Patent Application
20230006066
The present invention relates to semiconductor devices (e.g., diodes, transistors, etc.) having a region of undoped semiconductor material that is made effectively n-type or p-type by inducing a quantity of electrons or holes (respectively) on the semiconductor side of a metal-semiconductor junction by virtue of a negative Schottky barrier between the metal and the semiconductor.
With continued scaling of metal oxide semiconductor (MOS) field effect transistors (FETs), the area available for making electrical contacts to doped semiconductor source/drain regions is decreasing. As a consequence, the resistance of such contacts (which are typically metal-to-semiconductor contacts) is becoming an excessively large component of the total electrical resistance of a transistor when it is switched on. This undesired metal-semiconductor contact resistance is becoming a very significant performance limiting factor for such devices, contributing both to wasted energy and reduced switching speeds (clocking rates) in digital integrated circuits comprising such transistors. Furthermore, the decreasing volume of doped source and drain regions in state of the art transistors accommodates fewer dopant atoms, even at very high doping concentrations in excess of 1020 dopants/cm3. As a consequence, the variability in transistor performance that is a result of variance in doping species number and placement is predicted to become a significant problem in future, nanometer-scale MOS transistors, particularly in anticipated, gate-all-around nanowire transistors.
Contact resistance is conventionally calculated as “contact resistivity” divided by the area of the contact. Traditionally therefore contact resistance has been minimized by ensuring as low a contact resistivity and as large a contact area as technologically possible. According to Park et al., “Scaling effect on specific contact resistivity in nano-scale metal-semiconductor contacts”, Proc. Device Research Conference (2013), however, “initial results indicate that contact resistivity increases in the limit of very small contact areas and that the effect is stronger in the 3D wire case compared to the 2D wire case.” Contact resistance of metal-semiconductor contacts is expected to increase even more than a classical model would predict in the size regime of 10 nm and smaller due to the two-fold effects of increasing resistivity and decreasing contact area. There is then a serious metal-semiconductor contact resistance problem for present and future nanoscale transistors that have contact dimensions of approximately 10 nm or less.
Furthermore, in present day, state-of-the-art transistors, at nodes smaller than 20 nm, the semiconductor channel is fully depleted, whether the transistors are fully-depleted silicon-on-insulator (FDSOI) FETs, FinFETs, “tri-gate FETs”, nanowire FETs or gate-all-around FETs. Fully depleted implies that the thickness of the semiconductor body that includes the channel and the parts of the source/drain adjoining the channel are extremely thin, typically less than 12 nm or so. The parts of the source and drain that adjoin the channel may have a very small volume. When such transistors have conventional doped source/drain regions, the number of dopant atoms in the source and drain regions proximate to the channel may be of the order of ten or fewer and these dopants will have random placement. As such, the doping in any given transistor is stochastic rather than deterministic and this can lead to excessive variability in the electrical performance of a population of transistors that form an integrated circuit.
To explain this problem in more detail, even at high doping levels in excess of 1020 dopant/cm3, the dopants are sparse, at most comprising only 2% of the atoms present in the source/drain regions and more typically less than 1%. It has been recognized that when the volume of source/drain regions is small, the statistical variation of the number and location of the dopant atoms introduces a very large variance in the electrical responses of the transistors. See, e.g., Martinez et al., “Quantum-Transport Study on the Impact of Channel Length and Cross Sections on Variability Induced by Random Discrete Dopants in Narrow Gate-All-Around Silicon Nanowire Transistors,” IEEE Trans. Electron Devices, Vol. 58, No. 8, p. 2209 (2011). In this article, the authors point out that a transistor with an unfortunate configuration of dopant atoms in source/drain can have both an undesirably high “off” current (under zero gate bias) and an undesirably low “on” current (under high gate bias) relative to a transistor with a more favorable configuration of dopant atoms. In designing an integrated circuit, often comprising several billion transistors, it is the “weak” transistors that determine the performance of the whole circuit. That is, to obtain high yields of manufactured ICs it is necessary to design the circuit assuming transistors are the inferior or weak type. Stated differently, the performance of a circuit is determined by the weakest of the transistors rather than the strongest. In modern statistical design of circuits, the dependence is more nuanced but it is generally true that given a statistical distribution of device characteristics across a large population of transistors, the performance of a circuit is determined more by the low performance of the weaker transistors than the high performance of the stronger transistors. What is preferred is to have a population of transistors with the variance in their electrical performance as small as possible.
Quite apart from the severe contact resistance problem associated with nanoscale metal-semiconductor contacts, the statistical variance of source/drain doping thus presents another major challenge to further scaling of MOS transistors into the 7 nm node and beyond. Metal source/drain transistors provide a solution to the dopant variability problem in conventional doped source/drain technologies. Dopants can be eliminated if the source/drain regions are formed of a metal that adjoins the undoped channel region and provides carriers to the channel directly without any need for doped semiconductor. Such metal source/drain regions most desirably have a small Schottky barrier height in order for their performance to be competitive with doped source/drain counterparts.
