Material contrast in SEM: Fermi energy and work function effects.

'Is it possible to assign various grey levels of a scanning electron microscope (SEM) image to different components of a given sample? Among other instrumental effects, the answer is not only a function of the respective secondary electron emission (SEE) yields of the components, delta, but also of the angular fraction of the secondary electrons (SE)s being collected, k(alpha) and of a possible voltage contact effect between sample and detector, k(varphi). Expressed as a function of E(F), Fermi energy, and varphi, work function of the components of interest, equations of spectral, ( partial differentialdelta/ partial differentialE(k)), and angular, ( partial differentialdelta/ partial differentialalpha) distributions of the emitted SEs permit to evaluate k(alpha) and k(varphi) for Au and Si. It has been established that collected SE spectra, partial differentialdelta(alpha)/ partial differentialE(k), are distorted with respect to the emitted and fraction k(alpha) is material dependent for a solid angle of detection Omega degrees less than 2pi (or maximum semi-apex angle alpha(max)<90 degrees ) In particular, for coaxial detections around the normal incident beam the detected fraction of SEs from Au, k(alpha)(Au), is slightly larger than that for Si, k(alpha)(Si). For simple geometries in the vacuum gap, similar investigations show that parameter k(phi) is also larger for gold than for n-doped Si as well as for p-doped Si with respect to n-doped Si. Then Au is always quite brighter than n-doped Si in the SEM images while a doping contrast, C, due to a work function effect may reach approximately 15% for a Si p/n junction with N(p) approximately 10(16) and N(n) approximately 10(15)cm(-3). The present analysis may be extended to some metals such as Ag, Cu, Pb, Pd, Pt, and Zn that are expected to appear brighter than Si(n) and Ge in the SEM images. The influence of specimen surrounding in the vacuum gap and of detection conditions are outlined. The limitations of present approach are discussed and a strategy is suggested for the investigation of electronic devices where these components are in reduced number and are known a priori.