US20070020859A1 - Method of making non-volatile field effect devices and arrays of same - Google Patents

Abstract

Methods of making non-volatile field effect devices and arrays of same. Under one embodiment, a method of making a non-volatile field effect device includes providing a substrate with a field effect device formed therein. The field effect device includes a source, drain and gate with a field-modulatable channel between the source and drain. An electromechanically-deflectable, nanotube switching element is formed over the field effect device. Terminals and corresponding interconnect are provided to correspond to each of the source, drain and gate such that the nanotube switching element is electrically positioned between one of the source, drain and gate and its corresponding terminal, and such that the others of said source, drain and gate are directly connected to their corresponding terminals.

Description

CROSS REFERENCE TO RELATED APPLICATIONS
• This application is a continuation of and claims priority under 35 U.S.C. §120 to U.S. Patent Application Ser. No. 10/864,751, filed Jun. 9, 2004 and entitled Method of Making Non-Volatile Field Effect Devices and Arrays of Same, the entire contents of which are incorporated herein by reference, which claims priority under 35 U.S.C. §119(e) to U.S. Provisional Patent Application No. 60/476976, filed on Jun. 9, 2003, entitled Non-Volatile Electromechanical Field Effect Transistors and Methods of Forming Same, which is incorporated herein by reference in its entirety, and which is related to the following U.S. applications, the contents of which are incorporated herein in their entirety by reference:
• U.S. patent application Ser. No. 10/810,962, filed Mar. 26, 2004, entitled NRAM Bit Selectable Two-Device Nanotube Array;
• U.S. patent application Ser. No. 10/810,963, filed Mar. 26, 2004, entitled NRAM Byte/Block Released Bit Selectable One-Device Nanotube Array;
• U.S. patent application Ser. No. 10/811,191, filed Mar. 26, 2004, entitled Single Transistor With Integrated Nanotube (NT-FET); and
• U.S. patent application Ser. No. 10/811,373, filed Mar. 26, 2004, entitled Nanotube-On-Gate FET Structures and Applications.
BACKGROUND
• 1. Technical Field
• The present invention relates to field effect devices having non-volatile behavior as a result of control structures having nanotube components and to methods of forming such devices.
• 2. Discussion of Related Art
• Semiconductor MOSFET transistors are ubiquitous in modern electronics. These field effect devices possess the simultaneous qualities of bistability, high switching speed, low power dissipation, high-reliability, and scalability to very small dimensions. One feature not typical of such MOSFET-based circuits is the ability to retain a digital state (i.e. memory) in the absence of applied power; that is, the digital state is volatile.
• FIG. 1 depicts a prior artfield effect transistor10. Thetransistor10 includes agate node12, adrain node14, and asource node18. Typically, thegate node12 is used to control the device. Specifically, by applying an adequate voltage to thegate node12 an electric field is caused that creates a conductive path between thedrain14 andsource18. In this sense, the transistor is referred to as switching on.
• Currently, most memory storage devices utilize a wide variety of energy dissipating devices which employ the confinement of electric or magnetic fields within capacitors or inductors respectively. Examples of state of the art circuitry used in memory storage include FPGA, CPLD, ASIC, CMOS, ROM, PROM, EPROM, EEPROM, DRAM, MRAM and FRAM, as well as dissipationless trapped magnetic flux in a superconductor and actual mechanical switches, such as relays.
• An FPGA (Field Programmable Gate Array) or a CPLD (Complex Programmable Logic Device) is a programmable logic device (PLD), a programmable logic array (PLA), or a programmable array logic (PAL) with a high density of gates, containing up to hundreds of thousands of gates with a wide variety of possible architectures. The ability to modulate (i.e. effectively to open and close) electrical circuit connections on an IC (i.e. to program and reprogram) is at the heart of the FPGA (Field programmable gate array) concept.
• An ASIC (Application Specific Integrated Circuit) chip is custom designed (or semi-custom designed) for a specific application rather than a general-purpose chip such as a microprocessor. The use of ASICs can improve performance over general-purpose CPUs, because ASICs are “hardwired” to do a specific job and are not required to fetch and interpret stored instructions.
• Important characteristics for a memory cell in electronic device are low cost, nonvolatility, high density, low power, and high speed. Conventional memory solutions include Read Only Memory (ROM), Programmable Read only Memory (PROM), Electrically Programmable Memory (EPROM), Electrically Erasable Programmable Read Only Memory (EEPROM), Dynamic Random Access Memory (DRAM) and Static Random Access Memory (SRAM).
• ROM is relatively low cost but cannot be rewritten. PROM can be electrically programmed but with only a single write cycle. EPROM (Electrically-erasable programmable read-only memories) has read cycles that are fast relative to ROM and PROM read cycles, but has relatively long erase times and reliability only over a few iterative read/write cycles. EEPROM (or “Flash”) is inexpensive, and has low power consumption but has long write cycles (ms) and low relative speed in comparison to DRAM or SRAM. Flash also has a finite number of read/write cycles leading to low long-term reliability. ROM, PROM, EPROM and EEPROM are all non-volatile, meaning that if power to the memory is interrupted the memory will retain the information stored in the memory cells.
• DRAM (dynamic random access memory) stores charge on capacitors but must be electrically refreshed every few milliseconds complicating system design by requiring separate circuitry to “refresh” the memory contents before the capacitors discharge. SRAM does not need to be refreshed and is fast relative to DRAM, but has lower density and is more expensive relative to DRAM. Both SRAM and DRAM are volatile, meaning that if power to the memory is interrupted the memory will lose the information stored in the memory cells.
• Consequently, existing technologies are either non-volatile but are not randomly accessible and have low density, high cost, and limited ability to allow multiple writes with high reliability of the circuit's function, or they are volatile and complicate system design or have low density. Some emerging technologies have attempted to address these shortcomings.
• For example, magnetic RAM (MRAM) or ferromagnetic RAM (FRAM) utilizes the orientation of magnetization or a ferromagnetic region to generate a nonvolatile memory cell. MRAM utilizes a magnetoresistive memory element involving the anisotropic magnetoresistance or giant magnetoresistance of ferromagnetic materials yielding nonvolatility. Both of these types of memory cells have relatively high resistance and low-density. A different memory cell based upon magnetic tunnel junctions has also been examined but has not led to large-scale commercialized MRAM devices. FRAM uses circuit architecture similar to DRAM but which uses a thin film ferroelectric capacitor. This capacitor is purported to retain its electrical polarization after an externally applied electric field is removed yielding a nonvolatile memory. FRAM suffers from a large memory cell size, and it is difficult to manufacture as a large-scale integrated component. See U.S. Pat. Nos. 4,853,893; 4,888,630; 5,198,994, 6,048,740; and 6,044,008.
• Another technology having non-volatile memory is phase change memory. This technology stores information via a structural phase change in thin-film alloys incorporating elements such as selenium or tellurium. These alloys are purported to remain stable in both crystalline and amorphous states allowing the formation of a bi-stable switch. While the nonvolatility condition is met, this technology appears to suffer from slow operations, difficulty of manufacture and poor reliability and has not reached a state of commercialization. See U.S. Pat. Nos. 3,448,302; 4,845,533; and 4,876,667.
• Wire crossbar memory (MWCM) has also been proposed. See U.S. Pat. Nos. 6,128,214; 6,159,620; and 6,198,655. These memory proposals envision molecules as bi-stable switches. Two wires (either a metal or semiconducting type) have a layer of molecules or molecule compounds sandwiched in between. Chemical assembly and electrochemical oxidation or reduction are used to generate an “ON” or “OFF” state. This form of memory requires highly specialized wire junctions and may not retain non-volatilely owing to the inherent instability found in redox processes.
• Recently, memory devices have been proposed which use nanoscopic wires, such as single-walled carbon nanotubes, to form crossbar junctions to serve as memory cells. See WO 01/03208, Nanoscopic Wire-Based Devices, Arrays, and Methods of Their Manufacture; and Thomas Rueckes et al., “Carbon Nanotube-Based Nonvolatile Random Access Memory for Molecular Computing,” Science, vol. 289, pp. 94-97, 7 Jul. 2000. Electrical signals are written to one or both wires to cause them to physically attract or repel relative to one another. Each physical state (i.e., attracted or repelled wires) corresponds to an electrical state. Repelled wires are an open circuit junction. Attracted wires are a closed state forming a rectified junction. When electrical power is removed from the junction, the wires retain their physical (and thus electrical) state thereby forming a non-volatile memory cell.
• The use of an electromechanical bi-stable device for digital information storage has also been suggested (c.f. U.S. Pat. No. 4,979,149: Non-volatile memory device including a micro-mechanical storage element).
• The creation and operation of a bi-stable nano-electro-mechanical switches based on carbon nanotubes (including mono-layers constructed thereof) and metal electrodes has been detailed in a previous patent application of Nantero, Inc. (U.S. Pat. Nos. 6,574,130, 6,643,165, 6,706,402; U.S. patent application Ser. Nos. 09/915,093, 10/033,323, 10/033,032, 10/128,117, 10/341,005, 10/341,055, 10/341,054, 10/341,130, 10/776,059, and 10/776,572, the contents of which are hereby incorporated by reference in their entireties).
SUMMARY
• The invention provides methods of making non-volatile field effect devices and arrays of same.
• Under one aspect of the invention, a method of making a non-volatile field effect device includes providing a substrate with a field effect device formed therein. The field effect device includes a source, drain and gate with a field-modulatable channel between the source and drain. An electromechanically-deflectable, nanotube switching element is formed over the field effect device. Terminals and corresponding interconnect are provided to correspond to each of the source, drain and gate such that the nanotube switching element is electrically positioned between one of the source, drain and gate and its corresponding terminal, and such that the others of said source, drain and gate are directly connected to their corresponding terminals.
• Under another aspect of the invention, forming an electromechanically-deflectable, nanotube switching element includes forming an article from nanotube fabric.
• Under another aspect of the invention, the nanotube fabric is a porous nanotube fabric.
• Under another aspect of the invention, forming an electromechanically-deflectable, nanotube switching element includes forming a fourth terminal to control operation of said nanotube switching element.
• Under another aspect of the invention, the fourth terminal is covered with a dielectric material on a surface facing the article of nanotube fabric.
• Under another aspect of the invention, forming an electromechanically-deflectable, nanotube switching element includes forming an article of nanofabric that is positioned between the fourth terminal and the one of the source, drain and gate.
• Under another aspect of the invention, forming an electromechanically-deflectable, nanotube switching element includes forming an article of nanofabric that is suspended between the fourth terminal and the one of the source, drain and gate.
• Under another aspect of the invention, forming an electromechanically-deflectable, nanotube switching element includes forming an article of nanofabric that is positioned between the fourth terminal and the one of the source, drain and gate, and wherein there is a gap between the article and the fourth terminal.
• Under another aspect of the invention, forming an electromechanically-deflectable, nanotube switching element includes forming an article of nanofabric that is positioned between the fourth terminal and the one of the source, drain and gate, and wherein there is a gap between the one of the source, drain and gate.
• Under another aspect of the invention, forming an electromechanically-deflectable, nanotube switching element includes forming an article of nanofabric that is horizontally suspended relative to a horizontal substrate surface and wherein the article is electrically positioned between the fourth terminal and the one of the source, drain and gate.
• Under another aspect of the invention, the article is formed to have a suspended length that is sub-lithographic.
• Under another aspect of the invention, forming an electromechanically-deflectable, nanotube switching element includes forming an sacrificial layer and then forming a porous fabric of nanotubes thereover, said fabric subsequently being lithographically patterned and subsequent to that said sacrificial material being removed to suspend the article with a gap existing where the sacrificial material previously existed.
• Under another aspect of the invention, a second sacrificial layer is formed on a side of the patterned article opposite the side of the sacrificial layer, wherein both the sacrificial layer and the second sacrificial layer are removed to form gaps on either side of the article.
• Under another aspect of the invention, the sacrificial material is anisotropically etchable.
• Under another aspect of the invention, the sacrificial material is poly.
• Under another aspect of the invention, a method of making a non-volatile field effect device includes providing a substrate with a field effect device formed therein. The field effect device includes a source, drain and gate with a field-modulatable channel between the source and drain. Terminals and corresponding interconnect are provided to correspond to each of the source, drain and gate. The drain and gate directly connected to their corresponding terminals. An electromechanically-deflectable, nanotube switching element is formed over the field effect device. The nanotube switching element is connected such that it is electrically positioned in series between the source and the terminal corresponding to the source.
• Under another aspect of the invention, a method of making a non-volatile field effect device includes providing a substrate with a field effect device formed therein. The field effect device includes a source, drain and gate with a field-modulatable channel between the source and drain. Terminals and corresponding interconnect are provided to correspond to each of the source, drain and gate. The source and gate are directly connected to their corresponding terminals. An electromechanically-deflectable, nanotube switching element is formed over the field effect device. The nanotube switching element is connected such that it is electrically positioned in series between the drain and the terminal corresponding to the drain.
• Under another aspect of the invention, a method of making a non-volatile field effect device includes providing a substrate with a field effect device formed therein. The field effect device includes a source, drain and gate with a field-modulatable channel between the source and drain. Terminals and corresponding interconnect are provided to correspond to each of the source, drain and gate. The source and drain are directly connected to their corresponding terminals. An electromechanically-deflectable, nanotube switching element is formed over the field effect device. The nanotube switching element is connected such that it is electrically positioned in series between the gate and the terminal corresponding to the gate.
BRIEF DESCRIPTION OF THE DRAWINGS
• In the drawing,
• FIG. 1 is a schematic of a prior art field effect transistor;
• FIGS.2A-L illustrate schematics of three models of preferred embodiments of the invention;
• FIGS.3A-C illustrate the operation of field effect devices with controllable sources for two of the FED configurations;
• FIGS. 4-6 illustrate waveforms for exemplary operation of devices according to certain aspects of the invention;
• FIGS.7A-C illustrate the operation of field effect devices according to certain aspects of the invention;
• FIGS. 8 and 9 illustrate waveforms for exemplary operation of devices according to certain aspects of the invention;
• FIGS. 10A-12 illustrate the operational waveforms for field effect devices according to certain aspects of the invention;
• FIGS.13A-C illustrate schematic representations of preferred embodiments of the invention;
• FIG. 14 illustrates a cross section of one embodiment of the invention;
• FIG. 15 illustrates operational waveforms for field effect devices according to certain aspects of the invention;
• FIG. 16 illustrates electrical (I/V) characteristics of devices according to one aspect of the invention;
• FIGS.17A-D illustrate a schematic representation of devices according to one aspect of the invention along with depictions of memory states of such a device;
• FIG. 18 illustrates schematics of an NRAM system according to preferred embodiments of the invention;
• FIG. 19 illustrates operational waveforms for memory devices according to certain aspects of the invention;
• FIG. 20A illustrates a memory array flow chart according to one aspect of the invention;
• FIG. 20B illustrates a schematic of a switch amplifier/latch according to certain aspects of the invention;
• FIG. 21 illustrates waveforms for a memory system according to certain aspects of the invention;
• FIG. 22 is a flow chart of a method of manufacturing preferred embodiments of the invention;
• FIGS.23,23′ and23″ are flow charts illustrating acts performed in preferred methods of the invention;
• FIGS.24A-F illustrate exemplary structures according to aspects of the invention;
• FIGS.25A-GG illustrate exemplary intermediate structures according to certain aspects of the invention;
• FIG. 26 is a flow chart of a method of manufacturing preferred embodiments of the invention;
• FIGS.27,27′,28 and28′ are flow charts of method of manufacturing preferred embodiments of the invention;
• FIGS.29A-F illustrate intermediate structures according to certain aspects of the invention;
• FIGS.30A-P (exclusive of O) illustrate intermediate structures according to certain aspects of the invention;
• FIGS.31A-D illustrate intermediate structures according to certain aspects of the invention;
• FIGS.32A-B illustrate cross sections of an embodiment of the invention;
• FIG. 32C illustrates a plan view of an embodiment of the invention;
• FIGS.33A-C illustrate cross sections of an embodiment of the invention;
• FIG. 33D illustrates a plan view of an embodiment of the invention;
• FIGS.34A-D illustrate schematics of circuitry according to certain aspects of the invention;
• FIG. 35 illustrates schematics of memory arrays according to certain aspects of the invention;
• FIG. 36 illustrates operational waveforms of a memory array according to one aspect of the invention;
• FIG. 37A illustrates a diagram outlining a memory array system according to one aspect of the invention;
• FIG. 37B is a schematic of a cell according to once aspect of the invention;
• FIG. 38 illustrates operational waveforms of a memory array according to one aspect of the invention;
• FIG. 39A-D illustrate schematics of circuitry according to certain aspects of the invention;
• FIG. 40 illustrates a schematic of an NRAM system, according to one embodiment of the invention;
• FIG. 41 illustrates the operational waveforms of a memory array according to one aspect of the invention;
• FIG. 42A illustrates a diagram outlining a memory array system according to one aspect of the invention;
• FIG. 42B is a schematic of a cell according to once aspect of the invention;
• FIG. 43 illustrates the operational waveforms of a memory array according to one aspect of the invention;
• FIGS.44A-B illustrate cross sections of memory arrays according to aspects of the invention;
• FIG. 44C illustrates a plan view of a memory array structure according to one aspect of the invention;
• FIGS.45A-B illustrate cross sections of memory arrays according to aspects of the invention;
• FIG. 45C illustrates a plan view of a memory array structure according to one aspect of the invention;
• FIGS.46A-C illustrate cross sections of structures according to certain aspects of the invention;
• FIG. 46D illustrates a plan view of a memory array structure according to one aspect of the invention;
• FIGS.47A-C illustrate schematics of circuitry for a non-volatile field effect device according to aspects of the invention;
• FIG. 48 illustrates a schematic of an NRAM system according to one aspect of the invention;
• FIG. 49 illustrates operational waveforms of a memory array according to one aspect of the invention;
• FIG. 50A illustrates a diagram outlining a memory array system according to one aspect of the invention;
• FIG. 50B is a schematic of a cell according to once aspect of the invention;
• FIG. 51 illustrates operational waveforms of a memory array according to one aspect of the invention;
• FIGS.52A-G illustrate cross sections of exemplary structures according to aspects of the invention;
• FIG. 52H illustrates a plan view of an exemplary structure according to one aspect of the invention;
• FIGS.53A-C illustrate schematics of circuitry for two controlled source non-volatile field effect devices according to certain aspects of the invention;
• FIG. 54 illustrates a schematic of an NRAM system according to one aspect of the invention;
• FIG. 55 illustrates the operational waveforms of a memory array according to one aspect of the invention;
• FIG. 56A illustrates a diagram outlining a memory array system according to one aspect of the invention;
• FIG. 56B is a schematic of a cell according to once aspect of the invention;
• FIG. 57 illustrates the operational waveforms of a memory array according to one aspect of the invention;
• FIGS.58A-C illustrate cross sections of exemplary structures according to aspects of the invention;
• FIG. 58D illustrates a plan view of an exemplary structure according to one aspect of the invention.
DETAILED DESCRIPTION
• Preferred embodiments of the invention provide a field effect device that acts like a FET in its ability to create an electronic communication channel between a drain and a source node, under the control of a gate node. However, the preferred field effect devices further include a separate control structure to non-volatilely control the electrical capabilities of the field effect device. More specifically, the control structure uses carbon nanotubes to provide non-volatile switching capability that independently control the operation of the drain, source, or gate node of the field effect device. By doing so, the control structure provides non-volatile state behavior to the field effect device. Certain embodiments provide non-volatile RAM structures. Preferred embodiments are scalable to large memory array structures. Preferred embodiments use processes that are compatible with CMOS circuit manufacture. While the illustrations combine NMOS FETs with carbon nanotubes, it should be noted that based on the principle of duality in semiconductor devices, PMOS FETs may replace NMOS FETs, along with corresponding changes in the polarity of applied voltages
• Overview
• FIGS.2A-L illustrate schematics of three models of preferred embodiments of the invention. As will be explained further, below, a preferred field effect device includes a control structure using nanotubes to provide non-volatile behavior as a result of the control structure.
• Field Effect Devices (FEDs) with Controllable Sources
• Field effect devices (FEDs) with controllable sources may also be referred to as nanotube (NT)-on-Source.FIG. 2A illustrates a schematic for field effect device (FED1)20. TheFED1 device20 has a terminal T1 connected togate22, a terminal T2 connected to drain24, and acontrollable source26. Like a typical field effect device (e.g.,transistor10 ofFIG. 1) the gate node may be used to create a field to induce a conductive channel inchannel region27 between thedrain24 and a (controllable)source26. In this case, thesource26 is controllable so that it may be in open or closed communication as illustrated with theswitch30.Switch30, like all nanofabric articles referred to below, is fabricated using one or more carbon nanotubes (CNTs, or NTs) as described in incorporated references.Switch30 is preferably physically and electrically connected tocontrollable source26 bycontact28.Switch30 may be displaced to contact switch-plate (switch-node)32, which is connected to a terminal T3.Switch30 may be displaced to contact release-plate (release-node)34, which is connected to terminal T4. As will be explained below, the controllable gate utilizes nanotube components to create a non-volatile switching ability, meaning that the gate will retain its open or closed state even upon interruption of power to the circuit.
• FIG. 2B illustrates a schematic for second field effect device (FED2)40. TheFED2 device40 has a terminal T1 connected togate42, a terminal T2 connected to drain44, and acontrollable source46. Like a typical field effect device (e.g.,transistor10 ofFIG. 1) the gate node may be used to create a field to induce a conductive channel in channel region47 between thedrain44 and a (controllable)source46. In this case, thesource46 is controllable so that it may be in open or closed communication as illustrated with the depiction ofswitch50.Switch50 is fabricated using one or more carbon nanotubes (CNTs, or NTs).Switch50 is preferably physically and electrically connected to contact52, which is connected to a terminal T3.Switch50 may be displaced to contact a switch-plate48, which is connected to acontrollable source46.Switch50 may be displaced to contact release-plate54, which is connected to terminal T4. As will be explained below, the controllable gate utilizes nanotube components to create a non-volatile switching ability, meaning that the gate will retain its open or closed state even upon interruption of power to the circuit.
• FIG. 2C illustrates a schematic of third field effect device (FED3)60. TheFED3 device60 has a terminal T1 connected togate62, a terminal T2 connected to drain64, and acontrollable source66. Like a typical field effect device (e.g.,transistor10 ofFIG. 1) the gate node may be used to create a field to induce a conductive channel inchannel region67 between thedrain64 and a (controllable)source66. In this case, thesource66 is controllable so that it may be in open or closed communication as illustrated with the depiction ofswitch70.Switch70 is fabricated using one or more carbon nanotubes (CNTs, or NTs).Switch70 is preferably physically and electrically connected tocontrollable source66 bycontact68.Switch70 may be displaced to contact switch-plate72, which is connected to a terminal T3.Switch70 may be displaced to contact dielectric surface of release-plate76 on release-plate74, which is connected to terminal T4. As will be explained below, the controllable gate utilizes nanotube components to create a non-volatile switching ability, meaning that the gate will retain its open or closed state even upon interruption of power to the circuit, such non-volatilely is more fully described in incorporated references and will not be repeated here for the sake of brevity.
• FIG. 2D illustrates a schematic of fourth field effect device (FED4)80. TheFED4 device80 has a terminal T1 connected togate82, a terminal T2 connected to drain84, and acontrollable source86. Like a typical field effect device (e.g.,transistor10 ofFIG. 1) the gate node may be used to create a field to induce a conductive channel inchannel region87 between thedrain84 and a (controllable)source86. In this case, thesource86 is controllable so that it may be in open or closed communication as illustrated with by the depiction ofswitch90.Switch90 is fabricated using one or more carbon nanotubes (CNTs, or NTs) as described in incorporated references.Switch90 is preferably physically and electrically connected to contact92, which is connected to a terminal T3.Switch90 may be displaced to contact a switch-plate88, which is connected to acontrollable source86.Switch90 may be displaced to contact release-plate dielectric surface96 on release-plate94, which is connected to terminal T4. As will be explained below, the controllable gate utilizes nanotube components to create a non-volatile switching ability, meaning that the gate will retain its open or closed state even upon interruption of power to the circuit.
• Field Effect Devices (FEDs) with Controllable Drains
• Field effect devices (FEDs) with controllable drains may also be referred to as nanotube (NT)-on-Drain.FIG. 2E illustrates a schematic of fifth field effect device (FED5)100. TheFED5 device100 has a terminal T1 connected togate102, acontrollable drain104, and asource106 connected to a terminal T3. Like a typical field effect device (e.g.,transistor10 ofFIG. 1) the gate node may be used to create a field to induce a conductive channel inchannel region107 between the (controllable)drain104 and asource106. In this case, thedrain104 is controllable so that it may be in open or closed communication as illustrated by the depiction ofswitch110.Switch110 is fabricated using one or more carbon nanotubes (CNTs, or NTs).Switch110 is preferably physically and electrically connected tocontrollable drain104 bycontact108.Switch110 may be displaced to contact switch-plate112, which is connected to a terminal T2.Switch110 may be displaced to contact release-plate114, which is connected to terminal T4. As will be explained below, the controllable gate utilizes nanotube components to create a non-volatile switching ability, meaning that the gate will retain its open or closed state even upon interruption of power to the circuit.
• FIG. 2F illustrates a schematic of sixth field effect device (FED6)120. TheFED6 device120 has a terminal T1 connected togate122, acontrollable drain124, and asource126 connected to a terminal T3. Like a typical field effect device (e.g.,transistor10 ofFIG. 1) the gate node may be used to create a field to induce a conductive channel inchannel region127 between thedrain124 and a (controllable)source126. In this case, thedrain124 is controllable so that it may be in open or closed communication as illustrated by the depiction ofswitch130.Switch130 is fabricated using one or more carbon nanotubes (CNTs, or NTs).Switch130 is preferably physically and electrically connected to contact132, which is connected to terminal T2.Switch130 may be displaced to contact a switch-plate128, which is connected to acontrollable drain124.Switch130 may be displaced to contact release-plate134, which is connected to terminal T4. As will be explained below, the controllable gate utilizes nanotube components to create a non-volatile switching ability, meaning that the gate will retain its open or closed state even upon interruption of power to the circuit.
• FIG. 2G illustrates a schematic of seventh field effect device (FED7)140. TheFED7 device140 has a terminal T1 connected togate142, acontrollable drain144, and a source146 connected to a terminal T3. Like a typical field effect device (e.g.,transistor10 ofFIG. 1) the gate node may be used to create a field to induce a conductive channel inchannel region147 between the (controllable)drain144 and a source146. In this case, thedrain144 is controllable so that it may be in open or closed communication as illustrated by the depiction ofswitch150.Switch150 is fabricated using one or more carbon nanotubes (CNTs, or NTs).Switch150 is preferably physically and electrically connected tocontrollable drain144 bycontact148.Switch150 may be displaced to contact switch-plate152, which is connected to a terminal T2.Switch150 may be displaced to contact release-plate dielectric surface156 on release-plate154, which is connected to terminal T4. As will be explained below, the controllable gate utilizes nanotube components to create a non-volatile switching ability, meaning that the gate will retain its open or closed state even upon interruption of power to the circuit.
• FIG. 2H illustrates a schematic of eighth field effect device (FED8)160. TheFED8 device160 has a terminal T1 connected togate162, acontrollable drain164, and asource166 connected to a terminal T3. Like a typical field effect device (e.g.,transistor10 ofFIG. 1) the gate node may be used to create a field to induce a conductive channel inchannel region167 between the (controllable)drain164 and asource166. In this case, thedrain164 is controllable so that it may be in open or closed communication as illustrated by the depiction ofswitch170.Switch170 is fabricated using one or more carbon nanotubes (CNTs, or NTs).Switch170 is preferably physically and electrically connected to contact172, which is connected to terminal T2.Switch170 may be displaced to contact a switch-plate168, which is connected to acontrollable drain164.Switch170 may be displaced to contact release-plate dielectric surface176 on release-plate174, which is connected to terminal T4. As will be explained below, the controllable gate utilizes nanotube components to create a non-volatile switching ability, meaning that the gate will retain its open or closed state even upon interruption of power to the circuit.
• Field Effect Devices (FEDs) with Controllable Gates
• Field effect devices (FEDs) with controllable gates may also be referred to as nanotube (NT)-on-Gate.FIG. 2I illustrates a schematic of ninth field effect device (FED9)180. Thedevice180 has acontrollable gate182, adrain184 connected to terminal T2, and asource186 connected to a terminal T3. Like a typical field effect device (e.g.,transistor10 ofFIG. 1) the gate node may be used to create a field to induce a conductive channel inchannel region187 between adrain184 and asource186. In this case, thegate182 is controllable so that it may be in open or closed communication as illustrated by the depiction ofswitch190.Switch190 is fabricated using one or more carbon nanotubes (CNTs, or NTs).Switch190 is preferably physically and electrically connected tocontrollable gate182 bycontact188.Switch190 may be displaced to contact switch-plate192, which is connected to a terminal T1.Switch190 may be displaced to contact release-plate194, which is connected to terminal T4. As will be explained below, the controllable gate utilizes nanotube components to create a non-volatile switching ability, meaning that the gate will retain its open or closed state even upon interruption of power to the circuit.
• FIG. 2J illustrates a schematic of tenth field effect device (FED10)200. TheFED10 device200 has a terminalcontrollable gate202, adrain204 connected to a terminal T2, and asource206 connected to a terminal T3. Like a typical field effect device (e.g.,transistor10 ofFIG. 1) the gate node may be used to create a field to induce a conductive channel inchannel region207 between thedrain204 andsource206. In this case, thegate202 is controllable so that it may be in open or closed communication as illustrated by the depiction ofswitch210.Switch210 is fabricated using one or more carbon nanotubes (CNTs, or NTs).Switch210 is preferably physically and electrically connected to contact212, which is connected to terminal T1.Switch210 may be displaced to contact a switch-plate208, which is connected to acontrollable gate202.Switch210 may be displaced to contact release-plate214, which is connected to terminal T4. As will be explained below, the controllable gate utilizes nanotube components to create a non-volatile switching ability, meaning that the gate will retain its open or closed state even upon interruption of power to the circuit.
• FIG. 2K illustrates a schematic of eleventh field effect device (FED11)220. Thedevice220 has acontrollable gate222, adrain224 connected to a terminal T2, and asource226 connected to a terminal T3. Like a typical field effect device (e.g.,transistor10 ofFIG. 1) the gate node may be used to create a field to induce a conductive channel inchannel region227 between adrain224 and asource226. In this case, thegate222 is controllable so that it may be in open or closed communication as illustrated by the depiction ofswitch230.Switch230 is fabricated using one or more carbon nanotubes (CNTs, or NTs).Switch230 is preferably physically and electrically connected tocontrollable gate222 bycontact228.Switch230 may be displaced to contact switch-plate232, which is connected to a terminal T1.Switch230 may be displaced to contact release-plate dielectric surface236 on release-plate234, which is connected to terminal T4. As will be explained below, the controllable gate utilizes nanotube components to create a non-volatile switching ability, meaning that the gate will retain its open or closed state even upon interruption of power to the circuit.
• FIG. 2L illustrates a schematic of twelfth field effect device (FED12)240. TheFED12 device240 has acontrollable gate242, adrain244 connected to a terminal T2, and asource246 connected to a terminal T3. Like a typical field effect device (e.g.,transistor10 ofFIG. 1) the gate node may be used to create a field to induce a conductive channel inchannel region247 between the (controllable)drain244 and asource246. In this case, thegate242 is controllable so that it may be in open or closed communication as illustrated by the depiction ofswitch250.Switch250 is fabricated using one or more carbon nanotubes (CNTS, or NTs).Switch250 is preferably physically and electrically connected to contact252, which is connected to terminal T1.Switch250 may be displaced to contact a switch-plate248, which is connected to acontrollable gate242.Switch250 may be displaced to contact release-plate dielectric surface256 on release-plate254, which is connected to terminal T4. As will be explained below, the controllable gate utilizes nanotube components to create a non-volatile switching ability, meaning that the gate will retain its open or closed state even upon interruption of power to the circuit.
• As will be explained below, the controllable structures are implemented using nanotube technology. More specifically, non-volatile switching elements are made of ribbons of matted fabric of carbon nanotubes. These elements may be electromechanically deflected into an open or closed state relative to a respective source, drain, or gate node using electrostatic forces. Under preferred embodiments, the construction of the control structures is such that once switched “ON” inherent van der Waals forces are sufficiently large (relative to a restoring force inherent in the device geometry) so that the switching element will retain its non-volatilized state; that is, the element will retain its state even in the event of power interruption.
• Operation of Field Effect Devices with Controllable Sources
• Four schematics of field effect devices (FEDs) with controllable sources have been described (FIGS.2A-D).FIGS. 3A throughFIG. 9 illustrate the operation of field effect devices with controllable sources for two of the FED configurations, device80 (FIG. 2D) and device20 (FIG. 2A). FED devices with controllable sources are also referred to as NT-on-Source devices. For each of these two FED configurations, at least one switch-mode setting operation is described, followed by an example of full voltage swing circuit operation (digital switching), and an example of small signal analog circuit operation.
• FIG. 3A illustrates a first FED configuration;field effect device80 is combined withresistor302 of value R, such that one terminal ofresistor302 is attached toFED device80 terminal T2, and the other side ofresistor302 is attached topower supply terminal304 to formcircuit schematic300.FIG. 3B illustrates circuit schematic310 in which switch90 has been activated to position90′ to electrically connect switch-plate88 withcontact92 as illustrated inFIG. 3B.Controllable source86 is electrically connected to terminal T3 by means of the established continuous electrical path formed bysource86 connected to switch-plate88, switch-plate88 connected to one side ofswitch90′, the opposite side ofswitch90′ connected to contact92, and contact92 connected to terminal T3.