U.S. Pat. Nos. 6,833,556, 7,084,423, 7,112,478, 7,883,980, and 9,362,376, all assigned to the assignee of the present invention and each incorporated herein by reference, describe methods and structures that enable high performance metal source/drain field effect transistors. Briefly, an electrical junction includes an interface layer disposed between a contact metal and a semiconductor, and may comprise a passivation layer (which in some instances may be a monolayer) adjacent the semiconductor and, optionally, a separation layer disposed between the passivation layer and the metal. Various metals and semiconductors may be used, and the passivation layer may be an oxide of the semiconductor or other material. The separation layer, if present, may be a metal oxide. The very thin, interfacial dielectric layer between the metal and semiconductor acts to reduce the Schottky barrier at the junction from that which would exist in the absence of the interface layer, and at the same time has sufficient conductivity, despite being itself a dielectric with poor bulk electrical conduction, to provide a net improvement in the conductivity of the MIS junction. These devices overcome the statistical dopant variability problem by eliminating source/drain doping completely. However, these devices do have a remaining limitation in that the area of the metal-semiconductor interface, where a metal source or drain adjoins the semiconductor channel, is exceedingly small, being broadly comparable to the cross sectional area of the channel. U.S. Pat. No. 8,212,336 provides a solution that offers some relief to the area limitation by providing an interface that has an area exceeding the cross-sectional area of the channel.
It is known to induce “virtual” p-type and n-type regions using MOS capacitors. Such MOS capacitors are not conductive and do not provide a current to the semiconductor. The MOS capacitors induce variously (and optionally) p-type or n-type semiconductor regions. Electrical current into or out of these regions is provided by other (additional) electrical contacts. See e.g., André Heinzig et al., “Reconfigurable Silicon Nanowire Transistors”, Nano Letters, Vol. 12, pp. 119-124 (2012).
Regarding the work-functions of the overlap metals, the '234 patent states: “In one embodiment of the present invention, the conductor used to overlap the extension region is a metal possessing a low workfunction Φx in an n-channel FET. This effective workfunction is considered low when it is less than the electron affinity Xc of the semiconductor. It is generally advantageous to have Φx as low as possible. The lower the workfunction, the greater the amount of charge (in this case electrons) induced in the extension, which generally reduces the resistance of the extension region, generally advantageously increasing the drive-current capability of the transistor. In another embodiment of the present invention, the workfunction Φx of the metal is high in a p-channel FET, where Φx is greater than the hole affinity of the semiconductor (i.e., more than a bandgap greater than the semiconductor's electron affinity). The overlapping metal in this case induces holes in the extension region. It is generally advantageous to have a metal with as high a workfunction as possible. The workfunction of the metal lies outside of the semiconductor bandgap.”
Connelly et al., “Improved Short-Channel n-FET Performance with Virtual Extensions,” Abstracts of the 5th International Workshop on Junction Technology (2005) reports: “An alternative to purely doped S/D extensions is to form a charge layer electrostatically, of thickness comparable to the channel thickness of just a few nanometers. One approach, separately biased spacers, results in additional wiring complexity and capacitance. A better approach to electrostatically induced “virtual extensions” is . . . to overlay a metal of appropriate work function above the extension regions to induce such a mobile charge layer, a “virtual extension” . . . this creates a zero-bias MOS capacitor in the extension regions, where, for an n-FET, a negative VT results in a permanently induced charge layer that provides an ultra-shallow tip to conventional S/D doping profiles.” “[T]his “virtual extension” tip can reduce the electrostatic coupling between a S/D and the channel . . . The metal in the thin “overlap metal” had a work function of 3V (n-FET), comparable to Er or Yb. The virtual extension thus provides an ultra-thin sheet of charge.” In this paper, the exemplary virtual extension structure was modeled with an “extension oxide thickness” of 0.7 nm, an identical “gate oxide thickness” of 0.7 nm and an “overlap metal effective work-function” equal to 3 V. It is implied therefore that there is no current flow between the overlap metal and the semiconductor just as there is no current flow between the gate metal and the semiconductor.
U.S. Pat. Nos. 8,586,966 and 9,123,790 describe making contacts to FinFETs and nanowire source/drains. U.S. Pat. No. 8,586,966 states: “a nanowire field effect transistor (FET) device includes a channel region including a silicon nanowire portion having a first distal end extending from the channel region and a second distal end extending from the channel region, the silicon portion is partially surrounded by a gate stack disposed circumferentially around the silicon portion, a source region including the first distal end of the silicon nanowire portion, a drain region including the second distal end of the silicon nanowire portion, a metallic layer disposed on the source region and the drain region, a first conductive member contacting the metallic layer of the source region, and a second conductive member contacting the metallic layer of the drain region.” Doped source/drain regions are used: “The source and drain diffusion regions may include either N type (for NMOS) or P type (for PMOS) doped with, for example, As or P (N type) or B (P type) at a concentration level typically 1e19 atoms/cm3 or greater.”
