This invention relates generally to field-effect transistors (FETs), and more particularly to FETs employing buried gate structures and a channel width that varies with depth to engineer transfer characteristics which improve device linearity.
In most conventional field effect transistors (FETs), the gate electrode is formed on the top of a semiconductor surface to modulate the drain current by the use of a vertical gate electric field. This forms a region at the drain end of the gate where electric field strength becomes the highest during high voltage operation. This is a primary cause of critical problems for conventional FETs. For example, current collapse caused by electron trapping at the drain end of the gate on the semiconductor surface is a commonly known problem which degrades RF power performance (output power, efficiency, linearity, gain) of FETs when they are operated to provide a large voltage swing. For GaN-based high electron mobility transistors (HEMTs) with strong piezoelectricity, the high vertical electric field at the drain end of the gate can cause cracks in the epitaxial layers by inducing tensile stress in the top barrier material (so-called “inverse piezoelectric effect”). This effect limits the reliability of GaN-based HEMTs.
Further problems can arise when operating FETs at high frequencies. For example, for conventional FETs, electrostatic isolation is degraded when the gate length is scaled down for high frequency operations (“short channel effects”). This limits a FET's gain and can degrade its off-state leakage and breakdown voltage characteristics. Additionally, although a low knee voltage is preferred for high efficiency operation of power amplifiers (PAs) and low noise operation of low noise amplifiers (LNAs), knee voltage is largely limited by parasitic resistances arising from metal-semiconductor contact resistance and device access resistances. Ideally, when the drain-source voltage is low (in the ohmic region), all the voltage is applied only in the intrinsic active PET region for early channel pinch-off, without an excess voltage drop in parasitic resistances to decrease Vknee.
For high linearity operation, a gradual gin curve is preferred because abrupt change in the gin curve results in large gm derivatives (gm′ & gm″), which degrade the transistor's linearity performance. In conventional HEMTs with a top gate contact, the gm curve typically shows a peak, which is due to their operation principle that uses electron density modulation by the vertical gate field. In conventional metal-semiconductor field-effect transistors (MESFETs), the gm curve is more gradual due to their mode of operation, but the electron velocity is lower than HEMTs due to increased scattering by ionized impurities in the channel, limiting their operational frequencies.
During high power operation of a PET, dissipated power turns into heat (“self-heating”). This limits output power, gain, and efficiency of PAs. In addition, an increased junction temperature reduces the lifetime of the transistors. In a conventional PET, the junction temperature peaks at the drain end of the gate where the electric field is the highest, and the peak junction temperature increases with increasing device periphery, i.e., gate width.
MESFETs are known to be more linear than high electron mobility transistors (HEMTs). One device-level linearization technique that has been used in MESFETs is optimizing a doping profile for a linear gm profile (see, e.g., Jose Pedro, “MESFET linearity improvement by channel doping control”, 1527-1530, vol. 3, 10.1109/MWSYM (1995)). Drawbacks of MESFETs are their lower frequency and noise performances with respect to HEMTs.
A device known as a Superlattice Castellated Field Effect Transistor (SLCFET)-described, for example, in U.S. Pat. No. 9,202,906—employs a structure having multiple 2DEG channels and a tri-gate (i.e., with both top and lateral contacts), where the threshold voltage of the channels is determined by both vertical and lateral electric fields.
A FET with buried gate structures and a tapered channel layer is presented, in which the buried gate structures contact an epitaxial channel layer only from the sides. The gates do not contact the FET's top surface, and hence drain current is controlled only by channel width modulation by lateral gating of the channel layers by the buried gate structures.
The epitaxial channel layer comprises multiple channel segments, each of which has a width defined as the distance between adjacent ones of the buried gate structures. As described herein, the FET is fabricated such that the channel segment widths ‘taper’— i.e., the widths vary along the depth direction. By controlling the slope of the channel sidewalls and the distance between buried gate structures, the FET's transfer characteristics can be engineered to improve its linearity. For example, a FET has an associated transconductance (gm). As described herein, the slope of the channel sidewalls can be controlled to minimize gm″ and the slope of the gm″−Vgs curve near the point at which gm″=0. High gm/gm″ is known to improve a FET's linearity.
By having the channel segment widths vary as described herein, the threshold voltages of the channel segments also vary along the depth direction. This results in the FET having multiple transfer curves, which are superposed to provide the FET's overall transfer characteristic.
