1. Introduction
The characterization of suspended particles is necessary in different disciplines, such as environmental science [
1], oceanography [
2], planetary exploration [
3], and so forth. This characterization can be performed by measuring the light scattering of such particles and, more specifically, their inherent optical properties (IOPs).
This is the case of the dust sensor (DS) [
4], a mini-instrument whose scientific objective is the measurement of the size parameters of suspended dust and has been integrated into the meteorological suite METEO at the Surface Platform of the ExoMars’22 mission (currently canceled).
Suspended particles play a key role in the meteorology and climate of Mars. Thus, determining the effective dust parameters is important for radiometric modeling of the planet [
5]. Over the past 50 years, many instruments (both orbiters and in-situ instruments [
6]) have measured the opacity of the Martian atmosphere in different spectral ranges and conditions, resulting in multiple studies of dust size parameters [
7,
8,
9,
10].
Currently, there is an increasing interest in in situ instruments to study the surface–atmosphere interactions [
11,
12]. ESA is also considering a comprehensive network of weather landers on its flagship mission to Mars in the 2030s [
13]. One of the desirable requirements for the instruments of such networks would be low cost–weight–size, that is, their miniaturization.
The DS instrument is designed to provide local measurements of near-surface dust scattering, meeting those miniaturization requirements.
The physical principle of measurement is based on the fact that light scattered by a particle depends on its size, shape, the incident light wavelength, and the angle that light is scattered,
(
Figure 1). Therefore, by measuring the scattered light in different directions and wavelengths, it is possible to retrieve information about the dust size parameters.
According to the physical principle, our DS instrument consists of an IR source that emits infrared light and two detectors that measure the light scattered by the particle cloud in two spatial directions (
Figure 2): forward-like and backward-like.
One can describe the light scattered by a cloud of particles (in terms of intensity) by considering the individual scattering of each particle in a volume. However, some instruments can relate their signal to the volume scattering function (
[
]), an IOP describing the angular distribution of the scattered light by an infinitesimal volume element when an unpolarized plane wave illuminates it (
Figure 1).
where
E [
] is the incident irradiance, and (
[W/sr]) is the radiant intensity scattered by a volume element (
[
]).
Thus, to interpret the DS measurements, it is essential to relate the signal received by the detectors to the IOPs of the media, in this case
. Since these instruments span a wide range of scattering angles, we can express the optical flux received by the detectors,
, as an integral of the
multiplied by an angular weight function (
) over all scattering angles:
describes an instrument’s angular response (in terms of scattering angles) and it depends on the geometry and optics of the instrument. This
in a polar nephelometer, for example, can be approximated to a Dirac delta (
) [
14]. Another example would be an integrating nephelometer, which presents a sinusoidal response (
). Some studies include a corrected
(called angular sensitivity function) to solve the non-idealities [
15,
16,
17].
Regarding backward/forward instruments, their
depends mostly on the sampling volume. The sampling volume is the region defined by the particles that scatter light that is measured by the detector and is related to
because it indicates the range of scattering angles that the instrument ‘sees’. This volume is commonly defined by the intersection between two cones, the beam divergence of the source, and the detector’s field of view (FoV). Thus, the sampling volume defines an ideal region where any scattering contribution outside of it is not considered. Although this assumption is valid for collimated sources and detectors, it is less good when such elements are not highly collimated. In backscatter sensors, some studies compute
by partitioning the ideal sampling volume into elementary volume elements [
18,
19]. In any case, the definition of the sampling volume is the main handicap and the main source of uncertainty of all these methods to calculate
, either because they consider just an ideal volume or a correction to resolve non-idealities.
In this work, we present a new method to compute that overcomes the drawbacks associated with the volume definition. Indeed, this method avoids using the ideal sampling volume; instead, we defined the volume weighting function (VWF), a function which encompasses the instrumental response in each spatial location, and that considers the instrument geometry, the source emission pattern, and the angular sensitivity (rather than the divergence and FoV).