Similarly, U.S. Pat. No. 9,123,790 reports on “forming a contact coupled with the channel layer, the contact being configured to surround, in at least one planar dimension, material of the channel layer and to provide a source terminal or drain terminal for the transistor.” “In some embodiments, forming the contact further includes epitaxially depositing an epitaxial film on the channel layer prior to depositing the metal to form the contact, the epitaxial film being configured to surround, in the at least one planar dimension, the material of the channel layer and being disposed between the material of the channel layer and material of the contact.” In the specification, various doping methods are described: “The source and drain regions may be formed using either an implantation/diffusion process or an etching/deposition process. In the former process, dopants such as boron, aluminum, antimony, phosphorous, or arsenic may be ion-implanted into the substrate to form the source and drain regions. The ion implantation process is typically followed by an annealing process that activates the dopants and causes them to diffuse. In the latter process, materials of the stack of layers may first be etched to form recesses at the locations of the source and drain regions. An epitaxial deposition process may then be carried out to fill the recesses with a silicon alloy such as silicon germanium or silicon carbide, thereby forming the source and drain regions. In some implementations the epitaxially deposited silicon alloy may be doped in situ with dopants such as boron, arsenic, or phosphorous. In further implementations, alternate materials may be deposited into the recesses to form the source and drain regions, such as germanium or a group III-V material or alloy.”
Fischer, S. et al., “Dopant-free complementary metal oxide silicon field effect transistors,” Phys. Status Solidi A 213, No. 6, pp. 1494-1499 (2016), report on dopant-free CMOS devices utilizing ultrathin silicon nitrides and metals with appropriate work functions to provide n- and p-type semiconductor contacts. The reported silicon nitride layers are thicker than a monolayer (e.g., on the order of 7-27 Angstroms), and there is no mention of a negative Schottky barrier between the metal contact and the semiconductor.
In one embodiment of the invention, a nanowire transistor includes a nanowire disposed on a substrate, wherein a longitudinal length of the nanowire is made up of an undoped channel region of a first semiconductor material, an undoped semiconductor source region electrically coupled with a first end of the channel region, an undoped semiconductor drain region electrically coupled with a second end of the channel region, a gate stack including a gate insulator and a gate conductor coaxially wrapping completely around the channel region, a source stack including an interfacial layer and a source conductor that is electrically isolated from the gate conductor, coaxially wrapping completely around the semiconductor source region and extending along at least a portion of the semiconductor source region, and a drain stack comprising an interfacial layer and a drain conductor that is electrically isolated from the gate conductor, coaxially wrapping completely around the semiconductor drain region and extending along at least a portion of the semiconductor drain region. A Schottky barrier between the source conductor and the semiconductor source region is a negative Schottky barrier and a concentration of free charge carriers is induced in the semiconductor source region. A Schottky barrier between the drain conductor and the semiconductor drain region may, but need not necessarily, be a negative Schottky barrier such that a concentration of free charge carriers is induced in the semiconductor drain region. In some embodiments, the nanowire of the transistor is 20 nm or less thick. In some embodiments, the free charge carriers are electrons (in which case the interfacial layer of the source stack may comprise a monolayer of elements from group V or group VI), while in other embodiments the free charge carriers are holes (in which case the interfacial layer of the source stack may comprise a monolayer of elements from group III or group II).
In various instances of the nanowire transistor, the Schottky barrier between at least one of (a) the source conductor and the semiconductor source region, and (b) the drain conductor and the semiconductor drain region is between −0.1 eV and −0.5 eV. Further, in some instances, the interfacial layer of the source stack and interfacial layer of the drain stack each may include a material that would be an insulator or a semiconductor in its bulk state.
In some instances of the nanowire transistor, the semiconductor channel, the semiconductor source region, and the semiconductor drain region are all comprised of the same semiconductor material. In other instances, however, the semiconductor channel, the semiconductor source region, and the semiconductor drain region are not all comprised of the same semiconductor material. In general, the semiconductor source region may include silicon, germanium, silicon carbide, or an alloy comprising two or more of silicon, germanium, carbon and tin. The interfacial layer of the source stack and interfacial layer of the drain stack each may comprise a monolayer of elements from group V or group VI. A monolayer of group V or group VI atoms causes a negative Schottky barrier for electrons and consequentially a concentration of free electrons is induced in the semiconductor source and/or drain region. Alternatively, the interfacial layer of the source stack and interfacial layer of the drain stack each may comprise a monolayer of elements from group III. A monolayer of group III atoms causes a negative Schottky barrier for holes and consequentially a concentration of free holes is induced in the semiconductor source and/or drain region.
