Polymer-sorted high-density carbon nanotube (CNT) arrays have shown great potential to extend the silicon-based Moore's law. Imaging the CNT arrays on insulators like SiO2/Si using low-voltage scanning electron microscopy (LVSEM) to acquire array information like the alignment, density, and distribution of residual polymers is necessary. Such a task remains challenging due to the nanoscale CNT body (1-2 nm in diameter), nanoscale tube-to-tube separation (1-10 nm), the broadening of the apparent diameter, and the complex image contrast caused by the insulating substrate and polymer residues. In this study, the imaging mechanism for this system is investigated. Two methods are developed to separate the three dominant contrasts, i.e. topographic contrast, charge contrast, and material contrast, by selecting the take-off angle and energy of the emitted electrons as enabled by changing the working distance or the deceleration voltage. The contrast formation and separation mechanism is further confirmed by the dynamic contrast evolution due to the electron-beam-induced deposition of amorphous carbon. The contrast separation method is further applied to an individual CNT, reducing its apparent diameter from 36 nm to 6 nm. This result hints at the potential for LVSEM to count the density exceeding 150 CNTs/µm of CNT arrays. Finally, a comparative study of LVSEM and transmission electron microscopy confirms the failure of LVSEM to resolve CNTs in a bundle. The results suggest that the density of CNT arrays may be underestimated in reported SEM data. The proposed method can serve as a useful tool for further study and application of arrayed CNTs.