The definition of a specific coordinate system, where one of its coordinates is the scattering angle, allows us to compute using the VWF. This preserves a radiometric meaning; thus, a proper characterization of the optical elements (source and detector) avoids the instrument calibration. In this way, a direct model of the instrument signal is defined, where the instrumental part () is separated from the scattering of the particles ().
Furthermore, we used this method to calculate the of the DS instrument. The required geometric information and the characterization of the optical elements were obtained experimentally.
After calculating
for the DS instrument, it was validated with the results of the Monte Carlo method, and with
calculated by another method used for other commercial sensors [
19].
In
Section 2, we describe the method used to calculate
. In
Section 3, we present
obtained for the dust sensor. The validation, also in
Section 3, includes a comparison of the optical power with a Monte Carlo Method and a comparison of the normalized
obtained in [
19]. In
Section 4, we discuss some relevant aspects of this method.
2. Method and Data
This method relies on the fact that for a fixed position of a detector (D) and a source (S), each point (P) is associated with a single scattering angle (
Figure 3).
An appropriate choice of the coordinate system reduces the problem from a triple integral to a simple integral. In addition, this approach allows us to find the power received by each detector and quantify the distinct scattered angles detected by the instrument.
The reasoning followed in the next subsections should take into account the light spectral dependence. However, for the sake of clarity, this dependence was omitted.
2.1. Radiant Flux of the Detector
Using the definition of
(Equation (
1)), the radiant flux (
[
W]) reaching the detector can be calculated as
where
A is the area of the detector,
r is the distance between
and the detector, and
is the angular sensitivity of the detector.
The term [] is the solid angle underlying between and the detector and, in addition, takes into account the angular response of the detector. Thus, for an ideal detector, equals to and the term turns into the projected solid angle ().
Equation (
3) shows that the signal received by the detector depends on the
and the instrument’s characteristics (source irradiance, etc.). Thus, we can define a new function that encompasses all the parameters that appear in Equation (
3), which depend only on the instrument and not on the particles. This function, which we call volume weighting function (
VWF), is defined as follows:
Therefore, the total flux reaching the detector can be written as follows:
In the next section, we will see how an appropriate coordinate system selection allows us to find using the VWF.
2.2. Angular Weight Function
Comparing Equation (
2) and (
5), one can see that, by choosing an appropriate coordinate system in which one of the coordinates is
,
can be expressed as an exclusive function of
.
where
is the Jacobian determinant. The new coordinate system
is defined in
Appendix A.
This is why a suitable coordinate system allows one to find
from the
VWF. Since
VWF would be a function of
and two coordinates more,
(
) can be understood as the surface integral of the
VWF over a scattering angle isosurface:
where
is the surface that contains all the points with the same scattering angle.
Thus, the new coordinate system allows us to parameterize the surface,
, and therefore calculate
from the
VWF. Thus,
in this coordinate system is defined as
Moreover, the proposed method to calculate
also allows us to calculate an angular weighting function using the sampling volume (
) instead of the
VWF, simply by inserting a mask delimiting the volume. Precisely, to compare both angular weighting functions, a mask is introduced in Equation (8):
where
is the distance between S and P, and
is a logical function equal to one where the point is inside the sampling volume and equal to zero outside it:
In Equation (9), the term takes into account the quadratic decay of the source. However, it is a term that should not be considered with sources that do not show a quadratic decay (such as a laser). The term is the projected solid angle between the and the detector.
2.3. Data
We applied the method to calculate
explained in the previous subsection to the DS instrument [
4], consisting of an IR source (S) and a forward (D1) and backward (D2) detectors located at specified scattering angles (
Figure 4). The geometric parameters (positions and orientations of the emitter and the detector) for both configurations can be found in
Appendix B. Each detector encompasses two different spectral bands in the range of 1–3
and 3–5
. The thermal source emits in a similar way to a blackbody with a peak emission of 3.29
. To avoid thermal noise, the instrument implements lock-in amplification [
20].
and use different inputs for their calculation. That is, requires knowing the irradiance pattern of the emitter, E, and the angular sensitivity of the detector, . However, to compute , we need to know the sampling volume given by the detector FoV and the emitter divergence.
The irradiance pattern (
) and the angular sensitivity of the detector (
) were fitted to the parametric functions using empirical data. Equations (
A2) and (
A3) correspond to the
and
used to calculate
.