• FIG. 3C illustrates circuit schematic310′ in which switch90 has been activated to position90″ to electrically release-plate dielectric surface96.Controllable source86 is an electrically open circuited, and has no continuous electrical path to anyFED480 device terminals. The mode-setting electrical signals applied to the terminals T1, T2, T3, and T4 ofschematics300,310, and310′ to causeswitch90 to switch toposition90′ orposition90″ are illustrated inFIG. 4.
• FIG. 4 illustrates the operational mode-settingvoltage waveforms311 applied to terminals T1, T2, T3, and T4 to activateswitch90. Control signals are applied to terminals T1-T4 by a control circuit (not shown) using control lines (not shown). There is no electrical signal applied toelectrical terminal304 during mode-setting.Column1 illustrates the electrical signals used to changeswitch90 fromposition90″, (also referred to as the open (off) position), to position90′, (also referred to as the closed (on) position).Column2 illustrates the electrical signals used to changeswitch90 fromposition90′, (also referred to as the closed position), to position90″, (also referred to as the open position). The mode-setting waveforms are valid within the mode-setting time interval illustrated undercolumns1 and2 inFIG. 4. Other time intervals contain cross-hatched lines betweenvoltages0 and V_DD, indicating that these waveforms can be anywhere within this voltage range, and represent the circuit operating range. V_DDis selected to be less than the voltage switching voltage V_SWto ensure thatswitch90 is not activated (resulting in mode-change) during circuit operation.
• Mode-setting is based on electromechanical switching of carbon nanotube (NT) switch using electrostatic forces. The behavior of a NT fabric is similar to that of a single NT, see U.S. Pat. No. 6,643,165, where the electrostatic attractive force is due to oppositely chargedsurfaces1 and2, and where the electrostatic F_E=K (V₁−V₂)²/(R₁₂)². For an applied voltage, an equilibrium position of the NT, or NT fabric, is defined by the balance of the elastic, electrostatic, and van der Waals forces. As the NT, or NT fabric deflects, the elastic forces change. When the applied potential (voltage) difference between the nanotube and a reference electrode exceeds a certain voltage, the NT or NT fabric becomes unstable and collapses onto the reference electrode. The voltage difference between a NT or NT fabric, and a reference electrode that causes the NT or NT fabric to collapse, may be referred to as the pull-in voltage, or the collapse voltage, or the nanotube threshold voltage VNT-TH. The reference electrode may be a switch-plate, or a release-plate, or a release-plate with a dielectric layer. Once the NT or NT fabric is in contact with, or in very close proximity to, the reference electrode (in a region of strong van der Waals force), the electrostatic force FE may be reduced to zero by removing the voltage difference between NT or NT fabric and the reference electrode. Power may be removed, and the NT or NT fabric remains in contact, and thus stores information in a non-volatile mode.
• Column1 ofFIG. 4 illustrates the voltage and timing waveforms applied to terminals T1-T4 ofFED480 that force a transition of NT switch90 fromposition90″, in contact withinsulator surface96 on release-plate94 as illustrated inFIG. 3C, to position90′, in contact with switch-plate88 as illustrated inFIG. 3B.Switch90 transitions from open to closed. Voltage V_T4, applied to terminal T4, transitions to switching voltage V_SW. Voltage V_T2applied to terminal T2 transitions to zero (0) volts. V_T3applied to terminal T3 transitions to switching voltage V_SW. Terminal T1 (connected to gate82) transitions from zero to V_DDforming a channel inchannel region87, thereby drivingcontrollable source86 voltage V_SOURCEto zero. The electrostatic force betweenswitch90 inposition90″ and release-plate94 is zero. The electrostatic force betweenswitch90 inposition90″ and switch-plate88 is F_E=K (V_SW)²/(R₁₂)², where R₁₂is thegap separating switch90 from switch-plate88. Typical V_NT-THvoltages may range from 2 to 3 volts, for example, any appropriate potential difference however, is within the scope of the invention. V_NT-THis a function of the suspended length ofNT switch90 and the gap (separation) betweenNT switch90 and the switch-plate and release-plate electrodes. Typical NT switch suspended length is 130 to 180 nm, with gaps of 10 to 20 nm, for example, but other geometries are possible so long as the switching properties work appropriately.
• Column2 ofFIG. 4 illustrates the voltage and timing waveforms applied to terminals T1-T4 ofFED480 that force a transition of NT switch90 fromposition90′, in contact with switch-plate88 as illustrated inFIG. 3B, to position90″, in contact with release-plate dielectric surface96 on release-plate94 as illustrated inFIG. 3C.Switch90 transitions from closed to open. Voltage V_T4, applied to terminal T4, transitions to switching voltage V_SW. Voltage V_T2applied to terminal T2 transitions to zero (0) volts. V_T3applied to terminal T3 transitions to zero volts. Terminal T1 (connected to gate82) transitions from zero to V_DDforming a channel inchannel region87, thereby drivingcontrollable source86 voltage V_SOURCEto zero. The electrostatic force betweenswitch90 inposition90′ and switch-plate88 is zero. The electrostatic force betweenswitch90 inposition90′ and release-plate94 is F_E=K (V_SW)²/(R₁₂)², where R₁₂is thegap separating switch90 from release-plate94. Typical V_NT-THvoltages may range from 2 to 3 volts, for example. The threshold voltage forswitch90 transitions between open and closed, and closed and open positions may be different, without effecting the operation of the device. If V_SWexceeds V_NT-TH, then mode-setting will take place. Circuit operating voltages range from 0 to V_DD. In order to avoid unwanted mode-setting during circuit operation, V_DDis less than V_NT-TH.
• FIG. 5 illustrates the full signal (voltage)swing waveform313 operation ofcircuit300, with waveforms applied to terminals T1, T2, T3, and T4.Column1 illustrates the electrical signals applied to terminal T1-T4 for circuit schematic310 whenswitch90 is in theclosed position90′ as illustrated inFIG. 3B.Column2 illustrates the electrical signals applied to terminals T1-T4 for circuit schematic310′ whenswitch90 is in theopen position90″ as illustrated inFIG. 3C. Circuit schematic310 illustrates the FED used in a simple inverter configuration withload resistor302 of value R connected tovoltage terminal304 at voltage V=V_DD. For V_NT-THin the 2 to 3 volt range, for example, V_DDis selected as less than 2 volts, 1.0 to 1.8 volts, for example. The operation ofcircuit310 is as illustrated inFIG. 5,column1. Withswitch90 in the90′ position, the voltage V_T4on terminal T4 can be any value. Voltage V_T3applied to terminal T3 is set to zero volts. A pulse V_T1of amplitude V_DDis applied to terminal T1. When V_T1=0, no FET conductive path is activated, the electrical path between terminals T2 and T3 ofFED480 is open, current I=0, and V_OUT=V_DD. When V_T1=V_DD,FET80 channel of resistance R_FETis formed, in series with R_SWITCHofswitch90′, connecting terminals T2 and T3. The resistance of FED480 between terminals T2 and T3 is R_FED=R_FET+R_SWITCH. R_FETis the FET channel resistance, and R_SWITCHis the resistance of NT switch90′. R_SWITCHincludes the resistance between switch-plate88 andNT90′, theNT90′ resistance (typically much less than the contact resistances), and the contact resistance betweencontact92 andNT90′. R_FETis determined by the FET electrical parameters and the width to length ratio used in the FET design (Reference: Baker et al., “CMOS Circuit Design, Layout, and Simulation”, IEEE Press, 1998,Chapter 5 “the MOSFET”, pages 83-106). By selecting W/L ratio values, R_FETmay range from less than 10 Ohms to more than 10,000 Ohms. The quantum contact resistance between metal electrodes and the NT fabric varies as a function of the fabric density (number of NTs per unit area) and the width of the contact. The contact resistance per fiber may vary from less than 100 Ohms to more than 100,000 Ohms. When V_T1=V_DD, current I=V_DD/(R+R_FED), and V_T2=V_OUT=V_DD×(R_FED)/(R+R_FED). If R_FED<<R, then V_T2=V_OUT≈0 volts, illustrated inFIG. 5,column1.
• Circuit schematic310′ illustratesFED480 used in a simple inverter configuration withload resistor302 of value R connected tovoltage terminal304 at voltage V=V_DD. The full signal (voltage) swing operation ofcircuit310′ is as illustrated inFIG. 5,column2. Withswitch90 inposition90″, the FED electrical path between terminals T2 and T3 is open, terminal T4 is insulated, therefore current I=0, and V_T2=V_OUT=V_DDfor all applied voltages.
• FIG. 6 illustrates the small signal (voltage)swing waveforms315 operation ofcircuit300, with waveforms applied to terminals T1, T2, T3, and T4.Column1 illustrates the electrical signals applied to terminal T1-T4 for circuit schematic310 whenswitch90 is in theclosed position90′ as illustrated inFIG. 3B. Circuit schematic310 illustrates the FED used in a simple inverter configuration withload resistor302 of value R connected tovoltage terminal304 at voltage V=V_DD. For V_NT-THin the 2 to 3 volt range, for example, V_DDis selected as less than 2 volts, 1.0 to 1.8 volts, for example. The operation ofcircuit310 for small signal (analog) amplification is as illustrated inFIG. 5,column1. Withswitch90 inposition90′, the voltage V_T4on terminal T4 can be any value. Voltage V_T3applied to terminal T3 is set to zero volts. A signal V_T1of with amplitude exceeding FET threshold voltage V_FET-TH(V_FET-TH=0.3−0.7 volts, for example) is applied to terminal T1. Since V_T1>V_FET-TH, a path between terminals T2 and T3 is maintained. If R_SWITCHis less than R_FET, then the output VT₂=V_OUTofcircuit310 inverts the input signal and exhibits gain as illustrated inFIG. 6,column1. Circuit gain can be calculated as described in Baker et al., “CMOS Circuit Design, Layout, and Simulation”, IEEE Press, 1998, Chapter 9 “the MOSFET”, pages 165-181.
• Circuit schematic310′ illustratesFED480 used in a simple inverter configuration withload resistor302 of value R connected tovoltage terminal304 at voltage V=V_DD. The small signal (voltage) swing operation ofcircuit310′ is as illustrated inFIG. 6,column2. Withswitch90 inposition90″, the FED electrical path between terminals T2 and T3 is open, terminal T4 is insulated, therefore current I=0, and VT₂=V_OUT=V_DDfor all applied voltages.
• In the second FED configuration,field effect device20 is combined withfirst resistor324 of value R, such that one terminal ofresistor324 is attached toFED device20 terminal T2, and the other side ofresistor324 is attached topower supply terminal322 as illustrated inFIG. 7A. Asecond resistor328 of value R′ is attached toFED device20 terminal T4, and the other side ofresistor328 is attached topower supply326 to form the circuit schematic illustrated inFIG. 7A. Such configurations are exemplary and other working configurations are within the scope of the invention.
• FIG. 7B illustrates a schematic ofcircuit330 in which switch30 has been activated tofirst position30′ to electrically connectcontact28 to switch-plate32.Controllable source26 is electrically connected to terminal T3 by means of the established continuous electrical path formed bysource26 connected to contact28;contact28 connected to one side ofswitch30′; the opposite side ofswitch30′ connected to switch-plate32; switch-plate32 connected to terminal T3.FIG. 7C illustrates a schematic ofcircuit330′ in which switch30 has been activated tosecond position30″ and contacts release-plate34.Controllable source26 is electrically connected to FED120 device terminal T4. The mode-setting electrical signals applied to the terminals T1, T2, T3, and T4 ofschematics320,330, and330′ that causeswitch30 to switch tofirst position30′ orsecond position30″ are illustrated inFIG. 8.
• FIG. 8 illustrates the operational mode-settingwaveforms335 applied to terminals T1, T2, T3, and T4 to activateswitch30. Control signals are applied to terminals T1-T4 by a control circuit (not shown) using control lines (not shown). There is no electrical signal applied toelectrical terminals322 and326 during mode-setting.Column1 illustrates the electrical signals used to changeswitch30 fromposition30″, also referred to as the second position, to position30′, also referred to as the first position.Column2 illustrates the electrical signals used to changeswitch30 fromposition30′, also referred to as the first position, to position30″, also referred to as the second position. The mode-setting waveforms are valid within the mode-setting time interval illustrated undercolumns1 and2 inFIG. 8. Other time intervals contain cross-hatched lines betweenvoltages0 and V_DD, indicating that these waveforms can be anywhere within this voltage range, and represent the circuit operating range. V_DDis selected to be less than the voltage switching voltage V_SWto ensure thatswitch30 is not activated (resulting in mode-resetting) during circuit operation.
• Mode-setting is based on electromechanical switching of carbon nanotube (NT) switch using electrostatic forces. The behavior of a NT fabric is similar to that of a single NT, as stated above, where the electrostatic attractive force is due to oppositely charged surfaces.Column1 ofFIG. 8 illustrates the voltage and timing waveforms applied to terminals T1-T4 ofFED120 that force a transition of NT switch30 fromsecond position30″, in contact with release-plate94 as illustrated inFIG. 7C, tofirst position30′, in contact with switch-plate32 as illustrated inFIG. 7B. Voltage V_T4, applied to terminal T4, transitions to zero volts. Voltage V_T2applied to terminal T2 transitions to zero (0) volts. V_T3applied to terminal T3 transitions to switching voltage V_SW. Terminal T1 (connected to gate22) transitions from zero to V_DDforming a channel inchannel region27, thereby drivingcontrollable source26 voltage V_SOURCEto zero. The electrostatic force betweenswitch30 inposition30″ and release-plate34 is zero. The electrostatic force betweenswitch30 inposition30″ and switch-plate32 is F_E=K (V_SW)²/(R₁₂)², where R₁₂is thegap separating switch30 from switch-plate32. Typical V_NT-THvoltages may range from 2 to 3 volts, for example. Typical NT switch suspended length is 130 to 180 nm, with gaps of 10 to 20 nm, for example.
• Column2 ofFIG. 8 illustrates the voltage and timing waveforms applied to terminals T1-T4 ofFED20 that force a transition of NT switch30 fromfirst position30′, in contact with switch-plate32 as illustrated inFIG. 7B, tosecond position30″, in contact with release-plate34 as illustrated inFIG. 7C. Voltage V_T4, applied to terminal T4, transitions to switching voltage V_SW. Voltage V_T2applied to terminal T2 transitions to zero (0) volts. V_T3applied to terminal T3 transitions to zero volts, terminal T1 (connected to gate22) transitions from zero to V_DDforming a channel inchannel region27, thereby drivingcontrollable source26 voltage V_SOURCEto zero. The electrostatic force betweenswitch30 inposition30′ and switch-plate28 is zero. The electrostatic force betweenswitch30 inposition30′ and release-plate34 is F_E=K (V_SW)²/(R₁₂)², where R₁₂is thegap separating switch30 from release-plate34. Typical V_NT-THvoltages may range from 2 to 3 volts, for example. The threshold voltage forswitch30 transitions between second and first, and first and second positions may be different, without effecting the operation of the device. If V_SWexceeds V_NT-TH, then mode-setting will take place. Circuit operating voltages range from 0 to V_DD. In order to avoid unwanted mode-setting during circuit operation, V_DDis less than V_NT-TH.
• FIG. 9 illustrates the full signal (voltage)swing waveforms345 operation ofcircuit320, with waveforms applied to terminals T1, T2, T3, and T4.Column1 illustrates the electrical signals applied to terminal T1-T4 forcircuit330 whenswitch30 is in thefirst position30′ as illustrated inFIG. 7B.Column2 illustrates the electrical signals applied to terminals T1-T4 forcircuit330′ whenswitch30 is in thesecond position30″ as illustrated inFIG. 7C.Circuit330 illustrates a FED used in a simple inverter configuration withload resistor324 of value R connected tovoltage terminal322 at voltage V=V_DD. For V_NT-THin the 2 to 3 volt range, for example, V_DDis selected as less than 2 volts, 1.0 to 1.8 volts, for example. The operation ofcircuit330 is as illustrated inFIG. 9,column1. Withswitch30 in the30′ position, the voltage V_T4on terminal T4 can be any value. Voltage V_T3applied to terminal T3 is set to zero volts. A pulse V_T1of amplitude V_DDis applied to terminal T1. When V_T1=0, no FET conductive path is activated, the electrical path between terminals T2 and T3 ofFED20 is open, current I=0, and V_T2=V_OUT=V_DD. When V_T1=V_DD,FET channel27 of resistance R_FETis formed, in series with R_SWITCHofswitch30′, connecting terminals T2 and T3. The resistance ofFED20 between terminals T2 and T3 is R_FED=R_FET+^RSW_ITCH. R_FETis the FET channel resistance, and R_SWITCHis the resistance of NT switch30′. R_SWITCHincludes the resistance betweencontact28 andNT30′, theNT30′ resistance (typically much less than the contact resistances), and the resistance between switch-plate32 andNT30′. R_FETis determined by the FET electrical parameters and the width to length ratio used in the FET design (Reference: Baker et al., “CMOS Circuit Design, Layout, and Simulation”, IEEE Press, 1998,Chapter 5 “the MOSFET”, pages 83-106). By selecting W/L ratio values, R_FETmay range from less than 10 Ohms to more than 10,000 Ohms. The quantum contact resistance between metal electrodes and the NT fabric varies as a function of the fabric density (number of NTs per unit area) and the width of the contact. The contact resistance may vary from less than 100 Ohms to more than 100,000 Ohms. When V_T1=V_DD, current I=V_DD/(R+R_FED), and V_T2=V_OUT=V_DD×(R_FED)/(R+R_FED). If R_FED<<R, then V_T2=V_OUT≈0 volts, illustrated inFIG. 9,column1.
• The schematic ofcircuit330′ illustrates a FED used in a more complex circuit configuration withload resistor324 of value R connected tovoltage terminal322 at voltage V=V_DD, andresistor328 of value R′ connected tovoltage terminal326 at voltage zero. For V_NT-THin the 2 to 3 volt range, for example, V_DDis selected as less than 2 volts, 1.0 to 1.8 volts, for example. The operation ofcircuit330′ is as illustrated inFIG. 9,column2. Withswitch30 in the30′ position, the voltage V_T3on terminal T3 can be any value. A pulse V_T1of amplitude V_DDis applied to terminal T1. When V_T1=0, no FET conductive path is activated, the electrical path between terminals T2 and T4 ofFED120 is open, current I=0, and V_T2=V_OUT=V_DD, and V_T4=0. When V_T1=V_DD,FET channel27 of resistance R_FETis formed, in series with R_SWITCHofswitch30″, connecting terminals T2 and T4. The resistance ofFED20 between terminals T2 and T4 is R_FED=R_FET+R_SWITCH. R_FETis the FET channel resistance, and R_SWITCHis the resistance of NT switch30″. R_SWITCHincludes the resistance betweencontact28 andNT30″, theNT30″ resistance (usually much less than the contact resistances), and the resistance between release-plate34 andNT30″. R_FETis determined by the FET electrical parameters and the width to length ratio used in the FET design. By selecting W/L ratio values, R_FETmay range from less than 10 Ohms to more than 10,000 Ohms. The quantum contact resistance between metal electrodes and the NT fabric varies as a function of the fabric density (number of NTs per unit area) and the width of the contact. The contact resistance may vary from less than 100 Ohms to more than 100,000 Ohms When V_T1=V_DD, current I=V_DD/(R+R′+R_FED), VT₂=V_OUT=V_DD×(R′+R_FED)/(R+R′ +R_FED), and V_T4=V_DD×(R′)/(R+R′ +R_FED). If R_FED<<R, and R′=R, then V_T2=V_OUT=V_DD/2, and V_T4=V_DD/2, as illustrated inFIG. 9,column2.
• In the example of the operation of circuit320 (FIG. 7A), circuit operation for two switch-mode settings were described, one forswitch30 infirst position30′ as illustrated inFIG. 7B, and the other forswitch30 in thesecond position30″ as illustrated inFIG. 7C. The voltages on FED terminals T2 and T4 varied as a function of the switch-mode settings.FED120 may also be used in other applications. For example, a first network may be connected to terminal T2, a second network may be connected to terminal T3, and a third network may be connected to terminal T4. WhenFED120switch30 is in thefirst position30′ (FIG. 7B), a first network connected to terminal T2 is connected to a second network connected to terminal T3. WhenFED120switch30 is in thesecond position30″, a first network connected to terminal T2 is connected to a third network connected to terminal T4. Thus, in this application,FED120 is used to route signals from a first network to a second network, or instead, to a third network. The network configuration remains in place even if power is turned off becauseFED120 is a non-volatile device.
• Operation of Field Effect Devices with Controllable Drains
• Four schematics of field effect devices (FEDs) with controllable drains have been described (FIGS.2E-H).FIGS. 10A-12 illustrates the operation of field effect devices with controllable drains for one of the FED configurations, FED8 device160 (FIG. 2H). As stated above, FED devices with controllable drains are also referred to as NT-on-Drain devices. A switch-mode setting operation is described, followed by an example of full voltage swing circuit operation (digital switching).
• Fieldeffect device FED8160 is combined withresistor364 of value R, such that one terminal ofresistor364 is attached toFED8 device160 terminal T2, and the other side ofresistor364 is attached topower supply terminal362 to form circuit schematic360 as illustrated inFIG. 10A.
• FIG. 10B illustrates circuit schematic370 in which switch170 has been activated to position170′ to electrically connect switch-plate168 to contact172.Controllable drain164 is electrically connected to terminal T2 by means of the established continuous electrical path formed bydrain164 connected to switch-plate168; switch-plate168 connected to one side ofswitch170′; the opposite side ofswitch170′ connected to contact172; contact172 connected to terminal T2.
• FIG. 1C illustrates circuit schematic370′ in which switch170 has been activated to position170″ to contact release-plate dielectric surface176.Controllable drain164 is electrically open circuited, and has no continuous electrical path to any terminals ofFED8160 device. The mode-setting electrical signals applied to the terminals T1, T2, T3, and T4 ofschematics360,370, and370′ to causeswitch170 to switch toposition170′ orposition170″ are illustrated inFIG. 11.
• FIG. 11 illustrates the operational mode-settingwaveforms355 applied to terminals T1, T2, T3, and T4 to activateswitch170. Control signals are applied to terminals T1-T4 by a control circuit (not shown) using control lines (not shown). There is no electrical signal applied toelectrical terminal362.Column1 illustrates the electrical signals used to changeswitch170 fromposition170″, also referred to as the open position, to position170′, also referred to as the closed position.Column2 illustrates the electrical signals used to changeswitch170 fromposition170′, also referred to as the closed position, to position170″, also referred to as the open position. The mode-setting waveforms are valid within the mode-setting time interval illustrated undercolumns1 and2 inFIG. 11. Other time intervals contain cross-hatched lines betweenvoltages0 and V_DD, indicating that these waveforms can be anywhere within this voltage range, and represent the circuit operating range. V_DDis selected to be less than the voltage switching voltage V_SWto ensure thatswitch170 is not activated (resulting in mode-resetting) during circuit operation.
• Mode-setting is based on electromechanical switching of carbon nanotube (NT) switch using electrostatic forces. As stated above, the behavior of a NT fabric is similar to that of a single NT, where the electrostatic attractive force is due to oppositely charged surfaces.Column1 ofFIG. 11 illustrates the voltage and timing waveforms applied to terminals T1-T4 ofFED8160 that force a transition of NT switch170 fromposition170″, in contact withinsulator surface176 on release-plate174 as illustrated inFIG. 10C, to position170′, in contact with switch-plate168 as illustrated inFIG. 10B. Switch170 transitions from open to closed. Voltage V_T4, applied to terminal T4, transitions to switching voltage V_SW. Voltage V_T2applied to terminal T2 transitions switching voltage V_SW. V_T3applied to terminal T3 transitions to zero volts. Terminal T1 (connected to gate162) transitions from zero to V_DDforming a channel inchannel region167, thereby drivingcontrollable drain164 voltage V_DRAINto zero. The electrostatic force betweenswitch170 inposition170″ and release-plate174 is zero. The electrostatic force betweenswitch170 inposition170″ and switch-plate168 is F_E=K (V_SW)²/(R₁₂)², where R₁₂is thegap separating switch170 from switch-plate168. Typical V_NT-THvoltages may range from 2 to 3 volts, for example. V_NT-THis a function of the suspended length ofNT switch170 and the gap (separation) betweenNT switch170 and the switch-plate and release-plate electrodes. Typical, but non-exclusive exemplary ranges for NT switch suspended length is 130 to 180 nm, with gaps of 10 to 20 nm.
• Column2 ofFIG. 11 illustrates the voltage and timing waveforms applied to terminals T1-T4 ofFED8160 that force a transition of NT switch170 fromposition170′, in contact with switch-plate168 as illustrated inFIG. 10B, to position170″, in contact with release-plate dielectric surface176 on release-plate174 as illustrated inFIG. 10C. Switch170 transitions from closed to open. Voltage V_T4, applied to terminal T4, transitions to switching voltage V_SW. Voltage V_T2applied to terminal T2 transitions to zero (0) volts. V_T3applied to terminal T3 transitions to zero volts. Terminal T1 (connected to gate162) transitions from zero to V_DDforming a channel inchannel region167, thereby drivingcontrollable drain164 voltage V_DRAINto zero. The electrostatic force betweenswitch170 inposition170′ and switch-plate168 is zero. The electrostatic force betweenswitch170 inposition170′ and release-plate174 is F_E=K (V_SW)²/(R₁₂)², where R₁₂is thegap separating switch170 from release-plate174. Typical V_NT-THvoltages may range from 2 to 3 volts, for example. The threshold voltage forswitch170 transitions between open and closed, and closed and open positions may be different, without effecting the operation of the device. If V_SWexceeds V_NT-TH, then mode-setting will take place. Circuit operating voltages range from 0 to V_DD. In order to avoid unwanted mode-setting during circuit operation, V_DDis less than V_NT-TH.
• FIG. 12 illustrates the full signal (voltage)swing waveforms365 operation ofcircuit360, with waveforms applied to terminals T1, T2, T3, and T4.Column1 illustrates the electrical signals applied to terminal T1-T4 for circuit schematic370 whenswitch170 is in theclosed position170′ as illustrated inFIG. 10B.Column2 illustrates the electrical signals applied to terminals T1-T4 for circuit schematic370′ whenswitch170 is in theopen position170″ as illustrated inFIG. 10C. Circuit schematic370 illustrates the FED used in a simple inverter configuration withload resistor364 of value R connected tovoltage terminal362 at voltage V=V_DD. For V_NT-THin the 2 to 3 volt range, for example, V_DDis selected as less than 2 volts, 1.0 to 1.8 volts, for example. The operation ofcircuit370 is as illustrated inFIG. 12,column1. Withswitch170 in the170′ position, the voltage V_T4on terminal T4 can be any value. Voltage V_T3applied to terminal T3 is set to zero volts. A pulse V_T1of amplitude V_DDis applied to terminal T1. When V_T1=0, no FET conductive path is activated, the electrical path between terminals T2 and T3 ofFED8160 is open, current I=0, and V_OUT=V_DD. When V_T1=V_DD,FET167 channel of resistance R_FETis formed, in series with R_SWITCHofswitch170′, connecting terminals T2 and T3. The resistance ofFED8160 between terminals T2 and T3 is R_FED=R_FET+R_SWITCH. R_FETis the FET channel resistance, and R_SWITCHis the resistance ofNT switch170′. R_SWITCHincludes the resistance between switch-plate168 andNT170′, theNT170′ resistance (typically much less than the contact resistances), and the contact resistance betweencontact172 andNT170′. R_FETis determined by the FET electrical parameters and the width to length ratio used in the FET design. By selecting W/L ratio values, R_FETmay range from less than 10 Ohms to more than 10,000 Ohms. The quantum contact resistance between metal electrodes and the NT fabric varies as a function of the fabric density (number of NTs per unit area) and the width of the contact. The contact resistance may vary from less than 100 Ohms to more than 100,000 Ohms. When V_T1=V_DD, current I=V_DD/(R+R_FED), and V_T2=V_OUT=V_DD×(R_FED)/(R+R_FED). If R_FED<<R, then VT₂=V_OUT≈0 volts, illustrated inFIG. 12,column1.
• Circuit schematic370′ illustratesFED8160 used in a simple inverter configuration withload resistor364 of value R connected tovoltage terminal362 at voltage V=V_DD. The full signal (voltage) swing operation ofcircuit370′ is as illustrated inFIG. 12,column2. Withswitch90 inposition90″, the FED electrical path between terminals T2 and T3 is open, terminal T4 is insulated, therefore current I=0, and V_T2=V_OUT=V_DDfor all applied voltages.
• Operation of Field Effect Devices with Controllable Gates
• Four schematics of field effect devices (FEDs) with controllable gates have been described (FIGS.21-L).FIGS. 13A-16 illustrates the operation of field effect devices with controllable gates for one of the FED configurations, FED11 device240 (FIG. 2L). FED devices with controllable gates are also referred to as NT-on-Gate devices. A switch-mode setting operation is described, followed by an example of full voltage swing circuit operation (digital switching).
• FIG. 13A illustratesFED11240.FED11240 is combined withresistor886 of value R, such that one terminal ofresistor886 is attached toFED11 device240 terminal T2, and the other side ofresistor886 is attached topower supply terminal884 to form circuit schematic.FED11240 terminal T2 is connected to FET drain244; terminal T3 is connected toFET source246; terminal T4 is connected to releaseplate254.FIG. 13B illustrates circuit schematic390 in which switch250 has been activated to position250′ to electrically connect switch-plate248 to contact252.Controllable gate242 is electrically connected to terminal T1 by means of the established continuous electrical path formed bygate242 connected to switch-plate248; switch-plate248 connected to one side ofswitch250′; the opposite side ofswitch250′ connected to contact252; contact252 connected to terminal T1. The combination ofcontact252 area and NTfabric layer switch250 area may be referred to as the NT control gate, because the voltage applied to this control gate controls theFET channel region247 electrical characteristics.
• FIG. 13C illustrates circuit schematic390′ in which switch250 has been activated to position250″ to contact release-plate dielectric surface256.Controllable gate242 is electrically open circuited, and has no continuous electrical path to any FED249 device terminals.
• FIG. 13A also depicts aFED11240 with the coupling capacitances both inherent in the device and designed for the device, and corresponds toFIG. 14 which illustratescross section400 of theFED11240. Capacitance C_1Gis the capacitance betweencontact252 and switch250 combined areas (i.e., nanotube fabric-based switch250) and switch-plate248 area that connects topolysilicon gate242 using connecting contact (connecting stud, for example)243. C_G-CHis the capacitance between thepolysilicon gate242 and the channel region247 (FET gate oxide capacitance). C_CH-SUBis the depletion capacitance, in depletedregion402, between thechannel region247 andsubstrate382. Thesubstrate382 voltage is controlled usingsubstrate contact383, and is at zero volts in this example.Source diffusion246 is connected toFED11240 terminal T3, anddrain diffusion244 is connected toFED11240 terminal T2. The nanotube (NT)fabric layer switch250 is mechanically supported at both ends. Contact252 acts as both electrical contact and mechanical support, andsupport253 provides the other mechanical support (support253 may also provide an additional electrical connection as well) as illustrated inFIG. 14.
• Switch250 inclosed position250′ (FIG. 13B) is illustrated by the deflected NT fabric layer in contact with switch-plate248. The closed position is the “ON” state, thepolysilicon gate242 is in contact with the nanotube fabric layer switch250 (i.e., it is not floating) bycontact243. The polysilicon gate voltage is defined by the voltage of the nanotube control gate. The nanotube control gate includes thecontact252 area and the NT fabric-basedswitch250 area (not drawn to scale).
• Switch250 inopen position250″ is illustrated by the deflected NT fabric layer in contact withsurface256 ofinsulator404.FED11 device240 terminal T4 is connected to release-plate254 withinsulator404. The open position is the “OFF” state, the polysilicon gate is not in contact with the nanotube control gate. Thus, the polysilicon gate voltage floats, and the floating gate (FG) voltage has a value that depends on the capacitance coupling network in the device. The value of diffusion capacitance C_CH-SUBcan be modulated by the voltage applied to the drain244 (source246 may float, or may be at the voltage applied to drain244), and may be used to set the floating gate (FG) voltage whenswitch250 is inopen position250″. However, as used during write, drain244 voltage (V_DRAIN=0) and C_CH-SUBis not part of the network, and voltage V_T1is used to set the state ofswitch250. The principle of FET channel modulation using drain voltage is illustrated in U.S. Pat. No. 6,369,671.
• If voltage ondrain244 equals zero (V_DRAIN=0), thechannel247 remains as an inverted region, and capacitor C_CH-SUBis not part of the capacitor network. Capacitor C_G-CHholdspolysilicon gate242 at a relatively low voltage, which is transmitted to switchplate248 bycontact243. Therefore, a relatively high voltage appears betweenswitch250 and switchingplate248, across capacitor C_1G, and nanotubefabric layer switch250 switches from open (“OFF”)position250″ to closed (“ON”)position250′.
• FIG. 15 illustrates mode-setting electrical signals applied to the terminals T1, T2, T3, and T4 ofschematics380,390, and390′ to causeswitch250 to switch toposition250′ orposition250″.FIG. 15 illustrates the operational mode-settingwaveforms375 applied to terminals T1, T2, T3, and T4 ofFED11240 to activateswitch250. Control signals are applied to terminals T1-T4 by a control circuit (not shown) using control lines (not shown). There is no electrical signal applied toelectrical terminal884 during mode-setting.Column1 ofFIG. 15 illustrates the electrical signals used to changeswitch250 fromposition250″, also referred to as the open (“OFF”) position, to position250′, also referred to as the closed (“ON”) position.Column2 illustrates the electrical signals used to changeswitch250 fromposition250′, also referred to as the closed (“ON”) position, to position250″, also referred to as the open (“OFF”) position. The mode-setting waveforms are valid within the mode-setting time interval illustrated undercolumns1 and2 inFIG. 15. Other time intervals contain cross-hatched lines betweenvoltages0 and V_DD, indicating that these waveforms can be anywhere within this voltage range, and represent the circuit operating range. V_DDis selected to be less than the voltage switching voltage V_SWto ensure thatswitch250 is not activated (resulting in mode-resetting) during circuit operation.