Other embodiments of the invention include a finFET transistor, having a semiconductor fin disposed on a substrate, wherein the fin has two major faces and a longitudinal length of the fin includes: an undoped channel region of a first semiconductor material, an undoped semiconductor source region electrically coupled with a first end of the channel region, an undoped semiconductor drain region electrically coupled with a second end of the channel region, a gate stack comprising a gate insulator and a gate conductor wrapping around at least two sides of the channel region, a source stack comprising an interfacial layer and a source conductor wrapping around at least two sides of the semiconductor source region and extending along at least a portion of the semiconductor source region, and a drain stack comprising an interfacial layer and a drain conductor wrapping around at least two sides of the semiconductor drain region and extending along at least a portion of the semiconductor drain region. A Schottky barrier between the source conductor and the semiconductor source region is a negative Schottky barrier causing a concentration of free carriers to be induced in the semiconductor source region. A Schottky barrier between the drain conductor and the semiconductor drain region may, but need not necessarily, be a negative Schottky barrier such that a concentration of free carriers is induced in the semiconductor drain region. In some embodiments, the fin has a thickness as measured between the two major faces of 12 nm or less.
Still other embodiments of the invention provide a nanowire transistor, which includes: a gate circumferentially surrounding and displaced from a semiconductor nanowire channel by an electrically insulating gate oxide, the semiconductor nanowire channel having no intentional doping; a source at a first end of the nanowire channel, and a drain at a second end of the nanowire channel, the source and drain each comprising undoped semiconductor material; and a first metal contact circumferentially surrounding the source and providing an electrically conductive path to the source, and a second metal contact circumferentially surrounding the drain and providing an electrically conductive path to the drain. The first metal contact electrostatically induces free charge carriers in the source and, in some instances, the second metal contact may, but need not necessarily, electrostatically induce free charge carriers in the drain. The first metal contact is separated from the gate by an insulating material layer or a gap, and the second metal contact is separated from the gate by an insulating material layer or a gap. In some instances of this nanowire transistor the free charge carriers may be electrons, while in other instances the free charge carriers may be holes.
In some embodiments of the nanowire transistor, a Schottky barrier between the first metal contact and the source may have a negative Schottky barrier height. For example, the Schottky barrier between the first metal contact and the source may be between −0.1 eV and −eV.
In some embodiments of the nanowire transistor, the first metal contact is displaced from the source by a first interface layer, and the second metal contact is displaced from the drain by a second interface layer, the first and second interface layers each comprising a material that would be an insulator or a semiconductor in its bulk state. Also, in some embodiments a first interface layer at an interface between the first metal contact and the source and, optionally, a second interface layer at an interface between the second metal contact and the drain each may comprise a monolayer of elements from group V or group VI.
In some embodiments of the nanowire transistor, the semiconductor nanowire channel, the source, and the drain are may be comprised of the same semiconductor material. The semiconductor material may be silicon, germanium, silicon carbide, a compound semiconductor, a fullerene, or an alloy comprising two or more of silicon, germanium, carbon and tin. In other embodiments, the semiconductor nanowire channel, the source, and the drain are not all comprised of the same semiconductor material.
In still other embodiments of the invention, a nanowire device includes an undoped channel region of a first semiconductor material; an undoped semiconductor source region electrically coupled with a first end of the channel region; an undoped semiconductor drain region electrically coupled with a second end of the channel region; a gate stack comprising a gate insulator and a gate conductor coaxially wrapping completely around the channel region; a source stack electrically isolated from the gate conductor, coaxially wrapping completely around the semiconductor source region and extending along at least a portion of the semiconductor source region; and a drain stack electrically isolated from the gate conductor, coaxially wrapping completely around the semiconductor drain region and extending along at least a portion of the semiconductor drain region; wherein the source stack comprises a source conductor contacting an interfacial layer disposed over the semiconductor source region, the interfacial layer including at least one epitaxial bilayer of group III and group V atomic monolayers.
In such a nanowire device, the source conductor may be a degenerately doped n-type semiconductor, wherein the monolayer of group V atoms is adjacent to and in contact with the source region, the source region comprises a group IV semiconductor source region, and the monolayer of group III atoms is adjacent to and in contact with the degenerately n-type doped semiconductor. The group IV semiconductor and the degenerately doped n-type semiconductor may be different semiconductor materials, or may be the same semiconductor material. For example, where the interfacial layer includes a monolayer of gallium (Ga) atoms and a monolayer of arsenic (As) atoms, the group IV semiconductor and the degenerately doped n-type semiconductor may each comprise germanium (Ge).
In different embodiments of the nanowire device, the source conductor may be a degenerately doped p-type semiconductor, wherein the monolayer of group V atoms is adjacent to and in contact with the degenerately doped p-type semiconductor, the source region comprises a group IV semiconductor source region, and the monolayer of group III atoms is adjacent to and in contact with the group IV semiconductor. In such instances, the group IV semiconductor and the degenerately doped p-type semiconductor may be the same semiconductor material or different semiconductor materials. For example, where the interfacial layer includes a monolayer of gallium (Ga) atoms and a monolayer of arsenic (As) atoms, the group IV semiconductor and the degenerately doped p-type semiconductor may each comprise germanium (Ge).
