained by applying the Mie theory (see supplementary information S.I.−5, where the modal size parameters are provided).Data point #1234 Water droplets particle backscattering and extinction. is retrieved by applying Eq. (2) to targeted water droplets particles. Data point #2 corresponds to Fig. S.I.1 (see supplementary information S.I.−4), with other curves for other data points given in data availability statement. is obtained by applying the Mie theory (see supplementary information S.I.−5, where the modal size parameters are provided). Table 3Mineral dust particle backscattering and extinction. The dust lidar ratio is retrieved in laboratory by applying Eq. (2) to targeted dust particles. Data point #1 corresponds to Fig. 4b, with other curves for other data points given in data availability statement. Each datapoint refers to the same dust particle size distribution, given in Fig. S.I.2, with corresponding modal parameters.Data point #1234 Mineral dust particle backscattering and extinction. The dust lidar ratio is retrieved in laboratory by applying Eq. (2) to targeted dust particles. Data point #1 corresponds to Fig. 4b, with other curves for other data points given in data availability statement. Each datapoint refers to the same dust particle size distribution, given in Fig. S.I.2, with corresponding modal parameters. LR-lab validation experiment at different particle size and number concentration To validate our methodology, homogeneous spherical water droplets, which exactly follow the Mie theory, have been generated at different size distributions and number concentrations. Table 2 gathers the corresponding values of βwdp (for backscattering) and 1/τwdp – 1/τ0 (for extinction), together with the laboratory-retrieved value of , obtained by applying Eq. (2). Within experimental error bars, the laboratory-retrieved value of agrees with that obtained by applying the Mie theory, designated by in Table 2. Intensive property of LR-lab retrieved dust particle lidar ratios To confirm the intensive property of our laboratory-retrieved value of the dust particle lidar ratio, the dependence of with the dust particle concentration has been investigated, as presented in Table 3. Interestingly, all laboratory-retrieved values of agree within our experimental error bars, thus verifying the expected intensive property of the retrieved dust lidar ratio. Moreover, a precise laboratory evaluation of the dust lidar ratio can then be derived from the weighted-average of our set of four laboratory evaluations of the dust lidar ratio. We thus found . The corresponding dust particle depolarization ratio is equal to at 532 nm wavelength. The uncertainty on the retrieved dust lidar ratio is dominated by the uncertainty in the change in the cavity ring-down times and in the determination of the constant of our experimental set-up. The uncertainty on βp interestingly corresponds to the lowest uncertainty. Backscattering and extinction underlying assumptions Our methodology assumes that particle backscattering and extinction occur under the single-scattering approximation64. This approximation is actually valid and reliable in our experimental set-up, firstly due to the very small field of view of 0.2° (3.5 milliradians) of the backscatter experiments. Moreover, in case of multiple scattering, light depolarization is no longer an intrinsic single-particle property but becomes a bulk property depending on the particle density and hence on the particle concentration65,66. By contrapositive, the absence of any dependence of the PDR on the particle concentration in Tables 1, 2 and 3 implies that multiple scattering is negligible for all particles considered. Also, the observed variations of βref of βdust with the particle number concentration, to be seen in Fig. 3a, remain linear, as expected under the single-scattering approximation. Furthermore, similar linear relations are observed in Fig. 3b between the changes in ring down time rates and the particle number concentration for both reference and considered dust particles. Hence, while the change in ring down times rates from 1/τref – 1/τ0 to 1/τdust – 1/τ0 is key in the LR-lab methodology (see Results section), the assumption of linearity of the cavity ring down experiment is valid for the reference particles on the one hand, and for the considered dust particles on the other hand. It confirms key assumptions used in the cavity ring down experiment : the absence of significant multiple scattering, the linearity of losses, and the negligible recoupling into the fundamental cavity transverse mode37,40,42 (the fraction of the light intensity that can be forward scattered and re-trapped by the optical cavity, evaluated with Mie theory, was found negligible for both reference and considered dust particles). If the single scattering approximation was not fulfilled, the linearity of with the particle concentration observed in Fig. 3 would not be established down to the origin point corresponding to the absence of particles. To provide a more quantitative criterion, we followed the conclusions from Mishchenko et al. (2007)64: for the single-scattering approximation to be valid, the