BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to a method of producing ultrathin crystalline silicon nitride on Si (111) and formation of semiconductor devices using such ultrathin crystalline silicon nitride.
2. Brief Description of the Prior Art
The continued down-scaling of the geometries in VLSI technology has involved, as a result of such down-scaling, a reduction in component film thicknesses, examples being gate dielectrics for FET semiconductor devices and the capacitor dielectric for semiconductor memory devices. Thickness uniformity requirements for such films (about 0.14 nanometers in thickeness in present technology) requires extraordinary control over the silicon wafer surface morphology (i.e., subsequent interfacial roughness) to achieve necessary scaling. The acute sensitivity of interface roughness with ultrathin films is evident when one considers the control required over large (200 mm or 300 mm) wafers.
Conventional silicon semiconductor technology incorporates Si (100) substrates, largely because of interface trap density (Dit) considerations associated with oxide films on Si (100) which has been extensively researched over the past two decades. Moreover, it has been demonstrated that surface preparation methods currently developed, such as HF-last treatments, can result in a Si (100) hydrogen-terminated surface with a roughness which is unacceptable for prospective dielectric film thickness uniformity requirements. The use of alternative dielectric materials, such as silicon nitride, has been considered as a means to increase the gate dielectric constant and also to serve as a diffusion barrier to dopants in the gate material. However, the current silicon nitride fabrication techniques on Si (100) result in an amorphous nitride or oxynitride layer which may exhibit deleterious interface states (traps) which degrade ultimate device performance.
A further problem with silicon dioxide dielectrics over Si (100) substrates is that boron from boron-doped polysilicon gate structures can diffuse through the silicon dioxide, this problem increasing with decreased gate oxide thickness geometries, thereby degrading the properties of the device, particularly in the channel region. Boron, on the other hand, does not diffuise through silicon nitride, however, the interface between silicon nitride and Si (100) results in an amorphous silicon nitride and provides an inferior structure to that with silicon dioxide by causing a disruption of the electron flow in the channel of the active semiconductor devices.
A separate problem with silicon dioxide dielectrics is that the extremely small thicknesses allow unacceptable leakage currents as a result of electrons tunneling from the gate to the drain regions of transistors. Since silicon nitride has a larger bulk dielectric constant than silicon dioxide (˜7 compared to about 3.9), a thicker silicon nitride layer can be used which has the same capacitance density as a thinner silicon dioxide layer. Since electron tunneling currents depend exponentially on layer thickness, even an increase in dielectric thickness of about 10 to about 20 Angstroms could reduce leakage current by several orders of magnitude.
SUMMARY OF THE INVENTION
Recent work has demonstrated that the Dit from oxides on Si (111) can be made comparable to those on Si (100), thus making devices on such substrates feasible. The silicon (111) surface can be controlled to be made hydrogen-terminated and atomically flat from a careful control of the surface preparation solution pH. The resultant smooth surface can therefore result in a low roughness (<0.1 nm, rms) interface after subsequent film deposition. Recent research has also demonstrated the potential formation of an ordered silicon nitride film on the Si (111 ) surface from the reaction of NH3 with an atomically clean Si (111) surface, i.e., where no surface impurities are detected, at temperatures between 800° C. and 1130° C. under high vacuum conditions of from about 10−7 to about 10−5 Torr NH3 partial pressures. The cleaning process could include, for example, a standard semiconductor wet clean followed by oxidation (chemical or thermal), then followed by HF-last stripping of the oxide for H-termination. The hydrogen is then desorbed in the course of the temperature ramp-up for nitride deposition. Alternatively, the cleaning can take place by ultra high vacuum (UHV), from about 10−11 to about 10−9 Torr, “flash heating” to about 1100° C. and cooling to room temperature to form a well-ordered surface. Under the proper temperature conditions (850° C. to 1000° C.), the nitride film covered Si (111) surface is atomically flat, i.e., where only single height steps between nitride terraces exist. The resultant crystalline film is thus useful for an epitaxial nitride layer, or it will be useful for the purpose of surface passivation and subsequent crystalline or amorphous dielectric film overgrowth.
Interface state densities associated with such an epitaxial layer are low because the dangling bonds are consumed with the epitaxial growth process. Moreover, the smooth interface afforded by the Si (111) surface preparation results in an atomically flat nitride layer as well. Such sharp smooth interfaces result in enhanced electron mobility properties (less interface scattering) as well as a superior dopant diffuision barrier. Any residual dangling bonds can be satiated from a H2 or D2 sintering process.
In accordance with the present invention, the above described problems of the prior art are therefore minimized and there is provided an ultrathin crystalline silicon nitride layer on Si (111) for use, primarily though not exclusively as a gate dielectric for semiconductor devices and as a capacitor dielectric in semiconductor memory devices.
Briefly, by growing crystalline silicon nitride on Si (111), the barrier to boron diffusion is retained and, in addition, the channel is not disrupted as in the case of amorphous silicon nitride over a Si (100) substrate.
Another problem of the prior art that is minimized in accordance with the present invention is based upon the fact that as that the drive current is proportional to the capacitance between the gate electrode and the substrate. Therefore, for a given drive current as the contact area of the dielectric decreases the dielectric thickness must also decrease. The result is that electrons from the gate electrode are then capable of tunneling through the dielectric and add to the channel or drain current, resulting in lack of device control. Since the dielectric constant of silicon dioxide is about 3.9 and the dielectric constant of silicon nitride is about 7, a thicker layer of silicon nitride can be provided with the same capacitance and drive current properties, yet prevent electron tunneling through the dielectric.
To form a semiconductor device in accordance with the present invention, there is initially provided a surface of Si (111) which has been cleaned and is atomically flat as defined hereinabove. The Si(111) surface is placed in a standard reaction chamber and the chamber purged and filled with ammonia (NH3) at an ammonia partial pressure of from about 1×10−7 to about 1×10−5 Torr at a temperature of from about 850° C. to about 1000° C. for from about 5 seconds to about 5 minutes to provide a thin layer of crystalline silicon nitride on the Si (111) of from about 0.3 nm to about 3 nm. The remainder of the semiconductor device is then fabricated in standard manner by, for example, depositing a doped layer of polysilicon or a metal layer over the silicon nitride layer. In the case of a boron-doped polysilicon electrode, the boron will be prevented from diffusing through the dielectric due to the use of silicon nitride as the dielectric.
While silicon nitride has been discussed above as being the dielectric material, it should be understood that other materials can be used that have a higher dielectric constant than and are compatible with silicon nitride. When there is a lack of silicon compatibility resulting in SiOx formation at the interface, such as, for example, with tantalum pentoxide (Ta2O5), titanium dioxide (TiO2) or a perovskite material, a very thin layer of silicon nitride can be used to separate the dielectric material from the Si (111) substrate and/or the electrode over the dielectric, such thin layer having a thickness of about 2 monolayers.