next up previous contents
Next: 3.2.3 Extrinsic Diffusion Up: 3.2 Diffusion in Polysilicon Previous: 3.2.1 Fabrication and Morphology

3.2.2 Polysilicon Grain Growth

  It is known from experiments that the grain growth mechanism is significantly affected by the doping conditions [Wad87]. The grain growth is enhanced when the dopant concentration in the grains is above the solid solubility limit. It makes no conceptual difference, whether the polysilicon is doped in-situ or by ion implantation. In both cases there is a significant increase in the grain growth rates for n-type dopants and quite moderate enhancement for p-type dopants [Kim88] [Kim90].

Generally, grain growth can be characterized by two mechanisms termed primary recrystallization and secondary recrystallization. Recrystallization can only occur in materials with a high number of lattice defects or natural disorder. The process of primary recrystallization is driven by the defect energy, whereas the secondary recrystallization is directly deduced from the grain boundary energy. In the case of primary recrystallization the transition of the grains is continuous, which means that the grains are growing from a large amount of tiny grains to a small amount of large grains, without any interstates. Polysilicon, however, is primarily growing via secondary grain growth. Some existing primary grains (see I in Fig. 3.2-3) have preferred energetic properties and start to grow with an extraordinary growth rate. During the growth process they coalesce with adjacent smaller grains to form secondary grains (see II in Fig. 3.2-3).

   figure1108
Figure 3.2-3: Polysilicon film with thickness h composed of columnar primary grains I and an average surface energy tex2html_wrap_inline5463 and a single secondary grain II with a minimum surface energy tex2html_wrap_inline5465 .

When secondary grains are formed the surface energy is minimized. Now we want to calculate the migration of the grain boundaries from thermodynamic concepts of surface energy anisotropy and secondary grain growth [Tho85] [Kal90]. To model the growth of secondary grains, we assume a cylindrical grain with predefined radius tex2html_wrap_inline5467 and thickness h growing into a matrix of normal grains. The change in Helmholtz free energy tex2html_wrap_inline5471 for this transformation is given by (3.2-2), where tex2html_wrap_inline5473 is the surface energy anisotropy, tex2html_wrap_inline5475 the number of normal grains per unit volume, tex2html_wrap_inline5477 the grain boundary area and tex2html_wrap_inline5479 the average grain boundary energy.

  equation1121

Note, that the first two terms in (3.2-2) represent energies driving normal grain growth, while the third term represents a barrier for grain growth, due to the grain boundary energy. Surface energy and grain boundary energy can act as driving force of grain growth on the one hand or as energetic barrier on the other hand. All grains try to minimize their single surface energy, hence, this will lead to films composed of large secondary grains with uniform crystallographic texture.

From basic rate theory, the rate of atomic transfer of dopants from lattice sites of one grain to those of a neighbor grain is given by a complementary Arrhenius law in (3.2-3),where tex2html_wrap_inline5481 denotes a jump frequency for atoms at the boundary, and tex2html_wrap_inline5483 is the difference of the electrochemical potential on either side of the boundary which can be seen as an energetic limit for the dopants to diffuse into the neighboring grain.

  equation1132

Under constant pressure and volume the electrochemical potential is given by (3.2-4). The boundary migration G, which can be seen as a growth rate, is obtained from (3.2-5).

   eqnarray1139

N denotes Avogadro's number, tex2html_wrap_inline5489 the atomic volume, tex2html_wrap_inline5471 the change in Helmholtz free energy given by (3.2-2), and tex2html_wrap_inline5493 the thickness of the boundary. The jump frequency tex2html_wrap_inline5481 can be expressed in terms of the temperature dependent diffusivity. The growth rate is given by (3.2-6), where tex2html_wrap_inline5497 is the concentration dependent diffusivity of all mobile dopants.

  equation1149

The grain growth rate depends on the local doping conditions via the dopant diffusivity. Thus, the average grain size becomes non-uniform within the doped polysilicon film, which agrees with morphological investigations [Wad87]. It was also found, that polysilicon shows anisotropic diffusion properties. Due to the high diffusivity in the grain boundaries, the main diffusion path is given by the grain boundaries. The spatial arrangement of the grains and their shape determines the degree of anisotropy.

The grain growth process depends also on the local stress induced by the non-planar topology of the polysilicon film. In regions with high stress rates, like in film corners, grain growth is reduced [Jon91] [Mei82]. The stress information can be obtained from a previous deposition simulation. To model such phenomena, (3.2-6) has to be extended by the stress term S, as given in (3.2-7), where tex2html_wrap_inline5501 is a scaling factor.

  equation1164

Details of the induced stress and the initial setup obtained from the deposition process are presented in Section 4.3.5.


next up previous contents
Next: 3.2.3 Extrinsic Diffusion Up: 3.2 Diffusion in Polysilicon Previous: 3.2.1 Fabrication and Morphology

IUE WWW server
Fri Jul 5 17:07:46 MET DST 1996