particles ensemble must be sufficiently dilute (packing density < < 1) and sufficiently sparse (k1⟨d⟩≫1, where k1 is the wavevector in the surrounding medium and d is the average inter-particle distance). In our experiment, the packing densities are indeed low (7.4 × 10− 6 for reference particles, 7.4 × 10− 6 for dust), calculated from corresponding size distributions (see Fig. 5 and S.I.2b), by considering a 106 cm− 3 as a high estimate. The mean distance between neighbour particles was calculated from67 to get k1⟨d⟩ = 1 × 108 > > 1 for reference particles, k1⟨d⟩ = 5 × 107 > > 1 for dust. Comparison of LR-lab retrieved lidar ratios with lidar field experiments The laboratory-retrieved value of can be compared with field lidar measurements of dust lidar ratios at 532 nm wavelength4–13. In field, retrieved dust lidar ratios range from 30 to depending on the dust chemical composition and on the dust particle size4–10,12,13. The typical dust lidar ratio is equal to for most dust regions, except the western Sahara4. For Arizona dust, field measurements are more seldom but a value of was reported68,69 for the Arizona dust extinction-to-backscatter ratio at sizes comparable to ours (see Fig. S.I.−2), which agrees with our laboratory measurement within experimental error bars. Moreover, our combined evaluation of the dust lidar ratio and of the dust particle depolarization ratioin laboratory can be compared with lidar-based aerosol typing8–12. To our knowledge, it is the first time that such aerosol typing is verified in laboratory. Comparison of LR-lab retrieved lidar ratios with Mie and T-matrix numerical simulations The laboratory-retrieved value of can be likewise compared with light backscattering and extinction numerical models25–33 used to mimic the laboratory-retrieved dust lidar ratio. Application of the Mie theory to the modal parameters given in Fig. SI-2b led to a Mie-predicted dust lidar ratio of 14.3 sr at 180.0° lidar backscattering angle, using a for the refractive index of Arizona dust, from Di Biaggio et al70. The observed discrepancy with the laboratory-retrieved value of 47 sr may be questioned. To account for the uncertainty in the dust refractive index and in the bi-modal dust size distribution, a sensitivity study was performed over a nine-dimensional parameter space (6 for the double-mode size distribution, 2 for the dust refractive index, 1 for the dust dynamic shape factor), explored through a Sobol-based global sensitivity analysis. The Mie-predicted dust lidar ratio was then found equal to (14.4 ± 4.6) sr, (90 % confidency interval: [7.3–21.5] sr), driven almost exclusively by the real part of the refractive index. Likewise, integration over the experimental angular range (180.0 did not modify the retrieved value of the dust lidar ratio as the scattering phase function varied by no more than 0.05 % in between 179.8° and 180.0°. Therefore, the observed discrepancy with Mie theory can be safely attributed to the limitations of Mie theory: dust particles are irregular and heterogeneous and therefore their lidar ratio deviate from the homogeneous spherical-particle assumption. Further computations, such as T-matrix computations, are then highly desirable, though beyond the scope of this paper. To the best of our knowledge, no such numerical simulation exists on Arizona dust. However, numerically simulated dust lidar ratios at 532 nm wavelength lie in the same range as our laboratory finding, in agreement with our laboratory methodology. Moreover, the achieved accuracy achieved on the laboratory-retrieved value of the dust lidar ratio may be used to discuss on the applicability of such numerical models25–33. Such a comparison however requires intensive laboratory work and numerical simulations, which are far beyond the scope of this paper. The measured dust size distribution is however given in supplementary information S.I.−5 to initiate such a cooperation. Outlooks of the LR-lab methodology By proposing a combined evaluation of the particle lidar ratio and of the particle depolarization ratio in laboratory, we believe our methodology can meet the growing need for lidar-based aerosol typing8–12 and for validation of numerical models25–33 on complex-shaped particles. Outlooks of this work are then numerous, ranging from applying LR-lab to other particle sizes and complex refractive indices to EarthCARE satellite inversions at 355 nm wavelength10,12. Still, such promising future directions highlight the value of the proposed LR-lab methodology. Methodology This section provides extensive explanations on the LR-lab methodology, from the observation of particle light backscatter and extinction in laboratory up to the implementation of the LR-lab methodology through a concrete example. The dust lidar ratio is equal to . Its intensive character is discussed in Table 3 of the Discussion section while the methodology to retrieve is described in the dedicated section. Light backscatter by particles Laboratory set-up for particle light backscatter at 180.0° Our backscatter experimental set-up is schemed in Fig. 1. Light backscatter is observed with a ns-pulsed laser emitting at 532 nm wavelength