In still another embodiment of the invention, a nanowire device includes a nanowire disposed on a substrate, wherein a longitudinal length of the nanowire comprises: an undoped channel region of a first semiconductor material; an undoped semiconductor source region electrically coupled with a first end of the channel region; an undoped semiconductor drain region electrically coupled with a second end of the channel region; a gate stack comprising a gate insulator and a gate conductor coaxially wrapping completely around the channel region; a source stack electrically isolated from the gate conductor, coaxially wrapping completely around the semiconductor source region and extending along at least a portion of the semiconductor source region; and a drain stack electrically isolated from the gate conductor, coaxially wrapping completely around the semiconductor drain region and extending along at least a portion of the semiconductor drain region. The source stack includes a source conductor contacting the semiconductor source region and extending along at least a portion of the semiconductor source region, the source conductor comprising a degenerately p-type doped semiconductor and there is an offset in valence band energy between the degenerately doped p-type semiconductor and the semiconductor source region such that a valence band maximum in the degenerately doped p-type semiconductor is at a lower energy than a valence band maximum in the semiconductor source region. In such a nanowire device, the semiconductor source region may include germanium and the degenerately doped p-type semiconductor may include degenerately doped p-type silicon germanium.
The present invention is illustrated by way of example, and not limitation, in the figures of the accompanying drawings, in which:
The present inventors recognized a desire for a metal-semiconductor contact in a transistor that is as large as possible, unconstrained by the cross-sectional area of the channel, and such a contact is provided by the present invention by decoupling the metal-semiconductor contact interface from the cross-section of the semiconductor channel. The invention provides a solution that addresses some of the many challenges to the scaling of MOS transistors into the nanoscale, namely excessive random variability in source/drain doping and increasing metal-semiconductor contact resistance, by (i) eliminating source/drain dopants, and (ii) increasing the area of the source/drain metal-semiconductor contacts.
The present invention includes a metal-semiconductor junction that provides induced charge in the semiconductor region and also a direct path for the flow of electrical current into the semiconductor region. We call the induced charge region an “induced source/drain”. For an induced source/drain to be as effective as a doped source/drain, it should have a comparable concentration of free carriers, on the order of 1020 per cm3. A negative Schottky barrier height between the source/drain metal and the semiconductor is required to achieve such a carrier concentration. In various embodiments there is a deliberate interfacial layer between the metal and semiconductor at the metal-semiconductor junction, the interfacial layer serving the purpose of ensuring a required negative Schottky barrier height. The interfacial layer in certain embodiments is comprised of a monolayer of atoms. The atoms may be any of N, As, P, O, S, Se, or Te for n-channel transistors or B, Ga, Al, Zn, Cd or O for p-channel transistors. The interfacial layer in certain other embodiments is a “thin insulator” comprising a material that would be an insulator in its bulk state but which is conductive when very thin (in the thickness range 0.2 nm to 2 nm). In these embodiments, the interfacial layer at the metal-semiconductor junction is not comparable or equivalent to the thin insulator that separates the gate from the semiconductor channel. Specifically, the interfacial layer is selected to be highly conductive between the metal and the induced source/drain whereas the gate insulator being selected to be non-conductive between the gate and the channel.
The present invention overcomes many limitations of conventional contacts between metals and doped semiconductor bodies by inducing free carriers (electrons or holes) in the surface of a semiconductor body through electrostatic inducement rather than by doping the semiconductor with impurity atoms. Free carriers are induced in a semiconductor source/drain, close to an interface with an adjacent contacting metal by ensuring a negative Schottky barrier between the metal and the semiconductor.
It may be possible, in some embodiments to avoid the use of an I-layer if, for example, a contact metal or metals comprising the source/drain contact(s) and the semiconductor material comprising the semiconductor source/drain are paired such that the metal-semiconductor junction has a negative Schottky barrier. In other embodiments, the I-layer may be a multi-layer structure, including at least a passivation layer and a separation (or spacer) layer, where the passivation layer is adjacent the semiconductor material comprising the semiconductor source/drain.
In detail, for an “n-type” contact, a negative Schottky barrier means that the Fermi level of the metal is at a higher energy than the conduction band edge of the semiconductor, as illustrated in
For a “p-type” contact, a negative Schottky barrier means that the Fermi level of the metal is at a lower energy than the valence band edge of the semiconductor as depicted in
Other interfacial layers that can cause a negative Schottky barrier for holes on group IV semiconductors include dipole-inducing monolayers of group III elements such as boron, gallium or aluminum or bilayers of group V and group III elements where the group V atoms are on the metal side of the contact and group III atoms on the semiconductor side. Some preferred embodiments of p-type contacts have a high work function contact metal such as nickel, cobalt, iridium, rhodium, ruthenium, gold, osmium, palladium or platinum, or a high work function conductive metal oxide such as MoOx, WOx, CrOx (with x approximately equal to 3) or high work-function V2Ox (with x approximately equal to 5) with an interfacial layer between the high work function metal (or metal oxide) and the semiconductor.
