WO 3 nanostructures with various morphologies and crystallinity (such as cylinder-like WO 3 nanorod arrays (WNRs-1), porous WO 3 nanofibers, WO 3 nanoflake arrays, sharp cone-like nanorod arrays (WNRs-2) and annealed cylinder-like WNRs-1) were prepared on FTO substrates by using a hydrothermal method without organic additives. The effects of morphology, size and crystallinity on the electrochromic properties of WO 3 nanostructures were systematically investigated by scanning electron microscopy, X-ray diffraction, cyclic voltammetry and chronoamperometry. The WNRs-1 exhibited excellent cyclic stability, wide optical modulation (64%), and relatively high coloration efficiency (61 cm 2 C -1 ). A fast switching speed of 5 s and 6 s for bleaching and coloration are achieved for the WNRs-1 after annealing. Moreover, the relationship between the micro-morphologies/structures and electrochromic performance of WO 3 nanostructures was also discussed. the deintercalation) compared with the unannealed WNRs-1, which may be closely related to the different crystalline forms. The h-WO 3 could accommodate abundant Li-ions due to its three possible locations (trigonal cavity, hexagonal window and four-coordinated square window).However, when the h-WO 3 is translated to m-W 18 O 49 , a large amount of hexagonal windows disappear which result in low capacity for Li-ions. Moreover, the cyclic reversible value is only 71% calculated from the charge/discharge densities, which is mainly ascribed to the oxygen deficiency.Greater oxygen deficiency in W 18 O 49 could capture a certain amount of electrons in the deintercalation. Furthermore, this experiment has settled the disputes of electrochromic mechanism between the double injection model 33 and the small polaron model 34 . A major focus of the debate is the valence transformation of tungsten element in the color-changed process. The double injection model suggested that the valence transformation of tungsten element changed between plus five and plus six. While the small polaron model believed that it should be between plus five and plus four.From Fig. 8, we know that there indeed is a mixture of two kinds of valence changes in the annealed WNRs-1 during the charging and discharging process.