Physic nut (Jatropha curcas L.) is a highly promising species for biofuel production. Its descriptors are known and evaluation of its genetic variability has not been done yet. This is the first study on genetic variability in J. curcas accessions and its aim was to start the definition of morph-agronomical descriptors in juvenile phase, to evaluate the and to estimate genetic parameters in progenies and their matrix. Germplasm bank (GBA) of J. curcas at UFV contains 75 accessions from Brazil and three from Cambodia. Evaluations were done at eight and at 14 months of field. Morph-agronomical descriptors evaluated in progenies were plant height (PH) and height of branches (HEIB), crown diameter (CD) and diameter of stem (DS), number of branches (NB), length (LL) and width (WL) of the leaf, LL/WL ratio and petiole size (PS). Descriptors of seed evaluated in matrix were oil content in the seeds (OIL), weight of 100 seeds (WS), length and width of the seeds and LS/WS ratio. Estimates of genetic parameters, analysis of genotypic correlation among descriptors and Mahalanobis distance, which quantified genetic variability, were processed using softwares SELEGEN e SAS. On this distance matrix, it was applied Tocher and UPGMA clusters. Genetic divergence was also evaluated through canonical variables. The relative contribution from descriptors for divergence was evaluated based on matrix of distances and eigenvectors estimate associated to the later canonical variables. These last analyses were processed using software GENES. Average oil content in seeds was 31% ranging from 16 to 45%. It was found no genetic correlation among morph-agronomical descriptors and oil content. High coefficients of genetic variation were found for seed descriptors (PS and OIL) and for morph- agronomical at eight (HEIB, PH, NB and DS) and at 14 (HEIB and PS) months of field. The highest herdability coefficients in a strict sense were found, at eight months of field, for LL, WL and DS, and at 14 months of field, for descriptors LL, PS and WL. Tocher cluster made it possible, in both evaluation times, separation of accessions in three different groups. Dendogram by UPGMA was able to separate accesses, at eight and 14 months of field, in eight and 15 different groups, respectively. Analyses of canonical variable also evidenced divergence among accesses. The two first canonical variables explained 88.67 and 82.35% all variation, at eight and 14 months of field, respectively, forming four different groups for both ages of evaluation. The next step is choosing divergent groups and within them the identification of the most interesting accesses for the breeder. Thus, the existence of genetic diversity in GBA was proven which is important for continuity of genetic breeding works aiming at obtaining cultivars with high production of grain and high productivity of oil. Accessions with high oil content in the seeds were separately grouped, and crossing among them must be explored by the program. For registration of cultivar protection, the descriptors that contributed at most for divergence were PH and DS, the others vary in importance over time. Future evaluations involving the same descriptors and other related to reproductive cycle (inflorescences) and productive cycle (number of bunches, number of fruit per bunch, number of seeds per fruit and production) can increase even more the knowledge on the species and permit the advance on its breeding.