The invention also has great utility where a low resistance metal-contacted source or drain is required in a semiconductor material that might not be doped conveniently or for which doping might not be possible at all. Examples of such semiconductors may be two-dimensional semiconductors such as graphene, germanene, phosphorene, stannene and silicene or two-dimensional layered transition metal dichalcogenide (TMDC) semiconductors such as MoS and WSe which have recently been recognized as strong candidates as future transistor channel materials. Other materials that may not be amenable to conventional doping include organic semiconductors, polymer semiconductors, fullerenes such as carbon nanotubes, amorphous semiconductors, perovskites. All semiconductors in the form of a nanoscale thin film or nanowire benefit from this invention by virtue of the fact that free carriers are induced in the semiconductor material by an adjacent metal so long as there is a negative Schottky barrier between the metal and the semiconductor.
If the semiconductor is a thin film or two-dimensional semiconductor, typically having a thickness of 12 nm or less, with two primary faces, it is preferred to have metal contacts on both faces with both metal contacts having a negative Schottky barrier to the semiconductor. Note, however, that it is a characteristic of the present invention that the Schottky barrier between the metal source contact and the semiconductor source is a negative Schottky barrier, but that it is optional for the Schottky barrier between the metal drain contact and the semiconductor drain to be a negative Schottky barrier.
If the semiconductor is a “one-dimensional” semiconductor, either cylindrical, such as a carbon nanotube, or a semiconductor “nanowire” where the wire may have circular, square or any other cross-sectional shape with a wire width of approximately 20 nm or less, it is preferred to have the metal contact wrap around the outer surface of the nanowire, again with the metal contact having a negative Schottky barrier to the semiconductor.
In a transistor with a fully depleted channel (may be a FinFET, FDSOI FET or nanowire FET), a thin body of semiconductor includes a channel region and source and drain regions where the source/drain regions are undoped and the source/drain metal contacts surround the semiconductor on two sides (FDSOI-FET or FinFET) or completely (nanowire FET).
In other embodiments, the free carriers induced by the contact metal with negative Schottky barrier may be additional to free carriers introduced into the semiconductor by doping. As such it is not essential that the semiconductor be undoped for the invention to provide an advantage in a semiconductor device. For example, the source and drain regions of a nanowire or FinFET transistor maybe doped conventionally (e.g., by diffusion of atoms from an external solid source or ion implantation and thermal activation) and the randomness of the dopants tolerated, the induced carriers provided by this invention being additional to the carriers provided by doping and therefore reducing the variability while not eliminating it.
In one embodiment, a nanowire transistor comprises a nanowire disposed on a substrate. The nanowire further comprises, along a longitudinal length, an undoped channel region of a first semiconductor material, an undoped semiconductor source region electrically coupled with a first end of the channel region and an undoped semiconductor drain region electrically coupled with a second end of the channel region. A gate stack comprising a gate insulator and a gate conductor is wrapped coaxially completely around the channel region and controls electrical conduction through the semiconductor (channel) between the source and drain regions.
A source contact stack comprises an interfacial layer and a source conductor coaxially wrapping completely around the semiconductor source region and extending along at least a portion of the semiconductor source region. A drain stack comprises an interfacial layer and a drain conductor coaxially wrapping completely around the semiconductor drain region and extending along at least a portion of the semiconductor drain region. The Schottky barrier between the source conductor and the semiconductor source region is a negative Schottky barrier and a concentration of free carriers is induced in the semiconductor source region. The Schottky barrier between the drain conductor and the semiconductor drain region may, but need not necessarily, be a negative Schottky barrier, but if it is, a concentration of free carriers is induced in the semiconductor drain region. The thickness of the nanowire in the nanowire transistor is 20 nm or less. The first undoped semiconductor in the channel region and the source region may be comprised of the same semiconductor material. Alternatively, the source region may be comprised of a semiconductor material that is different from the first semiconductor material in the channel region.
In another embodiment, a finFET transistor comprises a semiconductor fin disposed on a substrate, wherein the fin has two major faces.
The source region is electrically contacted through a source contact stack comprising an interfacial layer and a source conductor 510 wrapping around at least two sides of the semiconductor source region and extending along at least a portion of the semiconductor source region. The drain region is electrically contacted through a drain contact stack comprising an interfacial layer and a drain conductor 512 wrapping around at least two sides of the semiconductor drain region and extending along at least a portion of the semiconductor drain region. The Schottky barrier between the source conductor and the semiconductor source region is a negative Schottky barrier causing a concentration of free carriers to be induced in the semiconductor source region 508. The Schottky barrier between the drain conductor and the semiconductor drain region may, but need not necessarily, be a negative Schottky barrier, but if it is, a concentration of free carriers is induced in the semiconductor drain region 509.
The source conductor 510 and the drain conductor 512 are electrically isolated from the gate 504 by insulating gate sidewall spacers 506. The thickness of the fin of the FinFET transistor as measured between the two vertical faces is 12 nm or less The first undoped semiconductor in the channel region and the source region may be comprised of the same semiconductor material. Alternatively, the source region may be comprised of a semiconductor material that is different from the first semiconductor material in the channel region. The first undoped semiconductor in the channel region and the drain region may be comprised of the same semiconductor material. Alternatively, the drain region may be comprised of a semiconductor material that is different from the first semiconductor material in the channel region.