as in most lidar applications. The laser source is a diode-pumped frequency doubled Nd: YAG laser source, with 7 ns pulse duration, as published in71.” The laser divergence is low (< 1 milliradian) and ns-laser pulses are emitted at a 1 kHz pulse repetition rate. The backscatter geometry is set by precisely aligning a well-specified polarizing beam splitter cube ( PBC, on the optical pathway from the laser source to the particle scattering medium with a precision of 1 mm out of 10 m, to cover the lidar exact backscattering angle with accuracy: . The experiment is carried out in laboratory ambient air to avoid using chamber walls and windows, with inherent AR-coatings, that may provoke a strong backscattering signal or/and modify the polarization state of the backscattered radiation, which is to be accurately analysed. Likewise, to deal with aerosol particles, the particles are not deposited on a substrate but are probed in their surrounding medium, the laboratory ambient air. As schemed in Fig. 1, the s-component of the backscattered electromagnetic radiation is detected after retroreflection on the PBC. Systematic errors affecting the particle backscattered light intensity were minimized58. For instance, polarization cross-talks were minimized by inserting a secondary PBC (not represented in Fig. 1) on the return path after the retroreflecting PBC. To gain in accuracy, the particle backscattered light intensity is recorded over a complete rotation of a quarter-wave plate ( QWP), labeled with the ψ-rotation angle introduced in Fig. 1(ψ is the angle between the fast axis of the QWP plate and the horizontal scattering plane). Retrieving the particle backscattering coefficient in terms of βp The elastic backscatter process is described in the framework of the scattering matrix formalism2,3 allowing to account for the light polarization property. At specific 180.0° backscattering angle, the particle scattering matrix reduces to the particle backscattering matrix [Sp(π)], which comprises only two non-vanishing elements, (π) and (π), where πdesignates the 180.0° backscattering angle. As explained in supplementary information ( SI.1), a laboratory experiment at specific 180.0° backscattering angle is mandatory for precise evaluations of the particle lidar ratio . Indeed, in the absence of analytical solution to the Maxwell’s equations for complex-shaped particles, no angular light scattering model exists to accurately derive the backscattering matrix from the scattering matrix at near backscattering angles (. In particular, the Mie theory cannot be applied. Following Fig. 1, the expression of the backscattered light intensity can then be obtained by projecting the unitary raw vector () = [1, 0, 0, 0] on the successive Mueller matrices encountered by the ns-laser pulse on its optical pathway from the polarizing beam-splitter cube ( PBC) to the particle scattering medium, then back to the backscatter photodetector providing . After a few calculations which are detailed in supplementary information ( SI.1), finally expresses as follows: where the and -coefficients only depend on (π) = (π)/(π), with(π) while 3(π) −1. Hence, the -curve exhibits null minima only when spherical particles (for which (π) =1) are involved, as for reference particles. In contrast, the -curve exhibits non-vanishing minima when non-spherical particles ((π) < 1) are considered, as for mineral dust. Recording the -curve over a complete -rotation of the QWP is then key for verifying that the size and the shape of our aerosol sample did not vary during the acquisition (otherwise, extrema would vary58–63. By adjusting the experimental datapoints with Eq. (3), the ratio , which is independent on the factor , could be accurately determined to retrieve the particle depolarization ratio in laboratory at 180.0° backscattering angle, as previously published58–63. Moreover, the factor appearing in Eq. (3) noticeably depends on the detector gain settings (see supplementary information ( SI.1) for detailed expression of ). It is hence difficult to determine with accuracy. As a result, the particle backscattering coefficient [in m−1 sr−1 could not be directly retrieved from by applying Eq. (3). Nevertheless, could be accurately retrieved from by noting that , so that: where the coefficients βp and βp express the coefficients and in βp-unit, as set in Eq. (3). By adjusting the variations of with Eq. (3), and can be directly retrieved to evaluate βp by applying Eq. (4). The uncertainty on βp is then calculated from the uncertainties affecting and . To reduce statistical uncertainty, the variations of were then recorded with a 4° angular step and 8 files per angular step. Each experimental data point was then weighted with its statistical error, to retrieve and with corresponding uncertainties. Applying the same methodology to the reference particles allowed to likewise evaluate βref with corresponding uncertainty determined from the recorded variations of . In all these backscatter experiments, the factor was therefore maintained constant, by operating at constant laser power density and detector gain settings, while ensuring a constant overlapping volume