The deposition conditions used to form polysilicon layers are determined to a large extent by the geometry of the reactor used to deposit the films. The structure of the films and the electrical behavior of devices incorporating the polysilicon layers are, in turn, influenced by the deposition conditions. Operation in the reaction-rate limited regime is necessary for film uniformity in a high-volume reactor; this is achieved by operating at low temperatures and low pressures. The temperature gradient often used in this reactor changes the structure of the deposited films, and cannot be used for devices which depend sensitively on the structure. Diffusion in polysilicon is influenced by the grain boundaries and, consequently, depends sensitively on the deposition conditions. The electrical properties of devices with their active layers within the polysilicon layer itself depend strongly on the structure of the polysilicon. Even conventional MOS transistors in which the polysilicon serves only as a conducting electrode can be affected by the structure of the polysilicon. Therefore, the deposition conditions of the polysilicon films and the resulting structure must be considered to obtain the desired electrical properties in the devices being fabricated.

A systematic series of films have been deposited on Si (111) wafers by the pyrolytic decomposition of disilane in a low pressure CVD reactor at temperatures below 765°C, at a variety of Si2H6 partial pressures. Usually, large columnar Si grains nucleate with the [110] fiber axis parallel to the [111] substrate normal, and exhibit a random arrangement of azimuthal orientations about this axis. Fine scale microtwinning and sub-grain boundaries which are often seen within individual columnar grains have been characterized by HREM. The facetted boundaries between individual columnar grains have also been characterized. The correlation between typical grain size, defect densities and Si2 H6 partial pressure is presented and the effect of changing partial pressure during growth is shown to be detrimental to film quality. Finally it is noted that randomly oriented polycrystalline Si films can be formed under certain growth conditions.

A selective nucleation based crystal-growth-technique over amorphous substrates is originated. The method manipulates nucleation sites and periods and hence, controls the grain boundary location by modifing the substrate surface. In Si, small Si3 N4 nucleation sites are formed, 1–2 pm in diameter, 100 μm in period, over Sio2. One Si nucleus is formed exclusively in the small area of Si3 N4 by CVD. The highly faceted and periodically located nuclei grow over SiO2 up to 100 μm in diameter before impingement. A MOS-FET fabricated inside the island operates comparably to the bulk Si control

Some typical microstructural studies of polycrystalline silicon using transmission electron microscopy (TEM) are described, including the application of this material for assisting TEM investigations themselves. Examples include oxidation and realignment of polysilicon thin films, the structure of polysilicon in EEPROM devices, polysilicon in trench capacitors and measurement of SiO2 layer thicknesses with polysilicon overlayers. It is also shown tha grain growth in heavily phosphorus doped polysilicon films can be followed by in situ heating in the TEM.

Phosphorus ions were implanted into silicon layers deposited by low pressure chemical vapor deposition onto thermally oxidized silicon substrates. Thermal anneals diffused the phosphorus and the resulting depth profiles were determined by secondary-ion mass spectrometry (SIMS). Transmission electron microscopy shows that the polysilicon layers have a multi-layer pattern of grains. The phosphorus profiles are fit by a Monte Carlo simulation technique that includes both grain and grain-boundary diffusion. The grain-boundary diffusion coefficient is found to be thermally activated with an activation energy of 3.3 eV.

It has been observed by several authors that metal-oxide-semiconductor devices with polycrystalline Si (polySi) gates behave differently depending on the doping species in polySi: the work-function difference between the silicon substrate and the gate (ø ps ) is higher when the gates are doped with arsenic than when they are doped with phosphorus.

We believe that the different behavior of ø ps . can be explained by different grain textures at the polySi/SiO2 interface. Our transmission electron microscoey of the films indicates that while P-doped material consists of large (≈3000Å) grains, As-doped polySi preserves its as-deposited columnar structure – even after a high temperature anneal. Moreover, at the interface with the gate oxide an as-deposited microstructure with very small (≈100Å) “embrionic” grains is preserved. On the basis of these observations, we suggest a model for the different behavior of ø ps . The model is based on a quantum-size effect which becomes important for such small grain dimensions at the interface in As-doped polySi. This effect drastically reduces the number of states available in the conduction band at low energies. The resulting shift of the Fermi level provides a qualitative explanation for the observed puzzling difference between the work-functions of Asand P- doped polySi.

