US 20050124125 A1
A method for forming a transistor includes forming a gate dielectric layer over a portion of a semiconductor substrate, the substrate being substantially free of silicon; defining a gate electrode over a portion of the gate dielectric layer; and introducing ions into the substrate proximate the gate electrode to define source and drain regions. A transistor includes a semiconductor substrate that is substantially free of silicon and a gate, dielectric layer over a portion of the substrate. The transistor can also include a gate electrode over a portion of the gate dielectric layer and introduce ions proximate the gate electrode, defining source and drain regions.
1. A method for forming a transistor, comprising:
forming a gate dielectric layer over a portion of a semiconductor substrate, the substrate being substantially free of silicon;
defining a gate electrode over a portion of the gate dielectric layer; and
introducing ions into the substrate proximate the gate electrode to define a source region and a drain region.
2. The method of
depositing an interlayer dielectric layer over at least part of the gate electrode, the source region, and the drain region;
defining first, second, and third openings in the interlayer dielectric layer over at least part of the gate electrode, the source region, and the drain region; and
depositing a metal into the first, second, and third openings to contact the gate electrode, the source region, and the drain region.
3. The method of
4. The method of
5. The method of
6. The method of
7. The method of
22. The method of
23. The method of
24. The method of
25. The method of
26. The method of
This application relates to metal oxide semiconductor field effect transistors (MOSFETS) formed with non-silicon semiconductors and high dielectric constant (k) gate dielectrics.
Silicon is commonly used as a substrate material for the fabrication of integrated circuits. Devices are continually being scaled down in size, including in the vertical direction by reducing gate oxide thickness and in the horizontal direction by reducing channel length. Device power supply voltage (Vdd) is also being reduced to reduce power consumption.
Silicon material properties, as well as integrated circuit processing capabilities, restrict the shrinking of silicon-based devices. These limiting properties include the intrinsic carrier mobility of silicon [μn=˜1450 centimeter2*volts−1*second−1(cm2V−1s−1) and μp=450 cm2V−1s−1, where μn=mobility of n-type carriers and μp=mobility of p-type carriers] which sets the achievable cutoff frequency to less than 160 Gigahertz (GHz) for a gate length of 30 nanometers (nm).
Power dissipation increases as threshold voltage (Vt) decreases. Two major components of power dissipation are dynamic capacitive switching and static, off-state leakage current. Dynamic power dissipation can be expressed as Pd=CV2f, where C=capacitance, V=operating voltage, and f=repetition frequency. Lowering V decreases dynamic power dissipation, but the effect is offset by higher operating frequency and increased C due to the vertical scaling down of gate dielectric thickness.
Leakage current primarily comprises subthreshold conduction in off-state(Isub), reverse bias pn junction conduction(ID), and tunneling through gate dielectrics(Ig). Subthreshold conduction occurs when a MOSFET device is operated with a Vg below Vt. Subthreshold conduction is proportional to the weak inversion carrier density ˜e−φs/kT, where φs=electric potential at semiconductor surface, k=Boltzmann's constant, and T=temperature, with φs being proportional to the difference between Vg and Vt. During the process of lowering the operating voltage, Vt should also be lowered to maintain the V/Vt ratio for sufficient current gain. An adverse consequence is that subthreshold leakage current increases exponentially with decreasing Vt.
Reverse bias leakage current occurs at reverse biased drain/well and source/well junction regions. It is caused by thermal generation in depleted regions and by diffusion of minority carriers across reverse biased junctions. This leakage is especially problematic at the source and channel well regions when the channel length is so short that the electric field of the drain to source voltage effectively lowers the barrier across the source/channel depletion region and causes large offstate leakage current. This is commonly called drain induced barrier lowering effect for short channel devices.
Tunneling leakage is due to quantum mechanical tunneling of electron wavefunction across a gate dielectric. Tunneling leakage is expected to increase as conventional silicon dioxide (SiO2) gate dielectrics shrink in a vertical dimension. This tunneling leakage current will become a dominant source of off-state leakage when conventional silicon dioxide layers are scaled down below an effective oxide thickness (Tox) of 1.6 nm.