of the aerosol sample with the backscattering laser beam, by injecting the particles through a fixed nozzle, located at a fixed distance from the detector. may vary due to fluctuations in laser power, beam alignment and optical losses. However, all the experiments were performed in laboratory cleaned air-conditioned and light backscattering from laboratory ambient air is subtracted to address particle backscattering. Moreover, fluctuations in the laser power are negligible as our laser source is a diode-pumped solid-state Nd: YAG laser. In between the reference particles experiment and that on the dust particles, fluctuations in the laser power density are likewise negligible: if not, the light intensity transmitted through the cavity end mirror would exhibit significant fluctuations. The beam alignment in the extinction experiment has been carefully checked by measuring the ring-down time τ0 of the empty cavity, while for the backscatter experiment, contra-propagating laser beams, aligned through a series of irises along the optical path, were used by observing diffraction rings. Light extinction by particles Laboratory experimental set-up for particle extinction using cavity ring down methodology Our cavity ring down experiment is schemed in Fig. 1. For accurate evaluate of the particle lidar ratio, the same ns-pulsed laser is used as for the backscatter experiment, using polarizing beam-splitter cube () and half-wave plate () to spread the laser power between the backscatter and extinction parts of the experimental setup. The optical cavity was placed along the x-axis, perpendicularly to the backscatter z-axis, to avoid obstruction of backscattering measurements. The optical cavity was composed of two highly reflective plane-concave mirrors and (Layertec, at nm, radius of curvature of 6 m), placed at a L=75 cm distance. To maximize coupling into the cavity transverse mode, a mode-matching optics composed of lenses and has been added to our experimental set-up. TEM₀,₀ is the fundamental transverse electromagnetic mode of an optical resonator, characterized by a Gaussian intensity profile with no transverse nodes. It is the lowest-loss, highest-quality resonant mode of the cavity. The two lenses allowed adjusting the laser waist and longitudinal position to coincide with the waist and location of the optical cavity mode. Two lenses were used for precise mode matching to provide two independent degrees of freedom, allowing simultaneous adjustment of the beam waist size and its position to match the cavity fundamental transverse mode TEM0,0. Their focal lengths (f1 = 50 mm, f2 = 100 mm) were adjusted using Gaussian beam optics, as detailed in reference literature, for example Svelto’s book, Chap. 472. To enable accurate statistical determination of the particle extinction, the particles were injected in the optical cavity through a continuous laminar flow and a buffer volume was used to stabilize the particle distribution. Finally, the light intensity transmitted through the cavity end mirror was recorded with a photomultiplier ( PM). Retrieving the particle extinction coefficient in terms of ring down time rates As detailed in36–44 and recalled in supplementary information ( SI.2), the particle extinction coefficient αp in m− 1] can be determined from the decrease in the cavity ring down times from an empty cavity (ring-down time τ0) to a cavity containing particles (ring-down time τp): where is the ratio of the optical cavity length to the sample length over which the cavity is filled with particles and c is the speed of the light. The sample length has been chosen to allow evaluating the particle backscatter and extinction on the same aerosol sample. Equation (5) assumes a linear relationship between the change in the ring down time rates 1/τp – 1/τ0 and the particle extinction coefficient αp, which is proportional to the particle number concentration. The validity of this assumption is discussed in the Discussion section, alongside with the possibility that a fraction of forward-scattered light recouple into the mode of the cavity and that scattered photons take multiple paths to partially recouple to this mode37,40. Under these assumptions, the change in the ring down time rates can be evaluated by fitting the exponential decay of the light intensity . Each individual exponential decay was acquired by first recording the baseline before the arrival of the laser pulse, then the exponential transient of the intensity until complete decay, using a sufficiently high sampling rate45, before adjusting the recorded data points with a functional form exp(-t/τp). To accurately retrieve the inverse 1/τp of the ring-down time, we followed the methodology published in36–44 by taking the natural logarithm of the base-line corrected signals, which allowed verifying the linearity obtained in Ln-scale and the absence of systematic structure in the residue plot, thus proving the single exponential nature of the ring-down decay time, in agreement with the use of a mode matching optics. As in41, the error in such fits remained small compared with the standard deviation of successive extinction measurements. Therefore, the