the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 31 Aug 2026
FTIR spectroscopy of desert dust: implications for complex refractive index spectra and dust sample diversity
Abstract. Mineral dust is a dominant natural aerosol that has a strong influence on Earth’s radiative budget. However, its radiative impacts remain poorly constrained due to limited knowledge of its complex refractive index (CRI), especially in the thermal infrared (TIR). We present a new application of Fourier Transform Infrared (FTIR) Attenuated Total Reflectance (ATR) spectroscopy to derive CRI spectra for six dust samples in the TIR, 2.5–25 μm. The CRI was estimated through two distinct methods. The first utilizes the Beer–Lambert Law to determine the imaginary component of the CRI, κ(λ), whereupon the real component, n(λ), is retrieved via the Kramers–Kronig relations. The second implements a direct ATR reflectance inversion approach where n(λ) and κ(λ) are simultaneously retrieved by fitting modeled Fresnel reflectance to measured ATR spectra. This methodology circumvents uncertainties associated with many sample preparation methods, thus enabling a more direct characterization of natural samples. Comparisons with literature CRI spectra for the six samples in question (and for mineral dust broadly) provide context for interpreting results from both retrieval approaches. For each sample, the two methods capture the same major absorption features as available literature does, but differ in retrieved absorption magnitude and long-wavelength behavior. Additionally, ATR measurements revealed significant variability in CRI spectral shape and magnitude, reflecting sample diversity linked to geographic origin. These results provide new constraints on desert dust optical properties and highlight sample-to-sample variability that can inform climate and radiative transfer models.
- Preprint
(12398 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 12 Oct 2026)
- CC1: 'Comment on egusphere-2026-2495', Lucas Mortier, 01 Sep 2026 reply
-
RC1: 'Comment on egusphere-2026-2495', Anonymous Referee #2, 20 Sep 2026
reply
The manuscript is clearly written and falls within the scope of Atmospheric Chemistry and Physics. However, I am not fully convinced of the methodological novelty or the scientific rigor of the proposed retrieval framework.
The primary objective of this work is to retrieve the complex refractive index of mineral dust from ATR-FTIR measurements using two approaches: (i) a Beer-Lambert law analysis combined with Kramers-Kronig relations, and (ii) a Fresnel-equation-based iterative inversion. While the application of these approaches to globally sourced dust samples is useful, the manuscript does not clearly establish what is fundamentally new about the methodology itself relative to the existing literature on optical-constant retrieval.
Optical constant retrieval from spectroscopic measurements is a well-established problem, and numerous inversion frameworks have previously been developed for this purpose. For example, Akbar and Blümel (10.1371/journal.pone.0320697 and 10.1021/acs.analchem.5c02453) reported a robust reconstruction framework for determining the complex refractive index (or equivalently the complex permittivity) from infrared extinction spectra. Although their framework was demonstrated for spherical scatterers, the authors explicitly describe it as modular, with the forward model being replaceable to accommodate different geometries. Furthermore, their framework is applicable both in the Beer-Lambert regime and in strongly scattering systems where the Beer-Lambert approximation breaks down. In the present manuscript, it is therefore unclear what conceptual or methodological advance distinguishes the proposed approach from existing inversion techniques beyond its application to ATR measurements of natural dust samples.
Similarly, the two retrieval methods employed here appear to rely on well-established techniques. The Beer-Lambert/Kramers-Kronig approach is essentially a direct calculation of the optical constants under the assumptions of the Beer-Lambert approximation, while the Fresnel-based retrieval is formulated as a conventional nonlinear fitting problem. The manuscript would benefit from a clearer discussion of how these approaches differ from existing commercial or published ATR optical-constant retrieval methods, and why the proposed workflow offers a meaningful methodological improvement.
The validation of the retrieval is also somewhat limited. The reconstructed refractive indices are compared only with literature reference spectra, and the agreement is qualitative rather than quantitative. While discrepancies may reasonably arise from experimental noise, imperfect sample-crystal contact, or differences in sample preparation, the manuscript does not provide sufficient evidence to distinguish these experimental effects from possible shortcomings of the inversion methodology itself. A controlled validation using synthetic spectra with known optical constants would greatly strengthen the conclusions by allowing the reconstruction accuracy of the method to be assessed independently of experimental uncertainties.
Finally, because the Fresnel-based retrieval constitutes a nonlinear inverse problem, its robustness should be demonstrated more rigorously. Such optimization problems are often sensitive to initialization and may contain multiple local minima. However, the manuscript does not specify the initial conditions, initialization strategy, parameter bounds, optimization algorithm, convergence criteria, or whether different initial guesses converge to the same solution. Without such information, it is difficult to assess the reproducibility and robustness of the proposed inversion framework.
