In order to select plant species that can efficiently adsorb atmospheric particles in air pollution, microstructures of leaf surface and morphology of particles on leaf surface of 16 urban tree species were observed by using scanning electron microscopy (SEM) and size fractions of particles were analyzed with Image J software in Xizhimen transportation hub of Beijing. The results showed that: 1) there were marked differences in leaf surface morphology. The morphology of the upper side of leaves of broad-leaved tree species was rather complex, containing grooves, folds, fluffs, waxes, etc. In contrast, the morphology of the lower side of the leaves was relatively simple, mainly including stomas, glandular hairs and fluffs. The leaf surface of conifers had a few folds and a lot of secretion. 2) Most of particles were rough and irregular. 3) Particles (2.5 μm) accounted for the largest proportion. The result of elementary components analysis indicated that leaf-accumulated particles were mainly comprised of C, N, O, Si, Ca, Fe and few Na, mg, Pb, Ni, Cd and Pb. The particles including mainly C, N,O, Na, Mg,Si and Ca might come from natural source, whereas particles containing Pb, Fe, Ni, Cd and Pd might be the result of human activity from diesel, coal and other anthropogenic sources. Comparing particle adhesion density on leaf surface of the 16 tree species, Sabina chinensis, Pinus tabuliformis, Koelreuteria paniculata, Sophora japonica and Amygdalus triloba were efficient in capturing particles on leaf surface. The study can provide scientific theories to reveal how trees absorb particles and thereby choose efficient greening tree species.