Petroleum is a complex mixture of hydrocarbons (50 to 95 percent) and nonhydrocarbons (mainly S, N, O and heavy metal compounds). Metallic traces are normally found in petroleum as naturally occurring elements associated to the formation process. Although present only in small amounts (Microgram −1 or ng g−1 levels), their determination is of great interest because of their high potential of environmental contamination and interference in the refining process. They may corrode refinery equipment, poison and foul catalysts and/or cause undesirable side reactions in refinery operations. Moreover, heavy metals present in petroleum derivates such as diesel, gasoline and naphtha are related to poor performance and reduction of their thermal stability. Thus, fast, simple and reliable procedures for the determination of trace metals in petroleum and its derivates by graphite furnace atomic absorption spectrometry were studied. For diesel, gasoline and naphtha samples, procedures for the determination of As and Mn (ng mL−1 levels) are proposed. Sample stabilization was necessary and achieved by the formation of three component system prepared by mixing appropriate volumes of the samples, propan-1-ol and nitric acid aqueous solution, resulting in a one-phase medium indefinitely stable (microemulsion). The use of matrix modification was also investigated. Multivariate optimization defined the optimum microemulsion composition as well as the temperature program. The limits of detection in the original samples for As were 1.8, 1.2 and 1.5 ng ml−1 while for Mn they were 0.6, 0.5 and 0.3 ng mL-1 for diesel, gasoline and naphtha, respectively. Methods accuracies were confirmed by recoveries tests and the analysis of a set of commercial samples by the proposed and independent comparative procedures. For petroleum samples, procedures for the direct determination of Ni, Cu, Fe and V using a solid sampling accessory, without any sample pre- treatment, are proposed. A Pd+Triton X-100 solution was used as chemical modifier. The pyrolysis and atomization temperatures, as well as the Pd mass were defined by multivariate optimization. The limits of detection at the optimized conditions were 230, 10, 200 and 800 pg for Ni, Cu, Fe and V, respectively. Methods` accuracies were confirmed by the analysis of oil certified reference materials as well as by comparison with independent methods. For all proposed procedures, calibration using aqueous analytical solutions was possible.