• Mode-setting is based on electromechanical switching of carbon nanotube (NT) switch using electrostatic forces.Column1 ofFIG. 15 illustrates the voltage and timing waveforms applied to terminals T1-T4 ofFED11240 that force a transition of NT switch250 fromposition250″, in contact withinsulator surface256 on release-plate254 as illustrated inFIGS. 13C and 14A, to position250′, in contact with switch-plate248 as illustrated inFIGS. 13B and 14A. Switch250 transitions from open to closed. Voltage V_T4, applied to terminal T4, transitions to switching voltage V_SW. Voltage V_T2applied to terminal T2 transitions to zero. V_T3applied to terminal T3 transitions to zero volts. Terminal T1 (connected toNT fabric switch250 through control gate contact252) transitions from zero to switching voltage V_SWforming a channel inchannel region247. The electrostatic force betweenswitch250 inposition250″ and release-plate254 is zero. The electrostatic force betweenswitch250 inposition250″ and switch-plate248 is F_E=K (V_SW−V_G)²/(R₁₂)², where R₁₂is thegap separating switch250 from switch-plate248. V_Gis determined by the relative values of capacitances C_1Gand C_G-CH(FIG. 14). C_1Gis typically designed to be 0.25 times the capacitance C_G-CH(C_1G=0.25 C_G-CH). Gate voltage V_G=V_SW×C_1G/(C_1G+C_G-CH); V_G=0.2 V_SW. If the voltage difference required betweenswitch250 and switch-plate248 to activateswitch250 is 2.5 volts, for example, then switching voltage V_SWgreater than approximately 3.2 volts is required.
• Column2 ofFIG. 15 illustrates the voltage and timing waveforms applied to terminals T1-T4 ofFED11240 that force a transition of NT switch250 fromposition250′, in contact with switch-plate248 as illustrated inFIGS. 13B and 14A, to position250″, in contact with release-plate dielectric surface256 on release-plate254 as illustrated inFIG. 13C. Switch250 transitions from closed to open. Voltage V_T4, applied to terminal T4, transitions to switching voltage V_SW. Voltage V_T2applied to terminal T2 transitions is between zero and 1 volt (as high as V_DDis acceptable). V_T3applied to terminal T3 transitions to zero to 1 volt (as high as V_DDis acceptable). Terminal T1 (connected to NT switch250 by contact252) transitions to zero volts. The electrostatic force betweenswitch250 inposition250′ and switch-plate248 is zero. The electrostatic force betweenswitch250 inposition250′ and release-plate254 is F_E=K (V_SW)²/(R₁₂)², where R₁₂is thegap separating switch250 from release-plate254. Typical V_NT-THvoltages may range from 2 to 3 volts, for example. The threshold voltage forswitch250 transitions between open (“OFF”) and closed (“ON”), and closed (“ON”) and open (“OFF”) positions may be different, without effecting the operation of the device. If V_SWexceeds V_NT-TH, then mode-setting will take place. Circuit operating voltages range from 0 to V_DD. In order to avoid unwanted mode-setting during circuit operation, V_DDis less than V_NT-TH.
• The threshold voltage V_FET-THof the FET device withgate242, drain244, andsource246 that forms a portion of FED II240 is modulated by the position ofNT fabric switch250.FIG. 16 illustrates the current—voltage (I-V)characteristic385 ofFED11240 forswitch250 in the closed (“ON”) state (switch250 inposition250′) and the open (“OFF”) state (switch250 inposition250″). Forswitch250 in the closed state, V_G=V_T1, current I flows when V_T1=V_Gis greater than FET threshold voltage V_FET-TH=0.4 to 0.7 volts. Current I flows between terminals T2 and T3 ofFED11240. Forswitch250 is in the open state, current I flows between terminals T2 and T3 ofFED11240 when V_T1is greater than 1.4 volts. At V_T1=1.4 volts, capacitive coupling raises FET gate voltage VG to greater than 0.7 volts, and current flows between terminals ofFED11240 device. The state ofFED11240 device may be detected by applying V_T1voltage of 1.2 volts. IfFED11240 is in the closed state (also referred to as the written or programmed state), then current I will flow when V_T1=1.2 volts. IfFED11240 is in the open state (also referred to as the released or erased state), then no current (I=0) will flow when V_T1=1.2 volts.
Nanotube Random Access Memory Using FEDs with Controllable Sources
• Nanotube Random Access Memory (NRAM) Systems and Circuits, with Same
• Non-volatile field effect devices (FEDs)20,40,60, and80 with controllable sources may be used as cells and interconnected into arrays to form non-volatile nanotube random access memory (NRAM) systems. The memory cells contain one select device (transistor) T and one non-volatile nanotube storage element NT (1T/1NT cells). By way of example, FED480 (FIG. 2D) is used to form a non-volatile NRAM memory cell that is also referred to as a NT-on-Source memory cell.
• NT-on-Source NRAM Memory Systems and Circuits with Parallel Bit and Reference Lines, and Parallel Word and Release Lines
• NRAM 1T/1NT memory arrays are wired using four lines. Word line WL is used to gate select device T, bit line BL is attached to a shared drain between two adjacent select devices. Reference line REF is used to control the NT switch voltage of storage element NT, and release line RL is used to control the release-plate of storage element NT. In this NRAM array configuration, REF is parallel to BL and acts as second bit line, and RL is parallel to WL and acts as a second word line. The NT-on-source with REF line parallel to BL and RL parallel WL is the preferred NT-on-source embodiment.
• FIG. 17A depicts non-volatile fieldeffect device FED480 with memory cell wiring to form NT-on-Source memory cell1000 schematic.Memory cell1000 operates in a source-follower mode. Word line (WL)1200 connects toterminal T11220 of FED480; bit line (BL)1300 connects toterminal T21320 of FED480; reference line (REF)1400 connects toterminal T31420 of FED480; and release line (RL)1500 connects toterminal T41520 of FED480.Memory cell1000 performs write and read operations, and stores the information in a non-volatile state. TheFED480 layout dimensions and operating voltages are selected to optimizememory cell1000.Memory cell1000 FET select device (T)gate1040 corresponds togate82;drain1060 corresponds to drain84; andcontrollable source1080 corresponds tocontrollable source86.Memory cell1000 nanotube (NT) switch-plate1120 corresponds to switch-plate88;NT switch1140 corresponds toNT switch90; release-plateinsulator layer surface1160 corresponds to release-plateinsulator layer surface96; and release-plate1180 corresponds to release-plate94. The interconnections between the elements ofmemory cell1000 schematic correspond to the interconnection of the corresponding interconnections of the elements ofFED480.BL1300 connects to drain1060 throughcontact1320;REF1400 connects toNT switch1140 throughcontact1420;RL1500 connects to release-plate1180 bycontact1520;WL1200 interconnects togate1040 bycontact1220. The non-volatileNT switching element1140 may be caused to deflect toward switch-plate1120 via electrostatic forces to closed (“ON”)position1140′ to store a logic “1” state as illustrated inFIG. 17B. The van der Waals force holdsNT switch1140 inposition1140′. Alternatively, the non-volatileNT switching element1140 may be caused to deflect toinsulator surface1160 on release-plate1180 via electrostatic forces to open (“OFF”)position1140″ to store a logic “0” state as illustrated inFIG. 17C. The van der Waals force holdsNT switch1140 inposition1140″. Non-volatileNT switching element1140 may instead be caused to deflect to an open (“OFF”) near-mid point position1140′″ between switch-plate1120 and release-plate1180, storing an apparent logic “0” state as illustrate inFIG. 17D. However, the absence of a van der Waals retaining force in this open (“OFF”) position is likely to result in a memory cell disturb that causesNT switch1140 to unintentionally transition to the closed (“ON”) position, and is not desirable. Sufficient switching voltage is needed to ensure that theNT switch1140 open (“OFF”) position isposition1140″. The non-volatile element switching via electrostatic forces is as depicted byelement90 inFIG. 2D.Voltage waveforms311 used to generate the required electrostatic forces are illustrated inFIG. 4.
• NT-on-Source schematic1000 forms the basis of a non-volatile storage (memory) cell. The device may be switched between closed storage state “1” (switched to position1140′) and open storage state “0” (switched toposition1140″), which means the controllable source may be written to an unlimited number of times to as desired. In this way, the device may be used as a basis for a non-volatile nanotube random access memory, which is referred to here as a NRAM array, with the ‘N’ representing the inclusion of nanotubes.
• FIG. 18 represents anNRAM memory array1700, according to preferred embodiments of the invention. Under this arrangement, an array is formed with m×n (only exemplary portion being shown) of non-volatile cells ranging from cell C0,0 to cell Cm−1,n−1.NRAM memory array1700 may be designed using one large m×n array, or several smaller sub-arrays, where each sub-array if formed of m×n cells. To access selected cells, the array uses read and write word lines (WL0, WL1, . . . WLn−1), read and write bit lines (BL0, BL1, . . . BLm−1), read and write reference lines (REF0, REF1, . . . REFm−1), and read and write release lines (RL0, RL1, . . . RLn−1). Non-volatile cell C0,0 includes a select device T0,0 and non-volatile storage element NT0,0. The gate of T0,0 is coupled to WL0, and the drain of T0,0 is coupled to BL0. NT0 is the non-volatilely switchable storage element where the NT0,0 switch-plate is coupled to the source of T0,0, the switching NT element is coupled to REF0, and the release-plate is coupled to RL0.Connection1720 connects BL0 to shared drain of select devices T0,0 and T0,1. Word, bit, reference, and release decoders/drivers are explained further below.
• Under preferred embodiments, nanotubes inNRAM array1700 may be in the “ON” “1” state or the “OFF” “0” state. The NRAM memory allows for unlimited read and write operations per bit location. A write operation includes both a write function to write a “1” and a release function to write a “0”. By way of example, a write “1” to cell C0,0 and a write “0” to cell C1,0 is described. For a write “1” operation to cell C0,0, select device T0,0 is activated when WL0 transitions from 0 to V_DD, BL0 transitions from V_DDto 0 volts, REF0 transitions from V_DDto switching voltage V_SW, and RL0 transitions from V_DDto switching voltage V_SW. The release-plate and NT switch of the non-volatile storage element NT0,0 are each at V_SWresulting in zero electrostatic force (because the voltage difference is zero). The zero BL0 voltage is applied to the switch-plate of non-volatile storage element NT0,0 by the controlled source of select device T0,0. The difference in voltage between the NT0,0 switch-plate and NT switch is V_SWand generates an attracting electrostatic force. If V_SWexceeds the nanotube threshold voltage V_NT-TH, the nanotube structure switches to “ON” state or logic “1” state, that is, the nanotube NT switch and switch-plate are electrically connected as illustrated inFIG. 17B. The near-Ohmic connection between switch-plate1120 andNT switch1140 inposition1140′ represents the “ON” state or “1” state. If the power source is removed, cell C0,0 remains in the “ON” state.
• For a write “0” (release) operation to cell C1,0, select device T1,0 is activated when WL0 transitions from 0 to V_DD, BL1 transitions from V_DDto 0 volts,REF1 transitions from V_DDto zero volts, and RL0 transitions from V_DDto switching voltage V_SW. The zero BL1 voltage is applied to the switch-plate of non-volatile storage element NT1,0 by the controlled source of select device T1,0, and zero volts is applied the NT switch by REF1, resulting in zero electrostatic force between switch-plate and NT switch. The non-volatile storage element NT1,0 release-plate is at switching voltage V_SWand the NT switch is at zero volts generating an attracting electrostatic force. If V_SWexceeds the nanotube threshold voltage V_NT-TH, the nanotube structure switches to the “OFF” state or logic “0” state, that is, the nanotube NT switch and the surface of the release-plate insulator are in contact as illustrated inFIG. 17C. The non-conducting contact betweeninsulator surface1160 on release-plate1180 andNT switch1140 inposition1140″ represents the “OFF” state or “0” state. If the power source is removed, cell C1,0 remains in the “OFF” state.
• An NRAM read operation does not change (destroy) the information in the activated cells, as it does in a DRAM, for example. Therefore the read operation in the NRAM is characterized as a non-destructive readout (or NDRO) and does not require a write-back after the read operation has been completed. For a read operation of cell C0,0, BL0 is driven high to V_DDand allowed to float. WL0 is driven high to V_DDand select device T0,0 turns on. REF0 is at zero volts, and RL0 is at V_DD. If cell C0,0 stores an “ON” state (“1” state) as illustrated inFIG. 17B, BL0 discharges to ground through a conductive path that includes select device T0,0 and non-volatile storage element NT0,0 in the “ON” state, the BL0 voltage drops, and the “ON” state or “1” state is detected by a sense amplifier/latch circuit (not shown) that records the voltage drop by switching the latch to a logic “1” state. BL0 is connected by the select device T0,0 conductive channel of resistance R_FETto the switch-plate of NT0,0. The switch-plate of NT0,0 in the “ON” state contacts the NT switch with contact resistance and the NT switch contacts reference line REF0 with contact resistance R_C. The total resistance in the discharge path is R_FET+R_SW+R_C. Other resistance values in the discharge path, including the resistance of the NT switch, are much smaller and may be neglected.
• For a read operation of cell C1,0, BL1 is driven high to V_DDand allowed to float. WL0 is driven high to V_DDand select device T1,0 turns on. REF1=0, and RL0 is at V_DD. If cell C1,0 stores an “OFF” state (“0” state) as illustrated inFIG. 17C, BL1 does not discharge to ground through a conductive path that includes select device T1,0 and non-volatile storage element NT1,0 in the “OFF” state, because the switch-plate is not in contact with the NT switch when NT1,0 is in the “OFF” state, and the resistance R_SWis large. Sense amplifier/latch circuit (not shown) does not detect a voltage drop and the latch is set to a logic “0” state.
• FIG. 19 illustrates theoperational waveforms1800 ofNRAM memory array1700 ofFIG. 18 during read, write “1”, and write “0” operations for selected cells, while not disturbing unselected cells (no change to unselected cell-stored logic states).Waveforms1800 illustrate voltages and timings to write logic state “1” in cell C0,0, write a logic state “0” in cell C1,0, read cell C0,0, and read cell C1,0.Waveforms1800 also illustrate voltages and timings to prevent disturbing the stored logic states (logic “1” state and logic “0” state) in partially selected (also referred to as half-selected) cells. Partially selected cells are cells inmemory array1700 that receive applied voltages because they are connected to (share) word, bit, reference, and release lines that are activated as part of the read or write operation to the selected cells. Cells inmemory array1700 tolerate unlimited read and write operations at each memory cell location.
• At the start of the write cycle, WL0 transitions from zero to V_DD, activating select devices T0,0, T1,0, . . . Tm−1,0. Word lines WL1, WL2 . . . WLn−1 are not selected and remain at zero volts. BL0 transitions from V_DDto zero volts, connecting the switch-plate of non-volatile storage element NT0,0 to zero volts. BL1 transitions from V_DDto zero volts connecting the switch-plate of non-volatile storage element NT1,0 to zero volts. BL2, BL3 . . . BLm−1 remain at V_DDconnecting the switch-plate of non-volatile storage elements NT2,0, NT3,0, . . . NTm−1,0 to V_DD. REF0 transitions from V_DDto switching voltage V_SW, connecting the NT switches of non-volatile storage elements NT0,0, NT0,1, NT0,n−2, NT0,n−1 to V_SW. REF1 transitions from V_DDto zero volts, connecting the NT switches of non-volatile storage elements NT1,0, NT1,1 . . . NT1,n−2,NT1,n−1 to zero volts. REF2, REF3, . . . REFm−1 remain at V_DD, connecting the NT switches of non-volatile storage elements NT3,0 to NTm−1,n−1 to V_DD. REL0 transitions from V_DDto switching voltage V_SW, connecting release-plates of non-volatile storage elements NT0,0, NT1,0, . . . NTm−1,0 to V_SW. RL1, RL2 . . . . RLn−1 remain at V_DD, connecting release-plates of non-volatile storage elements NT0,1 to NTn−1,n−1 to V_DD.
• NT0,0 may be in “ON” (“1” state) or “OFF” (“0” state) state at the start of the write cycle. It will be in “ON” state at the end of the write cycle. If NT0,0 in cell C0,0 is “OFF” (“0” state) it will switch to “ON” (“1” state) since the voltage difference between NT switch and release-plate is zero, and the voltage difference between NT switch and switch-plate is V_SW. If NT0,0 in cell C0,0 is in the “ON” (“1” state), it will remain in the “ON” (“1”) state. NT1,0 may be in “ON” (“1” state) or “OFF” (“0” state) state at the start of the write cycle. It will be in “OFF” state at the end of the write cycle. If NT1,0 in cell C1,0 is “ON” (“1” state) it will switch to “OFF” (“0” state) since the voltage difference between NT switch and switch-plate is zero, and the voltage difference between NT switch and release-plate is V_SW. If NT1,0 in cell C1,0 is “OFF” (“0” state), it will remain “OFF” (“0” state). If for example, V_SW=3.0 volts, V_DD=1.5 volts, and NT switch threshold voltage range is VN_T-TH=1.7 to 2.8 volts, then for NT0,0 and NT1,0 a difference voltage V_SW>V_NT-THensuring write states of “ON” (“1” state) for NT0,0 and “OFF” (“0” state) for NT1,0.
• Cells C0,0 and C1,0 have been selected for the write operation. All other cells have not been selected, and information in these other cells must remain unchanged (undisturbed). Since in an array structure some cells other than selected cells C0,0 and C1,0 inarray1700 will experience partial selection voltages, often referred to as half-select voltages, it is necessary that half-select voltages applied to non-volatile storage element terminals be sufficiently low (below nanotube activation threshold V_NT-TH) to avoid disturbing stored information. For storage cells in the “ON” state, it is also necessary to avoid parasitic current flow (there cannot be parasitic currents for cells in the “OFF” state because the NT switch is not in electrical contact with switch-plate or release-plate). Potential half-select disturb along activated array lines WL0 and RL0 includes cells C3,0 to Cm−1,0 because WL0 and RL0 have been activated. Storage elements NT3,0 to NTm−1,0 will have BL2 to BLm−1 electrically connected to the corresponding storage element switch-plate by select devices T3,0 to Tm−1,0. All release-plates in these storage elements are at write voltage V_SW. To prevent undesired switching of NT switches, REF2 to REFm−1 reference lines are set at voltage V_DD. BL2 to BLm−1 voltages are set to V_DDto prevent parasitic currents. The information in storage elements NT2,0 to NTm−1,0 in cells C2,0 to Cm−1,0 is not disturbed and there is no parasitic current. For those cells in the “OFF” state, there can be no parasitic currents (no current path), and no disturb because the voltage differences favor the “OFF” state. For those cells in the “ON” state, there is no parasitic current because the voltage difference between switch-plates (at V_DD) and NT switches (at V_DD) is zero. Also, for those cells in the “ON” state, there is no disturb because the voltage difference between corresponding NT switches and release-plate is V_SW−V_DD=1.5 volts, when V_SW=3.0 volts and V_DD=1.5 volts. Since this voltage difference of 1.5 volts is less than the minimum nanotube threshold voltage V_NT-THof 1.7 volts, no switching takes place.
• Potential half-select disturb along activated array lines REF0 and BL0 includes cells C0,1 to C0, n−1 because REF0 and BL0 have been activated. Storage elements NT0,1 to NT0, n−1 all have corresponding NT switches connected to switching voltage V_SW. To prevent undesired switching of NT switches, RL1 to RLn−1 are set at voltage V_DD. WL1 to WL n−1 are set at zero volts, therefore select devices T0,1 to T0,n−1 are open, and switch-plates (all are connected to select device source diffusions) are not connected to bit line BL0. All switch-plates are in contact with a corresponding NT switch for storage cells in the “ON” state, and all switch plates are only connected to corresponding “floating” source diffusions for storage cells in the “OFF” state. Floating diffusions are at approximately zero volts because of diffusion leakage currents to semiconductor substrates. However, some floating source diffusions may experience disturb voltage conditions that may cause the source voltage, and therefore the switch-plate voltage, to increase up to 0.6 volts as explained further below. The information in storage elements NT0,1 to NT0,n−1 in cells C0,1 to C0,n−1 is not disturbed and there is no parasitic current. For cells in both “ON” and “OFF” states there can be no parasitic current because there is no current path. For cells in the “ON” state, the corresponding NT switch and switch-plate are in contact and both are at voltage V_SW. There is a voltage difference of V_SW−V_DDbetween corresponding NT switch and release-plate. For V_SW=3.0 volts and V_DD=1.5 volts, the voltage difference of 1.5 volts is below the minimum V_NT-TH=1.7 volts for switching. For cells in the “OFF” state, the voltage difference between corresponding NT switch and switch-plate ranges from V_SWto V_SW−0.6 volts. The voltage difference between corresponding NT switch and switch-plate may be up to 3.0 volts, which exceeds the V_NT-THvoltage, and would disturb “OFF” cells by switching them to the “ON” state. However, there is also a voltage difference between corresponding NT switch and release-plate of V_SW−V_DDof 1.5 volts with an electrostatic force in the opposite direction that prevents the disturb of storage cells in the “OFF” state. Also very important is thatNT1140 is inposition1140″ in contact with the storage-plate dielectric, a short distance from the storage plate, thus maximizing the electric field that opposes cell disturb. Switch-plate1140 is far from theNT1140 switch greatly reducing the electric field that promotes disturb. In addition, the van der Waals force also must be overcome to disturb the cell.
• Potential half-select disturb along activated array lines REF1 and BL1 includes cells C1,1 to C1, n−1 because REF1 and BL1 have been activated. Storage elements NT1,1 to NT1, n−1 all have corresponding NT switches connected to zero volts. To prevent undesired switching of NT switches, RL1 to RLn−1 are set at voltage V_DD. WL1 to WL n−1 are set at zero volts, therefore select devices T1,1 to T1,n−1 are open, and switch-plates (all are connected to select device source diffusions) are not connected to bit line BL1. All switch-plates are in contact with a corresponding NT switch for storage cells in the “ON” state, and all switch plates are only connected to corresponding “floating” source diffusions for storage cells in the “OFF” state. Floating diffusions are at approximately zero volts because of diffusion leakage currents to semiconductor substrates. However, some floating source diffusions may experience disturb voltage conditions that may cause the source voltage, and therefore the switch-plate voltage, to increase up to 0.6 volts as explained further below. The information in storage elements NT1,1 to NT1,n−1 in cells C1,1 to C1,n−1 is not disturbed and there is no parasitic current. For cells in both “ON” and “OFF” states there can be no parasitic current because there is no current path. For cells in the “ON” state, the corresponding NT switch and switch-plate are in contact and both are at zero volts. There is a voltage difference of V_DDbetween corresponding NT switch and release-plate. For V_DD=1.5 volts, the voltage difference of 1.5 volts is below the minimum V_NT-TH=1.7 volts for switching. For cells in the “OFF” state, the voltage of the switch-plate ranges zero to 0.6 volts. The voltage difference between corresponding NT switch and switch-plate may be up to 0.6 volts. There is also a voltage difference between corresponding NT switch and release-plate of V_DD=1.5 volts. V_DDis less than the minimum V_NT-THof 1.7 volts the “OFF” state remains unchanged.
• For all remainingmemory array1700 cells, cells C2,1 to Cm−1,n−1, there is no electrical connection between NT2,1 to NTm−1,n−1 switch-plates connected to corresponding select device source and corresponding bit lines BL2 to BLm−1 because WL1 to WLn−1 are at zero volts, and select devices T2,1 to Tm−1,n−1 are open. Reference line voltages for REF2 to REFm−1 are set at V_DDand release line voltages for RL1 to RLn−1 are set at V_DD. Therefore, all NT switches are at V_DDand all corresponding release-plates are at V_DD, and the voltage difference between corresponding NT switches and release-plates is zero. For storage cells in the “ON” state, NT switches are in contact with corresponding switch-plates and the voltage difference is zero. For storage cells in the “OFF” state, switch plate voltages are zero to a maximum of 0.6 volts. The maximum voltage difference between NT switches and corresponding switch-plates is V_DD=1.5 volts, which is below the V_NT-THvoltage minimum voltage of 1.7 volts. The “ON” and “OFF” states remain undisturbed.
• Non-volatile NT-on-sourceNRAM memory array1700 with bit lines parallel to reference lines is shown inFIG. 18 contains 2^N×2^Mbits, is a subset of non-volatileNRAM memory system1810 illustrated asmemory array1815 inFIG. 20A.NRAM memory system1810 may be configured to operate like an industry standard asynchronous SRAM or synchronous SRAM because nanotubenon-volatile storage cells1000 shown inFIG. 17A, inmemory array1700, may be read in a non-destructive readout (NDRO) mode and therefore do not require a write-back operation after reading, and also may be written (programmed) at CMOS voltage levels (5, 3.3, and 2.5 volts, for example) and at nanosecond and sub-nanosecond switching speeds. NRAM read and write times, and cycle times, are determined by array line capacitance, and are not limited by nanotube switching speed. Accordingly,NRAM memory system1810 may be designed with industry standard SRAM timings such as chip-enable, write-enable, output-enable, etc., or may introduce new timings, for example. Non-volatileNRAM memory system1810 may be designed to introduce advantageous enhanced modes such as a sleep mode with zero current (zero power—power supply set to zero volts), information preservation when power is shut off or lost, enabling rapid system recovery and system startup, for example.NRAM memory system1810 circuits are designed to provide thememory array1700waveforms1800 shown inFIG. 19.
• NRAM memory system1810 accepts timinginputs1812, acceptsaddress inputs1825, and acceptsdata1867 from a computer, or providesdata1867 to a computer using a bidirectional bus sharing input/output (I/O) terminals. Alternatively, inputs and outputs may use separate (unshared) terminals (not shown). Address input (I/P)buffer1830 receives address locations (bits) from a computer system, for example, and latches the addresses. Address I/P buffer1830 provides word address bits toword decoder1840 viaaddress bus1837; address I/P buffer1830 provides bit addresses to bit decoder1850 viaaddress bus1852; and address bus transitions provided bybus1835 are detected by function generating, address transition detecting (ATD), timing waveform generator, controller (controller)1820.Controller1820 provides timing waveforms onbus1839 toword decoder1840.Word decoder1840 selects the word address location withinarray1815.Word address decoder1840 is used to decode both word lines WL and corresponding release lines RL (there is no need for a separate RL decoder) and drives word line (WL) and release line (RL)select logic1845.Controller1820 provides function and timing inputs onbus1843 to WL & RLselect logic1845, resulting inNRAM memory system1810 on-chip WL and RL waveforms for both write-one, write-zero, read-one, and read-zero operations as illustrated bywaveforms1800′ shown inFIG. 21.FIG. 21NRAM memory system1810waveforms1800′ correspond tomemory array1700waveforms1800 shown inFIG. 19.
• Bit address decoder1850 is used to decode both bit lines BL and corresponding reference lines REF (there is no need for a separate REF decoder) and drive bit line (BL) and reference (REF)select logic1855 viabus1856.Controller1820 provides timing waveforms onbus1854 tobit decoder1850.Controller1820 also provides function and timing inputs onbus1857 to BL & REF selectlogic1855. BL & REF selectlogic1855 uses inputs frombus1856 andbus1857 to generate data multiplexer select bits onbus1859. The output of BL and REFselect logic1855 onbus1859 is used to select control data multiplexers using combined data multiplexers & sense amplifiers/latches (MUXs & SAs)1860.Controller1820 provides function and timing inputs onbus1862 to MUXs &SAs1860, resulting inNRAM memory system1810 on-chip BL and REF waveforms for both write-one, write-zero, read-one, and read-zero operations as illustrated bywaveforms1800′ corresponding tomemory array1700waveforms1800 shown inFIG. 19. MUXs &SAs1860 are used to write data provided by read/write buffer1865 viabus1864 inarray1815, and to read data fromarray1815 and provide the data to read/write buffer1865 viabus1864 as illustrated inwaveforms1800′.
• Sense amplifier/latch1900 is illustrated inFIG. 20B.Flip flop1910, comprising two back-to-back inverters is used to amplify and latch data inputs fromarray1815 or from read/write buffer1865.Transistor1920 connectsflip flop1910 to ground when activated by a positive voltage supplied bycontrol voltage V_TIMING1980, which is provided bycontroller1820.Gating transistor1930 connects a bit line BL tonode1965 offlip flop1910 when activated by a positive voltage.Gating transistor1940 connects reference voltage V_REFto flipflop node1975 when activated by a positive voltage.Transistor1960 connects voltage V_DDto flipflop1910node1965,transistor1970 connects voltage V_DDto flipflop1910node1975, andtransistor1950 ensures that small voltage differences are eliminated whentransistors1960 and1970 are activated.Transistors1950,1960, and1970 are activated (turned on) when gate voltage is low (zero, for example).
• In operation, V_TIMINGvoltage is at zero volts whensense amplifier1900 is not selected.NFET transistors1920,1930, and1940 are in the “OFF” (non-conducting) state, because gate voltages are at zero volts.PFET transistors1950,1960, and1970 are in the “ON” (conducting) state because gate voltages are at zero volts. V_DDmay be 5, 3.3, or 2.5 volts, for example, relative to ground.Flip flop1910nodes1965 and1975 are at V_DD. If sense amplifier/latch1900 is selected, V_TIMINGtransitions to V_DD,NFET transistors1920,1930, and1940 turn “ON”,PFET transistors1950,1960, and1970 are turned “OFF”, andflip flop1910 is connected to bit line BL and reference voltage V_REF. V_REFis connected to V_DDin this example. As illustrated by waveforms BL0 and BL1 ofwaveforms1800′, bit line BL is pre-charged prior to activating a corresponding word line (WL0 in this example). Ifcell1000 of memory array1700 (memory system array1815) stores a “1”, then bit line BL inFIG. 20B corresponds to BL0 inFIG. 21, BL is discharged bycell1000, voltage droops below V_DD, and sense amplifier/latch1900 detects a “1” state. Ifcell1000 of memory array1700 (memory system array1815) stores a “0”, then bit line BL inFIG. 20B corresponds to BL1 inFIG. 21, BL is not discharged bycell1000, voltage does not droop below V_DD, and sense amplifier/latch1900 detect a “0” state. The time from sense amplifier select to signal detection by sense amplifier/latch1900 is referred to as signal development time. Sense amplifier/latch1900 typically requires 100 to 200 mV relative to V_REFin order to switch. It should be noted thatcell1000 requires a nanotube “OFF” resistance to “ON” resistance ratio of greater than about 10 to 1 for successful operation. A typical bit line BL has a capacitance value of 250 fF, for example. A typical nanotube storage device (switch) or dimensions 0.2 by 0.2 um typically has 8 nanotube filaments across the suspended region, for example, as illustrated further below. For a combined contact and switch resistance of 50,000 Ohms per filament, as illustrated further below, the nanotube “ON” resistance ofcell1000 is 6,250 Ohms. For a bit line of 250 fF, the time constant RC=1.6 ns. The sense amplifier signal development time is less than RC, and for this example, is between 1 and 1.5 nanoseconds.
• Non-volatileNRAM memory system1810 operation may be designed for high speed cache operation at 5 ns or less access and cycle time, for example.
• Non-volatileNRAM memory system1810 may be designed for low power operation at 60 or 70 ns access and cycle time operation, for example. For low power operation, address I/P buffer1830 operation requires 8 ns;controller1820 operation requires 16 ns;bit decoder1850 operation plus BL &select logic1855 plus MUXs &SA1860 operation requires 12ns (word decoder1840 operation plus WL & RLselect logic 1845 ns require less than 12 ns);array1815 delay is 8 ns; sensing1900 operation requires 8 ns; and read/write buffer1865 requires 12 ns, for example. The access time and cycle time of non-volatileNRAM memory system1810 is 64 ns. The access time and cycle time may be equal because the NDRO mode of operation of nanotube storage devices (switches) does not require a write-back operation after access (read).
• Method of Making Field Effect Device with Controllable Source and NT-on-Source Memory System and Circuits with Parallel Bit and Reference Array Lines and Parallel Word and Release Array Lines
• Non-volatile field effect devices (FEDs)20,40,60, and80 with controllable sources may be used as cells and interconnected into arrays to form non-volatile nanotube random access memory (NRAM) systems. The memory cells contain one select device (transistor) T and one non-volatile nanotube storage element NT (1T/1NT) cells). By way of example, FED480 (FIG. 2D) devices are fabricated and interconnected to form a non-volatile NRAM memory cell that is also referred to as a NT-on-Source memory cell with parallel bit and reference array lines, and parallel word and release array lines.
• FIG. 22 describes thebasic method3000 of manufacturing preferred embodiments of the invention. The following paragraphs describe such method in specific relation to an NRAM NT-on-source structure. However, this method is sufficient to cover the manufacturer of all the preferred field effect devices described.
• In general, preferred methods first form3002 a field effect device similar to a MOSFET, having drain, source, and gate nodes. Such a structure may be created with known techniques and thus is not described here. Such a structure defines a base layer on which a nanotube control structure may be created.
• Once the semiconductor structure is defined in the substrate, preferred methods then3004 a lower carbon nanotube intermediate control structure having nanotube electromechanical, non-volatile switches.FIGS. 24A, 24B,24C,24D, and24E depict five exemplary structures that are NT-on-source devices.