statistical distribution of retrieved 1/τp-values has been plotted in the form of a histogram. Following the central limit theorem, such histogram could be fitted with a Gaussian law, indicating that the observed variations were statistical, thus providing a reliable evaluation of 1/τp through the mean and standard deviation of the Gaussian histogram. However, over the eight minutes duration of the acquisition, fluctuations occurred from one extinction measurement to another, resulting in variations in the corresponding histograms. These fluctuations were attributed to fluctuations in the particle concentration. Fluctuations in the laser power density and in the laser mode stability were indeed negligible compared with the imperfect stability of the aerosol generator. To account for these fluctuations, the combined Gaussian distribution of 1/τp was considered, using the mean value and standard deviation of the combined histogram as ultimate evaluation of 1/τp. The same methodology was applied to the reference particles to likewise evaluate the change in ring-down times rates 1/τref – 1/τ0. In all these extinction experiments, the ratio was maintained constant, by operating at constant detector gain settings, while ensuring a constant overlapping volume of the aerosol particle sample with the intracavity laser beam, by injecting the particles into the optical cavity through a fixed nozzle, located at a fixed position. Backscattering and extinction acquisitions Following Fig. 1, the laser propagating in the ring-down optical cavity induces side scattering in the photodetector measuring the particle backscatter. Therefore, the particle extinction and backscatter measurements are interleaved in time. The decrease in time of the light intensity is then recorded before (at ) and after (at ) the backscatter experiment, as well as at mid-backscatter experiment (at . The HWP is then switched three times over the whole backscatter acquisition (from ψ = 0 to 360°), at positions ψ = 0, 180, 360° of the QWP, which provides the three extinction measurements plotted in Fig. 4. This alternate sequence is conducted through the rotation of the half-wave plate to be seen in Fig. 1. The extinction-to-backscatter switch takes about 5 s, governed by the PRM1/MZ8 rotation mount. The particle size distribution was measured over the eight minutes duration of the acquisition, as a control measurement, as particle extinction-to-backscatter ratios are size dependent. The acquisition is actually carried out twice, firstly in the absence of particles, then in their presence, to evaluate the change in cavity ring-down time rates from 1/τ0 to 1/τp and the particle backscattered light intensity . Backscattering and extinction acquisition curves for (a) reference particles, then (b) mineral dust. Left graphs: particle backscattering curves, with backscattered light intensity plotted as a function of the ψ-rotation angle of the QWP, fitted with Eq. (3) to retrieve and then βp . Right graphs: corresponding particle extinction curves, with statistical distribution of 1/τp before (at ψ in dark yellow) and after the backscatter experiment (at ψ in dark red), as well as at mid-backscatter experiment (at ψin dark orange). The ψ-angle specified in parentheses indicates that the extinction measurement was performed at different stages (before, during, after) the backscatter measurement. The 1/τp-values are retrieved by adjusting several individual exponential decays of the intensity (an example of such fit is shown in the insert). The ultimate value of 1/τp is given by the mean and standard deviation of the combined Gaussian distribution (in dark grey). Reference lidar ratio To accurately determine the reference lidar ratio αref/βref, an analytical solution of the Maxwell’s equations is valuable3, as it provides an exact evaluation of light backscattering and extinction. Among all analytical solutions, the Mie theory2,3 is the most widely used, with available numerical code3, but however requires rigorous homogeneous spheres. Reference particles should then be aerosolized without any solvent, to avoid forming a layered sphere, whose optical properties differ from those expected for a homogeneous particle of uniform refractive index42. Consequently, spheres usually available within a liquid solvent, such as polystyrene latex spheres, cannot be used as reference particles. Likewise, particle size standards, such as silica size standards, though monodisperse, should be also left out, as explained in the Discussion section. Therefore, to apply the Mie theory for evaluating , reference particles have to be homogeneous, perfect spheres, of known particle size and complex refractive index. The size distribution of the reference particles then needs to be accurately known, which requires a precise particle sizer. Implementation of the LR-lab methodology This section details the implementation of the LR-lab methodology, by presenting examples of recorded backscattering and extinction acquisition curves for both reference and mineral dust, used as an application case study of deviation to the Mie theory. Reference particles and mineral dust particles