These static power dissipation effects become a significant portion of the total power dissipation in increasingly smaller and highly packed logic products.
A high speed device, having a high cutoff frequency, e.g., >200 GHz with a 30 nm gate length, can be fabricated by using a semiconductor substrate with a narrow band gap and a carrier mobility higher than that of silicon. A high carrier mobility allows one to achieve a higher device speed than a silicon-based device with the same transistor gate length. Using a substrate with a high carrier mobility, therefore, allows one to achieve higher device speeds without requiring greater photolithographic capabilities.
A high carrier mobility can also allow lower operating voltage (V) for a given threshold voltage (VT). Because drain current is proportional to the product of carrier mobility and V−Vt, a high mobility semiconductor can provide equivalent current gain with a smaller difference V−Vt or, in another words, a smaller V/Vt ratio. A lower operating voltage, in turn, lowers power consumption. In an alternative embodiment, by using a non-silicon substrate with a high carrier mobility, one can maintain the threshold voltage of a transistor at a sufficiently high value to avoid excessive leakage current, while still achieving lower power consumption with a lower operating voltage without losing current gain.
A high mobility non-silicon semiconductor substrate is used to gain higher transistor operation speed. In some embodiments, the high mobility non-silicon semiconductor substrate can allow lower operating voltage without significantly lowering the threshold voltage. This avoids large subthreshold leakage in short channel devices.
To fabricate a non-silicon based transistor, a gate dielectric chemically compatible with the substrate is identified, which is analogous to the SiO2 used with silicon. An atomically smooth interface between the substrate and the gate dielectric is used to reduce surface recombination due to interface traps and electron hole pair generation at the substrate/gate dielectric interface. In comparison, in the Si/SiO2 system, less than one charge site in 105 interface atoms is achievable at the Si/SiO2 interface. The dielectric has a high dielectric constant that allows thicker gate dielectric thickness, thereby reducing gate leakage current.
A sacrificial oxide layer 12, such as a metal oxide, is formed on substrate 10 by, e.g., plasma enhanced chemical vapor deposition (PECVD). Sacrificial oxide layer 12 serves as a protective layer for substrate 10 during subsequent processing steps such as implant, clean, and polishing. Sacrificial oxide layer, 12 protects substrate 10 from contamination, provides a low defect interface, and can be etched away easily with a high selectivity to the underlying substrate 10. Sacrificial oxide layer 12 may have a thickness T1, of 10-2000 Ångstroms (Å). In an embodiment, sacrificial oxide layer 12 has a thickness T1, of 10-500 Å. A polishing stop layer 14 is deposited by PECVD over sacrificial oxide layer 12. Polishing stop layer 14 has a thickness T2 of, e.g., 1000-2000 Å, and is made of a hard material, such as silicon nitride, which can act as polishing stop layer-during subsequent processing.
Referring also to
The first photoresist layer is removed and a second photoresist layer (not shown) is applied and patterned, so that center region 26 is exposed and regions 28, 30 are covered. Ions are implanted in center region 26, which is unprotected by the photoresist layer, to form a p-well 36 by doping substrate 10. In the case of a group IV substrate 10, such as germanium, the dopant ions implanted to form p-well 36 can be an element with less than four valence electrons like a group III element such as boron, aluminum or gallium.
In selecting ions for doping both n-wells 32, 34 and p-well 36, the solubility of the dopant ions in substrate 10 may be taken into consideration. In this embodiment, each dopant is capable of forming a stable alloy phase with substrate 10, and is sufficiently soluble in substrate 10 to avoid cluster formation.
