Brazilian Eucalyptus plantations are some of the most productive forest plantations in the world, sustaining mean growth rates of 25 Mg ha_1 year_1 (50 m3 ha_1 year_1) over the 4.7 million hectares planted across the country. To better understand forest productivity, studies at the stand scale need to be coupled with tree level evaluations of the production ecology (the assessment of wood production as a function of crown light absorption and light use efficiency). The soil clay content (_20% to _40%), topography and historical land use of the experimental site generated a natural gradient in productivity. We measured (from 6 to 7 years after planting) stem wood dry biomass growth and estimated light absorption and light use efficiency at the tree level with a three-dimensional array model (MAESTRA) in 12 plots within a seed-origin Eucalyptus grandis plantation. We investigated the hypothesis that dominant trees (the 20% largest) are more productive than non-dominant trees (the 20% smallest) as a result of greater light absorption and light use efficiency; and that with increasing productivity across plots, dominant trees would show larger increases in light use and light use efficiency in comparison to non-dominant trees. The 20% smallest of the trees averaged 10.6 kg of stem wood dry biomass (1.6 kg of stem wood growth during the last year of the rotation), compared with 185 kg per stem wood in the 20% largest of trees (34 kg of stem wood growth over the same period). The smallest trees contained 7.2% of the leaf area as compared to the largest trees (3.0 m2 versus 41.7 m2), and they absorbed only 6.7% as much light (2.2 versus 32.8 GJ year_1). The smallest trees grew at about 4.7% of the rate of the largest trees, which is a smaller percentage than the difference in absorbed photosynthetically active radiation; therefore the light use efficiency was lower for the smallest trees (0.75 kg GJ_1 versus 1.03 kg GJ_1). Our results show the significant contribution of d