• FIG. 24A illustrates a cross section ofintermediate structure3103.Intermediate structure3103 includes anintermediate base structure3102′ (formed in step3002) with an intermediate nanotube control structure on top. Thebase structure3102′ includesN+ drain regions3126, and N+ dopedsource regions3124 in p-typemonocrystalline silicon substrate3128.Polysilicon gates3120 control the channel region between drain and source. Sharedconductive stud3118 contacts drain3126 incontact region3123. Contactstuds3122, one for each nanotube structure, physically and electrically connect thebase structure3102′ to the NT control structure. Specificallystud3122 connects to electrode3106 atcontact region3101, and tosource3124 at contactingregion3121.
• The NT structure is disposed over theplanar oxide region3116. The NT structure includes electrode (switch-plate)3106, a firstsacrificial gap layer3108 onelectrode3106, a nanotube fabric (porous)element3114 deposited on firstsacrificial gap layer3108, a nanotubeconductive contact layer3117 providing mechanical support (nanotube fabric element pinning betweenlayers3108 and3117) and electrical contact, andconductive layer3119 deposited onnanotube contact layer3117 for enhanced electrical conductivity, and to act as an etch mask forlayer3117. At this point, lower carbon nanotubeintermediate control structures3109 and3109′, illustrated inFIGS. 25E-25G and FIGS.25EE-25GG, respectively, have been formed. The material ofelectrode3106 may be tungsten, aluminum, copper, gold, nickel, chrome, platinum, palladium, or combinations of conductors such as chrome-copper-gold.Electrode3106 thickness is in the range of 25 to 200 nm. The material ofelectrode3106 is selected for reliable near-ohmic low contact resistance R_SWbetweenelectrode3106 andnanotube fabric layer3114, and cyclability (number or contact-release cycles) after gap formation (shown below), when switchingfabric layer3114 switches in-out-of contact withelectrode3106 during product operation. R_SWmay be in the range of 1,000 to 100,000 Ohms per contacted fiber infabric layer3114. For afabric layer3114 with 10 contacted fibers, for example, contact resistance R_SWmay be in the range of 100 to 10,000 Ohms, for example.
• Once the lower carbon nanotubeintermediate control structures3109 and3109′ are formed, then fabricate3006 an upper carbon nanotube electrode intermediate structure.Opening3136 defines the dimensions of thenanotube fabric element3114 to be suspended, including that portion of firstsacrificial gap layer3108 to be removed. The material from which nanotube fabricconductive contact layer3117 is chosen depends upon desiredelectrical contact3127 resistance R_Cproperties, such as a near-ohmic low resistance contact betweenconductor3117 andnanotube fabric element3114. Combinednanotube fabric element3114 below opening3136, and combinedelectrical conductors3117 and3119 in adjacent mechanical andelectrical contact region3127, form a low resistance R_Clocal NT toconductor contact3127 region. R_Cmay be in the range of 1,000 to 100,000 Ohms per contacted fiber infabric layer3114. For afabric layer3114 with 10 contacted fibers, for example, contact resistance R_Cmay be in the range of 100 to 10,000 Ohms, for example. This local conductor region surroundsopening3136 and may be referred to as a picture frame region, withnanotube contact layer3114 element pinned betweenconductor3117 and a portion of firstsacrificial gap layer3108 that remains in the final product structure. In a picture frame region as illustrated inFIG. 24A, each end of a fiber is electrically connected to the picture frame, such that the resistance connection to the switch is R_C/2. Combinedelectrical conductors3117 and3119 form a low resistance interconnect NT structure.
• At this stage of the method, electrode (release-plate)3205 is formed. A conformal secondsacrificial gap layer3201 deposited on patternedconductor3119, andelectrode3205 is deposited on secondsacrificial gap layer3201, planarized, and layers of material forelectrode3205 and3201 are patterned. The thickness of firstsacrificial gap layer3108 situated betweennanotube fabric layer3114 andelectrode3106 is typically in the range of 5 to 20 nm. The film thickness of secondsacrificial gap layer3201 situated betweennanotube fabric layer3114 andelectrode3205 is typically in the range of 5 to 40 nm. Film thicknesses are in the range of 100 to 200 nm, typical of 130 nm minimum dimension (half-period) semiconductor technology.Nanotube fabric layer3114 film thickness is on the order of 0.5-5 nm, for example.Nanotube fabric layer3114 minimum dimension is typically 130 nm. As will be explained below, once the sacrificial materials are removed, the suspended length of thenanotube fabric element3114 in the NT device region is on the order of 100 to 150 nm, but may be scaled to a suspended length of 20 to 40 nm, for example. The channel length betweendrain3126 andsource3124 can be on the order of 100 to 130 nm as defined bypolysilicon gate3120, but may be scaled to the 30 to 90 nm range, for example. The integrated semiconductor structure defines asurface3104′ on which the NT structure is formed.
• FIG. 24B illustrates a cross section ofintermediate structure3103′.Intermediate structure3103′ is similar tostructure3103 ofFIG. 24A, but adds additionalnanotube layer element3114 angled (non-horizontal) supports3112 (nanotube layer contact tosupports3112 is not visible in this cross sectional view).
• FIG. 24C illustrates a cross section ofintermediate structure3107.Intermediate structure3107 is similar tostructure3103 ofFIG. 24A, but has an additional insulatinglayer3203 between secondsacrificial gap layer3201 andelectrode3205. Insulatinglayer3201 thickness is typically in the range of 5 to 20 nm.Structure3107 with insulatinglayer3203 on the underside ofelectrode3205 forms a release-plate of the nanotube switch abovenanotube fabric layer3114 as discussed further below.Electrode3106 forms a switch-plate of the nanotube switch belownanotube fabric layer3114 as discussed further below.
• FIG. 24D illustrates a cross section ofintermediate structure3107′.Intermediate structure3107′ is similar tostructure3107 ofFIG. 24C, but addsadditional nanotube layer3114 element angled (non-horizontal) supports3112 (contact region is not visible in this cross sectional view).
• FIG. 24E illustrates a cross section ofintermediate structure3107X.Intermediate structure3107X is similar tostructure3107 ofFIG. 24C, except that firstsacrificial layer3108 insulator, Si₃N₄, for example, is replaced by firstsacrificial layer3108X semiconductor or conductor, silicon (Si), for example, and aninsulator border region3115, whereregion3115 may be SiO₂or Si₃N₄, for example. Firstsacrificial layer3108X dimensions correspond to the suspended region of the nanotube switch structure.Insulator border region3115 is used as part of a nanotube pinning structure (explained further below) under the nanotube fabric required to supportnanotube3114 when elongated during switching.
• FIG. 24F illustrates a cross section ofintermediate structure3107″.Intermediate structure3107″ is similar tostructure3103 ofFIG. 24A, but has an additional insulatinglayer3203′ between firstsacrificial gap layer3108 andelectrode3106. Insulatinglayer3203′ thickness is typically in the range of 5 to 20 nm.Structure3107″ with insulatinglayer3203′ on the topside ofelectrode3106 forms a release-plate of the nanotube switch belownanotube fabric3114 as discussed further below.Electrode3205 forms switch-plate of the nanotube switch abovenanotube fabric layer3114 as discussed further below. In other words, the roles of bottom and top electrodes inFIGS. 24C and 24E are reversed, however, after fabrication is completed and the nanotubes are released (gap regions are formed), both nanotube switches exhibit the same electrical operational characteristics. Fabrication methods used to fabricate the structures illustrated inFIGS. 24A-24D also may be used to fabricate structure24F, with slight modifications as discussed further below.
• FIG. 30F illustrates theintermediate structure3212, through completion ofmethod act3006.FIG. 30F showsstructure3212 much likestructure3103 inFIG. 24A which has been processed to include encapsulation over the nanotube structures in an insulator. Likewise, astructure3103′ ofFIG. 24B could be analogously encapsulated.FIG. 30F′ illustrates theintermediate structure3214, through completion ofStep3006.FIG. 30F′ showsstructure3214 much likestructure3107 inFIG. 24C which has been processed to include encapsulation over the nanotube structures in an insulator. Likewise, astructure3107′ ofFIG. 24D could be analogously encapsulated.FIG. 30FX illustrates theintermediate structure3212X, through completion ofmethod act3006.FIG. 30FX showsstructure3212X much likestructure3212 ofFIG. 30F, except that firstsacrificial layer3108 has been replaced with firstsacrificial layer3108X andco-planar border region3115.FIG. 30FX′ illustrates theintermediate structure3214X, through completion ofmethod act3006.FIG. 30FX′ showsstructure3214X much likestructure3214 ofFIG. 30F′, except that firstsacrificial layer3108 has been replaced with firstsacrificial layer3108X andco-planar border region3115. At this point, upper carbon nanotubeintermediate control structure3212 and3214 are formed. When encapsulated,FIG. 25E (not shown) is similar tostructure3214 ofFIG. 30F′, except thatinsulator layer3203 between secondsacrificial layer3201 andelectrode3205, but is instead between firstsacrificial layer3108 andelectrode3106.
• After the structure is completed through the pre-nanotube release (pre-suspend) level, preferred methods then create a gap region above and below the (carbon) nanotube element by etching to gap sacrificial layers and removing the sacrificial gap layer betweenelectrode3205 andconductor3119, and sacrificial gap layers in the NT switch region. The process of creating such a gap region is described below in connection withFIGS. 27 and 27′. Briefly, fluid communication paths are formed to the sacrificial gap material, see, e.g., opening3207′ ofFIG. 30H andopening3208′ ofFIG. 30H′. These paths are used to remove secondsacrificial gap material3201 and a segment of firstsacrificial gap material3108 of segment length defined by combinedconductor3119 and3117 opening e.g.,gap region3209A and3108A in FIGS.30K and30K′ to suspendsegment3114A ofnanotube elements3114. Alternatively, these paths are used to remove secondsacrificial gap material3201 and first sacrificialgap material layer3108X, leavingborder region3115. Afterwards the paths may be closed, see, e.g.,FIG. 30J andFIG. 30J′. A suspendedportion3114A ofnanotube elements3114 may be seen inpre-wiring level structure3213 illustrated inFIG. 30K andpre-wiring level structure3215 illustrated inFIG. 30K′.
• After sacrificial material has been removed, preferred embodimentcomplete fabrication3009 of the combined nanotube and semiconductor structure to the external contact and passivation layers (not shown). For example, after the fluid communication openings (paths) are closed (encapsulated), connections to drainnode3126 are made, seestructure3223 ofFIG. 30M andstructure3225 ofFIG. 30M′, prior to final wiring to terminal pads, passivation, and packaging. .
• FIGS.23,23′,23″ each describe methods (processes) of forming thenanotube switching structures3103,3103′ ofFIGS. 24A and 24B, respectively, andnanotube switching structures3107,3107′ ofFIGS. 24C and 24D, respectively. FIGS.23,23′, and23″ each describe methods (processes) of forming thenanotube switching structures3107X and3107″ ofFIGS. 24E and 24F, respectively.
• Referring to FIGS.23,23′ and23″, preferred methods inFlow Chart3004 start withact3010.Step3010 presumes that an intermediate structure has already been created, on top of which the nanotube control structure is to be formed. For example,FIGS. 24A, 24B,24C,24D,24E, and24F each illustrate anintermediate structure3102′ on which the control structure is to be formed.Structure3102′ already has many components of a field effect device, including drain, source, and gate nodes. The first step is to deposit a conductor layer onsurface3104intermediate structure3102. By way of example, conductor layer may be tungsten, aluminum, copper, gold, nickel, chrome, platinum, palladium, polysilicon, or combinations of conductors such as chrome-copper-gold. Alternatively, conductor layer may be formed of single-layers or multi-layers of single or multi-walled nanotube fabric with conductivities in the range of 0.1 to 100 Ohms per square as describe in incorporated patent references explained further below. Nanotube fabric may be used in vias and wiring in any array structure. Conductor thickness may be in the range of 50 to 200 nm.
• Then, preferred embodiments deposit3012 first sacrificial gap material layer on top of the conductor layer. Asacrificial layer3108′ of gap material such as insulator silicon nitride (Si₃N₄) or semiconductor silicon (Si) for example, is deposited onconductor layer3106′, as illustrated inFIG. 25A.Sacrificial layer3108′ may also be a conductor, such as TiW, for example. As will be explained below, the first sacrificial gap layer thickness controls the separation (or gap) between the nanotube fabric element (yet to be formed) andconductor layer3106′ in the nanotube switch region. In a preferred embodiment, this separation or gap dimension is approximately 1/10 of the suspended length of the nanotube element. For a nanotube switch design with suspended length of 130 nm, the gap is therefore chosen as about 13 nm.Sacrificial layer3108′ is deposited to a thickness of about 13 nm, for example. Alternatively, aftermethod act3010, but beforemethod act3012, insulatingfilm layer3203′ may be deposited as illustrated inFIG. 25A′. Insulatingfilm layer3203′ may be SiO₂, for example, ofthickness 5 to 20 nm, for example.Method3004 continues withstep3012. Adding insulatinglayer3203′ results instructure3107″ after completion ofmethods3004,3036, and3006 as described further below.
• Then, preferred embodiments deposit andimage3014 photoresist. Such patterning may be done using known techniques. This is done to define (in photoresist) the pattern for the electrode and sacrificial material, see, e.g.,electrode3106 and firstsacrificial gap layer3108 ofFIGS. 24A, 24B,24C,24D and24F.
• Alternatively, preferred embodiments step3014patterns layer3108′ resulting in firstsacrificial layer3108X as illustrated inFIG. 25AX, where firstsacrificial layer3108X is a conductor or semiconductor (silicon, for example), with dimensions corresponding to nanotube switching region suspended length L_SUSP, see e.g.,electrode3106 and firstsacrificial gap layer3108X ofFIG. 24E. The inventors envision that for certain applications, the ability to precisely control sacrificial layer removal may be advantageous for manufacturability. Specifically, to etch layers anisotropically has advantages over isotropic etching in defining the underlying gap,e.g. gap region3108A.
• Next, preferred embodiments deposit3015 insulatingmaterial layer3115′ such material may be SiO₂, Si₃N₄, Al₂O₃, or other insulating materials, for example, as illustrated inFIG. 25AX.
• Next, preferred embodiments CMP etch then directly etch3017 insulatinglayer3115′ exposing firstsacrificial layer3108X, silicon, for example, and forming coplanar insulatinglayer3115″, SiO₂or Si₃N₄, for example, as shown inFIG. 25AX′.
• Then, preferred methods etch3016conductor layer3106′ andsacrificial material layer3108′ to formelectrode structure3106 and sacrificialgap material layer3108 as follows.Sacrificial layer3108′ is etched. The photoresist layer (not shown) is removed. Etchedsacrificial layer3108 is used as the mask layer foretching conductor layer3106′. Alternatively, the photoresist layer is used to etch bothsacrificial gap layer3108′ andconductor layer3106′, and then the photoresist is removed (not shown). Alternatively, preferred methods etch3016conductor layer3106′ and insulatingmaterial3115″ ofcoplanar layer3115″ and firstsacrificial layer3108X using a photoresist layer, and then the photoresist is removed (not shown).
• After the electrode and sacrificial material region are formed, preferred methods deposit3018 a conformal sacrificial material layer. As shown inFIG. 25B, conformalsacrificial layer3110 is deposited over the combinedcontrol electrode3106 and firstsacrificial gap layer3108 structure. Alternatively, as shown inFIG. 25BX, conformalsacrificial layer3110 is deposited over the combinedcontrol electrode3106 and coplanar firstsacrificial layer3108X andborder layer3115.Conformal layer3110 may be formed using a variety of insulating materials such as SiO₂, Si₃N₄, Al₂O₃, and polyimide, or conducting materials such as aluminum, copper, nickel, chromium, tungsten, and silicon, for example. In a preferred implementation, SiO₂is selected. The SiO₂may be conformably deposited as spin-on-glass, or using Low Pressure Chemical Vapor Deposition (LPCVD), or by other conformal deposition techniques. The thickness of the deposited SiO₂layer depends on the thickness of the combinedcontrol electrode3106 and sacrificial layer3108 (or combinedcontrol electrode3106 and coplanar firstsacrificial layer3108X and border layer3115) and method of etchingconformal layer3110, and may range from 70 nm to 300 nm, for example.
• After the conformal sacrificial material is deposited, a first methods chemical-mechanical-polish etch3020 partially removessacrificial layer material3110 to top surface of firstsacrificial gap layer3108, leavingplanar support structure3110′ as illustrated inFIG. 25C. Alternatively, firstmethods CMP etch3020 partially removessacrificial layer material3110 to top surface of combinedcontrol electrode3106 and coplanar firstsacrificial layer3108X andborder layer3115, leavingsupport structure3110X′ as illustrated inFIG. 25CX. CMP etch applied to surface ofsacrificial layer3108 may result in surface damage to firstsacrificial gap layer3108. CMP etch applied to combinedcontrol electrode3106 and coplanar firstsacrificial layer3108X andborder layer3115 may result in damage to firstsacrificial layer3108X. Alternatively, asecond methods3020° CMP etch partially removessacrificial layer3110, then directional etch removes additionalsacrificial layer3110 exposing top surface of firstsacrificial gap layer3108, leavingplanar support structure3110′, or alternatively exposing top surface of firstsacrificial layer3108X, leaving support structure311OX′. Two-step etch3020′ method may be simplified to a single-step method without exposing the surface of firstsacrificial gap layer3108, or firstsacrificial gap layer3108X, to a CMP etch process. Alternatively,third etch3020″ directly etchessacrificial layer3110 material exposing top surface of firstsacrificial layer3108, leaving slopedsupport structure3112 as illustrated inFIG. 25CC. Conformalsacrificial layer3110 may be etched using sputter etching, reactive ion beam (RIE) etching, or other techniques.
• Next, preferred methods form3022 a porous layer of matted carbon nanotubes. This may be done with spin-on technique or other appropriate technique as described in U.S. Pat. Nos. 6,643,165 and 6,574,130 and U.S. patent application Ser. Nos. 09/915,093, 10/033,323, 10/033032, 10/128,118, 10/128,117, 10/341005, 10/341,055, 10/341,054, 10/341,130, 60/446,783 and 60/446,786, the contents of which are hereby incorporated by reference in their entireties (hereinafter and hereinbefore, the “incorporated patent references”). Under preferred embodiments, the carbon nanotube layer has a thickness of approximately 0.5-5 nm for devices using single-walled nanotubes and 5-20 nm and greater for devices using multi-walled nanotubes.
• Then, preferred methods deposit3023 a firstconductor material layer3117′ as shown inFIG. 25D andFIG. 25DX. The material ofconductor layer3117′ may be tungsten, aluminum, copper, gold, nickel, chrome, platinum, palladium, or combinations of conductors such as chrome-copper-gold.Conductor layer3117′ thickness is in the range of 25 to 100 nm. The material ofconductor layer3117′ is selected for reliable low contact resistance R_Cbetweenconductor layer3117′ andnanotube fabric layer3114′.
• Next, preferred methods deposit3025 a secondconductor material layer3119′ as shown inFIG. 25D andFIG. 25DX. The material ofconductor layer3119′ may be tungsten, aluminum, copper, gold, nickel, chrome, platinum, palladium, or combinations of conductors such as chrome-copper-gold.Conductor layer3119′ thickness is in the range of 50 to 200 nm. The material ofconductor layer3119′ is selected for good conductivity.
• Photoresist is then deposited and imaged inact3027 on secondconductor material layer3119′.
• Next,preferred methods3029 etchessecond conductor layer3119′ using appropriate known etch techniques to formelectrical conductor3119 as shown inFIGS. 25E, 25F,25EX, and25FX.
• Next,preferred methods3031 etches firstelectrical conductor3117 usingsecond conductor3119 as a masking layer using known etch techniques to formelectrical conductor3117. Combinedelectrical conductors3117 and3119 are shown inFIGS. 25E, 25F,25EX, and25FX.
• Next,preferred methods3035 etches the carbonnanotube fabric layer3114′ by using appropriate techniques as described in the incorporated patent applications, with combinedelectrical conductors3117 and3119 acting as a masking layer. Combinedelectrical conductors3117 and3119, and patternednanotube fabric layer3114 are shown inFIGS. 25E, 25F,25EX,25FX, and25G.
• Under certain embodiments, photoresist is deposited3027 and used to define an image ofelectrical conductor3119,electrical conductor3117, andnanotube fabric layer3114.
• FIG. 25G shows a plan view ofintermediate structure3109 andintermediate structure3109X. FIGS.25E and25EX show cross sectional views ofintermediate structure3109 and3109X, respectively, taken at AA-AA′ ofFIG. 25G, and FIGS.25F and25FX show cross sectional views ofintermediate structures3109 and3109X, respectively, taken at BB-BB′ ofFIG. 25G. Dimensions L_SUSPand L′_SUSPindicate orthogonal dimensions of firstsacrificial layer3108X and are typically at sub-minimum or minimum lithographic dimensions. Dimensions L and L′ indicate orthogonal dimensions ofelectrode3106. L and L′ and are typically at or greater than the minimum lithographic dimensions allowed for a technology.Intermediate structure3109 corresponds to a portion ofFIGS. 24A and 24C in whichelectrode3106, firstsacrificial gap layer3108 and combinedelectrical conductors3117 and3119 were formed using aplanar support structure3110′, but prior to the formation ofopening3136.Intermediate structures3109 and3109X were formed using methods as indicated inflow chart3004 shown in FIGS.23,23′, and23″, the steps used wereacts3010 through3018, next, acts3020 or3020′ to define theplanar support structure3110′ and3110X′, next, acts3022 through3035 to completesubstructures3109 and3109X.
• Referring tomethod3004 shown in FIGS.23,23′, and23″, a preferred method of forming anotherintermediate structure3109′ executes first,methods3010 through3018, next,method3020″ to define the slopedsupport structure3112, next,methods3022 through3035 to completesubstructure3109.
• FIG. 25GG shows a plan view ofintermediate structure3109′.FIG. 25EE shows a cross sectional view ofintermediate structure3109′ taken at AA-AA′ ofFIG. 25GG, andFIG. 25FF shows a cross sectional view ofintermediate structure3109′ taken at BB-BB′ ofFIG. 25GG. Dimensions L and L′ indicate orthogonal dimensions ofelectrode3106. L and L′ are typically at or greater than the minimum lithographic dimensions allowed for a technology.Intermediate structure3109′ corresponds to a portion ofFIGS. 24B and 24D in whichelectrode3106, firstsacrificial gap layer3114, and combinedelectrical conductors3117 and3119 were formed using a slopedsupport structure3112, but prior to the formation ofopening3136.
• When the suspended portion (structure not yet illustrated) of carbonnanotube fabric layer3114 shown schematically inFIG. 14 (position250′) andFIG. 17B (position1140′) storing logic state “1” (the same comments apply for a stored logic “0” state), carbon nanotube fibers in thenanotube fabric layer3114 are elongated and under strain (tension). The ends of carbon nanotube fibers in thenanotube fabric layer3114 that are supported (clamped, pinned) at the perimeter of the suspended region, apply a restoring force. The electrical and mechanical contact, support (clamping, pinning) region is illustrated bycontact3127 inFIGS. 24A-24F, with additional support in oxide layers beyondcontact region3127.Contacts3127 instructures3103 and3107 and on adjacent surfaces ofplanar support structure3110′ shown inFIGS. 24A, 24C,24E, and24F illustrated in corresponding FIGS.25F and25FX, are sufficient to provide the necessary restoring force without carbon nanotube fiber slippage. Layer3314 is thus pinned between3117 and3110′ inregion3127.Contacts3127 instructures3103′ and3107′ and on adjacentsloped support surfaces3112 illustrated inFIGS. 24B and 24D, with slopedsupport surface3112 overlap illustrated in correspondingFIG. 25FF, may tolerate still greater restoring forces without carbon nanotube fiber slippage.
• All preferred structures may be fabricated using lithographic minimum dimensions and greater than minimum lithographic dimensions for a selected generation of technology. Selective introduction of sub-minimum lithographic dimensions may be used to realize smaller cell size, lower carbon nanotube switching (threshold) voltages with tighter distributions through scaling (reducing) the carbon nanotube structure dimensions (combination of shorter suspended length and gap spacings), faster nanotube switching, and lower power operation. Carbon nanotubes fibers of 130 nm suspended length and 13 nm gaps typically switch in less than 350 ps. Selective introduction of sub-minimum lithographic dimensions may be used to form smaller fluid communication pipes used to remove sacrificial material, facilitating covering (sealing) the openings prior to deposition of the conductive wiring layers.
• Sub-minimum lithographic dimensions may be introduced on any planar surface at any step in the process.Flow chart3036 illustrated inFIG. 26 may be used to generate shapes with sub-minimum dimensions. Shapes having two opposite sides of sub-minimum dimension, and two orthogonal sides having minimum or greater than minimum dimension may be formed using well known sidewall spacer technology. Sidewall periodicity is at minimum or greater than minimum dimensions. Shapes having two opposite sides of sub-minimum dimensions, and two orthogonal sides also having sub-minimum dimensions may be formed using the intersection of two sub-minimum dimension sidewall spacers as described in U.S. Pat. Nos. 5,920,101 and 5,834,818. Sidewall periodicity is at minimum or greater than minimum dimensions in both orthogonal directions.
• Referring toFIG. 26, preferredmethods flow chart3036 start with methods step3042. Methods step3042 presumes that an intermediate base structure has already been created with a planar surface. Anintermediate base structure3102″ may include semiconductor and carbon nanotube structure elements, and may be at any step in a process that has a planar surface. The preferred methodsfirst step deposits3042sacrificial layer3131′ onintermediate structure3102″, havingsurface3104″, as illustrated inFIG. 29A.Sacrificial layer3131′ may be photo resist, an insulator such as Si₃N₄, a semiconductor, a conductor, and may be in the thickness range of 50 to 300 nm.Sacrificial layer3131′ is patterned to minimum or greater-than-minimum dimensions using photoresist (not shown).
• Then, preferred embodiments form3044 sub-lithographic sidewall spacer selectively etchable over sacrificial layer. Deposit a conformal layer of an insulator, or a conductor such as tungsten, for example, on patterned sacrificial layer of insulator Si₃N₄, for example. Tungsten thickness is selected to achieve a desired sidewall spacing dimension. For a technology of 130 nm minimum dimension, for example, a tungsten thickness is chosen that results in a sidewall lateral dimension in the range of 50 to 100 nm, for example. After deposition, the combined tungsten and Si₃N₄layer is planarized, forming thesidewall spacer structure3133 on the sidewalls ofsacrificial layer3131 illustrated inFIG. 29B
• Next,preferred methods3046 selectively etch sacrificial layer, leaving sub-minimum tungsten spacers on planarized surface. Sub-minimumtungsten spacer structure3133 of width in the range of 50 to 100 nm, for example, are shown inFIG. 29C. Alternatively, asecond methods3058 forms a second sidewall spacer structure above and orthogonal to sidewallspacer structure3133 as described in U.S. Pat. Nos. 5,920,101 and 5,834,818. For a technology of 130 nm minimum dimension, for example, a tungsten thickness is chosen that results in a shape of lateral dimension in the range of 50 to 100 nm in one dimension, and. a shape of lateral dimension in the range of 50 to 100 nm in an orthogonal dimension (not shown).
• Then, preferred methods deposit3048 asacrificial layer3130 and planarize. Thesacrificial layer3130 may be an insulator layer, or a photoresist layer, for example. Planarization exposes the spacer material.
• Next,preferred method3050 spacer material is etched leaving photoresist openings to the underlying planar surface having the dimensions of the spacer structures.Photoresist layer openings3134 may be shapes with one pair of minimum (or greater than minimum) shape W1, and sub-minimum pair of opposite dimensions of W2 as illustrated in plan viewFIG. 29D.Photoresist layer openings3132 may be shapes with one pair of sub-minimum opposite dimensions W2, and a second pair of orthogonal sub-minimum dimensions W3 as illustrated in plan viewFIG. 29E.FIG. 29F shows a cross sectional view of intermediatesacrificial structure3113 plan viewFIG. 29D intermediatesacrificial structure3113 taken at CC-CC′ ofFIG. 29D.FIG. 29F shows a cross sectional view of intermediatesacrificial structure3113′ plan viewFIG. 29E intermediatesacrificial structure3113′ taken at DD-DD′ ofFIG. 29E.
• FIGS. 27 and 27′ each describe methods (processes)3006 for completing the nanotube switch (control)structures3103 and3107 illustrated inFIGS. 24A and 24C, respectively.
• Referring toFIG. 27, preferred method preferred method acts inflow chart3006 start withstep3230.Step3230 presumes that a lower portion carbon nanotube intermediate structure3109 (FIGS. 25E, 25F, and25G) or nanotubeintermediate structure3109X (FIGS.25DX,25EX, and25FX) of dimension L have already been created on anintermediate substrate structure3102′.Structure3102′ already has many components of a field effect device, including drain, source, and gate nodes, andelectrode3106 ofstructure3109 or3109X is electrically connected to an FET source. The first step is to deposit and planarize an insulating layer that may be formed using a variety of insulating materials such as SiO₂, Si₃N₄, Al₂O₃. In a preferred implementation, SiO₂is selected. The SiO₂may be deposited as spin-on-glass, or using Low Pressure Chemical Vapor Deposition (LPCVD), or by other deposition techniques. The thickness of the deposited SiO₂layer depends on the thickness of the lower portion carbon nanotubeintermediate structure3109, and may range from 150 nm to 300 nm, for example, as illustrated inFIG. 25D orFIG. 25E. Method steps described fully below with respect toFIG. 30A also apply toFIG. 30AX
• Then, preferred methods deposit andimage3232 photoresist. Such patterning may be done using known techniques to produce images in the photoresist of minimum size L_MINor greater inphotoresist layer3129 shown inFIG. 30B. Alternatively, intermediatesacrificial structure3113 may be formed in lieu ofphotoresist layer3129, such that opening L_MINis reduced to sub-minimum dimension W2 (L_SUB-MIN=W2) as illustrated inFIGS. 29D and 29F. Lower portion carbon nanotubeintermediate structure3109 may be reduced in size, such that L is replaced by L_MIN, and L_MINis replaced by W2 (also referred to as L_SUB-MIN). For a 130 nm minimum feature technology, L may be reduced from 250 nm to 190 nm, with the opening reduced from L_MINof 130 nm to W2 (L_SUB-MIN) of 65 nm, for example. Alternatively, intermediateartificial structure3113′ may be formed in lieu ofphotoresist layer3129, such that opening L_MINis reduced to sub-minimum dimension W2, and orthogonal opening dimension (not shown) is reduced to sub-minimum dimension W3, as illustrated inFIGS. 29E and 29F. If W2=W3=65 nm, and, lower portion carbon nanotubeintermediate structure3109 dimensions L and L′ are equal (FIGS. 25E and 25F), then the dimension ofstructure3109 may reduced from 250×250 nm to 190×190 nm, with an opening reduced from 130×130 nm, to 65×65 nm, for example.
• Then, preferred methods etch3234 holes insecond conductor layer3119 to the top ofconductor3117. This etch can be done directly throughconductor3119 using RIE directional etch, for example, transferring the minimum or sub-minimum dimension of opening3136 intoconductor3119 as minimum orsub-minimum opening3151 as illustrated inFIG. 30C.Conductor3117 is used an etch stop for the RIE because RIE may destroy carbon nanotube fibers incarbon nanotube layer3114.
• Next, preferred methods etch3235 holes infirst conductor layer3117 to thecarbon nanotube layer3114. This etch can be done directly throughconductor3117, transferring the minimum or sub-minimum dimension of opening3151 intoopening3153 inconductor3117 as illustrated inFIG. 30D. A wet etch is used to createopening3153 inconductor3117. Wet etch is selected to prevent damage tonanotube layer3114 as described in the incorporated patent applications. Wet etch is selected not to etch firstsacrificial gap layer3108. Firstsacrificial gap layer3108 may consist of Si₃N₄or Si, for example.
• Then, preferred methods deposit3236 conformal layer of second sacrificial gap material overconductor3119, into opening3153′ contacting sidewalls ofconductors3119 and3117, and over thecarbon nanotube element3114 as illustrate inFIG. 30E. One example is thin conductor layer of TiW, of approximate thickness 5-50 nm. The actual thickness may vary depending upon the performance specifications required for the nanotube device.
• Next, preferred methods deposit3240 conductor layer, fill theopening3153′ illustrated inFIG. 30E, and planarized. Conductor layer may be composed of tungsten, aluminum, copper, gold, nickel, chrome, platinum, palladium, or combinations of conductors such as chrome-copper-gold, ofthickness 150 to 300 nm. Alternatively, preferred methods deposit3238 of aconformal insulator layer3203, layer3202 may be selected from materials such as SiO₂, Al₂O₃, or other suitable material with etch properties selective to Si₃N₄or Si, for example. SiO₂is preferred with approximate thickness 5-50 nm as illustrated inFIG. 30E′. Then, preferred methods deposit3240 conductor layer forelectrode3205 on insulator layer, fillopening3153. The actual thickness may vary depending upon the performance specifications required for the nanotube device.”
• Then,preferred methods3242 pattern conductor layer using photoresist. Next, pattern second sacrificial gap layer is patterned using the photoresist layer as a mask, or conductor layer as a mask. Alternatively,preferred methods3244 pattern conductor layer using photoresist. Next, pattern insulator layer using the photoresist layer as a mask, or conductor layer as a mask. Then, pattern second sacrificial gap layer is patterned using the photoresist layer as a mask, or combined metal and insulator as a mask.
