Raman spectroscopy: about chips and stress

Moore’s law* dictates microelectronics researchers to make integrated circuit (IC) devices smaller and to put them as close to each other as possible on a chip. This results in a better performance and a larger functionality of the chips. However, these devices also require a good electrical isolation from each other. This is in general done by the formation of a thick local oxide in the “field region” between the devices. In 1970, researchers from Philips1 invented the so-called LOCOS (LOCal Oxidation of Silicon) technique to achieve this isolation. Using a Si3N4 mask, the silicon is thermally oxidised in the nitride-free field regions. Figure 1 (left) shows a typical LOCOS structure. Although LOCOS seemed a perfect solution at that time, it came with a lot of problems, many of them related to mechanical stresses. Thermal oxidation of Si to SiO2 occurs together with a 125% volume expansion. As a result, the oxide grown in the field region, called the “field oxide”, exerts large forces on the surrounding silicon. Another major drawback of this technique is the so-called “bird’s beak”, caused by the lateral growth of the oxide under the nitride mask. This bird’s beak not only affects the intended device length, it also introduces large local mechanical stresses in the silicon, because of volume expansion, and it also deforms the nitride film. These stresses often resulted in the generation of dislocations in the silicon, which are quite harmful for the devices (see Figure 1, right). Measuring these local stresses using known techniques such as wafer bending or x-ray diffraction (XRD) was not possible at that time. The devices were too small (μm size). So, in the 1980s and beginning of the 1990s, a lot of effort was spent on simulation of the oxidation process, the LOCOS formation and the induced stresses. However, this simulation process was not straightforward. The different materials had to be treated as being non-linear visco-elasto-plastic, their material parameters could change with temperature and stress, and the oxidation rate depended on the (changing) silicon crystal orientation. Although this seemed a mission impossible, many researchers succeeded in simulating the LOCOS process quite well. But this did not really solve the questions on stress. How large is this stress, what is the effect of processing parameters and materials used, and what is the effect of geometry? Answering these questions is why Raman spectroscopy found much greater utilisation within the microelectronics industry.