The present FET may comprise multiple epitaxial channel layers stacked in the depth direction, each of which comprises a two-dimensional electron gas (2DEG) plane. Alternatively, the device may comprise a single uniformly doped epitaxial channel layer having a width that varies along the depth direction. The concepts descried herein can be applied to multiple FET types, including HEMTs and MESFETs.
These and other features, aspects, and advantages of the present invention will become better understood with reference to the following drawings, description, and claims.
An exemplary embodiment of a FET with a buried gate structure is illustrated in
As shown in
Active devices are a prime source of non-linearity in RF circuits. The present invention reduces the nonlinear behavior of FETs through material and device co-design, employing a new transistor structure that improves device-level linearity by engineering a desired transconductance (gm) profile. One possible embodiment of the present device structure is illustrated in the sectional view of
When so arranged, epitaxial channel layers 30 comprise multiple channel segments, such as channel segments 44, 46, 48. Each channel segment has a width defined as the distance between adjacent ones of buried gate structures 36. To engineer a desired transconductance (gm) profile, the FET is fabricated such that the channel segment widths vary along the depth direction. For example, in
By tapering the channel segments in this way, each channel has a slightly different threshold voltage. This is illustrated in the graph shown in
As shown in
One possible embodiment of the present FET is a HEMT in which the current-carrying portion of the epitaxial channel layer comprises a 2DEG plane. As shown in
Suitable materials for buried gate structures as described herein comprise metals, or p-type semiconductors (p-type NiO material, p-type GaN material, p-type CuS material, or a stack comprising a gate dielectric and a metal). Suitable metals include Pt, Ni, or Au. A suitable stack might include Al2O3/Pt or HfO2/Pt. Alternatively, the buried gate structures might comprise p-type NiO material, p-type GaN material, or p-type CuS material.
A FET with a tapered channel layer as described herein might also be employed in a device with a single uniformly doped channel layer. This is illustrated in the sectional views shown in
Just as the multi-channel devices described above can be arranged such that the channel segments get narrower or wider with depth (as shown in
The buried gate structures can have any of a number of shapes. For example, the structures may be cylindrical, as shown in
The buried gate structures preferably lie along a line which is parallel to and between the source and drain contacts, as illustrated in
HEMTs are depicted in the examples shown, though the buried gate and tapered channel layer arrangement described herein could be applied to other FET types, including MESFETs. A MESFET per the present invention would have an epitaxial buffer layer on a substrate, and an epitaxial channel layer on the buffer layer. The channel layer would be as described above—i.e., with a width that varies along the depth direction. In one possible embodiment, the buffer layer suitably comprises GaN, and the epitaxial channel layer comprises n-type or p-type GaN. Other possible materials for the epitaxial buffer layer include Al(Ga)N, (In)GaAs, InP, or Ga2O3.
The depth to which the buried gate structures should be buried depends on the device type. As noted above, for a HEMT, the bottoms of the gate structures should be buried to a depth at least equal to the bottom of the current-carrying 2DEG plane in the lowermost channel layer. If the device is a MESFET, the buried gate structure should be buried to a depth at least equal to the bottom of the channel layer. In general, the buried gate structures must be buried deep enough so that they contact the current-carrying portion of the channel layer or layers only from their sides, such that the FET's drain current is controlled by channel width modulation by lateral gating of the current-carrying layer by the buried gates structures.
As noted above, the gap between the top of the substrate and the bottom of the gate electrode head portion is filled with a dielectric, such as SiN, SiO2, BCB, or simply air. If a dielectric material is used, it should be thick enough so that the head portion does not modulate the FET current via the vertical gate field (through the dielectric).
A FET whose drain current is controlled only by channel width modulation by lateral gating of the current-carrying layer(s), and with channel layers having widths that vary along the depth direction, enables small variations and precise control in the threshold voltage for each channel, making it possible to achieve a very linear gm profile. Applications for the present FET include power amplifier (PA) MMICs with high linearity, and low noise amplifier (LNA) MMICs with high linearity.
The embodiments of the invention described herein are exemplary and numerous modifications, variations and rearrangements can be readily envisioned to achieve substantially equivalent results, all of which are intended to be embraced within the spirit and scope of the invention as defined in the appended claims.
This application claims the benefit of provisional patent application No. 63/352,564 to Keisuke Shinohara et al., filed Jun. 15, 2022.
Number | Date | Country | |
---|---|---|---|
63352564 | Jun 2022 | US |