Effectiveness of polycrystalline silicon diffusion sources

Abstract

Boron and arsenic concentration profiles, diffused from polycrystalline silicon (polysilicon) into the underlying-single crystalline silicon (mono) substrate, were analyzed by Rutherford backscattering spectrometry and secondary ion mass spectrometry for various levels of oxygen concentration at the poly/mono interface. In contrast with previous reports it was found that chemically grown interfacial oxide layers of about 1.4-nm thickness provide more effective diffusion sources than oxygen-free interfaces. This surprising phenomenon is caused by the strong correlation between the crystalline structure of the polysilicon layer and the diffusion rates of dopant species in that layer. It is shown that the small amount of oxide at the poly/mono interface prevents the epitaxial realignment of the polysilicon, thereby maintaining high diffusion rates in the polysilicon, without offering a significant barrier to the diffusion of boron and arsenic across the interface.

Diffusion and segregation of arsenic and boron in polysilicon/silicon systems during rapid thermal annealing

“…According to Batra et al [9], the thickness of the oxide layer is estimated to be 1.4 and 0.8 nm for RCA cleaning and that one in HF, respectively. This latter treatment was observed by other authors [6,10,11] to lead to epitaxial realignment of the polycrystalline silicon during heat treatment.…”

“…The boron profile shows a peak at the poly/mono interface in the case of diffusion which contains up to 2.6% of the implanted dose (up to 4.45% in the case of co-diffusion), and it appears that up to 11 and 10% of the implanted arsenic atoms segregated at the interface in the cases of diffusion and co-diffusion, respectively. It is important to compare the results obtained in this work and those obtained by Josquin et al [6]. These authors have noticed that 6% of the implanted dose have been segregated to the interface in polysilicon layers implanted with arsenic, after an anneal for 95 min at 1000 • C. Schaber et al [5] mention that 4% of arsenic atoms and 1% of boron atoms segregate to the interface (the doses used: 2 × 10 16 and 5 × 10 15 atoms/cm 3 for arsenic and boron, respectively).…”

“…For boron, the difference is much bigger, thus confirming the slowing down of boron diffusion in the presence of arsenic. Josquin et al [6] have also determined the percentages of arsenic and boron atoms from the polysilicon to Table 1 The percentage of the dose segregated at the interface (D 1 %) and that diffused in monosilicon (D 2 %) [21] on the co-diffusion of Ge with dopants (boron or arsenic) in Si 1−x Ge x /Si structures. These authors observe a pileup of Ge at the interface which has been seen to persist at high temperature with increased time indicating a possible Ge segregation.…”

“…This cleaning procedure makes it possible to avoid the epitaxial realignment of the polysilicon layer starting from the polycrystalline silicon/single crystal silicon interface [3,[5][6][7]. This realignment can modify the diffusion process in the polysilicon and makes its control more complicated.…”

Electrical and microstructural investigation of polysilicon emitter contacts for high-performance bipolar VLSI

“…Indeed, simulation of a single-crystal silicon layer with an ideal box-like doping profile results after the same anneal in a shallower but slightly more graded junction compared to a polysilicon diffusion source. This effect has been observed before [18] and is explained by epitaxial regrowth, which moves the fast grain-boundary diffusion sites away from the substrate.…”