After implantation of first and second n-wells 32, 34 and p-well 36, an implant activation anneal is performed by heating substrate 10 in a furnace at a temperature and time which depend on the elements comprising the dopants implanted in n-wells 32; 34 and p-well 36, as well as substrate 10. The anneal may be performed at a temperature which is roughly 70% of the melting point of substrate 10. For example, for a germanium substrate 10 having a melting point of 938° C., a suitable annealing temperature is approximately 658° C. The duration of the annealing depends on the type and dosage of the implanted species, and is a function of the mobility of the implanted dopant. This activation anneal activates dopants to increase the concentration of majority carriers.
Referring also to
The high dielectric constant of high-k gate dielectric 50 allows one to use a thicker gate dielectric layer than is possible with silicon dioxide, and thereby reduce gate leakage current. High-k gate dielectric 50 acts essentially as a capacitor, with capacitance C=(k*A)/thickness, where k=dielectric constant and A=area of capacitor. The thickness of a high-k gate dielectric having a capacitance equivalent to that of a SiO2 layer of a given thickness is determined by the equation
A gate electrode layer 54 is deposited over high-k gate dielectric 50. Gate electrode layer 54 is made of a material selected, in part, on the basis of its work function, i.e. the minimal energy required to move an electron from the Fermi level EF to vacuum. The work functions of gate electrode layer 54 and substrate 10 are matched, e.g., ideally, that the work function of metal together with the doping level (well doping) of the substrate, and the high k dielectric layer thickness 10 give the desired threshold voltage of the MOS transistors. Gate electrode layer 54 is made of, for example, titanium nitride, e.g., for pMOS germanium; tantalum for, e.g., nMOS germanium; tantalum nitride; titanium; nickel; platinum; polygermanium; polysilicon, etc. Gate electrode 54 has a thickness T4 of, for example, 50 Å-5000 Å.
To specify high-k dielectric layer 50 and gate electrode layer 54 materials, one determines the desired threshold voltage for the transistor that will be formed from these layers. An appropriate high-k dielectric layer 50 thickness is chosen, taking into account the amount of leakage that can be tolerated. The material for the gate electrode layer 54 is selected. The well doping levels of the corresponding n-type well regions 32, 34 for pMOS devices and p-type well regions 36 for NMOS devices are chosen, taking into account gate electrode layer 54 work functions and the high-k dielectric layer 50 thickness T3.
Referring also to
Transistor 90 includes source and drain regions 66, 68, gate electrode 56, and high-k gate dielectric 50, as well as source and drain shunting regions 70, 72, lightly doped drains 58, 60, and sidewall spacers 62, 64. Transistor 90 is fabricated on substrate 10 formed of, e.g., germanium, a material with a narrow bandgap of 0.66 eV and a carrier mobility higher than that of silicon. Because the cutoff frequency is directly proportional to carrier mobility, substrate 10's high carrier mobility enables transistor 90 to be designed with a cutoff frequency of >200 GHz, which is higher than that obtained with silicon.
After electrical contacts are made to the source region 66, drain region 68, and gate electrode 56, a Damascene interconnect scheme (not shown) is used to connect various transistors 90 to form an integrated circuit.
The application is not limited to the specific embodiments described above. For example, the substrate material can be any semiconducting material having a carrier mobility higher than that of silicon, in addition to the materials listed above. The semiconducting material can be an epitaxial layer on a substrate or it can be a layer bonded to a substrate. The sacrificial oxide layer may be formed by alternative methods, such as grown in a furnace or deposited by low pressure chemical vapor deposition (LPCVD). Instead of isolation trenches, isolation regions can be defined by ion implantation. N-wells and p-wells can be defined before the formation of isolation trenches or isolation regions. High-k gate dielectric can be one of many materials with a high capacitance, in addition to the materials listed above. The gate electrode can be a different material for n-channel and p-channel devices. The source and drain, as well as the lightly doped regions, can be formed by the implantation of various n-type ions. Alternatively, the source, drain, and lightly doped regions can be formed by introducing the ions with a CVD or a solid phase diffusion process. Source and drain shunting regions can be formed by a lift-off process, or a by forming a highly conductive metallic compound with the substrate.
Other embodiments not described herein are also within the scope of the following claims;