This invention relates generally to field effect transistors and, particularly, to such transistors having strained layers in the channel regions.
Strain may be used to enhance electron and hole mobility. For example, biaxial tensile strain may be produced in silicon to enhance the electron mobility by depositing a silicon cap layer on top of a silicon germanium base layer. Since the lattice structure of the silicon germanium base layer is larger than that normally found in silicon, the deposition of the silicon on top of the base layer stretches the silicon lattice in the plane parallel to the plane of the underlying silicon germanium containing layer. This induces biaxial strain and results in tensile stress higher than 1 GPa in the silicon cap layer.
One issue with bi-axially strained layers is called self-heating. Self-heating is the generation of heat by a transistor as it conducts current. As the transistor continues to conduct current, it heats itself, raising its temperature. This increased temperature reduces the performance of the transistor. The thick relaxed silicon germanium layer used in producing the tensile strain in the silicon layer, for example, has thermal conductivity ten times lower than a silicon layer, hence causing more serious self heating problem.
To produce a biaxial compressively strained channel, the same lattice mismatch property between silicon and silicon germanium can be used. In one case, a thin silicon germanium deposited on top of silicon. Where the larger silicon germanium accommodate to the underlying small lattice of silicon and results in compressive strain in the silicon germanium layer. In another case, silicon germanium is grown in a recessed source/drain region, exerts uniaxial compressive stress toward the channel region from both sides of the channel region, creating compressive stress in the channel region.
As complementary metal oxide semiconductor transistors become increasingly smaller, to improve density and electrical performance, the physical gate thickness has been aggressively scaled, along with the transistor gate length. However, reducing the physical thickness of the gate oxide, increases gate oxide leakage due to tunneling. A high dielectric constant gate oxide can be used in place of silicon dioxide to reduce the gate oxide leakage problem. Since a physically thicker high dielectric constant gate oxide can provide equivalent electrical control, at the channel surface, to a much thinner silicon dioxide gate oxide, the gate leakage through gate oxide may be reduced.
Referring to
Over the layer 14 may be a relaxed silicon germanium layer 16. The layer 16 may have the formula Si1-yGey. Over the layer 16 may be formed a biaxial tensile strained silicon layer 18. As described previously, the epitaxial deposition of the silicon layer 18 on the silicon germanium layer 16 results in biaxial tensile strain. Basically, the silicon layer 18 is strained in the x and y directions, parallel to the plane of the layer 16, because its lattice is forced to matches the size of the larger silicon germanium lattice of the layer 16. At the same time, in the direction transverse to the plane of the layer 16, the silicon layer 18 lattice may be shrunk.
A source and drain 20 may be formed through the silicon layer 18 and into the relaxed silicon germanium layer 16. Salicide contacts 22 may be formed on the source and drain 20 in one embodiment of the present invention.
A gate stack may include, in one embodiment, a salicide contact 30, a polysilicon layer 28, a metal layer 26, and a high dielectric constant gate oxide 24. The oxide 24 may have a gate dielectric constant greater than 10 in one embodiment of the present invention. A sidewall spacer 32 may be formed over the sides of the gate structure.
In one embodiment of the present invention, the thickness of the layer 18 is at least about 500 Angstroms. At thicknesses of about 500 Angstroms and greater, the thermal conductivity of the layer 18 is effective to reduce self-heating. In some embodiments of the present invention, the layer 18 may be as thick as 2000 Angstroms. The layer 18 may be as thick as 2000 Angstroms in some embodiments without releasing all the strain upon thermal processing.
The graded layer 14, in some embodiments, may be graded from zero percent germanium (x) at the silicon substrate 12 to y percent at the top where the relaxed layer 16 is formed. The grading of the layer 14 may reduce the dislocation density in the active channel area. Thus, the germanium concentration x may be gradually increased from the surface over the silicon substrate 12, up to that germanium concentration y, exhibited by the layer 16.
In some embodiments, the use of a high dielectric gate oxide 24 and a metal gate 26 may reduce the thermal anneal of the strained silicon layer 18, ensuring that more of the strain is retained after thermal processing. The thick silicon layer 18 may be an epitaxial layer having sufficient thermal conductivity to reduce the self-heating effect due to the poor thermal conductivity of silicon germanium graded and relaxed underlying layers. In other words, heat may be conducted upwardly out of the transistor despite the fact that the layers underlying the layer 18 may be relatively poor thermal conductors.
As a result, the electrical performance, including drive current, may be improved through mobility enhancement. The mobility enhancement arises from the biaxial strain, exerted by the silicon germanium relaxed layer 16, without suffering a loss of performance due to self-heating in some embodiments.
As one example, the strained silicon layer 18 may be grown on a relaxed layer 16, having the formula Si0.85Ge0.15 and a graded layer 14, which goes from 0 to 15% germanium. A 2100 Angstrom silicon layer 18 may have a thermal conductivity of about 1.5 W/(cm. deg.) based on simulated data.
Referring to
Referring to
Referring to
Then, referring to
The resulting structure is shown in
In some embodiments, a PMOS structure may be formed using any suitable process. It may be conventional in some embodiments of the present invention.
Referring to
The substrate 12 may be a substantially silicon substrate. A biaxial compressive strained silicon germanium epitaxial layer 30 can be formed thereover. The layer 30 may be extremely thin and may be coherently grown directly on the substantially silicon substrate 12 in one embodiment. As a result, the silicon germanium epitaxial layer 30 is biaxially compressively strained. The compressive strain is exerted in the channel area through global biaxial coherency stress from the silicon substrate 12 due to the lattice constant mismatch between the silicon substrate 12 and the thin silicon germanium epitaxial layer 30.