As reference particles, we chose DEHS-particles (di-ethyl-hexyl-sebacate, C26H50O), as DEHS is non-volatile liquid, known to form perfectly spherical droplets, which has been widely used in the literature in nephelometry73,74 and cavity ring-down experiments75. A pure DEHS liquid solution was hence nebulized to produce spherical DEHS-droplets with no layered sphere, using a commercial atomizer (TSI 3079). For the mineral dust sample, a natural mineral dust (Arizona Test Dust, ATD, Powder Technology Inc.) was considered as a proxy for mineral dust, as being composed of a heterogeneous mixture of complex-shaped particles. The dust particles were aerosolized with a solid aerosol generator (SAG 410), supplied with dried compressed air, then dispersed into a buffer volume to stabilize the particle size distribution before entering the extinction-to-backscatter experiment. Backscattering and extinction acquisition curves Figure 4 presents the recorded variations of , together with the corresponding extinction histograms measured at , for the reference particles (Fig. 4a) and for mineral dust (Fig. 4b). An example of individual exponential decay of is also shown in the insert, with corresponding exponential decay fit. The decay was found linear and remained linear in Ln -scale (corresponding fits systematically yield 0.995), thus proving the single exponential nature of the ring-down decay in agreement with our use of a mode-matching optics to promote coupling to the cavity transverse mode. While reference particles exhibited a constant spherical shape (see constant zero minima in Fig. 4a), the backscattering curve for mineral dust clearly exhibits non-zero minima, as a clear signature of the dust particle non-sphericity. In all backscattering curves, the particle size distribution remained stable over the whole duration of the acquisition (otherwise, maxima would have varied). After fitting the experimental backscatter data points with Eq. (3), and on the one hand, then and on the other hand were retrieved to derive βref and βdust . The ultimate values of (βref, 1/τref) and (βdust, 1/τdust) are given in Table 1 (datapoint #3) and 3 (datapoint #1) respectively with corresponding uncertainties. Mie computation of the reference lidar ratio We are confident that the reference particles are perfect spheres since Fig. 4a exhibits null minima backscattering curves. The Mie theory can then be safely applied to compute . The size distribution of reference particles was measured by following standard literature76 with a calibrated aerodynamic particle sizer (APS, TSI 3321), covering diameters from 0.5 to 20 μm, complemented with and a SMPS (TSI 3938), to cover mobility diameters from 0.01 to 0.7 μm. An aerosol diluter (TSI 3302 A, 100:1 ratio) allowed working with higher particle concentrations. The physical diameter of the spheres was retrieved by converting the aerodynamic diameter delivered by the APS into a particle diameter , where ρp/ρ0 is the particle density (0.912 for DEHS) and χ is the dynamical shape factor (= 1 for spheres). The SMPS measured the particle mobility diameter, which is equal to the physical diameter for spheres. As presented in Fig. 5, the particle size distribution was then adjusted with a triple-modal log-normal distribution to retrieve the median diameter and the geometric standard deviation of each mode, hereafter called modal parameters. Using m=1.4539+0i for the refractive index of DEHS reference particles at 532 nm wavelength (from Topas manufacturer), by applying Python PyMieScatt numerical code, we retrieved . The uncertainty in was derived from Monte Carlo propagation with parameter correlations, applied to the lognormal fit parameters and their covariance matrix. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements AcknowledgementsThe French National Center for Space Studies (CNES) is acknowledged for financial support, with remaining costs covered by institutional funds. The authors also acknowledge Dr. D. Cholleton for his help in initiating this work, as well as Dr. C. George and Y. Carreira Da Silva for leasing the SMPS instrument. Author contributions Author contributionsAM, CPB and AG developed the methodology and did the laboratory measurements. CPB and AM designed the figures and tables. AM wrote the main manuscript text. AM and PR obtained the funding. All authors developed the experimental design and reviewed the manuscript. Funding This work was supported in part by the French National Center for Space Studies (CNES), with remaining costs covered by institutional funds. Data availability Data availability statementAll backscattering and extinction acquisitions curves and underlying numerical data are available at the following website for the sake of clarity: https://doi.org/10.5281/zenodo.20734559 Corresponding raw data are available upon request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement Data availability statementAll backscattering and extinction acquisitions curves and underlying numerical data are available at the following website for the sake of clarity: https://doi.org/10.5281/zenodo.20734559 Corresponding raw data are available upon request.