The electronic states at silicon grain boundaries trap preferentially majority carriers, thus cause a potential barrier impeding current flow. We here summarize measurement techniques for the energy distribution of these grain boundary traps. The analysis reveals band tails due to disorder in the boundary plane as well as interface state continua at midgap. The spatial distribution of trapped charges results in significant electrostatic potential fluctuations which are observable in ac-admittance and noise experiments; the charge fluctuations cause in particular 1/f-like noise. The naive model of dangling bonds to describe grain boundary charges is insufficient, impurity atoms, especially oxygen, play essential roles in determining electronic behavior.

A near Σ25 grain boundary has been studied electrically in FZ-silicon at various degrees of n-doping. The boundary is electrically active without previous heat treatment. A unique density-of-states can be derived from dc and ac measurements at all dopings and temperatures investigated. The model of the charged g.b. includes potential fluctuations in the g.b. plane. Admittance spectroscopy allows to differentiate between g.b. states and deep centers its the compensating space charge zone.

It is shown that the grain boundary (GB) in polycrystalline-silicon (poly-Si) films need not be modeled as a temperature-dependent potential barrier or as an amorphous region to explain the temperature (T) dependence of resistivity (ρ) in p-type poly-Si films at low T. Specifically, we consider that QB defect states allow for the tunneling component of current to occur by a two-step process. Incorporation of the two-step process in a numerical calculation of ρ vs. T results in excellent agreement with available data from 100 K to 300 K.

The structural and electrical properties have been investigated of antimony doped polycrystalline silicon films obtained by molecular beam deposition on oxidized silicon substrates. We show that low resistivity films with smooth morphology are obtained by Solid Phase Crystallization of antimony doped amorphous silicon layers deposited at 250°C. A resistivity of 4.3 mΩ cm is obtained by crystallizing the films at temperatures as low as 650°C for 15 minutes. Similar resistivities are typically obtained by Chemical Vapor Deposition at temperature of at least 850 °C. In-situ crystallization of the amorphous silicon is needed to obtain low resistivity polysilicon. We also show that direct deposition above 650 ° C gives rise to polycrystalline silicon with much higher resistivities.

A high performance light-beam-induced-current (LBIC) analyser has been used to determine the recombination velocity at the grain boundary (S) and the minority-carrier diffusion length (L). For this purpose a Schottky diode (Cr/Si) was fabricated using a p-type silicon bicrystal (1Ω cm, Σ13 grain boundary). The contacts were obtained by a “cold” technology. The diffusion length, determined by the method proposed by Ioannou, was subsequently fitted into the model proposed by Marek to evaluate the recombination velocity by the curve-fitting of the experimental and theoretical photocurrent profiles. A value of S = 2.104 cm/s was thus obtained. The influence of the thin oxide layer at the Cr/Si interface is also discussed.

In large grained polycrystalline silicon, the recombination activity of G.B.'s and their passivation by hydrogen is found to be dependent on the decoration by dislocations. Dislocations appear to be preferential paths for in-diffusion, at a depth of a few hundreds of pim's. Similar enhancements of diffusion and passivation exist in grains around dislocations.

Silicon bicrystals containing Σ =3(60°/[111]1,2) twin boundaries have been fabricated by sintering together single crystal (IMi) :vafers (which deviated from the exact (111) by upto ±4′). Transmission electron microscopic investigation revealed the boundaries to be near-twin boundaries having a superposed tilt component. The grain boundary dislocation structure of these boundaries has been studied in a transmission electron microscope (TEM) and interpreted as arising from the interaction of a large tilt component with a smaller twist component, to give the observed low energy configuration.

Slightly silicon rich oxides or off-stoichiometry oxides (OSO) have been prepared by LPCVD. When R, the nitrous oxide/silane ratio, equals or exceeds 30 these films have the refractive index and CV characteristics of stoichiometric CVD oxide. Injection into these oxides by oxides richer in silicon has been compared with injection into thermal oxides. We find that silicon rich oxides with higher silicon content having the same Research-article values provide the same enhancement of electron injection into thermal oxides and off-stoichiometry oxides. Conduction in off-stoichiometry oxides depends on preparation, but can be accounted for by tunneling between islands.