In other embodiments a source region is electrically contacted through a source contact stack comprising an interfacial layer and a source conductor wrapping around at least two sides of the semiconductor source region and extending along at least a portion of the semiconductor source region and having a negative Schottky barrier for electrons and a drain region is electrically contacted through a drain contact stack comprising an interfacial layer and a drain conductor wrapping around at least two sides of the semiconductor drain region and extending along at least a portion of the semiconductor drain region and, optionally, having a negative Schottky barrier for holes such that the carriers induced in the source region are of opposite type to the carriers induced in the drain region. Such a configuration of “n-type” source and “p-type” drain may be useful for example for providing a gated diode electrical function.
In other embodiments, a source region is electrically contacted through a source contact stack comprising an interfacial layer and a source conductor wrapping around at least two sides of the semiconductor source region and extending along at least a portion of the semiconductor source region and having a negative Schottky barrier for holes and a drain region is electrically contacted through a drain contact stack comprising an interfacial layer and a drain conductor wrapping around at least two sides of the semiconductor drain region and extending along at least a portion of the semiconductor drain region and, optionally, having a negative Schottky barrier for electrons with the result that the carriers induced in the source region are of opposite polarity to the carriers induced in the drain region. Such a configuration of “p-type” source and “n-type” drain may be useful for example for providing a gated diode electrical function.
In still further embodiments, a source contact stack comprises a source conductor contacting a semiconductor source region and extending along at least a portion of the semiconductor source region wherein the source conductor is a degenerately n-type doped semiconductor and there is an offset in conduction band energy between the degenerately doped semiconductor and the semiconductor source region such that the conduction band minimum in the degenerately doped n-type semiconductor is at a higher energy than the conduction band minimum in the semiconductor source region. As a consequence of the conduction band offset, electrons from the degenerately doped n-type semiconductor populate the semiconductor source region.
Still another embodiment of the invention provides an interfacial layer that causes the offset in the conduction band energies referred to above. A preferred interfacial layer that causes the desired offset in the conduction band energies comprises at least one epitaxial bilayer of group III and group V atomic monolayers, such interfacial layer causing an electronic dipole that induces the conduction band offset (as detailed in the present applicant's U.S. Pat. No. 9,362,376, incorporated herein by reference).
One example of a source contact stack in which a source conductor is a degenerately n-type doped semiconductor is a source contact stack in which the source material is silicon and the degenerately doped n-type semiconductor that contacts the source is degenerately doped n-type gallium phosphide (GaP). The desired conduction band offset arises naturally between GaP and silicon with the conduction band edge in the GaP being at a higher energy than the conduction band edge in the silicon. The use of a source contact stack in which a source conductor is a degenerately n-type doped semiconductor in accordance with embodiments of the present invention is not limited to n-type GaP contacting silicon but should be understood to include a coupling of any two semiconductors that have a naturally occurring conduction band offset at their heterointerface. These include, in addition to GaP, the following: Ge; AlAs; AlSb, ZnS; ZnSe and ZnTe.
Alternatively, in another embodiment, an interfacial layer between a group IV semiconductor source region and a degenerately n-type doped semiconductor contact region causes an additional offset in the conduction band energies. A preferred interfacial layer that causes the desired offset in the conduction band energies comprises at least one epitaxial bilayer of group III and group V atomic monolayers, with the monolayer of group V atoms being adjacent to and in contact with the group IV semiconductor source region and the monolayer of group III atoms being adjacent to and in contact with the degenerately n-type doped semiconductor contact region. In this embodiment, the group IV semiconductor source region and the degenerately doped semiconductor contact region may be formed of different semiconductor materials or of the same semiconductor material. In one example, a junction between two regions of germanium (Ge) are separated by an interface layer comprising a monolayer of gallium (Ga) and a monolayer of arsenic (As) atoms, the Ge conduction band edge on the As side of the junction is at a lower energy (of the order of 0.35-0.45 eV) than the Ge conduction band edge on the Ga side of the junction. Such an interfacial layer causes an electronic dipole that induces the valence band offset is described in U.S. Pat. No. 9,362,376 and in a 1991 article by McKinley et al. entitled “Control of Ge homojunction band offsets via ultrathin Ga—As dipole layers,” J. Vac. Sci. Technol. A 9 (3), May/Jun 1991, and in a similar article by McKinley et al. in 1992, entitled “Control of Ge homojunction band offsets via ultrathin Ga—As dipole layers,” Applied Surface Science Vol. 56 -58, pp. 762-765 (1992).
Equivalent embodiments for instances in which holes are the charge carries are also embodiments of the present invention. For example, a source contact stack may comprise a source conductor contacting a semiconductor source region and extending along at least a portion of the semiconductor source region, wherein the source conductor is a degenerately p-type doped semiconductor and there is an offset in valence band energy between the degenerately doped p-type semiconductor and the semiconductor source region such that the valence band maximum in the degenerately doped semiconductor is at a lower energy than the valence band maximum in the semiconductor source region. As a consequence of the conduction band offset, holes from the degenerately doped p-type semiconductor populate the semiconductor source region.