Citation: https://doi.org/10.5194/egusphere-2026-2495-RC1 -
RC2: 'Comment on egusphere-2026-2495', Anonymous Referee #1, 20 Sep 2026
reply
General comments
The manuscript presents an interesting spectroscopic study of the infrared optical properties of mineral and mineral-dust samples using ATR measurements and a retrieval of the complex refractive index. The study covers a broad spectral range, including the far-infrared, and provides potentially valuable information on spectral regions that are less extensively documented than the conventional mid-infrared range. The experimental dataset is interesting, and several of the comparisons presented in the manuscript, particularly those involving particle-size fractions and retrieved refractive indices, could provide useful understanding into the spectral behavior of mineral dust.
However, the methodology and its validation require further clarification before the retrieved optical constants can be fully evaluated. In particular, some assumptions used in the retrieval are insufficiently justified, including the imposed boundary conditions on the imaginary refractive index and the choice of the reference/anchor value for the real refractive index. The retrieval workflow presented in Figure 3 should also be simplified and described more explicitly in the text. In addition, the far-infrared spectral features deserve more attention, as they appear to contain significant and sample-dependent information that is currently under-discussed.
My specific comments are detailed below
Specific comments:
1. Organization of the Background and Methods sections
I recommend removing the separate Background section, since much of its content is directly related to the methodology and could be incorporated into the Methods section.
More specifically:
- Section 2.1 Measurement technique, including the discussion of the Fresnel equations, could be moved to approximately line 188, where the measurement methodology is introduced.
- Section 2.2 Sample description and preparation, could be moved to approximately line 156.
- However, the discussion in lines 146–154 explaining why ATR was selected is more appropriately part of the methodological motivation and could be moved to approximately line 85.
This restructuring would improve the logical progression of the manuscript and make the distinction between methodological motivation, sample preparation, and data analysis clearer.
2. Boundary condition imposed on the imaginary refractive index
The initialization/boundary condition imposed on the imaginary refractive index in the flowchart of Figure 3, where k(ν) appears to be constrained to values less than or equal to 1, requires justification. From a theoretical perspective, there is no general requirement that the imaginary part of the refractive index must remain below 1. Based on the dielectric response and dispersion relations, k can exceed unity, particularly in the vicinity of strong resonances. Please explain the physical or numerical motivation for imposing this boundary condition. Is it based on previously reported optical constants for the considered minerals, introduced only to stabilize the inversion, or required by another aspect of the retrieval algorithm?
Please clarify this point around line 197.
3. Retrieval methodology and Figure 3
The retrieval methodology still requires clarification, particularly regarding the validation of the retrieval procedure. In addition, Figure 3 is currently too detailed and should be simplified. All important steps appearing in the flowchart should first be properly explained in the main text.
In particular:
- The Hann taper smoothing appearing in the flowchart is not explained in the text. Please describe its implementation and purpose, for example around line 204.
- The notation used in Equation (5) should be made consistent with that used in the flowchart.
- It is not necessary to reproduce all equations explicitly inside the flowchart if they are already properly introduced in the text. Expressions such as n(w)/A_ATR could therefore be removed from the figure.
- The least-squares expression should instead be explicitly defined in the methodological text, including the quantity being minimized and, where relevant, the associated weighting.
- Numerical settings, intermediate variables, and algorithmic details that are not essential for understanding the workflow could be removed from the flowchart and described in the text instead.
The objective of Figure 3 should be to provide a clear overview of the retrieval sequence rather than reproduce the complete mathematical implementation.
4. Far-infrared spectral features
The absorption bands between approximately 15 and 25 µm are not clearly identified in Figure 4 and receive relatively little discussion in the manuscript. These far-infrared modes appear to have amplitudes comparable in order of magnitude to several features observed in the mid-infrared as can also be seen in Figure 5. Moreover, this spectral region contains rich and strongly sample-dependent structures; the Atacama sample is one notable example! I recommend highlighting these bands more clearly in Figure 4 and expanding their discussion. This could represent an important added value of the study since previously it is more commonly discussed in terms of the well-known mid-infrared features.
5. Interpretation of Figure 5 and particle-size effects
Figure 5 provides a particularly interesting comparison between the different size fractions. However, direct comparison of the absolute intensities may be misleading because the overall absorbance level can vary between measurements. I suggest additionally comparing normalized spectra, for example by normalizing each spectrum to a common maximum, potentially the main feature around 10 µm. This would facilitate comparison of spectral shape independently of absolute intensity. Particle size would be expected to affect not only intensity but also the apparent width and shape of the absorption bands. A normalized comparison could therefore help identify whether spectral broadening or changes in band shape occur between the different size fractions.