The detector FoV and emitter divergence are calculated by applying a full width at half-maximum criterion to and . Both of them are used to calculate the sampling volume mask, , used to calculate .
To compute both angular weighting functions, we define a mesh of 200 × 200 × 200 nodes for the coordinates (, , ). Then, we perform the numerical integration of the Equations (8) and (9). The software developed to calculate and for any geometry, and can be found in the data section of this paper.
As this work is about determining a method, and its theoretical application, all the radiometric constants involved, such as the power or the sensitivity of the source or detector, were normalized.
4. Discussion
In this section, we discuss the results, the validation, and some relevant aspects related to this method, such as light attenuation, multiple scattering, geometrical constraints, or instrument calibration.
4.1. Results and Validation
The analysis of reveals that their peaks are very similar to the nominal scattering angles. Nevertheless, the forward peak is smaller than the nominal scattering angle. This leads us to think that the calculation can also be useful for detecting behaviors of the instrument elements that are diverted from the nominal. The comparison of and makes it clear that the last one neglects scattering angles, , that really contribute to the detector’s signal.
The comparison of the signal obtained with the Monte Carlo method shows that maintains a radiometric sense, justifying the use of volume elements rather than particles. The advantage of this method over Monte Carlo is its ability to group the instrumental response into one term, . The advantage of using is twofold. First, it allows us to develop inversion algorithms to estimate b, and second, it can be used to calibrate the system. In addition, being able to separate the geometrical part is very valuable in the design and optimization of the instrument.
In addition to the validation by the Monte Carlo method, we compared the angular function with that of another method when a sampling volume is taken into account (). Furthermore, we saw that the use of a sampling volume results in a truncated angular weighting function. In this method, since it is not necessary to use a sampling volume, we obtain a that includes all the scattering angles.
4.2. Light Attenuation
The algorithm proposed in this work to calculate does not take into account the attenuation suffered by the light along its path. This is because the DS is intended to operate on Mars, where the expected particle density is low. However, this factor may become important when the medium where the particles are located significantly attenuates the light, either because of high particle densities or because of the presence of other elements that may attenuate the light (gases, water, etc.).
This factor, in principle, can be taken into account during the
calculation. Taking into account the Beer–Lambert law, which describes the attenuation of light traveling through a medium, one can write Equation (
3) as
where
c is the absorption coefficient of the medium. The term
indicates the total light path from the emitter to the detector.
In this way,
can be calculated taking into account the attenuation of the light:
The fact that depends on the absorption coefficient of the medium is definitely a disadvantage since the idea of is to separate the instrumental response from the . When the attenuation of the medium can be neglected, only depends on the instrument. However, when the attenuation of the medium cannot be neglected, becomes dependent on the attenuation coefficient, which makes it a function that depends not only on the instrument, but also on the medium.
Figure 8 is an example of this effect. In it, we can observe how an increase in the attenuation coefficient decreases the value of
. In addition, we can observe how the peak value of the function shifts toward smaller scattering angles because, in general, the points associated with a smaller scattering angle have a shorter optical path.
4.3. Spectral Dependence
In the calculation of W in this work, its spectral dependence is not taken into account. When the emitter is not monochromatic, it is necessary to consider this dependence. The optical flux reaching the detector would be given by
It should be noted that this W implies the emission pattern of the source, as well as the spectral dependence of the source. In principle, this will cause changes only in the magnitude of
, but not in its shape. In such cases,
We can express as the product of a spectral function, g and an angular function, f; in other words, the instrument “sees” the same scattering angles for all wavelengths but with different magnitudes.
There are other situations where this is not true, for example, the existence of a medium where the absorption coefficient has a spectral dependence or interferential filters, where the spectral sensitivity depends on the viewing angle.
4.4. Single Scattering
The method for finding
presented in this document was computed using a first-order scattering approximation [
25]. This assumption implies that the particles only scatter the light coming from the emitter (ignoring the scattered light from other particles). Thus, the lower the particle density of the medium, the more reliable the
computed by this method [
26]. A multiple-scattering regime can produce secondary peaks in the
or produce a change in the shape of the primary peak, as shown in [
19].
