The Rietveld method was used for the quantification and qualification of minerals in the clay fraction of some highly weathered soils developed under tropical climate. Twenty one samples of 13 profiles of soils were collected (B horizons and fragipans) in different Brazilian environments, including areas with pronounced soil moisture deficit and marked dry season and areas with soil moisture surplus and high precipitation. The soils, which belong to Oxisol, Ultisol and Spodosol orders, were formed under the influence of different parent materials such as gneiss, granites and sandy-clayey sediments. 

The identification of the components of the clay fraction was accomplished by X-ray diffractometry (XRD). The data was obtained in a SIEMMENS D-5000 difractometer using CoKα radiation. The instrument, equipped with a graphite monochromator, was operated at 40kV and 25 mA. The samples were ground in agate mortar, together with 5% of NaCl (MERCK, ACS), used as internal standard. The samples were mounted in a glass slide, after weak pressure on a paper filter to minimize preferential orientation of the particles. The scans ranged from 10 to 70 º2θ, with increment steps of 0.02 º2θ for every 6 seconds. The same support was used for all samples. 

The software RIETICA (www.rietica.org) was used and the crystallographic data for kaolinite, goethite, hematite, quartz, rutile and anatase were obtained from MINCRYST (www.database.iem.ac.ru). The function to model the peak shape was the pseudo-Voigt. The width of the half height of the peak was modeled by a quadratic function with three refinable parameters. All parameters of the unit cell were selectively refined and the baseline was modeled by a polynomial function with four parameters. To describe the quality of the refinement R indexes and the difference line between the observed and calculated diffraction patterns were observed.  To validate the use of the Rietveld method, the amounts of goethite (Gt) and hematite (Hm) obtained by this method were compared with those obtained by the combined use of extraction by dithionite-citrate-bicarbonate (DCB) and diffuse reflectance spectrophotometry (DRE). The extraction by the DCB allowed to obtain the amount of Fe in oxide form, while DRE was used to determine the hematite : goethite ratio. 

The adjustment of the model was initially attempted taking into consideration the presence of the triclinic kaolinite. The results for the R indexes and the estimate of the amount of internal standard obtained were unsatisfactory. Therefore, monoclinic kaolinite was considered as an additional phase. By doing so, better adjustments were obtained with smaller Rp values and, in some cases, the error in the estimate of the amount of the standard was about of 1% by weight.

The possible coexistence of monoclinic and triclinic kaolinites in these soils can be related to the soil formation environment, which may have shown variations in the amount of free Fe with time. It is suggested that the triclinic kaolinite would have been formed more recently at low levels of Fe, what would have lead to a better structural organization. On the other hand, the monoclinic kaolinite would have been formed in past times, at higher levels of Fe, what would had favored larger content of Fe in kaolinite structure and, consequently, causing a larger degree of structural disorder.  The correlations between the amounts of Gt and Hm obtained by the Rietveld method and obtained by the DCB + DRE were significant at p<0.01 suggesting that the Rietveld method can provide a good estimate of those minerals in soils.