The objective of this work was the development of methods for simultaneous determinations of As and Hg by hydride and vapor generation and simultaneous determinations of Cd, Cr and Pb by direct nebulization of the solution in mineral fertilizers by microwave induced plasma-OES. During development mineral fertilizers samples from the interlaboratory program and also reference material of inorganic fertilizer (SRM ® 695-NIST) were used. Optimizations were made in the operation conditions of the MIP OES to obtain better conditions of the interest analytes determination. The method by chemical vapor generation was also optimized, and the arsenic pre-reduction step was taken, so that simultaneous determination of As and Hg was possible, the test evaluated the observed result in standard solutions of As(III), As(V) and a standard solution containing the two species in the same concentration, in the presence of different concentrations of potassium iodide and also in its absence. The results showed that there wasn't significant influence with the usage of potassium iodide, and the pre-reduction stage could be eliminated without impairing the results. The analytical responses obtained at different concentrations of the reducing solution (NaBH4) were also evaluated in order to obtain the best hydride and vapor formation conditions. The chosen concentration was 0.5% (m v-1) of NaBH4 in 0.25% (m v-1) NaOH. A matrix effect test with mineral fertilizer sample containing 10% Zn (m v-1), Cu at 5% (m v-1) and Mn at 9.8% (m v-1) was carried out to evaluate the effect of these elements on the reaction of chemical vapor generation, and the results showed that in the adopted conditions, the elements present in the sample didn't cause significant interferences. Matrix effect tests were also performed for the Cd, Cr and Pb determinations, and for these analytes, significant interferences weren't observed due to the presence of high concentrations of Mn, Zn and Cu. The methods developed were validated according to the Manual of Quality Assurance (MAPA), presenting good linearity for the working range employed, presenting r> 0.99 and the residual plot showed a fairly random pattern, the linear model provides a decent fit to the data, good repeatability and intermediate precision, presenting lower RSD than the RSD calculated by the Horwitz equation and obeying the hypothesis tests (ANOVA), good accuracy, with relative errors between-1 and-12%, that is, less than 20%, according to the acceptability established criterion. The methods developed are comparable to the official ones.