Experimental observations of recrystallization, normal grain growth and secondary grain growth in silicon films are reviewed. Normal grain growth leads to grain sizes which are approximately equal to the film thickness. Secondary grain growth can lead to larger grains with restricted crystallographic textures. These procesess are affected by the as-deposited or as-crystallized grain structures and orientations. The rate of grain growth has been shown to be higher in phosphorous or arsenic doped films. Ion bombardment, oxidation, and interactions with silicides also lead to increased grain growth rates. Grain growth enhancement has been related to increased point defect concentrations or dopant redistribution.

Diffusion-induced grain boundary migration (DIGM) is a common, but only recently discovered low temperature phenomenon that results in high rates of both chemical mixing (or unmixing) and grain boundary migration. DIGM is found in many situations where chemical heterogeneities lead to diffusion. For example, DIGM is observed during diffusion and compound formation in polycrystalline multilayer contact systems produced by low temperature deposition techniques. The diffusional mixing along the moving grain boundary is high, localized, and results in a distinctive composition profile behind the moving interface. Theory has indicated, and experiments have confirmed, which conditions lead to DIGM and which conditions suppress it. The microstructural changes can result in either a grain refinement as seen in many metallic systems or in enhanced grain growth as seen in polysilicon. In either case these microstructural and compositional changes are controllable in a way that may allow fabrication of unique devices.

The diffusion behavior of arsenic and the grain growth of Si in arsenic doped poly-Si were investigated by MeV4 He2+ backscattering techniques and transmission electron microscopy. By implanting arsenic ions into poly-Si films the surface portion was made amorphous and crystallized upon annealing. In-situ mssurements showed crystal nucleation and growth at temperatures of 650 – 700° C with a dimension comparable to the thickness of the amorphous layer. Annealing at temperatures up to 850°C increased the number of the large grains, but the average grain size did not change significantly. In the unimplanted region grains retained their initial size until 885°C, although implanted arsenic was found to diffuse into this region along grain boundaries. At 885°C penetration of arsenic into the interior of grains caused significant grain growth. We also found that single implants of boron somewhat increased grain size, whereas boron codoped with arsenic appeared to reduce the effect of arsenic doping. These observations support the hypothesis that the enhanced growth rate and the electrical activity of Si near the grain boundary are closely interrelated.

In previous work it has been shown that doping of silicon with P or As leads to enhanced rates of grain growth while doping with B has little effect, except in compensation of the effect of P or As. Here we report a detailed study of the effects of P doping on normal grain growth in silicon films. We also outline a kinetic model for grain growth which is consistent with the various observed effects of dopants. This model is based on the assumption that dopants primarily affect grain boundary mobilities and that grain boundary motion occurs through parallel diffusive and non-diffusive processes. It is further assumed that the rate of the diffusive process is proportional to the vacancy concentration which is a known function of the electron concentration.

Analytical results obtained from detailed Secondary Ion Mass Spectrometry (SIMS), Rutherford Backscattering Spectrometry (RBS), Auger analysis, Scanning Electron Microscopy (SEM), and Transmission Electron Microscopy (TEM) of ∼ 2000Å TiSi2/n+ polysilicon interfaces are reported for thermally annealed silicide samples and silicide samples subjected to further high temperature processing. The LPCVD polysilicon was heavily POCI3 doped at 900°C and TiSi was formed by rf sputtering 1000Å Ti and forming the silicide using ?wo successive thermal anneals at 600°C and 800°C in forming gas resulting in a silicide sheet resistance R of 1.45 Ω/□. The high temperature process stability of the silicide – polysilicon interface was investigated by systematically stressing the polycide at process temperatures in the range of 700° C to 1100° C. The silicide was stable for temperatures up to 900° C; however, significant degradation in the silicide sheet resistance, phosphorus, silicon, and titanium redistribution, and agglomeration and film breakage of TiSi2 were observed at higher process temperatures.

The thermal stability of silicide fine lines on undoped CVD polycrystalline silicon was investigated. Heat treatments were in vacuum at temperatures up to 950 ° C. We observed that fine silicide lines on undoped polysilicon degrade during vacuum annealing. Voids and hillocks are formed as silicon diffuses out from the fine grained poly-Si, undergoes long-range transport through the silicide, and recrystallizes into large grains. The phenomenon is similar to that seen previously in planar samples, but fine lines were found to degrade at lower temperatures and especially along line edges. Lines of two refractory metal silicides, TiSi2 and CrSi2, and one near-noble silicide, CoSi2 were examined.