Then, as shown in
The gate electrode 34 may be a metal which is workfunction matched to the silicon germanium epitaxial layer 30, or a combination of workfunction matched metal at the bottom and a polysilicon in the upper portion of layer 34. The gate dielectric 36 can be a high dielectric constant material. The high dielectric constant gate oxide 36 may reduce gate oxide leakage and may reduce the gate oxide to silicon germanium epitaxial layer 30 interface charge.
In some embodiments, the resulting transistor may exhibit enhanced hole mobility at both low and high electric fields because of the coherently strained thin silicon germanium channel. The silicon germanium channel can be operated as a surface channel with the gate dielectric 36 grown directly on the silicon germanium epitaxial layer 30.
The structure shown in
In one embodiment, the silicon germanium epitaxial layer 30 may have the formula Si1-xGex where x is from 10 to 50 percent. The layer 30 may be from 25 to 1000 Angstroms thick in one embodiment of the present invention.
The thin silicon germanium epitaxial layer 30 is grown on a silicon substrate 12 with the germanium concentration between 10 and 50 percent and a thickness between 25 Angstroms and 1000 Angstroms. The gate dielectric 36 may be a high dielectric constant gate dielectric such as hafnium dioxide or zirconium dioxide. The gate electrode 34 may be a metal gate, such as titanium nitride or tantalum nitride, with or without a polysilicon cap. The workfunction of the gate electrode 34 may be matched to the silicon germanium channel. The dielectric 36 may be formed directly on the silicon germanium epitaxial layer 30.
Salicide (not shown) may be formed on top of the source/drain regions 20 and/or the gate region in some embodiments. The silicon germanium epitaxial layer 30 may be removed from the NMOS active region by a wet chemical etching down to the silicon bulk surface and then the NMOS devices can be fabricated using standard NMOS process flows.
In accordance with another embodiment of the present invention, shown in
The silicon capping layer 38 creates a quantum well, as well as a surface on which can be grown the gate dielectric 36. The quantum well confines the carriers to the well and so there is less scattering, especially surface scattering. The capping 38, made of silicon, has a wider band gap than the hole quantum well formed by the silicon germanium epitaxial layer 30. The capping layer 38 is not strained since the layer 30 is compressively stained to match the lattice structure of the underlying silicon substrate 12. Thus, the layer 38, of silicon, has the lattice structure which matches the lattice structure of the layer 30 without strain.
A thin silicon germanium epitaxial layer 30 is grown on a silicon substrate 12 with a germanium concentration between about 10 to 50 percent and a thickness between 25 to 1000 Angstroms. This is followed by a thin epitaxial silicon cap growth to form the silicon germanium quantum well structure. The silicon capping layer 38 may have a thickness of between 25 and 200 Angstroms in some embodiments.
The gate dielectric 36, shown in
Referring to
The overlying silicon capping layer 38 may provide a layer from which to grow a gate dielectric 42 if desired.
The structure shown in
The narrow band gap layer 40 effectively increases the sensitivity of the channel threshold voltage to source-substrate bias. The rate of change of threshold voltage, with respect to source-substrate bias may be determined as a function of band gap of the layer 40. For a given change in source-substrate bias V, a narrow band gap layer 40 may result in a higher reduction in “on” threshold voltage, resulting in higher drive current.
Using a narrower band gap layer 40 reduces the built-in potential of the source-channel junction. Thus, devices may have a higher off current. In order to match the off current, the halo doping concentration may be made higher.
In one embodiment, using silicon germanium in layer 40, the threshold gain may be higher in PMOS transistors than NMOS transistors. The higher gain in PMOS transistors may be due to the fact that band offset at the silicon germanium interface lies mostly in the valence band. The valence band discontinuity tends to confine the holes in the narrow band gap layer 40, increasing the peak density of the inversion charge. However, using a different narrow band gap material, such as silicon carbon with band offsets mostly in the conduction band, may result in higher gain in NMOS devices as well.
In one embodiment, shown in
In some embodiments of the present invention, the narrow band gap layer 40 may have a thickness of below 300 Angstroms. This thin, narrow band gap layer 40 reduces self-heating. Generally, the germanium concentration of the narrow band gap layer 40 improves performance. Germanium concentrations of greater than 40 percent may be advantageous, but germanium concentrations beyond 10 percent may be effective in some cases.
Advantageously, the silicon capping layer 38 is made sufficiently thin to avoid unduly spacing the layer 40 from the gate electrode 44. For example, the silicon capping layer 38 may be approximately 10 Angstroms thick in one embodiment. In general, it is desired that the band gap of the layer 40 be less than the band gap of the underlying material, such as the substrate 12, in the configuration shown in
Referring to
In some embodiments, it may be advantageous to provide an interface (not shown) between the substrate 12 and the layer 40. The interface may exhibit a decreasing germanium concentration to reduce biaxial strain.
In another embodiment of the present invention, the layer 40, shown in
In the embodiments of
While the present invention has been described with respect to a limited number of embodiments, those skilled in the art will appreciate numerous modifications and variations therefrom. It is intended that the appended claims cover all such modifications and variations as fall within the true spirit and scope of this present invention.