Brian Carter-Stiglitz

Unmixing magnetic assemblages and the magnetic behavior of bimodal mixtures

Stable single-domain (SSD) grains were mixed separately with superparamagnetic, pseudosingle-domain, and multidomain (MD) magnetite/maghemite particles in order to test the linearity of various magnetic parameters as a function of mixing ratio. Hysteresis loops, isothermal remanent magnetization acquisition curves, DC demagnetization curves, and low-temperature thermal demagnetization curves were measured on the mixtures. The experiments demonstrate that magnetization parameters are linearly dependent on the mixing ratio, while more complex parameters, e.g., coercivities, do not behave linearly as a function of mixing ratio. Armed with linearity, we apply a mathematical technique which, given a database of type curves, uses singular value decomposition to solve for the various concentrations of the magnetic phases in the mixture and a Monte Carlo simulation to determine the error in the inversion. We then test the technique on numerical mixtures, on the physical mixtures, and on a small set of natural samples from Lake Pepin, Minnesota. Finally, the magnetic behavior of the mixture of MD and SSD grains is considered, and two more mixture strains of MD and SSD grains (numerically produced) are considered to facilitate this discussion.

Day plots of mixtures of superparamagnetic, single‐domain, pseudosingle‐domain, and multidomain magnetites

[1] We test how well a few hysteresis parameters (saturation remanence M rs, coercive force He and remanent coercivity Her) serve to determine the proportions of end-members in binary mixtures. Our end-members are six magnetites whose grain sizes are within the superparamagnetic (SP), stable single-domain (SD, three samples), pseudo-SD (PSD), and multidomain (MD) ranges (Carter-Stiglitz et al., 2001). The three SD magnetites have contrasting origins and properties: (1) bacterial magnetite crystals of a single size and coercivity, arranged in chains; (2) natural volcanic magnetites with a narrow distribution of coercivities; and (3) synthetic magnetites precipitated in glass, with a broader coercivity distribution. Our parameter mixing theory assumes linear magnetization curves of the end-members between zero field and the largest coercive force He (that of the SD phase, if present). Similarly remanent hysteresis curves should be linear up to the maximum remanent coercive force Her. Three of our mixtures (SP plus bacterial SD, PSD plus bacterial SD, MD plus volcanic SD) had acceptable agreement between predicted and measured dependences of H c, H cr and the curve of M rs /M s versus. H cr /H c (Day plot) on end-member concentrations. A nonlinear approximation to remanent hysteresis curves gave a reasonable fit to MD plus glass SD results. In this case, H cr /H c for the most MD-rich mixture is larger than H cr /H c of either end-member. Such behavior is characteristic of bimodal mixtures in which Her is largely determined by the hard (SD) phase and He by the soft (MD) phase. The only mixture that could not be modeled by linear or nonlinear parameter theory was MD plus bacterial SD. The bacterial SD hysteresis loop descends almost vertically at -H c because of the extremely narrow range of particle sizes and coercivities. In general, linear and nonlinear parameter mixing models are adequate if only an approximate fit to real data is needed. An inversion method using complete magnetization curves as end-member basis functions is preferable as an unmixing technique. However, comparison of measured data to type curves, for example, on a Day plot, gives a quick indication of what end-member phases might be involved in the mix and provides additional insight before beginning an inversion.