The two-dimensional (2D) profiling of the dopant concentration is of great importance not only for the development of new semiconductor devices (see e.g. ITRS99) but also for the failure analysis of state-of-the-art semiconductor devices. The scanning capacitance microscope (SCM [2-5]) seems to be the most promising tool concerning 2D-dopant profiling. SCM profits from three considerable advantages: it is commercially available, it offers a high spatial resolution and it is easy to use. On the other hand, many SCM operators complain about a difficult interpretation of the result of SCM measurements. The objective of this work is to give a detailed insight into scanning capacitance microscopy and to work out some guidelines for the usage of SCM in semiconductor failure analysis. The starting point is the explanation of the influence of the capacitance sensor. The demand for a high spatial resolution and therefore a small probing area (r < 50nm), pushes the required sensitivity of the sensor to a value of more than 100μV/aF. This condition leads to a large probing voltage of more than 5000mV and therefore to a reduction of the lateral resolution. Only an optimum adjustment of the sensor gives us an optimum lateral resolution. In addition, SCM differs strongly from conventional CV measurements, because of the 3D geometry of the sample and the probe. Whereas the conventional method relies on the accurate determination of the probing area and the capacitance, SCM-CV curves show the influence of a large stray capacitance and an effective voltage-dependent area, due to the edge effects of the 3D geometry. The influence of the shape of the probe was examined by using a 3D device simulator (TCAD). Figure 7b) shows the SCM output versus the doping concentration. A comparison of experimental and theoretical results shows an excellent correspondence, assuming a proper sample preparation technique. This technique must offer an extremely smooth surface (sample roughness of less than 0.5nm{sub}(RMS)), a high quality oxide with a thickness on the order of 3nm and a surface and oxide with a low charge density (less than 10{sup}10cm{sup}(-2)), Such a preparation method was developed and will be presented as a guideline for proper SCM measurements. Moreover, we like to present recent insights into the dependence of the lateral resolution on the dopant concentration and the probing area. The combination of an optimal preparation and an optimal choice of voltages enables us to use a simple deconvolution procedure that works even for pn-junctions.
展开▼