Resolving the Interpretation of Magnetic Coercivity Components From Backfield Isothermal Remanence Curves Using Unmixing of Non‐Linear Preisach Maps: Application to Loess‐Paleosol Sequences.

Unmixing of remanent magnetization curves, either isothermal remanent magnetization (IRM) or backfield IRM, is widely used in rock magnetic and environmental magnetic studies to discriminate between magnetic coercivity components of different origins. However, the wide range of physical properties of natural magnetic particles gives rise to an ambiguous interpretation of these components. To reduce this ambiguity and provide a straightforward interpretation of coercivity components in terms of domain state, interactions, and constituent magnetic phases, we combined backfield IRM unmixing with unmixing of nonlinear Preisach maps for two typical mid‐latitude northern hemisphere loess‐paleosol sequences. Both backfield IRM and nonlinear Preisach maps unmixing are based on the same non‐parametric algorithm, and provide similar endmembers (EMs) in the two sections studied. The first EM (EM1) has a low median coercivity (∼21 mT) and is a non‐interacting single domain (SD) magnetite/maghemite of pedogenic origin. The second EM (EM2) has a moderate median coercivity (∼60 mT) and is a mixture of pseudo‐single domain/multidomain, SD magnetite/maghemite and non‐interacting SD hematite, all of eolian origin. The same EM1 found in both sections suggests that this component's grain size and coercivity are independent of pedogenesis intensity. The same EM2 indicates that a similar magnetic population is being transported and deposited, irrespective of the dust source area and loess granulometry. The approach outlined here provides strong evidence that non‐parametric backfield IRM unmixing isolates physically realistic EMs. Unmixing nonlinear Preisach maps elucidates these EMs in terms of domain states and their constituent magnetic phases. Plain Language Summary: Magnetic minerals occurring in natural materials carry crucial information about environmental processes. However, since a given material typically contains several magnetic components, the environmental signals it encodes are convoluted. Due to different physical properties of natural magnetic particles, recovering information about each individual process can be a challenging task. One popular method used to differentiate magnetic populations with different origins is numerical unmixing of their coercivity distributions because they are sensitive to characteristics of individual magnetic populations. Coercivity distributions represent the first derivatives of IRM acquisition curves, which are built by exposing a sample to a series of magnetic fields in the laboratory. Generally, there is ambiguity in the interpretation of coercivity components. Therefore, we devised an unmixing method using data sets called non‐linear Preisach maps, built from a sequence of IRM curves that are processed into two‐dimensional diagrams. These data sets give us additional information to that gleaned from coercivity distributions, such as magnetic domain states and particle interaction signatures. We tested this approach on two loess‐paleosol sections located in Southeastern Europe, because they developed in different climatic conditions, and had different dust source areas. Key Points: Non‐parametric remanent magnetization curves unmixing provides physical, though ambiguous, componentsNon‐parametric Preisach maps unmixing resolves this ambiguity by allowing the identification of mineral phases and domain statesIn loess‐paleosol deposits the pedogenic and eolian components are respectively independent of pedogenesis intensity and dust source areas [ABSTRACT FROM AUTHOR]