• Then,preferred methods3246 deposit insulating layer and planarize to formintermediate structure3212 as illustrated inFIG. 30F.Insulator3116overcoats electrode3205. Secondsacrificial gap layer3201 separates electrode3205 fromconductors3119 and3117, and carbonnanotube fabric layer3114. Alternatively,preferred methods3246 deposit insulating layer and planarize to formintermediate structure3214 as illustrated inFIG. 30F′.Insulator3116overcoats electrode3205.Conformal insulator layer3203 separateselectrode3205 and secondsacrificial gap layer3201, and remains on the lower surface ofelectrode3205 after the removal of second sacrificial gap layer3201 (a later step). Secondsacrificial gap layer3201 separates electrode3205 fromconductors3119 and3117, and carbonnanotube fabric layer3114 formingintermediate structure3212. Alternatively,preferred methods3232 through preferredmethods3246 applied toFIG. 30AX result in thestructure3212X shown inFIG. 30FX andstructure3214X shown inFIG. 30FX′.
• FIGS. 28 and 28′ describe processes for removing sacrificial layers around the switching portion (region) of carbonnanotube fabric layer3114 so that gaps are formed around the nanotube element so that the element may be suspended and switched in response to electrostatic forces. Each method presumes an intermediate structure such as3212 or3214 (FIGS.30F and30F′, respectively) has already been formed.
• FIGS. 28 and 28′ describe processes for removing sacrificial layers around the switching portion (region) of carbonnanotube fabric layer3114 so that gaps are formed around the nanotube element so that the element may be suspended and switched in response to electrostatic forces. Each method presumes an intermediate structure such as3212X or3214X (FIGS.30FX and30FX′, respectively) has already been formed. While preferred methods are described further below with respect tostructures3212 and3214 (FIGS.30F and30F′, respectively), it is understood that these preferred methods may also be applied tostructure3212X shown inFIG. 30FX andstructure3214X shown inFIG. 30FX′.
• With reference toflow chart3008 ofFIGS. 28 and 28′ and tointermediate structures3212 and3214 of FIGS.30F and30F′, respectively, preferred methods form3250 minimum images in photoresist maskingsacrificial layer3130. Alternatively, intermediatesacrificial structure3113 may be formed in lieu of a photoresist layer, providing an opening of sub-minimum dimension W2 as illustrated inFIGS. 29D and 29F.
• Then, preferred methods directionally etch3252 insulator form, via holes and expose a top surface of a top electrode. Via holes are located outside nanotube switching regions. Viahole3207 throughinsulator3116 totop electrode3205 illustrated inFIG. 30G is taken at EE-EE′ as shown inFIG. 30F. No insulating layer is present betweenelectrode3205 and secondsacrificial gap layer3201. Alternatively, viahole3208 throughinsulator3116 totop electrode3205 illustrated inFIG. 30G′ is taken at FF-FF′ as shown inFIG. 30F′. Insulatinglayer3203 is present betweenelectrode3205 and secondsacrificial gap layer3201.
• Next, preferred methods directionally etch3254 conductor electrode to top of second sacrificial gap layer.Openings3207′ provide fluid communication paths to secondsacrificial layers3201 as illustrated inFIG. 30H. Alternatively, preferred methods directionally etch3256 conductor electrode to top of insulating layer between conductor electrode and second sacrificial gap layer. Next, methods directionally etch3254 insulator layer to top of second sacrificial layer.Openings3208′ provide fluid communication paths to secondsacrificial gap layers3201 as illustrated inFIG. 30H′
• Then, preferred methods etch (remove)3258 second sacrificial gap layer material creating a gap and extending fluid communication paths to the exposed top portion (region) of first sacrificial gap layers inside openings in conductors in contact with carbon nanotube fabric layers. At this point in the process a gap exists above a portion of the carbon nanotube film, which may also be referred to as a single-gap nanotube switch structure, and switched as described further down.
• Next, preferred methods etch (remove)3260 through porous carbon nanotube fabric layer without damaging carbon nanotube fibers by using appropriate techniques as descried in the incorporated patent applications, to exposed portion (region) of first sacrificial gap layers inside openings in conductors in contact with carbon nanotube fabric layer. Portions (regions) of first sacrificial gap layers exposed to the etch are removed and carbon nanotube fibers are suspended (released) in the switching region. Firstsacrificial layer3108 is partially removed using industry standard wet etches for Si₃N₄, for example. Alternatively, firstsacrificial layer3108X is removed using industry standard wet etches for a silicon layer, for example. At this point a gap exists above and below a portion of the carbon nanotube, which may be referred to as a dual-gap switch structure, and switched as described further down. Carbon nanotube fibers in the peripheral region outside a switching region remain mechanically pinned and electrically connected, sandwiched between a conductor layer and the remaining (unetched) portion of the first sacrificial layer. A switching region is defined by openings in conductors in contact with carbon nanotube fabric layers.Gap regions3209,3209A, and3108A forintermediate structure3213 with no insulating layer abovegap3108A are illustrated inFIGS. 30I and 30K.Gap regions3211,3209A, and3108A forintermediate structure3215 with insulating layer abovegap3108A are illustrated inFIG. 30K′.Gap regions3209,3209A, and3108A forintermediate structure3215′ with insulatinglayer3203′ belowgap3108A are illustrated inFIG. 30K″.Insulator3203′ was deposited as illustrated inFIG. 25A′.
• Next, preferred methods deposit3262 insulating layer to fill (seal) openings (via holes) that provide a fluid communication path (or fluid conduit) used to release (suspend) carbon nanotube fibers. Insulator surface is planarized. Openings (via holes) that provide fluid communication paths are sealed as illustrated by sealedopening3207″ inFIG. 30J and by sealedopening3208″ inFIG. 30J′.
• Next, preferred methods etch3264 via holes to reach buried studs in contact with FET drain regions. Via holes are filled with a conductor and planarized.FIG. 30K illustratesstructure3213 withelectrode3205, combinedmetal conductors3119 and3117, andcarbon nanotube region3114A separated bygap regions3209A and3108A.Stud3118A contacts stud3118 that connects to drain3126 throughcontact3123.Structure3213 is ready for first wiring layer.FIG. 30K′ illustratesstructure3215 with combinedelectrode3205 andbottom insulator layer3203, combinedmetal conductors3119 and3117, andcarbon nanotube region3114A separated bygap regions3209A and3108A.Stud3118A contacts stud3118 that connects to drain3126 throughcontact3123.Structure3215 is ready for a first wiring layer.
• FIG. 30KK illustrates thenanotube switch portion3217 of integrated dual-gap structure3215 ofFIG. 30K′, where the suspendedportion3114A ofnanotube3114 has been switched to the open position “OFF” state, with the elongated suspendedportion3114A′ in contact withinsulator3203 on release-plate3216, and held in the open position by van der Waals forces betweeninsulator3203 andcarbon nanotube portion3114A′.Switch portion3217 corresponds to switch90 illustrated in the schematic ofFIG. 3A switched to position90″, as illustrated in the schematic ofFIG. 3C. Nanotube elongated suspendedportion3114A′ ofFIG. 30KK corresponds to nanotubeelongated portion1140″ of the memory cell schematic illustrated inFIG. 17C.FIG. 30KK′ illustrates thenanotube switch portion3217′ of integrated dual-gap structure3215 ofFIG. 30K′, where the suspendedportion3114A ofnanotube3114 has been switched to the closed position “ON” state, with the elongated suspendedportion3114A″ in contact with switch-plate3206, and held in the closed position by van der Waals forces between switch-plate3206 andcarbon nanotube portion3114A″.Switch portion3217′ corresponds to switch90 illustrated in the schematic ofFIG. 3A switched to position90′, as illustrated in the schematic ofFIG. 3B. Nanotube elongated suspendedportion3114A″ ofFIG. 30KK′ also corresponds to nanotubeelongated portion1140′ of the memory cell schematic illustrated inFIG. 17B.
• FIG. 30L illustrates a cross section of an alternate integrated nanotube structure that uses a single gap region above the nanotube switching region to form integrated single-gapnanotube switching structure3219, instead of a dual-gap nanotube structure that uses a gap region above and below the switching region of the nanotube.Structure3219 is referred to as a single-gap structure becausesegment3114B ofnanotube3114 only has asingle gap3209A.Dielectric layer3108 belownanotube segment3114B is not removed by etching.Structure3219 is fabricated using the steps as illustrated byflow chart3008 inFIG. 28′, and corresponds to the method of fabrication described above for fabricating cross section ofstructure3213 ofFIG. 30K, except that method steps3260 are omitted, such that the first sacrificial gap layer is not removed.Electrode3106 shown belownanotube3114 in dual-gapintegrated structure3215 ofFIG. 30K′ performs a switch-plate function, as doeselectrode3205 shown abovenanotube3114 in single-gapintegrated structure3219 ofFIG. 30L. In other words, thebottom electrode3106 ofFIG. 30K′ and thetop electrode3205 ofFIG. 30L each performs a switch-plate function.Electrode3205 with insulatinglayer3203 shown abovenanotube3114 in dual-gapintegrated structure3215 ofFIG. 30K′ performs a release-plate function, as doeselectrode3106 with insulatinglayer3108 shown belownanotube3114 in single-gapintegrated structure3219 ofFIG. 30L. In other words, the insulatedtop electrode3205 ofFIG. 30K′ and theinsulated bottom electrode3106 ofFIG. 30L each performs a release-plate function.Source3124 is connected toelectrode3106 as illustrated inFIG. 30K′, such thatsource3124 controls the voltage applied toelectrode3106, which is used a switch-plate instructure3215 shown inFIG. 30K′.Source3124 controls the voltage ofinsulated electrode3106, which is used as a release-plate instructure3219 shown inFIG. 30L.
• FIG. 30L′ illustrates thestructure3219′ in whichstructure3219 ofFIG. 30L has been modified so thatsource3124 controls the voltage of switch-plate electrode3205. In operation,structure3215 ofFIG. 30K′ andstructure3219′ ofFIG. 30L′ operate in the same way, except that the position of corresponding switch plates have been interchanged, such that the switch-plate is below the nanotube layer instructure3215, and above the nanotube layer instructure3219′.
• FIG. 30L″ illustrates thenanotube switch portion3221 of integrated single-gap structure3219 ofFIG. 30L, and single-gap structure3219′ ofFIG. 30L′, where the suspendedportion3114B ofnanotube3114 is in the open position “OFF” state. In the open position,nanotube3114 remains in contact withinsulator layer3108, in an approximately non-elongated state, with van der Waals force betweennanotube3114 andinsulator layer3108.FIG. 30L′″ illustrates thenanotube switch portion3221′ of integrated single-gap structure3219 ofFIG. 30L, and single-gap structure3219′ ofFIG. 30L′, where the suspendedportion3114B ofnanotube3114 has been switched to the closed position “ON”state3114B′. In the closed position,nanotube3114 has been switched in contact with switch-plate3205, and remains incontact electrode3205, in an elongated state, with van der Waals force betweennanotube3114B segment andelectrode3205. A single-gap structure may be used in lieu of a dual-gap structure to fabricate field effect devices with controllable sources and memories using NT-on-Source arrays.
• Continuing the fabrication process using a dual-gap nanotube structure such as illustrated inFIG. 30K,bit line3138 is then deposited and patterned; the resultingcross section3223 is illustrated inFIG. 30M.Wiring layer3138contacts stud3118A atcontact region3140 ofintermediate structure3223. Final processing to the passivation layer is not shown. Alternatively, continuing the fabrication process using a dual-gap nanotube structure such as illustrated inFIG. 30K′,bit line3138 is then deposited and patterned; the resultingcross section3225 is illustrated inFIG. 30M′.Wiring layer3138contacts stud3118A atcontact region3140 ofintermediate structure3225. Final processing to the passivation layer is not shown.
• FIG. 30M′ illustrates cross section A-A′ ofarray3225 taken at A-A′ of the plan view ofarray3225 illustrated inFIG. 30P, and showsFET device region3237 in the FET length direction,nanotube switch structure3233, interconnections and insulators.FIG. 30N illustrates cross section B-B′ ofarray3225 taken at B-B′ of plan view ofarray3225 illustrated inFIG. 30P, and shows arelease array line3205, areference array line3119/3117 composed of combinedconductors3119 and3117, and aword array line3120.FIG. 30P illustrates a plan view ofarray3225 includingexemplary cell3165 region, bitarray line3138 contactingdrain3126 throughcontact3140 tostud3118A, tostud3118, to contact3123, and to drain3126, (studs3118,3118A, andcontact3123 not shown in plan view3225).Reference array line3119/3117 is parallel tobit line3138, is illustrated in cross section inFIG. 30N, and contacts a corresponding reference line segment in the picture frame region formed by combinedconductors3117 and3119, in contact withnanotube3114, as shown inFIG. 30M′.Release array line3205 is parallel toword array line3120.Release line3205 contacts and forms a portion ofrelease electrode3205 as illustrated in the nanotube switching region ofFIG. 30M′. This nanotube switching region is illustrated asnanotube switch structure3233 inarray3225 ofFIG. 30P. In terms of minimum technology feature size, NT-on-source cell3165 is approximately 12 to 13 F². Nanotube-on-source array3225 structures illustrated in FIGS.30M′,30N, and30P correspond to nanotube-on-source array1700 schematic representations illustrated inFIG. 18.Bit line3138 structures correspond to any of bit lines BL0 to BLm−1 schematic representations;reference line3119/3117 structures correspond to any of reference lines REF0 to REFm−1 schematic representations;word line3120 structures correspond to any of word lines WL0 to WLn−1 schematic representations;release line3205 structures correspond to any of release lines RL0 to RLn−1 schematic representations;source contact3140 structures correspond to any ofsource contacts1720 schematic representations;nanotube switch structures3233 correspond to any of NT0,0 to NTm−1,n−1 schematic representations;FET3237 structures correspond to any of FETs T0,0 to Tm−1, n−1 schematic representations; andexemplary cell3165 corresponds to any of cells C0,0 to cell Cm−1,n−1 schematic representations.
• It is desirable to enhancearray3225 illustrated in plan viewFIG. 30P by enhancing wireability, for example, or cell density, for example. In order to minimize the risk of shorts caused by misaligned via (vertical) connections between conductive layers, it is desirable to coat the top and sides of some selected conductors with an additional insulating layer that is not etched when etching the common insulator (common insulator SiO₂, for example) between conductive layers as illustrated bystructure3227 inFIG. 31D. Amethod3144 of coating a conductive layer with an additional insulating layer to forminsulated conductor structure3227 is described with respect to structures illustrated inFIGS. 31A-31D.
• FIG. 31A presumes that an intermediate structure has already been created and insulated withinsulator layer3116, SiO₂for example. Then, preferred methodsdeposit conductor layer3139′ oninsulator3116. By way of example,conductor layer3139′ may be tungsten, aluminum, copper, gold, nickel, chrome, platinum, palladium, polysilicon, or combinations of conductors such as chrome-copper-gold deposited by evaporation, sputtering, CVD, and other methods. Conductor thickness may be in the range of 50 to 200 nm.
• Then, preferred methods deposit insulatinglayer3143′ on top ofconductor layer3139′ as illustrated inFIG. 31A. Insulator material maybe silicon nitride, alumina, or polyimide, for example. Insulator thickness may be 20 to 100 nm, for example.
• Then, preferred methods deposit and image photoresist using known techniques. This is done to define a pattern in the photoresist that corresponds to the electrode and insulating layer.
• Then, preferred methods etch defineconductor3139 and insulatinglayer3143 as illustrated inFIG. 31B. The photoresist layer (not shown) is removed.
• After theconductor3139 and insulatinglayer3143 are defined, preferred methods deposit conformal insulatinglayer3147 as illustrated inFIG. 31C. Insulatinglayer3147 may be of the same material as insulatinglayer3143. Insulating thickness may be 20 to 100 nm, for example.
• Next, preferred methods directionally etch (reactive ion etch, for example) insulatinglayer3147, resulting inconductor3139 having insulatinglayer3148 on top and on the sides and forminginsulated conductor structure3227 as illustrated inFIG. 31D. Method3144 (or comparable methods) of insulating a conductor as illustrated inFIGS. 31A-31D may be applied to various conductive layers, such as those inmemory array3225.
• It is desirable to enhance the wireability ofarray3225 illustrated inFIG. 30P by formingreference array line3138′ on the same wiring level and at the same time asbit line3138.Reference array line3138′ contacts referenceline segments3119/3117 composed of combinedconductors3119 and3117 as illustrated further below.Line segments3119/3117 are not required to span relatively long sub-array regions and may be optimized for contact to nanotubelayer3114.
• FIG. 32A illustrates cross section A-A′ ofarray3229 taken at A-A′ of the plan view ofarray3229 illustrated inFIG. 32C, and showsFET device region3237 in the FET length direction,nanotube switch structure3233, interconnections and insulators.FIG. 32B illustrates cross section B-B′ ofarray3229 taken at B-B′ of plan view ofarray3229 illustrated inFIG. 32C, and shows arelease array line3205 with insulatinglayer3149 corresponding to insulatinglayer3148 in structure3227 (FIG. 31D), areference array line3138′ in contact withconductor3119 of combinedconductors3119 and3117, and aword array line3120.Reference array line3138′contacts conductor3119 throughcontact3155, tostud3157, throughcontact3159, toconductor3119.Insulator3149 is used to prevent contact betweenrelease line electrode3205 andstud3157 in case ofstud3157 misalignment.FIG. 32C illustrates a plan view ofarray3229 including exemplary cell3167 region, withbit array line3138 contactingdrain3126 throughcontact3140 tostud3118A, tostud3118, to contact3123, and to drain3126, (stud3118A,stud3118 andcontact3123 not shown in plan view3229).Reference array line3138′ is on the same array wiring layer and parallel tobit line3138, as is illustrated in plan view ofarray3229 inFIG. 32C, andreference line3138′ contacts a correspondingreference line segment3119, as shown inFIG. 32B.Release array line3205 is parallel toword array line3120.Release line3205 contacts and forms a portion ofrelease electrode3205 as illustrated in the nanotube switching region ofFIG. 32A. This nanotube switching region is illustrated asnanotube switch structure3233 inarray3229 ofFIG. 32C. In terms of minimum technology feature size, NT-on-source cell3167 is approximately 12 to 13 F². Nanotube-on-source array3229 structures illustrated inFIGS. 32A, 32B, and32C correspond to nanotube-on-source array1700 schematic representation illustrated inFIG. 18.Bit line3138 structures correspond to any of bit lines BL0 to BLm−1 schematic representations;reference line3138′ structures correspond to any of reference lines REF0 to REFm−1 schematic representations;word line3120 structures correspond to any of word lines WL0 to WLn−1 schematic representations;release line3205 structures correspond to any of release lines RL0 to RLn−1 schematic representations;source contact3140 structures correspond to any ofsource contacts1720 schematic representations;nanotube switch structure3233 correspond to any of NT0,0 to NTm−1,n−1 schematic representations; andFET3237 structures correspond to any of FET T0,0 to Tm−1, n−1 schematic representations; and exemplary cell3167 corresponds to any of cells C0,0 to cell Cm−1,n−1 schematic representations.
• It is desirable to enhance the density ofarray3225, illustrated inFIG. 30P, to reduce the area of each bit in the array, resulting in higher performance, lower power, and lower cost due to smaller array size. Smaller array size results in the same number of bits occupying a reduced silicon chip area, resulting in increased productivity and therefore lower cost, because there are more chips per wafer. Cell area is decreased by reducing the size (area) ofnanotube switch region3233, thereby reducing the periodicity betweennanotube switch regions3233 and correspondingly reducing the spacing betweenbit lines3138 andreference lines3119/3117.
• FIG. 33A illustrates cross section A-A′ ofarray3231 taken at A-A′ of the plan view ofarray3231 illustrated inFIG. 33D, and showsFET device region3237 in the FET length direction, reduced area (smaller)nanotube switch structure3239, interconnections and insulators. A smaller picture frame opening is formed in combinedconductors3119 and3117 by applyingsub-lithographic method3036 shown inFIG. 26 and corresponding sub-lithographic structures shown inFIGS. 29D, 29E, and29F during the fabrication ofnanotube switch structure3239.FIG. 33B illustrates cross section B-B′ ofarray3231 taken at B-B′ of plan view ofarray3231 illustrated inFIG. 33D, and showsreference line3163 comprisingconductive layers3117 and3119, and conformal insulatinglayer3161.Conductive layers3117 and3119 ofreference line3163 are extended to form the picture frame region ofnanotube device structure3239; however, insulatinglayer3161 is not used as part of thenanotube switch structure3239.FIG. 33B also illustratesrelease line3205, andword array line3120.FIG. 33C illustrates cross section C-C′ ofarray3231 taken at C-C′ of the plan view ofarray3231 illustrated inFIG. 33D.Bit line3138 is connected to draindiffusion3126 throughcontact3140, tostud3118A, and throughcontact3123. In order to achieve greater array density, there is a small spacing betweenstud3118A andreference line3163.Insulator3161 is used to prevent electrical shorting betweenstud3118A andreference line3163conductors3119 and3117 ifstud3118A is misaligned.FIG. 33D illustrates a plan view ofarray3231 includingexemplary cell3169 region, withbit array line3138 contactingdrain3126 as illustrated inFIG. 33C,reference array lines3163 parallel tobit line3138 but on a different array wiring level (wiring plane).Release array line3205 is parallel toword array line3120.Release line3205 contacts and forms a portion ofrelease electrode3205 as illustrated in the nanotube switching region ofFIG. 33A.Exemplary cell3169 area (region) is smaller (denser) than exemplary cell3167 area shown inFIG. 32C andexemplary cell3165 area shown inFIG. 30P, and therefore correspondingarray3231 is denser (occupies less area) than corresponding array areas ofarray3229 and3225. The greater density ofarray3231 results in higher performance, less power, less use of silicon area, and therefore lower cost as well. In terms of minimum technology feature size, NT-on-source cell3169 is approximately 10 to 11 F². Nanotube-on-source array3231 structures illustrated inFIGS. 33A-33D correspond to nanotube-on-source array1700 schematic representation illustrated inFIG. 18.Bit line3138 structures correspond to any of bit lines BL0 to BLm−1 schematic representations;reference line3163 structures correspond to any of reference lines REF0 to REFm−1 schematic representations;word line3120 structures correspond to any of word lines WL0 to WLn−1 schematic representations;release line3205 structures correspond to any of release lines RL0 to RLn−1 schematic representations;source contact3140 structures correspond to any ofsource contacts1720 schematic representations;nanotube switch structure3239 correspond to any of NT0,0 to NTm−1,n−1 schematic representations; andFET3237 structures correspond to any of FET T0,0 to Tm−1, n−1 schematic representations; andexemplary cell3169 corresponds to any of cells C0,0 to cell Cm−1,n−1 schematic representations.
NT-on-Source NRAM Memory Systems and Circuits with Parallel Bit and Release Lines, and Parallel Word and Reference Lines
• NRAM 1T/1NT memory arrays are wired using four lines. Word line WL is used to gate select device T, bit line BL is attached to a shared drain between two adjacent select devices. Reference line REF is used to control the NT switch voltage of storage element NT, and release line RL is used to control the release-plate of storage element NT. In this NRAM array configuration, RL is parallel to BL and acts as second bit line, and REF is parallel to WL and acts as a second word line.
• FIG. 34A depicts a structure comprising non-volatile field effect device.FED480 with memory cell wiring to form NT-on-Source memory cell2000 schematic.Memory cell2000 operates in a source-follower mode. Word line (WL)2200 connects to terminal T1 ofFED480; bit line (BL)2300 connects to terminal T2 ofFED480; reference line (REF)2400 connects to terminal T3 ofFED480; and release line (RL)2500 connects to terminal T4 of FED480 (T1-T4 shown inFIG. 2D).Memory cell2000 performs write and read operations, and stores the information in a non-volatile state. TheFED480 layout dimensions and operating voltages are selected to optimizememory cell2000.Memory cell2000 FET select transistor (T)gate2040 corresponds togate82;drain2060 corresponds to drain84; andcontrollable source2080 corresponds tocontrollable source86.Memory cell2000 nanotube (NT) switch-plate2120 corresponds to switch-plate88;NT switch2140 corresponds toNT switch90; release-plateinsulator layer surface2160 corresponds to release-plateinsulator layer surface96; and release-plate2180 corresponds to release-plate94. The interconnections between the elements ofmemory cell2000 schematic correspond to the interconnection of the corresponding interconnections of the elements ofFED480.BL2300 connects to drain2060 throughcontact2320;REF2400 connects toNT switch2140 throughcontact2420;RL2500 connects to release-plate2180 bycontact2520;WL2200 interconnects togate2040 bycontact2220. The non-volatileNT switching element2140 may be caused to deflect toward switch-plate2120 via electrostatic forces to closed (“ON”)position2140′ to store a logic “1” state as illustrated inFIG. 34B. The van der Waals force holdsNT switch2140 inposition2140′. Alternatively, the non-volatileNT switching element2140 may be caused to deflect toinsulator surface2160 on release-plate2180 via electrostatic forces to open (“OFF”)position2140″ to store a logic “0” state as illustrated inFIG. 34C. The van der Waals force holdsNT switch2140 inposition2140″. Non-volatileNT switching element2140 may instead be caused to deflect to an open (“OFF”) near-mid point position2140′″ between switch-plate2120 and release-plate2180, storing an apparent logic “0” state as illustrate inFIG. 34D. However, the absence of a van der Waals retaining force in this open (“OFF”) position is likely to result in a memory cell disturb that causesNT switch2140 to unintentionally transition to the closed (“ON”) position, and is not desirable. Sufficient switching voltage is needed to ensure that theNT switch2140 open (“OFF”) position isposition2140″. The non-volatile element switching via electrostatic forces is as depicted byelement90 inFIG. 2D.Voltage waveforms311 used to generate the required electrostatic forces are illustrated inFIG. 4.
• NT-on-Source schematic2000 forms the basis of a non-volatile storage (memory) cell. The device may be switched between closed storage state “1” (switched to position2140′) and open storage state “0” (switched toposition2140″), which means the controllable source may be written to an unlimited number of times to as desired. In this way, the device may be used as a basis for a non-volatile nanotube random access memory, which is referred to here as a NRAM array, with the ‘N’ representing the inclusion of nanotubes.
• FIG. 35 represents anNRAM system2700, according to preferred embodiments of the invention. Under this arrangement, an array is formed with m×n (only exemplary portion being shown) of non-volatile cells ranging from cell C0,0 to cell Cm−1,n−1.NRAM system2700 may be designed using one large m×n array, or several smaller sub-arrays, where each sub-array is formed of m×n cells. To access selected cells, the array uses read and write word lines (WL0, WL1, . . . WLn−1), read and write bit lines (BL0, BL1, . . . BLm−1), read and write reference lines (REF0, REF1, . . . REFm−1), and read and write release lines (RL0, RL1, . . . RLn−1). Non-volatile cell C0,0 includes a select device T0,0 and non-volatile storage element NT0,0. The gate of T0,0 is coupled to WL0, and the drain of T0,0 is coupled to BL0. NT0 is the non-volatilely switchable storage element where the NT0,0 switch-plate is coupled to the source of T0,0, the switching NT element is coupled to REF0, and the release-plate is coupled to RL0.Connection2720 connects BL0 to shared drain of select devices T0,0 and T0,1. Word, bit, reference, and release decoders/drivers are explained further below.
• Under preferred embodiments, nanotubes inarray2700 may be in the “ON” “1” state or the “OFF” “0” state. The NRAM memory allows for unlimited read and write operations per bit location. A write operation includes both a write function to write a “1” and a release function to write a “0”. By way of example, a write “1” to cell C0,0 and a write “0” to cell C1,0 is described. For a write “1” operation to cell C0,0, select device T0,0 is activated when WL0 transitions from 0 to V_SW, BL0 transitions from V_DDto 0 volts, RL0 transitions from V_DDto switching voltage V_SW, and REF0 transitions from V_DDto switching voltage V_SW. The release-plate and NT switch of the non-volatile storage element NT0,0 are each at V_SWresulting in zero electrostatic force (because the voltage difference is zero). The zero BL0 voltage is applied to the switch-plate of non-volatile storage element NT0,0 by the controlled source of select device T0,0. The difference in voltage between the NT0,0 switch-plate and NT switch is V_SWand generates an attracting electrostatic force. If V_SWexceeds the nanotube threshold voltage V_NT-TH, the nanotube structure switches to “ON” state or logic “1” state, that is, the nanotube NT switch and switch-plate are electrically connected as illustrated inFIG. 34B. The near-Ohmic connection between switch-plate2120 andNT switch2140 inposition2140′ represents the “ON” state or “1” state. If the power source is removed, cell C0,0 remains in the “ON” state.
• For a write “0” (release) operation to cell C1,0, select device T1,0 is activated when WL0 transitions from 0 to V_SW, BL1 transitions from V_DDto V_SWvolts,RL1 transitions from V_DDto zero volts, and REF0 transitions from V_DDto switching voltage V_SW. The V_SWBL1 voltage is applied to the switch-plate of non-volatile storage element NT1,0 by the controlled source of select device T1,0, and switching voltage V_SWis applied to the NT switch by REF0, resulting in zero electrostatic force between switch-plate and NT switch. The non-volatile storage element NT1,0 release-plate is at switching voltage zero and the NT switch is at switching voltage V_SWgenerating an attracting electrostatic force. If V_SWexceeds the nanotube threshold voltage V_NT-TH, the nanotube structure switches to the “OFF” state or logic “0” state, that is, the nanotube NT switch and the surface of the release-plate insulator are in contact as illustrated inFIG. 34C. The non-conducting contact betweeninsulator surface2160 on release-plate2180 andNT switch2140 inposition2140″ represents the “OFF” state or “0” state. If the power source is removed, cell C1,0 remains in the “OFF” state.
• An NRAM read operation does not change (destroy) the information in the activated cells, as it does in a DRAM, for example. Therefore the read operation in the NRAM is characterized as a non-destructive readout (or NDRO) and does not require a write-back after the read operation has been completed. For a read operation of cell C0,0, BL0 is driven high to V_DDand allowed to float. WL0 is driven high to V_DDand select device T0,0 turns on. REF0 is at zero volts, and RL0 is at V_DD. If cell C0,0 stores an “ON” state (“1” state) as illustrated inFIG. 34B, BL0 discharges to ground through a conductive path that includes select device T0,0 and non-volatile storage element NT0,0 in the “ON” state, the BL0 voltage drops, and the “ON” state or “1” state is detected by a sense amplifier/latch circuit (not shown) that records the voltage drop by switching the latch to a logic “1” state. BL0 is connected by the select device T0,0 conductive channel of resistance R_FETto the switch-plate of NT0,0. The switch-plate of NT0,0 in the “ON” state contacts the NT switch with contact resistance R_SWand the NT switch contacts reference line REF0 with contact resistance R_C. The total resistance in the discharge path is R_FET+R_SW+R_C. Other resistance values in the discharge path, including the resistance of the NT switch, are much small and may be neglected
• For a read operation of cell C1,0, BL1 is driven high to V_DDand allowed to float. WL0 is driven high to V_DDand select device T1,0 turns on. REF0=0, and RL1 is at V_DD. If cell C1,0 stores an “OFF” state (“0” state) as illustrated inFIG. 34C, BL1 does not discharge to ground through a conductive path that includes select device T1,0 and non-volatile storage element NT1,0 in the “OFF” state, because the switch-plate is not in contact with the NT switch when NT1,0 is in the “OFF” state, and the resistance R_Cis large. During read, BL2 to BLm−1 is at zero volts. Sense amplifier/latch circuit (not shown) does not detect a voltage drop and the latch is set to a logic “0” state.
• FIG. 36 illustrates theoperational waveforms2800 ofmemory array2700 ofFIG. 35 during read, write “1”, and write “0” operations for selected cells, while not disturbing unselected cells (no change to unselected cell stored logic states).Waveforms2800 illustrate voltages and timings to write logic state “1” in cell C0,0, write a logic state “0” in cell C1,0, read cell C0,0, and read cell C1,0.Waveforms2800 also illustrate voltages and timings to prevent disturbing the stored logic states (logic “1” state and logic “0” state) in partially selected (also referred to as half-selected) cells. Partially selected cells are cells inmemory array2700 that receive applied voltages because they are connected to (share) word, bit, reference, and release lines that are activated as part of the read or write operation to the selected cells. Cells inmemory array2700 tolerate unlimited read and write operations at each memory cell location.
• At the start of the write cycle, WL0 transitions from zero to V_SW, activating select devices T0,0, T1,0, . . . Tm−1,0. Word lines WL1, WL2, . . . WLn−1 are not selected and remain at zero volts. BL0 transitions from V_DDto zero volts, connecting the switch-plate of non-volatile storage element NT0,0 to zero volts. BL1 transitions from V_DDto V_SWconnecting the switch-plate of non-volatile storage element NT1,0 to V_SWvolts. BL2, BL3, . . . BLm−1 transition to V_SWconnecting the switch-plate of non-volatile storage elements NT2,0, NT3,0 . . . NTm−1,0 to V_WS. RL0 transitions from V_DDto switching voltage V_SW, connecting the release-plates of non-volatile storage elements NT0,0, NT0,1, . . . NT0,n−2, NT0,n−1 to V_SW. RL1 transitions from V_DDto zero volts, connecting the release-plates of non-volatile storage elements NT1,0, NT1,1 . . . NT1,n−2,NT1,n−1 to zero volts. RL2, RL3, . . . RLm−1 remain at V_DD, connecting the release-plates of non-volatile storage elements NT3,0 to NTm−1,n−1 to V_DD. REF0 transitions from V_DDto switching voltage V_SW, connecting NT switches of non-volatile storage elements NT0,0, NT1,0, . . . NTm−1,0 to V_SW. REF1, REF2 . . . REFn−1 remain at V_DD, connecting NT switches of non-volatile storage elements NT0,1 to NTn−1,n−1 to V_DD.