Another point, the variability between size fractions appears to differ considerably among the samples. For example, for the Atacama sample fractions above approximately 20 µm show significantly greater spectral variability than fractions below 20 µm. Can the authors explain the origin of this behavior?
As discussed around line 250, penetration depth can affect the measured ATR spectrum. Could the observed size-dependent behavior caused by the penetration depth therefore represent, at least partly, a limitation of the ATR measurement for these samples ?
It would be useful to discuss whether this effect introduces a measurable uncertainty or bias and, if possible, whether its magnitude can be estimated.
6. Absolute magnitude of the retrieved real refractive index
Figure 6 shows noticeable differences in the absolute magnitude of the retrieved real refractive index between samples. The origin of these differences is not sufficiently discussed, particularly since the uncertainty ranges presented in Figure 10 appear to overlap in several cases. Similar differences are observed for the dust samples in Figure 8.
One possible contribution could arise from the reference or anchor value used in the Kramers–Kronig retrieval. The manuscript indicates that this value is fixed at n=1.5, but the spectral position associated with this reference value should be explicitly stated. Please also clarify whether the Kramers–Kronig calculation is implemented using a single-sided or double-sided formulation and how the unmeasured spectral regions are treated.
Since the absolute level of the retrieved real refractive index can depend on the reference condition used in the dispersion calculation, the choice of n=1.5 requires stronger justification. Where possible, I recommend using or comparing against literature values at the corresponding wavelength or wavenumber, for example values reported for mineral dust in studies such as Di Biagio et al. (2019).
The sensitivity of the retrieved real index spectra to this assumed value should also be evaluated or at least discussed. The interpretation of these differences should consequently be expanded in Section 4.2.
Technical corrections
- For the sample labels used throughout the figures, I recommend using either a sample number or a shortened sample-location name. The current labels make some figures unnecessarily crowded.
- Please ensure that the notation is consistent throughout the manuscript and between equations, text and figures. You use in equation (5) R_reference while R_air in the flowchart.
- Please check that all quantities and variables appearing in Figure 3 are defined in the main text before or immediately after the figure is introduced.
Citation: https://doi.org/10.5194/egusphere-2026-2495-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 395 | 117 | 34 | 546 | 36 | 38 |
- HTML: 395
- PDF: 117
- XML: 34
- Total: 546
- BibTeX: 36
- EndNote: 38
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
1. Representativeness of the ATR sample
The manuscript emphasises that ATR-FTIR preserves the physical and mineralogical characteristics of the dust sample. However, I am not sure to what extent the optical properties retrieved from particles in direct contact with the ATR crystal are representative of airborne dust. Could particle–crystal contact, packing density, aggregation, or particle morphology affect the retrieved CRI? The manuscript itself notes differences between size fractions and attributes some of these differences to sample–crystal contact. It would be useful if the authors could discuss whether the retrieved CRI should be considered an intrinsic material property, or rather an effective property of the ATR sample configuration, and how transferable these values are to atmospheric aerosols.
2. Uncertainty and robustness of the Fresnel inversion
The Fresnel-based approach avoids the wavelength-independent penetration-depth assumption of the Beer–Lambert method, but the inversion itself seems to be an ill-posed problem. The retrieved CRI may therefore depend on the initial conditions, parametrisation, and optimisation procedure. Could the authors provide some quantitative assessment of how sensitive the retrieved CRI is to these choices? In particular, do different initial guesses or parametrisations converge to the same solution? This would give a better indication of how well constrained the reported CRI values actually are.
3. Effective CRI of heterogeneous mineral mixtures
The dust samples contain mixtures of several mineral phases with different optical properties. I therefore wonder how the retrieved CRI should be interpreted for these heterogeneous samples. Is it intended to represent an effective optical property of the mineral mixture? If so, how might it depend on the relative mineral abundances, particle size, or mixing state? I think this is particularly relevant if the reported values are intended to be used as optical constants for atmospheric dust.
4. Relevance for climate models
The manuscript motivates the retrieval of dust CRI partly by its importance for climate and radiative-transfer modelling. However, it is not entirely clear to me how much the differences in CRI reported here would actually matter in an atmospheric context. Could the authors provide a simple sensitivity estimate showing how the different retrieved CRIs affect quantities such as extinction, absorption, single-scattering albedo, or radiative forcing? Even a relatively simple or order-of-magnitude estimate could help demonstrate the relevance of the differences for climate modelling.