One example is a source contact stack in which the source material is germanium and the degenerately doped p-type semiconductor that contacts the source is degenerately doped p-type silicon germanium (SiGe) alloy. The desired valence band offset arises naturally between germanium and SiGe with the valence band edge in the SiGe being at a lower energy than the valence band edge in the germanium. Embodiments of the invention are not limited to p-type SiGe contacting germanium but should be understood to include a coupling of any two semiconductors that have a naturally occurring valence band offset at their heterointerface.
Alternatively, in another embodiment, an interfacial layer between a group IV semiconductor source region and a degenerately doped semiconductor contact region causes an additional offset in the valence band energies. A preferred interfacial layer that causes the desired offset in the valence band energies comprises at least one epitaxial bilayer of group III and group V atomic monolayers, with the monolayer of group III atoms being adjacent to and in contact with the group IV semiconductor source region and the monolayer of group V atoms being adjacent to and in contact with the degenerately p-type doped semiconductor contact region. In this embodiment, the group IV semiconductor source region and the degenerately doped semiconductor contact region may be formed of different semiconductor materials of the same semiconductor material. In an exemplary junction between two regions of germanium (Ge) separated by an interface layer comprising a monolayer of gallium (Ga) and a monolayer of arsenic (As) atoms, the Ge valence band edge on the As side of the junction is at a lower energy (of the order of 0.35-0.45 eV) than the Ge valence band edge on the Ga side of the junction. Such an interfacial layer causes an electronic dipole that induces the valence band offset is described in U.S. Pat. No. 9,362,376, and in the McKinley articles cited above.
In the above description, the nanowire has been described (at least in the accompanying illustrations) as having a circular or approximately circular cross-section. However, the invention is not limited to such geometries and nanowires of the present invention may have other shapes, such as square, rectangular, oval, or other cross-sections. Such geometries may be recognized as “nanosheets” and as used herein the term nanowire should be read as including nanosheets. Stated differently, the cross-sectional shape of the nanowire is not critical to the present invention.
Similarly, the foregoing description refers to a gate stack wrapping completely around a channel region; a source stack coaxially wrapping completely around a semiconductor source region; and a drain stack coaxially wrapping completely around a semiconductor drain region. Such a geometry is true for a three-dimensional nanowire—that is, a nanowire having a three-dimensional cross-section. However, the present invention is also applicable to nanowires comprised of two-dimensional semiconductors, such as graphene, hexagonal boron nitride, or transition metal dichalocogenides (e.g., MoS2, MoSe2, MOTe2, WS2, WSe2, WTe2, etc.). In such instances, “wrapping completely around” should be understood as including instances where contact is made to both (opposing) sides of the two-dimensional semiconductor. In some cases, contacts may be made to only a single face of a two-dimensional semiconductor, but such instances would not be considered as a stack “wrapping completely around” a region.
In various embodiments, the present invention further provides wraparound MIS contacts to FinFETs and/or stacked nanowire FETs. One benefit provided by the present wrap around MIS contacts over silicide contacts is that it avoids the need to deposit sacrificial silicon (or other semiconductor material), e.g., in the case of forming contacts for FinFETs, for silicide contacts. In the case of nanowires, the use of wraparound MIS contacts allows for the use of a surrounding metal contact rather than an epitaxial silicon contact, which improves current load balancing across a stack of nanowires.
Thus, semiconductor devices having a region of undoped semiconductor material that is made effectively n-type or p-type by inducing a quantity of electrons or holes (respectively) on the semiconductor side of a metal-semiconductor junction by virtue of a negative Schottky barrier between the metal and the semiconductor have been described.
This is a CONTINUATION of U.S. patent application Ser. No. 17/091,959, filed Nov. 6, 2020, which is a CONTINUATION of U.S. patent application Ser. No. 16/693,143, filed Nov. 22, 2019, now U.S. Pat. No. 10,833,199, which is a CONTINUATION of U.S. patent application Ser. No. 16/202,507, filed Nov. 28, 2018, now U.S. Pat. No. 10,505,047, which is a DIVISIONAL of U.S. patent application Ser. No. 15/816,231, filed Nov. 17, 2017, now U.S. Pat. No. 10,170,627, which is a NONPROVISIONAL of: (a) U.S. Provisional Application No. 62/424,176, filed Nov. 18, 2016, and (b) U.S. Provisional Application No. 62/456,437, filed Feb. 8, 2017, each of which is incorporated by reference in its respective entirety.
| Number | Date | Country | |
|---|---|---|---|
| 62424176 | Nov 2016 | US | |
| 62456437 | Feb 2017 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 15816231 | Nov 2017 | US |
| Child | 16202507 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 17091959 | Nov 2020 | US |
| Child | 17931052 | US | |
| Parent | 16693143 | Nov 2019 | US |
| Child | 17091959 | US | |
| Parent | 16202507 | Nov 2018 | US |
| Child | 16693143 | US |