• NT0,0 may be in “ON” (“1” state) or “OFF” (“0” state) state at the start of the write cycle. It will be in “ON” state at the end of the write cycle. If NT0,0 in cell C0,0 is “OFF” (“0” state) it will switch to “ON” (“1” state) since the voltage difference between NT switch and release-plate is zero, and the voltage difference between NT switch and switch-plate is V_SW. If NT0,0 in cell C0,0 is in the “ON” (“1” state), it will remain in the “ON” (“1”) state. NT1,0 may be in “ON” (“1” state) or “OFF” (“0” state) state at the start of the write cycle. It will be in “OFF” state at the end of the write cycle. If NT1,0 in cell C1,0 is “ON” (“1” state) it will switch to “OFF” (“0” state) since the voltage difference between NT switch and switch-plate is zero, and the voltage difference between NT switch and release-plate is V_SW. If NT1,0 in cell C1,0 is “OFF” (“0” state), it will remain “OFF” (“0” state). If for example, V_SW=3.0 volts, V_DD=1.5 volts, and NT switch threshold voltage range is V_NT-TH=1.7 to 2.8 volts, then for NT0,0 and NT1,0 a difference voltage V_SW>V_NT-THensuring write states of “ON” (“1” state) for NT0,0 and “OFF” (“0” state) for NT1,0.
• Cells C0,0 and C1,0 have been selected for the write operation. All other cells have not been selected, and information in these other cells must remain unchanged (undisturbed). Since in an array structure some cells other than selected cells C0,0 and C1,0 inarray2700 will experience partial selection voltages, often referred to as half-select voltages, it is necessary that half-select voltages applied to non-volatile storage element terminals be sufficiently low (below nanotube activation threshold V_NT-TH) to avoid disturbing stored information. For storage cells in the “ON” state, it is also necessary to avoid parasitic current flow (there cannot be parasitic currents for cells in the “OFF” state because the NT switch is not in electrical contact with switch-plate or release-plate). Potential half-select disturb along activated array lines WL0 and REF0 includes cells C3,0 to Cm−1,0 because WL0 and REF0 have been activated. Storage elements NT3,0 to NTm−1,0 will have BL2 to BLm−1 electrically connected to the corresponding storage element switch-plate by select devices T3,0 to Tm−1,0. All NT switches in these storage elements are at write voltage V_SW. To prevent undesired switching of NT switches, RL2 to RLm−1 reference lines are set at voltage V_DD. BL2 to BLm−1 voltages are set to V_SWto prevent parasitic currents. The information in storage elements NT2,0 to NTm−1,0 in cells C2,0 to Cm−1,0 is not disturbed and there is no parasitic current. For those cells in the “OFF” state, there can be no parasitic currents (no current path), and no disturb because the voltage differences favor the “OFF” state. For those cells in the “ON” state, there is no parasitic current because the voltage difference between switch-plates (at V_DD) and NT switches (at V_DD) is zero. Also, for those cells in the “ON” state, there is no disturb because the voltage difference between corresponding NT switches and release-plate is V_SW−V_DD=1.5 volts, when V_SW=3.0 volts and V_DD=1.5 volts. Since this voltage difference of 1.5 volts is less than the minimum nanotube threshold voltage V_NT-THof 1.7 volts, no switching takes place.
• Potential half-select disturb along activated array lines RL0 and BL0 includes cells C0,1 to C0, n−1 because RL0 and BL0 have been activated. Storage elements NT0,1 to NT0, n−1 all have corresponding switch-plates connected to switching voltage V_SW. To prevent undesired switching of NT switches, REF1 to REFn−1 are set at voltage V_DD. WL1 to WL n−1 are set at zero volts, therefore select devices T0,1 to T0,n−1 are open, and switch-plates (all are connected to select device source diffusions) are not connected to bit line BL0. All switch-plates are in contact with a corresponding NT switch for storage cells in the “ON” state, and all switch plates are only connected to corresponding “floating” source diffusions for storage cells in the “OFF” state. Floating diffusions are at approximately zero volts because of diffusion leakage currents to semiconductor substrates. However, some floating source diffusions may experience disturb voltage conditions that may cause the source voltage, and therefore the switch-plate voltage, to increase up to 0.6 volts as explained further below. The information in storage elements NT0,1 to NT0,n−1 in cells C0,1 to C0,n−1 is not disturbed and there is no parasitic current. For cells in both “ON” and “OFF” states there can be no parasitic current because there is no current path. For cells in the “ON” state, the corresponding NT switch and switch-plate are in contact and both are at voltage V_SW. There is a voltage difference of V_SW−V_DDbetween corresponding NT switch and release-plate. For V_SW=3.0 volts and V_DD=1.5 volts, the voltage difference of 1.5 volts is below the minimum V_NT-TH=1.7 volts for switching. For cells in the “OFF” state, the voltage difference between corresponding NT switch and switch-plate ranges from V_DDto V_DD−0.6 volts. The voltage difference between corresponding NT switch and switch-plate may be up to 1.5 volts, which is less than V_NT-THminimum voltage of 1.7 volts, and does not disturb the “OFF” cells by switching them to the “ON” state. There is also a voltage difference between corresponding NT switch and release-plate of V_SW−V_DDof 1.5 volts with an electrostatic force that supports the “OFF” state.
• Potential half-select disturb along activated array lines RL1 and BL1 includes cells C1,1 to C1, n−1 because RL1 and BL1 have been activated. Storage elements NT1,1 to NT1, n−1 all have corresponding NT release-plates connected to zero volts. To prevent undesired switching of NT switches, REF1 to REFn−1 are set at voltage V_DD. WL1 to WL n−1 are set at zero volts, therefore select devices T1,1 to T1,n−1 are open, and switch-plates (all are connected to select device source diffusions) are not connected to bit line BL1. All switch-plates are in contact with a corresponding NT switch for storage cells in the “ON” state, and all switch plates are only connected to corresponding “floating” source diffusions for storage cells in the “OFF” state. Floating diffusions are at approximately zero volts because of diffusion leakage currents to semiconductor substrates. However, some floating source diffusions may experience disturb voltage conditions that may cause the source voltage, and therefore the switch-plate voltage, to increase up to 0.6 volts as explained further below. The information in storage elements NT1,1 to NT1,n−1 in cells C1,1 to C1,n−1 is not disturbed and there is no parasitic current. For cells in both “ON” and “OFF” states there can be no parasitic current because there is no current path. For cells in the “ON” state, the corresponding NT switch and switch-plate are in contact and both are at voltage V_DD. There is a voltage difference of V_DDbetween corresponding NT switches and release-plates. For V_DD=1.5 volts, the voltage difference of 1.5 volts is below the minimum V_NT-TH=1.7 volts for switching. For cells in the “OFF” state, the voltage of the switch-plate ranges zero to 0.6 volts. The voltage difference between corresponding NT switch and switch-plate may be up to V_DD. There is also a voltage difference between corresponding NT switch and release-plate of V_DD=1.5 volts. V_DDis less than the minimum V_NT-THof 1.7 volts the “OFF” state remains unchanged.
• For all remainingmemory cells2700, C2,1 to Cm−1,n−1, there is no electrical connection between NT2,1 to NTm−1,n−1 switch-plates connected to corresponding select device source and corresponding bit lines BL2 to BLm−1 because WL1 to WLn−1 are at zero volts, and select devices T2,1 to Tm−1,n−1 are open. Release line voltages for RL2 to RLm−1 are set at V_DDand reference line voltages for REF1 to REFn−1 are set at V_DD. Therefore, all NT switches are at V_DDand all corresponding release-plates are at V_DD, and the voltage difference between corresponding NT switches and release-plates is zero. For storage cells in the “ON” state, NT switches are in contact with corresponding switch-plates and the voltage difference is zero. For storage cells in the “OFF” state, switch-plate voltages are zero to a maximum of 0.6 volts. The maximum voltage difference between NT switches and corresponding switch-plates is V_DD=1.5 volts, which is below the V_NT-THvoltage minimum voltage of 1.7 volts. The “ON” and “OFF” states remain undisturbed.
• Non-volatile NT-on-sourceNRAM memory array2700 with bit lines parallel to release lines is shown inFIG. 35 contains 2^N×2^Mbits, is a subset of non-volatileNRAM memory system2810 illustrated asmemory array2815 inFIG. 37A.NRAM memory system2810 may be configured to operate like an industry standard asynchronous SRAM or synchronous SRAM because nanotubenon-volatile storage cells2000 shown inFIG. 34A, inmemory array2700, may be read in a non-destructive readout (NDRO) mode and therefore do not require a write-back operation after reading, and also may be written (programmed) at CMOS voltage levels (5, 3.3, and 2.5 volts, for example) and at nanosecond and sub-nanosecond switching speeds. NRAM read and write times, and cycle times, are determined by array line capacitance, and are not limited by nanotube switching speed. Accordingly,NRAM memory system2810 may be designed with industry standard SRAM timings such as chip-enable, write-enable, output-enable, etc., or may introduce new timings, for example. Non-volatileNRAM memory system2810 may be designed to introduce advantageous enhanced modes such as a sleep mode with zero current (zero power—power supply set to zero volts), information preservation when power is shut off or lost, enabling rapid system recovery and system startup, for example.NRAM memory system2810 circuits are designed to provide thememory array2700waveforms2800 shown inFIG. 36.
• NRAM memory system2810 accepts timinginputs2812, acceptsaddress inputs2825, and acceptsdata1867 from a computer, or providesdata2867 to a computer using a bidirectional bus sharing input/output (I/O) terminals. Alternatively, inputs and outputs may use separate (unshared) terminals (not shown). Address input (I/P)buffer2830 receives address locations (bits) from a computer system, for example, and latches the addresses. Address I/P buffer2830 provides word address bits toword decoder2840 viaaddress bus2837; address I/P buffer2830 provides bit addresses to bit decoder2850 viaaddress bus2852; and address bus transitions provided bybus2835 are detected by function generating, address transition detecting (ATD) timing waveform generator, controller (controller)2820.Controller2820 provides timing waveforms onbus2839 toword decoder2840.Word decoder2840 selects the word address location withinarray2815.Word address decoder2840 is used to decode both word lines WL and corresponding reference lines REF (there is no need for a separate REF decoder) and drives word line (WL) and reference line (REF)select logic2845.Controller2820 provides function and timing inputs onbus2843 to WL & REF selectlogic2845, resulting inNRAM memory system2810 on-chip WL and REF waveforms for both write-one, write-zero, read-one, and read-zero operations as illustrated bywaveforms2800′ shown inFIG. 38.FIG. 38NRAM memory system2810waveforms2800′ correspond tomemory array2700waveforms2800 shown inFIG. 36.
• Bit address decoder2850 is used to decode both bit lines BL and corresponding release lines RL (there is no need for a separate RL decoder) and drive bit line (BL) and release (RL)select logic2855 viabus2856.Controller2820 provides timing waveforms onbus2854 tobit decoder2850.Controller2820 also provides function and timing inputs onbus2857 to BL & RLselect logic2855. BL & RLselect logic2855 uses inputs frombus2856 andbus2857 to generate data multiplexer select bits onbus2859. The output of BL and RLselect logic2855 onbus2859 is used to select control data multiplexers using combined data multiplexers & sense amplifiers/latches (MUXs & SAs)2860.Controller2820 provides function and timing inputs onbus2862 to MUXs &SAs2860, resulting inNRAM memory system2810 on-chip BL and RL waveforms for both write-one, write-zero, read-one, and read-zero operations as illustrated bywaveforms2800′ corresponding tomemory array2700waveforms2800 shown inFIG. 36. MUXs &SAs2860 are used to write data provided by read/write buffer2865 viabus2864 inarray2815, and to read data fromarray2815 and provide the data to read/write buffer2865 viabus2864 as illustrated inwaveforms2800′.
• Sense amplifier/latch2900 is illustrated inFIG. 37B.Flip flop2910, comprising two back-to-back inverters is used to amplify and latch data inputs fromarray2815 or from read/write buffer2865.Transistor2920 connectsflip flop2910 to ground when activated by a positive voltage supplied bycontrol voltage V_TIMING2980, which is provided bycontroller2820.Gating transistor2930 connects a bit line BL tonode2965 offlip flop2910 when activated by a positive voltage.Gating transistor2940 connects reference voltage V_REFto flipflop node2975 when activated by a positive voltage.Transistor2960 connects voltage V_DDto flipflop2910node2965,transistor2970 connects voltage V_DDto flipflop2910node2975, andtransistor2950 ensures that small voltage differences are eliminated whentransistors2960 and2970 are activated.Transistors2950,2960, and2970 are activated (turned on) when gate voltage is low (zero, for example).
• In operation, V_TIMINGvoltage is at zero volts whensense amplifier2900 is not selected.NFET transistors2920,2930, and2940 are in the “OFF” (non-conducting) state, because gate voltages are at zero volts.PFET transistors2950,2960, and2970 are in the “ON” (conducting) state because gate voltages are at zero volts. V_DDmay be 5, 3.3, or 2.5 volts, for example, relative to ground.Flip flop2910nodes2965 and2975 are at V_DD. If sense amplifier/latch2900 is selected, V_TIMINGtransitions to V_DD,NFET transistors2920,2930, and2940 turn “ON”,PFET transistors2950,2960, and2970 are turned “OFF”, andflip flop2910 is connected to bit line BL and reference voltage V_REF. V_REFis connected to V_DDin this example. As illustrated by waveforms BL0 and BL1 ofwaveforms2800′, bit line BL is pre-charged prior to activating a corresponding word line (WL0 in this example). Ifcell2000 of memory array2700 (memory system array2815) stores a “1”, then bit line BL inFIG. 37B corresponds to BL0 inFIG. 38, BL is discharged bycell2000, voltage droops below V_DD, and sense amplifier/latch2900 detects a “1” state. Ifcell2000 of memory array2700 (memory system array2815) stores a “0”, then bit line BL inFIG. 37B corresponds to BL1 inFIG. 38, BL is not discharged bycell2000, voltage does not droop below V_DD, and sense amplifier/latch2900 detect a “0” state. The time from sense amplifier select to signal detection by sense amplifier/latch2900 is referred to as signal development time. Sense amplifier/latch2900 typically requires 100 to 200 mV relative to V_REFin order to switch. It should be noted thatcell2000 requires a nanotube “OFF” resistance to “ON” resistance ratio of greater than 10 to 1 for successful operation. A typical bit line BL has a capacitance value of 250 fF, for example. A typical nanotube storage device (switch) or dimensions 0.2 by 0.2 um typically has 8 nanotube filaments across the suspended region, for example, as illustrated further below. For a combined contact and switch resistance of 50,000 Ohms per filament, as illustrated further below, the nanotube “ON” resistance ofcell2000 is 6,250 Ohms. For a bit line of 250 fF, the time constant RC=1.6 ns. The sense amplifier signal development time is less than RC, and for this example, is between 1 and 1.5 nanoseconds.
• Non-volatileNRAM memory system2810 operation may be designed for high speed cache operation at 5 ns or less access and cycle time, for example. Non-volatileNRAM memory system2810 may be designed for low power operation at 60 or 70 ns access and cycle time operation, for non-limiting example. For low power operation, address I/P buffer2830 operation typically requires 8 ns;controller2820 operation requires 16 ns;bit decoder2850 operation plus BL & RLselect logic2855 plus MUXs &SA2860 operation requires 12 ns (word decoder2840 operation plus WL & RLselect logic 2845 ns require less than 12 ns);array2815 delay is 8 ns; operation ofsense latch2900 requires 8 ns; and read/write buffer2865 requires 12 ns, for non-limiting example. The access time and cycle time of non-volatileNRAM memory system2810 is 64 ns. The access time and cycle time may be equal because the NDRO mode of operation of nanotube storage devices (switches) does not require a write-back operation after access (read).
• NT-on-source arrays with bit lines BL parallel to release lines RL and reference lines REF parallel to word lines WL may be fabricated by applying methods illustrated previously illustrated above to fabricate preferred NT-on-source arrays with BLs parallel to REF lines and WLs parallel to RLs. Examples of preferred NT-on-source arrays with BLs parallel to REF lines and WLs parallel to RLs are illustrated byarray3225 in FIGS.30M′,30N, and30P;array3229 shown inFIGS. 32A-32C, andarray3231 shown inFIGS. 33A-33D. The methods used to fabricatearrays3225,3229, and3231 may be used to fabricate NT-on-source arrays with BLs parallel to RLs, and WLs parallel to REF lines. These methods includemethods3000 shown inFIG. 22 and corresponding figures and structures;methods3004 shown inFIGS. 23 and 23′ and corresponding figures and structures;methods3036 shown inFIG. 26 and corresponding figures and structures;methods3006 shown inFIGS. 27 and 27′ and corresponding figures and structures;methods3008 shown inFIGS. 28 and 28′ and corresponding figures and structures; and other methods and structures illustrated in fabricatingarrays3225,3229, and3231 as described above.
Nanotube Random Access Memory Using FEDs with Controllable DrainsNanotube Random Access Memory (NRAM) Systems and Circuits, with Same
• Non-volatile field effect devices (FEDs)100,120,140, and160 with controllable drains may be used as cells and interconnected into arrays to form non-volatile nanotube random access memory (NRAM) systems. The memory cells contain one select device (transistor) T and one non-volatile nanotube storage element NT (1T/1NT cells). By way of example, FED8160 (FIG. 2H) is used to form a non-volatile NRAM memory cell that is also referred to as a NT-on-Drain memory cell.
NT-on-Drain NRAM Memory Systems and Circuits with Parallel Bit and Reference Lines, and Parallel Word and Release Lines
• NRAM 1T/1NT memory arrays are wired using four lines. Word line WL is used to gate select device T, reference line REF is attached to a shared source between two adjacent select devices. Bit line BL is used to control NT switch voltage of storage element NT, and release line RL is used to control the release-plate of storage element NT. In this NRAM array configuration, REF is parallel to BL and acts as second bit line, and RL is parallel to WL and acts as a second word line.
• FIG. 39A depicts non-volatilefield effect device160 with memory cell wiring to form NT-on-Drain memory cell4000 schematic. Word line (WL)4200 connects to terminal T1 ofFED8160; bit line (BL)4400 connects to terminal T2 orFED8160; reference line (REF)4300 connects to terminal T3 ofFED8160; and release line (RL)4500 connects to terminal T4 ofFED8160.Memory cell4000 performs write and read operations, and stores the information in a non-volatile state. TheFED8160 layout dimensions and operating voltages are selected to optimizememory cell4000.Memory cell4000 FET select device (T)gate4040 corresponds togate162;controllable drain4080 corresponds tocontrollable drain164; andsource4060 corresponds to source166.Memory cell4000 nanotube (NT) switch-plate4120 corresponds to switch-plate168;NT switch4140 corresponds toNT switch170; release-plateinsulator layer surface4160 corresponds to release-plateinsulator layer surface176; and release-plate4180 corresponds to release-plate174. The interconnections between the elements ofmemory cell4000 schematic correspond to the interconnection of the corresponding interconnections of the elements ofFED8160.REF4300 connects to source4060 throughcontact4320;BL4400 connects toNT switch4140 throughcontact4420;RL4500 connects to release-plate4180 bycontact4520;WL4200 interconnects togate4040 bycontact4220. The non-volatileNT switching element4140 may be caused to deflect toward switch-plate4120 via electrostatic forces to closed (“ON”)position4140′ to store a logic “1” state as illustrated inFIG. 39B. The van der Waals force holdsNT switch4140 inposition4140′. Alternatively, the non-volatileNT switching element4140 may be caused to deflect toinsulator surface4160 on release-plate4180 via electrostatic forces to open (“OFF”)position4140″ to store a logic “0” state as illustrated inFIG. 39C. The van der Waals force holdsNT switch4140 inposition4140″. Non-volatileNT switching element4140 may instead be caused to deflect to an open (“OFF”) near-mid point position4140′″ between switch-plate4120 and release-plate4180, storing an apparent logic “0” state as illustrate inFIG. 24D. However, the absence of a van der Waals retaining force in this open (“OFF”) position is likely to result in a memory cell disturb that causesNT switch4140 to unintentionally transition to the closed (“ON”) position, and is not desirable. Sufficient switching voltage is needed to ensure that theNT switch4140 open (“OFF”) position isposition4140″. The non-volatile element switching via electrostatic forces is as depicted byelement170 inFIG. 2H.Voltage waveforms355 used to generate the required electrostatic forces are illustrated inFIG. 11.
• NT-on-Drain memory cell schematic4000 forms the basis of a non-volatile storage (memory) cell. The device may be switched between closed storage state “1” (switched to position4140′) and open storage state “0” (switched toposition4140″), which means the controllable drain may be written to an unlimited number of times to as desired. In this way, the device may be used as a basis for a non-volatile nanotube random access memory, which is referred to here as a NRAM array, with the ‘N’ representing the inclusion of nanotubes.
• FIG. 40 represents anNRAM system4700, according to preferred embodiments of the invention. Under this arrangement, an array is formed with m×n (only exemplary portion being shown) of non-volatile cells ranging from cell C0,0 to cell Cm−1,n−1.NRAM system4700 may be designed using one large m×n array, or several smaller sub-arrays, where each sub-array if formed of m×n cells. To access selected cells, the array uses read and write word lines (WL0, WL1, . . . WLn−1), read and write bit lines (BL0, BL1, . . . BLm−1), read and write reference lines (REF0, REF1, . . . REFm−1), and read and write release lines (RL0, RL1, . . . RLn−1). Non-volatile cell C0,0 includes a select device T0,0 and non-volatile storage element NT0,0. The gate of T0,0 is coupled to WL0, and the source of T0,0 is coupled to REF0. NT0 is the non-volatilely switchable storage element where the NT0,0 switch-plate is coupled to the drain of T0,0, the switching NT element is coupled to BL0, and the release-plate is coupled to RL0.Connection4720 connects REF0 to shared source of select devices T0,0 and T0,1. Word, bit, reference, and release decoders/drivers are explained further below.
• Under preferred embodiments, nanotubes inarray4700 may be in the “ON” “1” state or the “OFF” “0” state. The NRAM memory allows for unlimited read and write operations per bit location. A write operation includes both a write function to write a “1” and a release function to write a “0”. By way of example, a write “1” to cell C0,0 and a write “0” to cell C1,0 is described. For a write “1” operation to cell C0,0, select device T0,0 is activated when WL0 transitions from 0 to V_DD, REF0 transitions from V_DDto 0 volts, BL0 transitions from V_DDto switching voltage V_SW, and RL0 transitions from V_DDto switching voltage V_SW. The release-plate and NT switch of the non-volatile storage element NT0,0 are each at V_SWresulting in zero electrostatic force (because the voltage difference is zero). The zero REF0 voltage is applied to the switch-plate of non-volatile storage element NT0,0 by the controlled drain of select device T0,0. The difference in voltage between the NT0,0 switch-plate and NT switch is V_SWand generates an attracting electrostatic force. If V_SWexceeds the nanotube threshold voltage V_NT-TH, the nanotube structure switches to “ON” state or logic “1” state, that is, the nanotube NT switch and switch-plate are electrically connected as illustrated inFIG. 39B. The near-Ohmic connection between switch-plate4120 andNT switch4140 inposition4140′ represents the “ON” state or “1” state. If the power source is removed, cell C0,0 remains in the “ON” state.
• For a write “0” (release) operation to cell C1,0, select device T1,0 is activated when WL0 transitions from 0 to V_DD, REF1 transitions from V_DDto 0 volts, BL1 transitions from V_DDto zero volts, and RL0 transitions from V_DDto switching voltage V_SW. The zero REF1 voltage is applied to the switch-plate of non-volatile storage element NT1,0 by the controlled drain of select device T1,0, and zero volts is applied the NT switch by BL1, resulting in zero electrostatic force between switch-plate and NT switch. The non-volatile storage element NT1,0 release-plate is at switching voltage V_SWand the NT switch is at zero volts generating an attracting electrostatic force. If V_SWexceeds the nanotube threshold voltage V_NT-TH, the nanotube structure switches to the “OFF” state or logic “0” state, that is, the nanotube NT switch and the surface of the release-plate insulator are in contact as illustrated inFIG. 39C. The non-conducting contact betweeninsulator surface4160 on release-plate4180 andNT switch4140 inposition4140″ represents the “OFF” state or “0” state. If the power source is removed, cell C1,0 remains in the “OFF” state.
• An NRAM read operation does not change (destroy) the information in the activated cells, as it does in a DRAM, for example. Therefore the read operation in the NRAM is characterized as a non-destructive readout (or NDRO) and does not require a write-back after the read operation has been completed. For a read operation of cell C0,0, BL0 is driven high to V_DDand allowed to float. WL0 is driven high to V_DDand select device T0,0 turns on. REF0 is at zero volts, and RL0 is at V_DD. If cell C0,0 stores an “ON” state (“1” state) as illustrated inFIG. 39B, BL0 discharges to ground through a conductive path that includes select device T0,0 and non-volatile storage element NT0,0 in the “ON” state, the BL0 voltage drops, and the “ON” state or “1” state is detected by a sense amplifier/latch circuit (not shown) that records the voltage drop by switching the latch to a logic “1” state. REF0 is connected by the select device T0,0 conductive channel of resistance R_FETto the switch-plate of NT0,0. The switch-plate of NT0,0 in the “ON” state contacts the NT switch with contact resistance R_SWand the NT switch contacts bit line BL0 with contact resistance R_C. The total resistance in the discharge path is R_FET+R_SW+R_C. Other resistance values in the discharge path, including the resistance of the NT switch, are much small and may be neglected
• For a read operation of cell C1,0, BL1 is driven high to V_DDand allowed to float. WL0 is driven high to V_DDand select device T1,0 turns on. REF1=0, and RL0 is at V_DD. If cell C1,0 stores an “OFF” state (“0” state) as illustrated inFIG. 39C, BL1 does not discharge to ground through a conductive path that includes select device T1,0 and non-volatile storage element NT1,0 in the “OFF” state, because the switch-plate is not in contact with the NT switch when NT1,0 is in the “OFF” state, and the resistance R_Cis large. Sense amplifier/latch circuit (not shown) does not detect a voltage drop and the latch is set to a logic “0” state.
• FIG. 41 illustrates theoperational waveforms4800 ofmemory array4700 ofFIG. 40 during read, write “1”, and write “0” operations for selected cells, while not disturbing unselected cells (no change to unselected cell stored logic states).Waveforms4800 illustrate voltages and timings to write logic state “1” in cell C0,0, write a logic state “0” in cell C1,0, read cell C0,0, and read cell C1,0.Waveforms4800 also illustrate voltages and timings to prevent disturbing the stored logic states (logic “1” state and logic “0” state) in partially selected (also referred to as half-selected) cells. Partially selected cells are cells inmemory array4700 that receive applied voltages because they are connected to (share) word, bit, reference, and release lines that are activated as part of the read or write operation to the selected cells. Cells inmemory array4700 tolerate unlimited read and write operations at each memory cell location.
• At the start of the write cycle, WL0 transitions from zero to V_DD, activating select devices T0,0, T1,0, . . . Tm−1,0. Word lines WL1, WL2 . . . WLn−1 are not selected and remain at zero volts. REF0 transitions from V_DDto zero volts, connecting the switch-plate of non-volatile storage element NT0,0 to zero volts. REF1 transitions from V_DDto zero volts connecting the switch-plate of non-volatile storage element NT1,0 to zero volts. REF2, REF3 . . . REFm−1 remain at V_DDconnecting the switch-plate of non-volatile storage elements NT2,0, NT3,0 . . . NTm−1,0 to V_DD. BL0 transitions from V_DDto switching voltage V_SW, connecting the NT switches of non-volatile storage elements NT0,0, NT0,1 . . . NT0,n−2, NT0,n−1 to V_SW. BL1 transitions from V_DDto zero volts, connecting the NT switches of non-volatile storage elements NT1,0, NT1,1 . . . NT1,n−2,NT1,n−1 to zero volts. BL2, BL3 . . . BLm−1 remain at V_DD, connecting the NT switches of non-volatile storage elements NT3,0 to NTm−1,n−1 to V_DD. RL1, RL2 . . . RLn−1 remain at V_DD, connecting release-plates of non-volatile storage elements NT0,1 to NTn−1,n−1 to V_DD.
• NT0,0 may be in “ON” (“1” state) or “OFF” (“0” state) state at the start of the write cycle. It will be in “ON” state at the end of the write cycle. If NT0,0 in cell C0,0 is “OFF” (“0” state) it will switch to “ON” (“1” state) since the voltage difference between NT switch and release-plate is zero, and the voltage difference between NT switch and switch-plate is V_SW. If NT0,0 in cell C0,0 is in the “ON” (“1” state), it will remain in the “ON” (“1”) state. NT1,0 may be in “ON” (“1” state) or “OFF” (“0” state) state at the start of the write cycle. It will be in “OFF” state at the end of the write cycle. If NT1,0 in cell C1,0 is “ON” (“1” state) it will switch to “OFF” (“0” state) since the voltage difference between NT switch and switch-plate is zero, and the voltage difference between NT switch and release-plate is V_SW. If NT1,0 in cell C1,0 is “OFF” (“0” state), it will remain “OFF” (“0” state). If for example, V_SW=3.0 volts, V_DD=1.5 volts, and NT switch threshold voltage range is V_NT-TH=1.7 to 2.8 volts, then for NT0,0 and NT1,0 a difference voltage V_SW>V_NT-THensuring write states of “ON” (“1” state) for NT0,0 and “OFF” (“0” state) for NT1,0.
• Cells C0,0 and C1,0 have been selected for the write operation. All other cells have not been selected, and information in these other cells must remain unchanged (undisturbed). Since in an array structure some cells other than selected cells C0,0 and C1,0 inarray4700 will experience partial selection voltages, often referred to as half-select voltages, it is necessary that half-select voltages applied to non-volatile storage element terminals be sufficiently low (below nanotube activation threshold V_NT-TH) to avoid disturbing stored information. For storage cells in the “ON” state, it is also necessary to avoid parasitic current flow (there cannot be parasitic currents for cells in the “OFF” state because the NT switch is not in electrical contact with switch-plate or release-plate). Potential half-select disturb along activated array lines WL0 and RL0 includes cells C3,0 to Cm−1,0 because WL0 and RL0 have been activated. Storage elements NT3,0 to NTm−1,0 will have REF2 to REFm−1 electrically connected to the corresponding storage element switch-plate by select devices T3,0 to Tm−1,0. All release-plates in these storage elements are at write voltage V_SW. To prevent undesired switching of NT switches, BL2 to BLm−1 reference lines are set at voltage V_DD. REF2 to REFm−1 voltages are set to V_DDto prevent parasitic currents. The information in storage elements NT2,0 to NTm−1,0 in cells C2,0 to Cm−1,0 is not disturbed and there is no parasitic current. For those cells in the “OFF” state, there can be no parasitic currents (no current path), and no disturb because the voltage differences favor the “OFF” state. For those cells in the “ON” state, there is no parasitic current because the voltage difference between switch-plates (at V_DD) and NT switches (at V_DD) is zero. Also, for those cells in the “ON” state, there is no disturb because the voltage difference between corresponding NT switches and release-plate is V_SW−V_DD=1.5 volts, when V_SW=3.0 volts and V_DD=1.5 volts. Since this voltage difference of 1.5 volts is less than the minimum nanotube threshold voltage V_NT-THof 1.7 volts, no switching takes place.
• Potential half-select disturb along activated array lines REF0 and BL0 includes cells C0,1 to C0, n−1 because REF0 and BL0 have been activated. Storage elements NT0,1 to NT0, n−1 all have corresponding NT switches connected to switching voltage V_SW. To prevent undesired switching of NT switches, RL1 to RLn−1 are set at voltage V_DD. WL1 to WL n−1 are set at zero volts, therefore select devices T0,1 to T0,n−1 are open, and switch-plates (all are connected to select device drain diffusions) are not connected to bit line REF0. All switch-plates are in contact with a corresponding NT switch for storage cells in the “ON” state, and all switch plates are only connected to corresponding “floating” drain diffusions for storage cells in the “OFF” state. Floating diffusions are at approximately zero volts because of diffusion leakage currents to semiconductor substrates. However, some floating source diffusions may experience disturb voltage conditions that may cause the source voltage, and therefore the switch-plate voltage, to increase up to 0.6 volts as explained further below. The information in storage elements NT0,1 to NT0,n−1 in cells C0,1 to C0,n−1 is not disturbed and there is no parasitic current. For cells in both “ON” and “OFF” states there can be no parasitic current because there is no current path. For cells in the “ON” state, the corresponding NT switch and switch-plate are in contact and both are at voltage V_SW. There is a voltage difference of V_SW−V_DDbetween corresponding NT switch and release-plate. For V_SW=3.0 volts and V_DD=1.5 volts, the voltage difference of 1.5 volts is below the minimum V_NT-TH=1.7 volts for switching. For cells in the “OFF” state, the voltage difference between corresponding NT switch and switch-plate ranges from V_SWto V_SW−0.6 volts. The voltage difference between corresponding NT switch and switch-plate may be up to 3.0 volts, which exceeds the V_NT-THvoltage, and would disturb “OFF” cells by switching them to the “ON” state. However, there is also a voltage difference between corresponding NT switch and release-plate of V_SW−V_DDof 1.5 volts with an electrostatic force in the opposite direction that prevents the disturb of storage cells in the “OFF” state. Also very important is thatNT switching element4140 is inposition4140″ in contact with the storage-plate dielectric, a short distance from the storage plate, thus maximizing the electric field that opposes cell disturb. Switch-plate4140 is far from theNT switching element4140 switch greatly reducing the electric field that promotes disturb. In addition, the van der Waals force also must be overcome to disturb the cell.