4.5. Geometric Constraints
As we saw in
Section 3.2, an
can be calculated with the concept of sampling volume. To do so, we added a binary mask in Equation (9), indicating the space where a scattering event can or cannot be measured. However, this is just an example of an advantage of our method to calculate
. In general, many changes in the geometrical configuration can be modeled with a geometrical mask.
For instance, it is possible to add a geometric constriction mask to the
VWF that limits the areas where scattering may or may not be measured. Specifically, the emitters and detectors of such instruments are usually contained within an encapsulation (which makes it impossible to find particles in those locations). Thus, it is possible to obtain truncated
s by adding a mask,
, to Equation (8):
Figure 9 shows a truncated
of the forward DS configuration. For this purpose, a geometric mask was placed for all positions z < 10 mm. We can see how a geometric constraint to the particles translates to a reduction of the scattering angles that the instrument can ‘see’.
4.6. Calibration
Although this method theoretically does not require calibration (as long as the emitter and detector are well-characterized radiometrically), in practice, an experimental calibration encompasses the optical and electronic response in a constant factor. According to the literature, there are two ways to calibrate this kind of sensor [
18].
One method relies on the use of different media with a well-characterized phase function (or ). This approach makes sense for marine sensors, as a tank can be filled with a liquid with a given . However, this method can be a problem for atmospheric sensors, where it is more difficult to create a controlled environment.
The other method calibrates the response of the sensor using a Lambertian target with a known reflectivity. Maffione and Dana [
27] present a mathematical derivation of this method using Lambertian planes at different profundities and relate the profundity with
. This approach is valid for backscattering sensors, where the sampling volume comprises a narrow range of scattering angles. Instruments with large conical sections comprise a range of scattering angles. The relationship between the scattering angle and the Cartesian coordinates (Equation (
A1)) is given by the inscribed angle theorem. Giving a fixed position of S and D (
Figure A1), all the scattering angles with the same value are contained in a convoluted arc circle. Thus, to correctly relate the profundity weight function to the
, it would be necessary to use convoluted arc circles surfaces rather than flat plaques.
Figure 10 shows the different scattering angle isosurfaces (for a fixed S and D) and the ideal (conical) FoV of the emitter and the detector. The intersection volume between the detector and emitter cones shows the volume where the
is measured. The smaller the volume of the intersection, the more similar the isosurfaces will be to the planes (within that volume).
5. Conclusions
In this work, we developed a method to compute a that allows us to obtain the radiant flux that an instrument would receive from the light scattered by a distribution of suspended dust particles. In addition, we applied the method to a dust sensor designed to measure the properties of the dust particles suspended in the Martian atmosphere. Finally, we compared our method with others based on different principles (Monte-Carlo-based) and applied it to other commercial sensors.
The main advantage of our proposal is that the method does not require the previous definition of a sampling volume. On the contrary, our method considers the contribution of all particles, thus avoiding uncertainties due to idealizations and approximations made when defining the sampling volume.
Another strength of our method lies in the fact that, thanks to an appropriate choice of coordinates, it is able to isolate the contribution of the geometric parameters of the instrument from the optical properties of the particles. This feature is of vital importance when designing scattering instruments since it allows to simulate the signal received by the detector in a much more agile, efficient and fast way. Thus, the method proposed in this work reduces the computational and time cost of simulating different geometric designs of the instrument, especially when compared to methods based on particle-by-particle calculations.
The efficiency of the method is especially relevant to obtain physical information from the signal provided by the DS instrument. Indeed, the simulations carried out for different dust particle size distributions will provide relationships between the characteristics of the dust distributions and the signals received in the DS instrument. The derived algorithms will lead us to achieve the ultimate goal of the DS instrument: to determine the properties of the dust near the Martian surface from the signal of two detectors located in the forward and backward directions.
However, the scope of the calculation methodology goes beyond the DS instrument, as it can be applied to other scattering instruments and dust sensors in the Earth’s atmosphere and contribute to the knowledge of the meteorology and climatology of the planet.