• Potential half-select disturb along activated array lines REF1 and BL1 includes cells C1,1 to C1, n−1 because REF1 and BL1 have been activated. Storage elements NT1,1 to NT1, n−1 all have corresponding NT switches connected to zero volts. To prevent undesired switching of NT switches, RL1 to RLn−1 are set at voltage V_DD. WL1 to WL n−1 are set at zero volts, therefore select devices T1,1 to T1,n−1 are open, and switch-plates (all are connected to select device drain diffusions) are not connected to reference line REF1. All switch-plates are in contact with a corresponding NT switch for storage cells in the “ON” state, and all switch plates are only connected to corresponding “floating” drain diffusions for storage cells in the “OFF” state. Floating diffusions are at approximately zero volts because of diffusion leakage currents to semiconductor substrates. However, some floating source diffusions may experience disturb voltage conditions that may cause the source voltage, and therefore the switch-plate voltage, to increase up to 0.6 volts as explained further below. The information in storage elements NT1,1 to NT1,n−1 in cells C1,1 to C1,n−1 is not disturbed and there is no parasitic current. For cells in both “ON” and “OFF” states there can be no parasitic current because there is no current path. For cells in the “ON” state, the corresponding NT switch and switch-plate are in contact and both are at zero volts. There is a voltage difference of V_DDbetween corresponding NT switch and release-plate. For V_DD=1.5 volts, the voltage difference of 1.5 volts is below the minimum V_NT-TH=1.7 volts for switching. For cells in the “OFF” state, the voltage of the switch-plate ranges zero to 0.6 volts. The voltage difference between corresponding NT switch and switch-plate may be up to 0.6 volts. There is also a voltage difference between corresponding NT switch and release-plate of V_DD=1.5 volts. V_DDis less than the minimum V_NT-THof 1.7 volts the “OFF” state remains unchanged.
• For all remaining cells ofmemory array4700, cells C2,1 to Cm−1,n−1, there is no electrical connection between NT2,1 to NTm−1,n−1 switch-plates connected to corresponding select device drain and corresponding reference lines REF2 to REFm−1 because WL1 to WLn−1 are at zero volts, and select devices T2,1 to Tm−1,n−1 are open. Bit line voltages for BL2 to BLm−1 are set at V_DDand release line voltages for RL1 to RLn−1 are set at V_DD. Therefore, all NT switches are at V_DDand all corresponding release-plates are at V_DD, and the voltage difference between corresponding NT switches and release-plates is zero. For storage cells in the “ON” state, NT switches are in contact with corresponding switch-plates and the voltage difference is zero. For storage cells in the “OFF” state, switch plate voltages are zero to a maximum of 0.6 volts. The maximum voltage difference between NT switches and corresponding switch-plates is V_DD=1.5 volts, which is below the V_NT-THvoltage minimum voltage of 1.7 volts. The “ON” and “OFF” states remain undisturbed.
• Non-volatile NT-on-drainNRAM memory array4700 with bit lines parallel to reference lines is shown inFIG. 40 contains 2^N×2^Mbits, is a subset of non-volatileNRAM memory system4810 illustrated asmemory array4815 inFIG. 42A.NRAM memory system4810 may be configured to operate like an industry standard asynchronous SRAM or synchronous SRAM because nanotube non-volatile storage cells of memory cell schematic4000 shown inFIG. 39A, inmemory array4700, may be read in a non-destructive readout (NDRO) mode and therefore do not require a write-back operation after reading, and also may be written (programmed) at CMOS voltage levels (5, 3.3, and 2.5 volts, for example) and at nanosecond and sub-nanosecond switching speeds. NRAM read and write times, and cycle times, are determined by array line capacitance, and are not limited by nanotube switching speed. Accordingly,NRAM memory system4810 may be designed with industry standard SRAM timings such as chip-enable, write-enable, output-enable, etc., or may introduce new timings, for example. Non-volatileNRAM memory system4810 may be designed to introduce advantageous enhanced modes such as a sleep mode with zero current (zero power—power supply set to zero volts), information preservation when power is shut off or lost, enabling rapid system recovery and system startup, for example.NRAM memory system4810 circuits are designed to provide thememory array4700waveforms4800 shown inFIG. 41.
• FigureNRAM memory system4810 accepts timinginputs4812, acceptsaddress inputs4825, and acceptsdata4867 from a computer, or providesdata4867 to a computer using a bidirectional bus sharing input/output (I/O) terminals. Alternatively, inputs and outputs may use separate (unshared) terminals (not shown). Address input (I/P)buffer4830 receives address locations (bits) from a computer system, for example, and latches the addresses. Address I/P buffer4830 provides word address bits toword decoder4840 viaaddress bus4837; address I/P buffer4830 provides bit addresses to bit decoder4850 viaaddress bus4852; and address bus transitions provided bybus4835 are detected by function generating, address transition detecting (ATD), timing waveform generator, controller (controller)4820.Controller4820 provides timing waveforms onbus4839 toword decoder4840.Word decoder4840 selects the word address location withinarray4815.Word address decoder4840 is used to decode both word lines WL and corresponding release lines RL (there is no need for a separate RL decoder) and drives word line (WL) and release line (RL)select logic4845.Controller4820 provides function and timing inputs onbus4843 to WL & RLselect logic4845, resulting inNRAM memory system4810 on-chip WL and RL waveforms for both write-one, write-zero, read-one, and read-zero operations as illustrated bywaveforms4800′ shown inFIG. 43.FIG. 43NRAM memory system4810waveforms4800′ correspond tomemory array4700waveforms4800 shown inFIG. 41.
• Bit address decoder4850 is used to decode both bit lines BL and corresponding reference lines REF (there is no need for a separate REF decoder) and drive bit line (BL) and reference (REF)select logic4855 viabus4856.Controller4820 provides timing waveforms onbus4854 tobit decoder4850.Controller4820 also provides function and timing inputs onbus4857 to BL & REF selectlogic4855. BL & REF selectlogic4855 uses inputs frombus4856 andbus4857 to generate data multiplexer select bits onbus4859. The output of BL and REFselect logic4855 onbus4859 is used to select control data multiplexers using combined data multiplexers & sense amplifiers/latches (MUXs & SAs)4860.Controller4820 provides function and timing inputs onbus4862 to MUXs &SAs4860, resulting inNRAM memory system4810 on-chip BL and REF waveforms for both write-one, write-zero, read-one, and read-zero operations as illustrated bywaveforms4800′ corresponding tomemory array4700waveforms4800 shown inFIG. 41. MUXs &SAs4860 are used to write data provided by read/write buffer4865 viabus4864 inarray4815, and to read data fromarray4815 and provide the data to read/write buffer4865 viabus4864 as illustrated inwaveforms4800′ ofFIG. 43A.
• Sense amplifier/latch4900 is illustrated inFIG. 42B.Flip flop4910, comprising two back-to-back inverters is used to amplify and latch data inputs fromarray4815 or from read/write buffer4865.Transistor4920 connectsflip flop4910 to ground when activated by a positive voltage supplied bycontrol voltage V_TIMING4980, which is provided bycontroller4820.Gating transistor4930 connects a bit line BL tonode4965 offlip flop4910 when activated by a positive voltage.Gating transistor4940 connects reference voltage V_REFto flipflop node4975 when activated by a positive voltage.Transistor4960 connects voltage V_DDto flipflop4910node4965,transistor4970 connects voltage V_DDto flipflop4910node4975, andtransistor4950 ensures that small voltage differences are eliminated whentransistors4960 and4970 are activated.Transistors4950,4960, and4970 are activated (turned on) when gate voltage is low (zero, for example).
• In operation, V_TIMINGvoltage is at zero volts whensense amplifier4900 is not selected.NFET transistors4920,4930, and4940 are in the “OFF” (non-conducting) state, because gate voltages are at zero volts.PFET transistors4950,4960, and4970 are in the “ON” (conducting) state because gate voltages are at zero volts. V_DDmay be 5, 3.3, or 2.5 volts, for example, relative to ground.Flip flop4910nodes4965 and4975 are at V_DD. If sense amplifier/latch4900 is selected, V_TIMINGtransitions to V_DD,NFET transistors4920,4930, and4940 turn ON,PFET transistors4950,4960, and4970 are turned “OFF”, andflip flop4910 is connected to bit line BL and reference voltage V_REF. V_REFis connected to V_DDin this example. As illustrated by waveforms BL0 and BL1 ofwaveforms4800′, bit line BL is pre-charged prior to activating a corresponding word line (WL0 in this example). Ifmemory cell4000 of memory array4700 (memory system array4815) stores a “1”, then bit line BL inFIG. 42B corresponds to BL0 inFIG. 43, BL is discharged bycell4000, voltage droops below V_DD, and sense amplifier/latch4900 detects a “1” state. Ifcell4000 of memory array4700 (memory system array4815) stores a “0”, then bit line BL inFIG. 42B corresponds to BL1 inFIG. 43, BL is not discharged bycell4000, voltage does not droop below V_DD, and sense amplifier/latch4900 detect a “0” state. The time from sense amplifier select to signal detection by sense amplifier/latch4900 is referred to as signal development time. Sense amplifier/latch4900 typically requires 100 to 200 mV relative to V_REFin order to switch. It should be noted thatcell4000 requires a nanotube “OFF” resistance to “ON” resistance ratio of greater than 10 to 1 for successful operation. A typical bit line BL has a capacitance value of 250 fF, for example. A typical nanotube storage device (switch) or dimensions 0.2 by 0.2 um typically has 8 nanotube filaments across the suspended region, for example, as illustrated further below. For a combined contact and switch resistance of 50,000 Ohms per filament, as illustrated further below, the nanotube “ON” resistance ofcell1000 is 6,250 Ohms. For a bit line of 250 fF, the time constant RC=1.6 ns. The sense amplifier signal development time is less than RC, and for this example, is between 1 and 1.5 nanoseconds.
• Non-volatileNRAM memory system4810 operation may be designed for high speed cache operation at 5 ns or less access and cycle time, for example. Non-volatileNRAM memory system4810 may be designed for low power operation at 60 or 70 ns access and cycle time operation, for example. For low power operation, address I/P buffer4830 operation requires 8 ns;controller4820 operation requires 16 ns;bit decoder4850 operation plus BL &select logic4855 plus MUXs &SA4860 operation requires 12 ns (word decoder4840 operation plus WL & RLselect logic 4845 ns require less than 12 ns);array4815 delay is 8 ns; sensing operation ofsense amplifier latch4900 requires 8 ns; and read/write buffer4865 requires 12 ns, for example. The access time and cycle time of non-volatileNRAM memory system4810 is 64 ns. The access time and cycle time may be equal because the NDRO mode of operation of nanotube storage devices (switches) does not require a write-back operation after access (read).
Method of Making Field Effect Device with Controllable Drain and NT-on-Drain Memory System and Circuits with Parallel Bit and Reference Array Lines, and Parallel Word and Release Array Lines
• Methods of fabricating NT-on-drain memory arrays are the same as those used to fabricate NT-on-source memory arrays.Methods3000 shown inFIG. 22 and associated figures;methods3004 shown inFIGS. 23 and 23′ and associated figures;methods3036 shown inFIG. 26 and associated figures;methods3006 shown inFIGS. 27 and 27′ and associated figures;methods3008 shown inFIGS. 28 and 28′ and associated figures; andmethods3144 as illustrated inFIGS. 31A-31D. Conductors, semiconductors, insulators, and nanotubes are formed in the same sequence and are in the same relative position in the structure. Length, widths, thickness dimensions may be different, reflecting differences in design choices. Also, conductor materials may be different, for example. The function of some electrodes may be different for NT-on-source and NT-on-drain memory arrays. For example, bit array lines and reference lines connect to different electrodes in the nanotube structure as may be seen further below. Also, connections to source and drain diffusions are different. For NT-on-source memory arrays, the switch-plate of the nanotube structure is connected to the source diffusion of the FET device. However, for NT-on-drain memory arrays, the switch-plate of the nanotube structure is connected to the drain diffusion of the FET device, as may be seen further below. Differences between NT-on-source and NT-on-drain memory arrays may be seen by comparing figures:30M′ and44A;FIGS. 30N and 44B;FIGS. 30P and 44C;FIGS. 32A and 45A;FIGS. 32B and 45B;FIGS. 32C and 45C;FIGS. 33A and 46A;FIGS. 33B and 46B;FIGS. 33C and 46C; andFIGS. 33D and 46D.
• FIG. 44A illustrates cross section A-A′ ofarray4725 taken at A-A′ of the plan view ofarray4725 illustrated inFIG. 44C, and showsFET device region3237′ in the FET length direction,nanotube switch structure3233′, interconnections and insulators.FIG. 44B illustrates cross section B-B′ ofarray4725 taken at B-B′ of plan view ofarray4725 illustrated inFIG. 44C, and shows arelease array line3205′, abit array line3119′/3117′ composed of combinedconductors3119′ and3117′, and aword array line3120′.FIG. 44C illustrates a plan view ofarray4725 including exemplary cell4765 region,reference array line3138″ contactingsource3126′ throughcontact3140′ tostud3118A′, tostud3118′, to contact3123′ (3118A′, tostud3118′, to contact3123′ not shown in plan view4725), and to source3126′.Bit array line3119′/3117′ is parallel toreference line3138″, is illustrated in cross section inFIG. 44B, and contacts a corresponding bit line segment in the picture frame region formed by combinedconductors3117′ and3119′, in contact withnanotube3114′, as shown inFIG. 44A.Release array line3205′ is parallel toword array line3120′.Release line3205′ contacts and forms a portion ofrelease electrode3205′ as illustrated in the nanotube switching region ofFIG. 44A. This nanotube switching region is illustrated asnanotube switch structure3233′ inarray4725 ofFIG. 44C. In terms of minimum technology feature size, NT-on-drain cell4765 is approximately 12 to 13 F². Nanotube-on-drain array4725 structures illustrated inFIGS. 44A, 44B, and44C correspond to nanotube-on-drain array4700 schematic representations illustrated inFIG. 40.Bit line3119′/3117′ structures correspond to any of bit lines BL0 to BLm−1 schematic representations;reference line3138″ structures correspond to any of reference lines REF0 to REFm−1 schematic representations;word line3120′ structures correspond to any of word lines WL0 to WLn−1 schematic representations;release line3205′ structures correspond to any of release lines RL0 to RLn−1 schematic representations;source contact3140′ structures correspond to any ofsource contacts4720 schematic representations;nanotube switch structures3233′ correspond to any of NT0,0 to NTm−1,n−1 schematic representations;FET3237′ structures correspond to any of FETs T0,0 to Tm−1, n−1 schematic representations; and exemplary cell4765 corresponds to any of cells C0,0 to cell Cm−1,n−1 schematic representations.Switch plate3106′ is connected to drain3124′ throughcontact3101′,conductive stud3122′, andcontact3121′.Drain3124′ is insubstrate3128′.
• It is desirable to enhancearray4725 illustrated in plan viewFIG. 44C by enhancing wireability, for example, or cell density, for example. In order to minimize the risk of shorts caused by misaligned via (vertical) connections between conductive layers, it is desirable to coat the top and sides of some selected conductors with an additional insulating layer that is not etched when etching the common insulator (common insulator SiO₂, for example) between conductive layers as illustrated bystructure3227 inFIG. 31D. A method such asMethod3144 of coating a conductive layer with an additional insulating layer to forminsulated conductor structure3227 as described with respect to structures illustrated inFIGS. 31A-31D may be applied to structures as illustrated further below.
• It is desirable to enhance the wireability ofarray4725 illustrated inFIG. 44C by formingbit array line3138′″ on the same wiring level and at the same time asreference line3138″.Bit array line3138′″ contacts bitline segments3119′/3117′ composed of combinedconductors3119′ and3117′ as illustrated further below.Line segments3119′/3117′ are not required to span relatively long sub-array regions and may be optimized for contact to nanotubelayer3114′.
• FIG. 45A illustrates cross section A-A′ ofarray4729 taken at A-A′ of the plan view ofarray4729 illustrated inFIG. 45C, and showsFET device region3237′ in the FET length direction,nanotube switch structure3233′, interconnections and insulators.FIG. 45B illustrates cross section B-B′ ofarray4729 taken at B-B′ of plan view ofarray4729 illustrated inFIG. 45C, and shows arelease array line3205′ with insulatinglayer3149′ corresponding to insulatinglayer3148 in structure3227 (FIG. 31D), abit array line3138′″ in contact withconductor3119′ of combinedconductors3119′ and3117′, and aword array line3120′.Bit array line3138′″contacts conductor3119′ throughcontact3155′, tostud3157′, throughcontact3159′, toconductor3119′.Insulator3149′ is used to prevent contact betweenrelease line conductor3205′ andstud3157′ in case ofstud3157′ misalignment.FIG. 45C illustrates a plan view ofarray4729 including exemplary cell4767 region, withreference array line3138″ contactingsource3126′ throughcontact3140′ tostud3118A′, tostud3118′, to contact3123′, (stud3118A′,stud3118′ andcontact3123′ not shown in plan view4725) and to source3126′.Reference array line3118″ is on the same array wiring layer and parallel tobit line3138′″, as is illustrated in plan view ofarray4729 inFIG. 45C, andbit line3138′″ contacts a correspondingbit line segment3119′, as shown inFIG. 45B.Release array line3205′ is parallel toword array line3120′. Portions ofrelease line3205′ act asrelease electrode3205′ as illustrated in the nanotube switching region ofFIG. 45A. This nanotube switching region is illustrated asnanotube switch structure3233′ inarray4729 ofFIG. 45C. In terms of minimum technology feature size, NT-on-drain cell4767 is approximately 12 to 13 F². Nanotube-on-drain array4729 structures illustrated inFIGS. 45A, 45B, and45C correspond to nanotube-on-drain array4700 schematic representation illustrated inFIG. 40.Bit line3138′″ structures correspond to any of bit lines BL0 to BLm−1 schematic representations;reference line3138″ structures correspond to any of reference lines REF0 to REFm−1 schematic representations;word line3120′ structures correspond to any of word lines WL0 to WLn−1 schematic representations;release line3205′ structures correspond to any of release lines RL0 to RLn−1 schematic representations;source contact3140′ structures correspond to any ofsource contacts4720 schematic representations;nanotube switch structure3233′ correspond to any of NT0,0 to NTm−1,n−1 schematic representations; andFET3237′ structures correspond to any of FET T0,0 to Tm−1, n−1 schematic representations; and exemplary cell4767 corresponds to any of cells C0,0 to cell Cm−1,n−1 schematic representations.
• It is desirable to enhance the density ofarray4725 illustrated inFIG. 44C to reduce the area of each bit in the array, resulting in higher performance, lower power, and lower cost due to smaller array size. Smaller array size results in the same number of bits occupying a reduced silicon chip area, resulting in increased productivity and therefore lower cost, because there are more chips per wafer. Cell area is decreased by reducing the size ofnanotube switch region3233′, thereby reducing the periodicity betweennanotube switch regions3233′, and correspondingly reducing the spacing betweenreference lines3138″ andbit lines3119′/3117′.
• FIG. 46A illustrates cross section A-A′ ofarray4731 taken at A-A′ of the plan view ofarray4731 illustrated inFIG. 46D, and showsFET device region3237′ in the FET length direction, reduced area (smaller)nanotube switch structure3239′, interconnections and insulators. A smaller picture frame opening is formed in combinedconductors3119′ and3117′ by applyingsub-lithographic method3036 shown inFIG. 26 and corresponding sub-lithographic structures shown inFIGS. 29D, 29E, and29F during the fabrication ofnanotube switch structure3239′.FIG. 46B illustrates cross section B-B′ ofarray4731 taken at B-B′ of plan view ofarray4731 illustrated inFIG. 46D, and showsreference line3163′ comprisingconductive layers3117′ and3119′, and conformal insulatinglayer3161′.Conductive layers3117′ and3119′ ofreference line3163′ are extended to form the picture frame region ofnanotube device structure3239′, however, insulatinglayer3161′ is not used as part of thenanotube switch structure3239′.FIG. 46B also illustratesrelease line3205′, andword array line3120′.FIG. 46C illustrates cross section C-C′ ofarray4731 taken at C-C′ of the plan view ofarray4731 illustrated inFIG. 46D.Reference line3138″ is connected to sourcediffusion3126′ throughcontact3140′, tostud3118A′, and throughcontact3123′. In order to achieve greater array density, there is a small spacing betweenstud3118A′ andreference line3163′.Insulator3161′ is used to prevent electrical shorting betweenstud3118A′ andreference line3163′conductors3119′ and3117′ ifstud3118A′ is misaligned.FIG. 46D illustrates a plan view ofarray4731 includingexemplary cell4769 region, withreference array line3138″ contactingsource3126′ as illustrated inFIG. 46C,bit array lines3163′ parallel toreference line3138″ but on a different array wiring level (wiring plane).Release array line3205′ is parallel toword array line3120′.Release line3205′ contacts and forms a portion ofrelease electrode3205′ as illustrated in the nanotube switching region ofFIG. 46A.Exemplary cell4769 area (region) is smaller (denser) than exemplary cell4767 area shown inFIG. 45C and exemplary cell4765 area shown inFIG. 44C, and therefore correspondingarray4731 is denser (occupies less area) than corresponding array areas ofarray4729 and4725. The greater density (smaller size) ofarray4731 results in higher performance, less power, less use of silicon area, and therefore lower cost as well. In terms of minimum technology feature size, NT-on-drain cell4769 is approximately 10 to 11 F². Nanotube-on-drain array4731 structures illustrated inFIGS. 46A-46D correspond to nanotube-on-drain array4700 schematic representation illustrated inFIG. 40.Bit line3163′ structures correspond to any of bit lines BL0 to BLm−1 schematic representations;reference line3138″ structures correspond to any of reference lines REF0 to REFm−1 schematic representations;word line3120′ structures correspond to any of word lines WL0 to WLn−1 schematic representations;release line3205′ structures correspond to any of release lines RL0 to RLn−1 schematic representations;source contact3140′ structures correspond to any ofsource contacts4720 schematic representations;nanotube switch structure3239′ correspond to any of NT0,0 to NTm−1,n−1 schematic representations; andFET3237′ structures correspond to any of FET T0,0 to Tm−1, n−1 schematic representations; andexemplary cell4769 corresponds to any of cells C0,0 to cell Cm−1,n−1 schematic representations.
Nanotube Random Access Memory Using FEDs with Controllable GatesNanotube Random Access Memory (NRAM) Systems and Circuits, with Same
• Non-volatile field effect devices (FEDs)180,200,220, and240 with controllable gates may be used as cells and interconnected into arrays to form non-volatile nanotube random access memory (NRAM) systems. The memory cells contain a single element that combines both select and storage functions, and is referred to as a nanotube transistor (NT-T). By way of example, FED12240 (FIG. 2L) is used to form a non-volatile NRAM memory cell that is also referred to as a NT-on-Gate memory cell.
NT-on-Gate NRAM Memory Systems and Circuits with Parallel Bit and Release Lines, and Parallel Word and Reference Lines
• NRAM1NT-T memory arrays are wired using four lines. Word line WL is used to gate combined nanotube/select device NT-T, bit line BL is attached to a shared drain between two adjacent combined nanotube/select devices. Reference line REF is attached to a shared source between two adjacent nanotube/select devices and is grounded. Release line RL is used to control a release-plate of a combined nanotube/select device. In this NRAM array configuration, RL is parallel to BL and acts as second bit line, and REF is parallel to WL, and REF is grounded.
• FIG. 47A depicts non-volatilefield effect device240 with memory cell wiring to form NT-on-Gate memory cell5000 schematic. Word line (WL)5200 connects to terminal T1 ofFED12240; bit line (BL)5300 connects to terminal T2 ofFED12240; reference line (REF)5400 connects to terminal T3 ofFED12240; and release line (RL)5500 connects to terminal T4 ofFED12240.Memory cell5000 performs write and read operations, and stores the information in a non-volatile state. TheFED12240 layout dimensions and operating voltages are selected to optimizememory cell5000.Memory cell5000 FET combined nanotube/select devicecontrollable gate5120 corresponds to a combination ofgate242 andswitch plate248;drain5080 corresponds to drain244; andsource5060 corresponds to source246.Memory cell5000 combined nanotube/select device control gate andNT switch5140 corresponds toNT switch250; release-plateinsulator layer surface5160 corresponds to release-plateinsulator layer surface256; and release-plate5180 corresponds to release-plate254. The interconnections between the elements ofmemory cell5000 schematic correspond to the interconnection of the corresponding inteconnections of the elements ofFED12240.BL5300 connects to drain5080 throughcontact5320;REF5400 connects to source5060 throughcontact5420;RL5500 connects to release-plate5180 bycontact5520;WL5200 interconnects to combined nanotube/select device NTswitch control gate5140 bycontact5220. The non-volatileNT switching element5140 may be caused to deflect toward combined switch-platecontrollable gate5120 via electrostatic forces to closed (“ON”)position5140′ to store a logic “1” state as illustrated inFIG. 47B. The van der Waals force holdsNT switch5140 inposition5140′. Inposition5140′ combined switch platecontrollable gate5120 is at the same voltage as NTswitch control gate5140′. Alternatively, the non-volatileNT switching element5140 may be caused to deflect toinsulator surface5160 on release-plate5180 via electrostatic forces to open (“OFF”)position5140″ to store a logic “0” state as illustrated inFIG. 47C. The van der Waals force holdsNT switch5140 inposition5140″. Inposition5140″ combined switch-platecontrollable gate5120 is floating (not connected). When combined switch platecontrollable gate5120 is not connected to a terminal, its voltage is determined by the internal capacitance network as illustrated inFIG. 13A andFIG. 14. Combined switch platecontrollable gate5120 is a combination ofelements242,243, and248 as illustrated in more detail incross section400 inFIG. 14. C_CH-SUBis not in the internal device capacitance network because bit lines BL and reference lines REF are held at zero volts during the write operation. When combined switch platecontrollable gate5120 is floating, its voltage V_Gmay be calculated as V_G=V_CG×C_1G/(C_1G+C_G-CH), where V_CGis the voltage of NTswitch control gate5140. Capacitance C_1Gis designed for a desired capacitance ratio relative to device gate capacitance C_G-CH. For C_1G=0.25×C_G-CH, V_G=0.2×V_CG. The non-volatile element switching via electrostatic forces is as depicted byelement250 inFIG. 2L.Voltage waveforms375 used to generate the required electrostatic forces are illustrated inFIG. 15.
• NT-on-Gate schematic ofmemory cell5000 forms the basis of a non-volatile storage (memory) cell. The device may be switched between closed storage state “1” (switched to position5140′) and open storage state “0” (switched toposition5140″), which means the controllable gate may be written to an unlimited number of times as desired. In this way, the device may be used as a basis for a non-volatile nanotube random access memory, which is referred to here as a NRAM array, with the ‘N’ representing the inclusion of nanotubes. In the NT-on-gate structure, no dc current flows through the switch-plate to NT fabric contact, maximizing cyclability (maximum number of ON/OFF cycles).
• FIG. 48 represents anNRAM system5700, according to preferred embodiments of the invention. Under this arrangement, an array is formed with m×n (only exemplary portion being shown) of non-volatile cells ranging from cell C0,0 to cell C2,2.NRAM system5700 may be designed using one large m×n array, or several smaller sub-arrays, where each sub-array is formed of m×n cells. Non-volatile cell C0,0 contains a single combined nanotube/select device NT-T0,0. To access selected cells, the array uses read and write word lines (WL0, WL1, WL2), read bit lines (BL0, BL1, BL2), grounded reference lines (REF0, REF1), and write release lines (behave as write bit lines) (RL0, RL1, RL2). The NT switch control gate of NT-T0,0 is coupled to WL0, the drain of NT-T0,0 is coupled to BL0, the source of NT-T0,0 is coupled to REF0, and the release-plate of NT-T0,0 is coupled to RL0.Connection5720 connects BL0 to shared drain of select devices NT-T0,0 and NT-T0,1.Connection5740 connects REF1 to shared source of select devices NT-T0,1 and NT-T0,2. Word, bit, reference, and release decoders/drivers are explained further below.
• Under preferred embodiments, nanotubes inarray5700 may be in the “ON” “1” state or the “OFF” “0” state. The NRAM memory allows for unlimited read and write operations per bit location. A write operation includes both a write function to write a “1” and a release function to write a “0”. By way of example, a write “1” to cell C0,0 and a write “0” to cell C1,0 is described. For a write “1” operation to cell C0,0, combined nanotube/select device NT-T0,0 is activated when WL0 transitions from 0 to V_SW, BL0 has transitioned from V_DDto 0 volts prior to WL0 activation, RL0 transitions from V_DDto switching voltage V_SW, and REF0 remains at zero. The release-plate and combined NT-switch-control-gate of the non-volatile combined nanotube/select device NT0,0 are each at V_SWresulting in zero electrostatic force (because the voltage difference is zero). The zero BL0 voltage is applied to the drain, and zero REF0 reference is applied to the source of combined nanotube/select device NT-T0,0. The difference in voltage between the NT0,0 combined NT-switch-control-gate and the combined switch-plate-gate is V_SW−0.2 V_SW, and generates an attracting electrostatic force. If V_SW−0.2 V_SWexceeds the nanotube threshold voltage V_NT-TH(V_SW>1.25 V_NT-TH), then the nanotube structure switches to “ON” state or logic “1” state, that is, combined NT-switch-control-gate and combined switch-plate-gate are electrically connected as illustrated inFIG. 47B. If NT-T0,0 was in the “1” state at the onset of the write “1” cycle, it remains in the “1” state. The near-Ohmic connection between combined switch-plate-gate5120 and combined NT-switch-control-gate5140 inposition5140′ represents the “ON” state or “1” state. If the power source is removed, cell C0,0 remains in the “ON” state.
• For a write “0” (release) operation to cell C1,0, combined nanotube/select device NT-T1,0 is activated when WL0 transitions from 0 to V_SWand drives combined NT-switch-control-gate to V_SW. BL1 transitioned from V_DDto 0 volts prior to WL0 activation, RL1 transitions from V_DDto zero volts, and REF0 remains at zero volts. If cell C1,0 is in the “1” state, then switching voltage V_SWis applied to the combined switch-plate-gate of NT-T1,0. There is no electrostatic force between combined switch-plate-gate and combined NT-switch-control-gate. The non-volatile storage element NT1,0 release-plate is at switching voltage zero and the combined NT-switch-control-gate is at switching voltage V_SWgenerating an attracting electrostatic force. If V_SWexceeds the nanotube threshold voltage V_NT-TH, the nanotube structure switches to the “OFF” state or logic “0” state, that is, the nanotube NT switch and the surface of the release-plate insulator are in contact as illustrated inFIG. 47C. If NT-T1,0 was in the “0” state at the onset of the write “0” cycle, it remains in the “0” state. The non-conducting contact betweeninsulator surface5160 on release-plate5180 and combined NT-switch-control-gate5140 inposition5140″ represents the “OFF” state or “0” state. If the power source is removed, cell C1,0 remains in the “OFF” state.
• An NRAM read operation does not change (destroy) the information in the activated cells, as it does in a DRAM, for example. Therefore the read operation in the NRAM is characterized as a non-destructive readout (or NDRO) and does not require a write-back after the read operation has been completed. In this example, Cell C0,0 combined nanotube/select device NT-T0,0 stores a “1” state as illustrated inFIG. 47B. The electrical characteristics (source-drain current ISD VS combined switch-plate-gate) depend on the stored logic state (“1” state or “0” state). Combined nanotube/select device NT-T0,0 is field effect device (FED)240 (FIG. 2L) with structure400 (FIG. 14) used incell5000, andmemory array5700, and exhibits electrical characteristic385, as illustrated inFIG. 16.FED12240,NT switch250 andposition250′, correspond to NT-T0,0 combined NT-switch-control-gate5140position5140′. NT-switch-control-gate5140 is connected to WL0 (which corresponds to V_T1inFIG. 16). During read, BL0 is precharged to V_DDand allowed to float. WL0 transitions from zero to V_DD(1.2 volts, for example). For a stored logic “1” state, the FET threshold voltage V_FET-TH=0.4 volts is exceeded by 0.8 volts and BL0 is discharged. The change in BL0 voltage is detected by a sense amplifier/latch, and a logic “1” state is stored in the latch. BL0, in contact with NT-T0,0drain5080, discharges through conductive channel of resistance R_FETto the groundedsource terminal5060. The combined NT-switch-control-gate5140 contacts combined switch-plate-gate5120 of NT-T0,0 through conductor to NT contact resistances R_Cand NT switch to switch-plate resistance R_SW. R_C+R_SWare not in the discharge path for a NT-on-gate cell.
• In this example, cell C1,0 combined nanotube/select device NT-T1,0 stores a “0” state as illustrated inFIG. 47C. For a read operation of cell C1,0, BL1 is precharged high to V_DDand allowed to float. WL0 is driven high to V_DD(1.2 volts, for example). WL0 voltage V_DDis capacitively coupled to combined switch-plate-gate5120 by the internal capacitance network illustrated inFIG. 14, resulting in an FET-gate voltage of 0.24 volts (0.2×1.2 volts). Since the FET gate voltage is less than VFET-TH=0.4 volts (electrical characteristic385,FIG. 16), there is no conductive path betweendrain5080 andsource5060, and BL1 is not discharged. Sense amplifier/latch circuit (not shown) does not detect a voltage drop and the latch is set to a logic “0” state.
• FIG. 49 illustrates theoperational waveforms5800 ofmemory array5700 ofFIG. 48 during read, write “1”, and write “0” operations for selected cells, while not disturbing unselected cells (no change to unselected cell stored logic states).Waveforms5800 illustrate voltages and timings to write logic state “1” in cell C0,0, write a logic state “0” in cell C1,0, read cell C0,0, and read cell C1,0.Waveforms5800 also illustrate voltages and timings to prevent disturbing the stored logic states (logic “1” state and logic “0” state) in partially selected (also referred to as half-selected) cells. Partially selected cells are cells inmemory array5700 that receive applied voltages because they are connected to (share) word, bit, reference, and release lines that are activated as part of the read or write operation to the selected cells. Cells inmemory array5700 tolerate unlimited read and write operations at each memory cell location.
• At the start of the write cycle, BL0 transitions from V_DDto zero volts, connecting the drain to ground. REF0 is at zero volts connecting source to ground. BL1 and BL2 transition from V_DDto zero volts connecting all drains to ground. REF1 and REF2 are also at zero volts connecting all sources to ground. WL0 transitions from zero to V_SW, activating select devices NT-T0,0, NT-T1,0, NT-T2,0. Word lines WL1, WL2 are not selected and remain at zero volts. RL0 transitions from V_DDto switching voltage V_SW, connecting the release-plates combined nanotube/select device NT-T0,0, NT-T0,1, and NT-T0,2 to V_SW. RL1 transitions from V_DDto zero volts, connecting the release-plates of combined nanotube/select devices NT-T1,0, NT-T1,1, and NT-T1,2, to zero volts. RL2 remains at V_DD, connecting the release-plates of NT-T3,0 to V_DD. REF0 transitions from V_DDto switching voltage V_SW, connecting NT switches of non-volatile storage elements NT0,0, NT1,0 . . . NTm−1,0 to V_SW. REF1, REF2 . . . REFn−1 remain at V_DD, connecting NT switches of non-volatile storage elements NT0,1 to NTn−1,n−1 to V_DD.
• NT-T0,0 may be in “ON” (“1” state) or “OFF” (“0” state) state at the start of the write cycle. It will be in “ON” state at the end of the write cycle. If NT-T0,0 in cell C0,0 is “OFF” (“0” state) it will switch to “ON” (“1” state) since the voltage difference between combined NT-switch-control-gate and release-plate is zero, and the voltage difference between combined NT-switch-control-gate and combined switch-plate-gate is V_SW−0.2 V_SWbecause of the internal device capacitance coupling network. Therefore, V_SWmust be sufficiently elevated to ensure nanotube switching occurs. For V_NT-THin the range of 1.7 to 2.2 volts, V_SW−0.2 V_SWmust exceed 2.2 volts, therefore V_SW>2.75 volts. V_SW=2.8 volts is used in this example to ensure an “OFF” to “ON” transition. If NT-T-T0,0 in cell C0,0 is in the “ON” (“1” state), it will remain in the “ON” (“1”) state. NT-T1,0 may be in “ON” (“1” state) or “OFF” (“0” state) state at the start of the write cycle. It will be in “OFF” state at the end of the write cycle. If NT-T1,0 in cell C1,0 is “ON” (“1” state) it will switch to “OFF” (“0” state) since the voltage difference between combined NT-switch-control-plate and combined switch-plate-gate is zero, and the voltage difference between NT-switch-control-plate and release-plate is V_SW. If NT-T1,0 in cell C1,0 is “OFF” (“0” state), it will remain “OFF” (“0” state). If for example, V_SW=2.4 volts, V_DD=1.2 volts, and NT switch threshold voltage range is V_NT-TH=1.7 to 2.2 volts, then for NT-T0,0 and NT-T1,0 a difference voltage V_SW>V_NT-THensuring write states of “ON” (“1” state) for NT0,0 and “OFF” (“0” state) for NT1,0. Although VSW=2.4 volts ensures an “ON” to “OFF” transition, VSW=2.8 volts is used in this example to ensure “OFF” to “ON” transition.
• Cells C0,0 and C1,0 have been selected for the write operation. All other cells have not been selected, and information in these other cells must remain unchanged (undisturbed). Since in an array structure some cells other than selected cells C0,0 and C1,0 inarray5700 will experience partial selection voltages, often referred to as half-select voltages, it is necessary that half-select voltages applied to non-volatile storage element terminals be sufficiently low (below nanotube activation threshold V_NT-TH) to avoid disturbing stored information. It is also necessary to avoid parasitic current flow. For NT-on-Gate memory cells during write operations, all bit lines (connected to drain) and reference lines (connected to sources) are at zero volts, so no disturb currents flow for write “1” or write “0” operations. Release lines are used as write bit lines in NT-on-Gate memory arrays. Potential half-select disturb along activated array lines WL0 (REF0 voltage is zero) includes cell C2,0 because WL0 has been activated. Storage element NT-T2,0 will have BL2 at zero volts. To prevent undesired switching of NT-T2,0, RL2 is set at voltage V_DD. The information in storage elements NT-T2,0 in cell C2,0 is not disturbed, and there is no parasitic current. Since corresponding source and drain voltages are zero, there can be no parasitic current. If cell C2,0 is in the “ON” state, there is no disturb because the voltage difference between corresponding combined NT-switch-control-gates and corresponding release-plate is V_SW−V_DD=1.2 volts, when V_SW=2.8 volts and V_DD=1.2 volts. Since this voltage difference of 1.6 volts is less than the minimum nanotube threshold voltage V_NT-THof 1.7 volts, no switching takes place. If C2,0 is in the “OFF” state, then the difference in voltage between combined NT-switch-control-gate and combined switch-pate-gate is V_SW−0.2 V_SW=2.2 volts. However, for NT-T0,1 and NT-T0,2 release-plate at V_SW=2.8 volts, corresponding combined NT-switch-control-gate at V_DD, and corresponding combined switch-plate-gate at V_DD(for ON) and 0.2 V_DD(equals 0.24 volts for OFF), and with minimum V_NT-TH=1.7 volts, no disturb occurs.
• Potential half-select disturb along activated array lines RL0 and BL0 includes cells C0,1 and C0,2 because RL0 and BL0 have been activated. RL0 drives combined nanotube/select device NT-T0,1 and NT-T0,2 release-plates to switching voltage V_SW, and WL1 and WL2 drive corresponding combined NT-switch-control-gates to V_DD. Combined nanotube/select devices NT-T0,1 and NT-T0,2 have corresponding release-plates at V_SWand combined NT-switch-control-gates at V_DD. For a stored “1” (“ON”) state, combined switch-plate-gate is at V_DD. The voltage difference V_SW−V_DD=1.6 volts, less than minimum V_NT-TH=1.7 volts, and the stored “1” (“ON”) state is not disturbed. For a stored “0” (“OFF”) state, combined switch-plate-gate is at 0.2 V_DDdue to internal device capacitance network coupling. The electrostatic attractive force due to V_DD−0.2 V_DD=1 volt and cannot overcome a much stronger electrostatic force due to the V_SW−V_DD=1.6 volts and close proximity between release-plate and corresponding combined NT-switch-control-gate, and the “0” (“OFF”) state is not disturbed.
• Potential half-select disturb along activated array lines RL1 and BL1 includes cells C1,1 and C1,2 because RL1 and BL1 have been activated. RL1 drives combined nanotube/select device NT-T0,1 and NT-T0,2 release-plates to zero volts, and WL1 and WL2 drive corresponding combined NT-switch-control-gates to V_DD. Combined nanotube/select devices NT-T1,1 and NT-T1,2 have corresponding release-plates at zero volts and combined NT-switch-control-gates at V_DD. For a stored “1” (“ON”) state, combined switch-plate-gate is at V_DD. The voltage difference V_DD−0=1.2 volts, less than minimum V_NT-TH=1.7 volts, and the stored “1” (“ON”) state is not disturbed. For a stored “0” (“OFF”) state, combined switch-plate-gate is at 0.2 V_DDdue to internal device capacitance network coupling. The electrostatic attractive force due to V_DD−0.2 V_DD=1 volt causes a counter-balancing electrostatic, and the “0” (“OFF”) state is not disturbed.
• For all remainingmemory array5700 cells C2,1 and C2,2 BL2 andREL1 and REL2 voltages are zero, so no parasitic currents can flow between drains and sources of combined nanotube/select devices NT-T2,1 and NT-T2,2. RL2 drives combined nanotube/select device NT-T2,1 and NT-T2,2 release-plates to V_DD, and WL1 and WL2 drive corresponding combined NT-switch-control-gates to V_DD. Combined nanotube/select devices NT-T2,1 and NT-T2,2 have corresponding release-plates at V_DDand corresponding combined NT-switch-control-gates at V_DD, for a voltage difference of zero. For a stored “1” (“ON”) state, combined switch-plate-gate is at V_DD, all voltage differences are zero, and the stored “1” (“ON”) state is not disturbed. For a stored “0” (“OFF”) state, combined switch-plate-gate is at 0.2 V_DDdue to internal device capacitance network coupling. The electrostatic attractive force due to V_DD−0.2 V_DD=1 volt is much less than V_NT-TH=1.7 volts, and the “0” (“OFF”) state is not disturbed.
• Non-volatile NT-on-gateNRAM memory array5700 with bit lines parallel to release lines is shown inFIG. 48 contains 2^N×2^Mbits, is a subset of non-volatileNRAM memory system5810 illustrated asmemory array5815 inFIG. 50A.NRAM memory system5810 may be configured to operate like an industry standard asynchronous SRAM or synchronous SRAM because nanotubenon-volatile storage cells5000 shown inFIG. 47A, inmemory array5700, may be read in a non-destructive readout (NDRO) mode and therefore do not require a write-back operation after reading, and also may be written (programmed) at CMOS voltage levels (5, 3.3, and 2.5 volts, for example) and at nanosecond and sub-nanosecond switching speeds. NRAM read and write times, and cycle times, are determined by array line capacitance, and are not limited by nanotube switching speed. Accordingly,NRAM memory system5810 may be designed with industry standard SRAM timings such as chip-enable, write-enable, output-enable, etc., or may introduce new timings, for example. Non-volatileNRAM memory system5810 may be designed to introduce advantageous enhanced modes such as a sleep mode with zero current (zero power—power supply set to zero volts), information preservation when power is shut off or lost, enabling rapid system recovery and system startup, for example.NRAM memory system5810 circuits are designed to provide thememory array5700waveforms5800 shown inFIG. 49.
• NRAM memory system5810 accepts timinginputs5812, acceptsaddress inputs5825, and acceptsdata5867 from a computer, or providesdata5867 to a computer using a bidirectional bus sharing input/output (I/O) terminals. Alternatively, inputs and outputs may use separate (unshared) terminals (not shown). Address input (I/P)buffer5830 receives address locations (bits) from a computer system, for example, and latches the addresses. Address I/P buffer5830 provides word address bits toword decoder5840 viaaddress bus5837; address I/P buffer5830 provides bit addresses to bit decoder5850 viaaddress bus5852; and address bus transitions provided bybus5835 are detected by function generating, address transition detecting (ADT), timing waveform generator, controller (controller)5820.Controller5820 provides timing waveforms onbus5839 toword decoder5840.Word decoder5840 selects the word address location withinarray5815 and provides WL waveforms for both write-one, write-zero, read-one, and read-zero operations as illustrated bywaveforms5800′ shown inFIG. 51.FIG. 51NRAM memory system5810waveforms5800′ correspond tomemory array5700waveforms5800 shown inFIG. 49. Reference lines REF are grounded.
• Bit address decoder5850 is used to decode both bit lines BL and corresponding release lines RL (there is no need for a separate RL decoder) and drive bit line (BL) and release (RL)select logic5855 viabus5856.Controller5820 provides timing waveforms onbus5854 tobit decoder5850.Controller5820 also provides function and timing inputs onbus5857 to BL & RLselect logic5855. BL & RLselect logic5855 uses inputs frombus5856 andbus5857 to generate data multiplexer select bits onbus5859. The output of BL and RLselect logic5855 onbus5859 is used to select control data multiplexers using combined data multiplexers & sense amplifiers/latches (MUXs & SAs)5860.Controller5820 provides function and timing inputs onbus5862 to MUXs &SAs5860, resulting inNRAM memory system5810 on-chip BL and RL waveforms for both write-one, write-zero, read-one, and read-zero operations as illustrated bywaveforms5800′ corresponding tomemory array5700waveforms5800 shown inFIG. 49. MUXs &SAs5860 are used to write data provided by read/write buffer5865 viabus5864 inarray5815, and to read data fromarray5815 and provide the data to read/write buffer5865 viabus5864 as illustrated inwaveforms5800′, ofFIG. 51.
• Sense amplifier/latch5900 is illustrated inFIG. 50B.Flip flop5910, comprising two back-to-back inverters is used to amplify and latch data inputs fromarray5815 or from read/write buffer5865.Transistor5920 connectsflip flop5910 to ground when activated by a positive voltage supplied bycontrol voltage V_TIMING5980, which is provided bycontroller5820.Gating transistor5930 connects a bit line BL tonode5965 offlip flop5910 when activated by a positive voltage.Gating transistor5940 connects reference voltage V_REFto flipflop node5975 when activated by a positive voltage.Transistor5960 connects voltage V_DDto flipflop5910node5965,transistor5970 connects voltage V_DDto flipflop5910node5975, andtransistor5950 ensures that small voltage differences are eliminated whentransistors5960 and5970 are activated.Transistors5950,5960, and5970 are activated (turned on) when gate voltage is low (zero, for example).
• In operation, V_TIMINGvoltage is at zero volts whensense amplifier5900 is not selected.NFET transistors5920,5930, and5940 are in the “OFF” (non-conducting) state, because gate voltages are at zero volts.PFET transistors5950,5960, and5970 are in the “ON” (conducting) state because gate voltages are at zero volts. V_DDmay be 5, 3.3, or 2.5 volts, for example, relative to ground.Flip flop5910nodes5965 and5975 are at V_DD. If sense amplifier/latch5900 is selected, V_TIMINGtransitions to V_DD,NFET transistors5920,5930, and5940 turn “ON”,PFET transistors5950,5960, and5970 are turned “OFF”, andflip flop5910 is connected to bit line BL and reference voltage V_REF. V_REFis connected to V_DDin this example. As illustrated by waveforms BL0 and BL1 ofwaveforms5800′,bit line BL is pre-charged prior to activating a corresponding word line (WL0 in this example). Ifcell5000 of memory array5700 (memory system array5815) stores a “1”, then bit line BL inFIG. 50B corresponds to BL0 inFIG. 51, BL is discharged bycell5000, voltage droops below V_DD, and sense amplifier/latch5900 detects a “1” state. Ifcell5000 of memory array5700 (memory system array5815) stores a “0”, then bit line BL inFIG. 50B corresponds to BL1 inFIG. 51, BL is not discharged bycell5000, voltage does not droop below V_DD, and sense amplifier/latch5900 detect a “0” state. The time from sense amplifier select to signal detection by sense amplifier/latch5900 is referred to as signal development time. Sense amplifier/latch5900 typically requires 100 to 200 mV relative to V_REFin order to switch. It should be noted thatcell5000 requires a nanotube “OFF” resistance to “ON” resistance ratio of greater than 10 to 1 for successful operation. A typical bit line BL has a capacitance value of 250 fF, for example. A typical nanotube storage device (switch) or dimensions 0.2 by 0.2 um typically has 8 nanotube filaments across the suspended region, for example, as illustrated further below. For a combined contact and switch resistance of 50,000 Ohms per filament, as illustrated further below, the nanotube “ON” resistance ofcell5000 is 6,250 Ohms. For a bit line of 250 fF, the time constant RC=1.6 ns. The sense amplifier signal development time is less than RC, and for this example, is between 1 and 1.5 nanoseconds.
• Non-volatileNRAM memory system5810 operation may be designed for high speed cache operation at 5 ns or less access and cycle time, for example. Non-volatileNRAM memory system5810 may be designed for low power operation at 60 or 70 ns access and cycle time operation, for example. For low power operation, address I/P buffer5830 operation requires 8 ns;controller5820 operation requires 16 ns;bit decoder5850 operation plus BL &select logic5855 plus MUXs &SA5860 operation requires 12 ns (word decoder5840 operation requires less than 12 ns)array5815 delay is 8 ns; operation ofsense amplifier5900 requires 8 ns; and read/write buffer5865 requires 12 ns, for example. The access time and cycle time of non-volatileNRAM memory system5810 is 64 ns. The access time and cycle time may be equal because the NDRO mode of operation of nanotube storage devices (switches) does not require a write-back operation after access (read).
Method of Making Field Effect Device with Controllable Gate and NT-on-Gate Memory System and Circuits with Parallel Bit and Release Array Lines, and Parallel Word and Reference Array Lines
• NT-on-Gate memory cells are based onFED12240 devices shown inFIG. 2L.Switch250 may be displaced to contact a switch-plate248, which is connected to acontrollable gate242.Switch250 may be displaced to contact release-plate dielectric surface256 on release-plate254, which is connected to terminal T4.FED12240 devices are interconnected to fabricate a NT-on-gate memory array.
• FIG. 22 describes abasic method3000 of manufacturing preferred embodiments of the invention. In general, preferred methods first form3002 a base structure including field effect device similar to a MOSFET, having drain, source, gate nodes, and conductive studs on source, drain, and gate structures for connecting to additional layers above the MOSFET device used to fabricate the nanotube switch.Base structure3102′ shown inFIG. 24A-24E is used when fabricating NT-on-source memory arrays. The nanotube switch structure is fabricated onplanar surface3104′.Base structure3102′″ shown inFIG. 44A is used when fabricating NT-on-drain memory arrays. The nanotube switch structure is fabricated onplanar surface3104′″ using the same methods as used to fabricate the NT-on-source memory array.Base structure6002 shown inFIG. 52B is used when fabricating NT-on-gate memory arrays. The nanotube switch structure is fabricated onplanar surface6004 using the same methods as used to fabricate the NT-on-source and NT-on-drain memory arrays.
• Preferred methodsfirst form3002base structure6002 in two steps. First, MOSFET devices are formed using well known industry methods having a polysilicon (or metallic)gate6120, for example, andsource diffusion6124 anddrain diffusion6126 insemiconductor substrate6128, for example, as illustrated inFIG. 52A. Then studs (tungsten, aluminum, etc., for example) are embedded in dielectric6116 (SiO₂, for example) using well known industry methods, and the surface is planarized.Stud6129 contacts source6124 atcontact6121,stud6118′ contacts drain6126 atcontact6123, andstud6122′contacts gate6120 atcontact6125.
• Next, reference array line (REF)6163 is deposited and patterned using standard semiconductor process techniques, andcontact stud6129 atcontact6101 as illustrated inFIG. 52B. Standard semiconductor process methods insulatereference array line6163. Next, standard semiconductor processes are used to open via holes tostuds6122′ and6118′, fill via holes with metal, planarize, and pattern. Standard semiconductor processes deposit and insulator, such as SiO2, for example, and planarize.Stud6122′ andstud6118′ are thus extended in length above the top ofreference array line6163 to surface6004 ofbase structure6002 as illustrated inFIGS. 52A and 52B.
• Oncebase structure6002 is defined, then methods of fabricating NT-on-gate memory arrays are the same as those used to fabricate NT-on-source memory arrays.Preferred methods3004 shown inFIGS. 23 and 23′ and associated figures;methods3036 shown inFIG. 26 and associated figures;methods3006 shown inFIGS. 27 and 27′ and associated figures;methods3008 shown inFIGS. 28 and 28′ and associated figures; andmethods3144 as illustrated inFIGS. 31A-31D. Conductors, semiconductors, insulators, and nanotubes are formed in the same sequence and are in the same relative position in the structure. Length, widths, thickness dimensions may be different, reflecting differences in design choices. Also, conductor materials may be different, for example. The function of some electrodes may be different for NT-on-source and NT-on-gate memory arrays. For example, reference array lines are connected to source diffusions. Alternatively, source diffusions may be used as reference array lines without a separate conductor layer, however, performance may be slower. Word array lines connect to different electrodes in the nanotube structure, the nanotube switch for example, as may be seen further below. For NT-on-gate memory arrays, the switch-plate of the nanotube structure is connected to the gate diffusion of the FET device. However, for NT-on-drain memory arrays, the switch-plate of the nanotube structure is connected to the drain diffusion of the FET device, and for NT-on-source memory arrays, the switch-plate of the nanotube structure is connected to the source diffusion of the FET device, as may be seen further below.
• The nanotube switch region of the NT-on-gate cross section illustrated inFIG. 52C corresponds to the nanotube switch region of the NT-on-source cross section illustrated inFIG. 30F′ after the formation of first and second gap regions, sealing of the fluid communication paths, and planarizing as discussed with respect toFIG. 30J′. Switch-plate6106 is in electrical communication withFET gate6120 by means ofcontact6127,stud6122, andcontact6125, (seeFIGS. 52A and 52B).Insulator6108 is between switch-plate6106 andnanotube fabric6114.Conductors6117 and6119form composite conductor6325, with an opening to form a picture frame opening used to suspendnanotube fabric6114.Gap region6209 is between the top ofconductor6119 andinsulator6203 on the bottom of release-plate6205, in the combined nanotube/device switching region6301.Reference array line6263 is in electrical contact withsource6124 by means ofcontact6101 andstud6129.Insulator6116, with a planarized surface, encapsulates the nanotube switch structure and array wiring.
• FIG. 52D illustrates the structure ofFIG. 52C with extendedstud6118A contacting stud6118 and reaching the planarized top surface ofinsulator6116.Extended stud6118A is surrounded byinsulator6310 to ensure thatstud6118A does not connect to regions of combined nanotube/device structure6301 ifstud6118A is misaligned.Insulator6310 is a conformal insulating layer deposited in the via hole reaching the top surface ofstud6118. A directional etch (RIE, for example) removes the insulator region in contact with6118. The via hole is filled with a conductor, and the top surface is planarized as illustrated inFIG. 52D.Bit line6138 is deposited and patterned formingstructure6000 illustrated inFIG. 52E. Differences between NT-on-source and NT-on-gate memory arrays may be seen by comparingFIGS. 33A and 52E;FIGS. 33B and 52F;FIGS. 33C and 52G; andFIGS. 33D and 52H.
• FIG. 52E illustrates cross section A-A′ ofarray6000 taken at A-A′ of the plan view ofarray6000 illustrated inFIG. 52H, and shows reduced area (smaller) combined nanotube/device switch region6301 in the FET length, interconnections and insulators. A smaller picture frame opening is formed in combinedconductors6119 and6117 by applyingsub-lithographic method3036 shown inFIG. 26 and corresponding sub-lithographic structures shown inFIGS. 29D, 29E, and29F during the fabrication of combined nanotube/device switch structure6301.FIG. 52F illustrates cross section B-B′ ofarray6000 taken at B-B′ of plan view ofarray6000 illustrated inFIG. 52H, and showsword line6325 comprisingconductive layers3117 and3119.Conductive layers6117 and6119 ofword line6325 are extended to form the picture frame region of nanotube device structure.FIG. 52F also illustratesrelease line6205, andreference array line6263.FIG. 52G illustrates cross section C-C′ ofarray6000 taken at C-C′ of the plan view ofarray6000 illustrated inFIG. 52H.Bit line6138 is connected to draindiffusion6126 throughcontact6140, tostud6118A, tostud6118, and throughcontact6123. In order to achieve greater array density, there is a small spacing betweenstud6118A andrelease line6205.Insulator6310 is used to prevent electrical shorting betweenstud6118A andrelease line6205 ifstud6118A is misaligned.FIG. 52H illustrates a plan view ofarray6000 includingexemplary cell6400 region, withbit array line6138 contactingdrain6126 as illustrated inFIG. 52G,release array line6205 parallel tobit line6138 but on a different array wiring level (wiring plane).Reference array line6263 is parallel toword array line6325.Release line6205 contacts and forms a portion ofrelease electrode6205 as illustrated in the nanotube switching region ofFIG. 52E. NT-on-gateexemplary cell6400 area (region) is smaller (denser) than corresponding exemplary nanotube-on-source cell3169 area shown inFIG. 33D and corresponding NT-on-drainexemplary cell4769 area shown inFIG. 46D, and therefore correspondingarray6000 is denser (occupies less area) than corresponding array areas ofarray3231 and4731. The greater density ofarray6000 results in higher performance, less power, less use of silicon area, and therefore lower cost as well. In terms of minimum technology feature size, NT-on-gate cell6400 is approximately 7 to 9 F². Nanotube-on-gate array6000 structures illustrated inFIGS. 52E-52H correspond to nanotube-on-gate array5700 schematic representation illustrated inFIG. 48.Bit line6138 structures correspond to any of bit lines BL0 to BLm−1 schematic representations;reference line6263 structures correspond to any of reference lines REF0 to REFm−1 schematic representations;word line6325 structures correspond to any of word lines WL0 to WLn−1 schematic representations;release line6205 structures correspond to any of release lines RL0 to RLn−1 schematic representations;source contact6140 structures correspond to any ofsource contacts5740 schematic representations; combined nanotube/device switch structure6301 correspond to any of combined nanotube/select devices NT0,0 to NTm−1,n−1 schematic representations; andexemplary cell6400 corresponds to any of cells C0,0 to cell Cm−1,n−1 schematic representations.
• Nanotube Random Access Memory Using more than One FED Per Cell with Controllable Sources
Nanotube Random Access Memory (NRAM) Systems and Circuits, with Same
• Non-volatile field effect devices (FEDs)20,40,60, and80 with controllable sources may be used as the cells of one FED device and interconnected into arrays to form non-volatile nanotube random access memory (NRAM) systems as illustrated further above. In operation, cells with a single FED require a partial (or half-select) mode of operation as illustrated byarray1700 shown inFIG. 18 andcorresponding waveforms1800 shown inFIG. 19, for example. Memory cells that contain two select device (transistors) T and T′, and two non-volatile nanotube storage element NT and NT′ (2T/2NT cells) use full cell select operation, and do not require nanotube partial (or half-select) operation. By using full select operation, nanotube electrical characteristics such as threshold voltage and resistance may be operated over a wider range of values, and sensing may be faster because true and complement bit lines BL and BLb, respectively, are used in a differential signal mode. Cell size (area), however, is increased significantly (by more than two times single FED cell area). By way of example, two FED480 (FIG. 2D) devices are used to form a non-volatile NRAM memory cell that is also referred to as a two device NT-on-Source memory cell. Two FED device NT-on-drain cells using non-volatile field effect devices (FEDs)100,120,140, and180 and two FED device NT-on-gate cells using non-volatile field effect devices (FEDs)180,200,220, and240 may also be used (not shown). More than two non-volatile field effect devices (FEDs) per cell may be used for additional performance advantages, for example. Four devices, for example, with separate (non-shared) read and write cell terminals may be used (not shown), however, cell size (area) is increased significantly (by more than four times single FED cells).
• Two FED Device NT-on-Source NRAM Memory Systems and Circuits with Parallel Bit and Reference Lines, and Parallel Word and Release Lines
• NRAM 2T/2NT memory arrays are wired using three sets of unique array lines (a set of word lines and two sets of complementary bit lines), and one group of shared reference lines all at the same voltage, zero (ground) in this example. Read and write word line WL is used to gate select devices T and T′, read and write bit line BL is attached to a shared drain between two adjacent select T devices, and read and write complementary bit line BLb (or BL′) is attached to a shared drain between two adjacent select T′ devices. Reference line REF is used to control the NT switch voltage of storage element NT and NT′ and is grounded (zero volts). Voltages applied to the switch-plates and release-plates of NT and NT′ are controlled by transistor T and T′ sources. True bit array lines BL and complementary bit array lines BLb (bit line bar) are parallel to each other, and orthogonal to array word lines WL. Reference array lines may be parallel to bit lines or to word lines, or alternatively, a conductive layer (plane) may be used.
• FIG. 53A depicts two controlled source non-volatile field effect devices, FED480 (FIG. 2D) and memory cell wiring to form non-volatile 2T/2NT NT-on-Source memory cell7000 schematic. A first FED device and associated elements and nodes is referred to asFED4 device80, and a second FED device and associated elements and nodes is referred to as FED4′device80′.Memory cell7000 is configured as two controlled source FED devices sharing a common gate input provided by a common word line WL, with two independent drain connections each connected to complementary bit lines. Word line (WL)7200 connects to terminal T1 ofFED480 and also to terminal T1′ of FED480′; bit line (BL)7300 connects to terminal T2 ofFED480 and complementary bit line (BLb)7300′ connects to terminal T2′ of FED480′; reference line (REF)7400 connects to terminal T3 ofFED480 and terminal T3′ of FED480′.Memory cell7000 performs write and read operations, and stores the information in a non-volatile state. TheFED480 and FED480′ layout dimensions and operating voltages are selected to optimizememory cell7000.Memory cell7000 FET select device (T)gate7040 and select device (T′)gate7040′ correspond togate82;drains7060 and7060′ correspond to drain84; andcontrollable sources7080 and7080′ correspond tocontrollable source86.Memory cell7000 nanotube (NT) switch-plates7120 and7120′ correspond to switch-plate88; NT switches1140 and1140′ correspond toNT switch90; release-plateinsulator layer surfaces7184 and7184′ correspond to release-plateinsulator layer surface96; and release-plates7180 and7180′ correspond to release-plate94. The interconnections between the elements ofmemory cell7000 schematic correspond to the interconnection of the corresponding interconnections of the elements ofFED480.BL7300 connects to drain7060 throughcontact7320 andBLb7300′ connects to drain7060′ throughcontact7320′;REF7400 connects toNT switch7140 and in parallel to NT′ switch7141′ throughconnector7145;WL7200 interconnects togate7040 bycontact7220 and interconnects togate7040′ bycontact7220′. Alternatively,WL7200 may form andinterconnect gates1040 and1040′, requiring no separate contacts, as shown further below.Transistor T source7080 connects to nanotube NT switch-plate7120 and connects to nanotube NT′ release-plate7180′ throughconnector7190. Transistor T′source7080′ connects to nanotube NT release-plate7180 and connects to nanotube NT′ switch-plate7120′ throughconnector7190′.
• In operation, the non-volatileNT switching element7140 may be caused to deflect to switch-plate surface7120 via electrostatic forces to closed (“ON”) position7140S1, and non-volatile NT′ switchingelement7140′ may be caused to deflect toinsulator7184′ on release-plate7180′ via electrostatic forces to open (“OFF”)position7140′S2, to store a logic “1” state as illustrated inFIG. 53B. That is, a logic “1” state for the twoFED cell7000 consists of NT in closed (“ON”) position7140S1 and NT′ in open (“OFF”)position7140′S2, as illustrated inFIG. 53B. The van der Waals forces holdnanotube switches7140 and7140′ in positions7140S1 and7140′S2, respectively. Alternatively, the non-volatile NT switching element7140-may be caused to deflect toward release-plate7180 via electrostatic forces to open (“OFF”) position7140S2, andnon-volatile switching element1140′ may be caused to deflect toward switch-plate7120′ to closed (“ON”)position7140′S1, to store a logic “0” state as illustrated inFIG. 53C. That is, a logic “0” state for the twoFED cell7000 consists of NT in open (“OFF”) position7140S2 and NT′ in closed (“ON”)position7140′S1, as illustrated inFIG. 53C. The van der Waals forces holdnanotube switches1140 and1140′ in positions7140S2 and7140′S1, respectively. The non-volatile element switching via electrostatic forces is as depicted byelement90 inFIG. 2D withvoltage waveforms311 used to generate the required electrostatic forces illustrated inFIG. 4.
• NT-on-Source schematic7000 forms the basis of a non-volatile 2T/2NT storage (memory) cell. The non-volatile 2T/2NT memory cell may be switched between storage state “1” and storage state “0”, which means the controllable sources may be written to an unlimited number of times as desired, and that the memory cell will retain stored information if power is removed (or lost). In this way, the device may be used as a basis for a non-volatile nanotube random access memory, which is referred to here as a NRAM array, with the ‘N’ representing the inclusion of nanotubes.
• FIG. 54 represents anNRAM array system7700, according to preferred embodiments of the invention. Under this arrangement, an m×n cell array is formed, with only an exemplary 3×2 potion of non-volatile cells ranging from cell C0,0 to cell C2,1 being shown. To access selected cells,array7700 uses read and write word lines (WL0 and WL1), read and write bit lines (BL0, BL1, and BL2) and read and write complementary bit lines (BLb0, BLb1, and BLb2. Reference lines REF are all at the same reference voltage, zero volts in this example. Non-volatile cell C0,0 includes select devices T0,0 and T′0,0, and non-volatile storage elements NT0,0 and NT′0,0. The gates of T0,0 and T′0,0 are coupled to WL0, the drain of T0,0 is coupled to BL0, and the drain of T′0,0 is coupled to BLb0. NT0,0 is the non-volatilely switchable storage element where the NT0,0 switch-plate is coupled to the source of T0,0, the switching NT element is coupled to REF, and the release-plate is coupled to the source of T′0,0. NT′0,0 is the non-volatilely switchable storage element where the NT′0,0 switch-plate is coupled to the source of T′0,0, the switching NT element is coupled to REF, and the release-plate is coupled to the source of T0,0. Word and bit decoders/drivers, sense amplifiers, and controller circuits are explained further below.
• Under preferred embodiments, nanotubes inarray7700 may be in the “ON”, “1” state or the “OFF”, “0” state. The NRAM memory allows for unlimited read and write operations per bit location. A write operation includes both a write function to write a “1” and a release function to write a “0”. By way of example, a write “1” to cell C0,0 and a write “0” to cell C1,0 is described. For a write “1” operation to cell C0,0, select devices T0,0 and T′0,0 are activated when WL0 transitions from 0 to V_SW+V_FET-TH, after BL0 has transitioned to V_SWvolts and after BL0b has transitioned to zero volts. REF voltage is at zero volts. The NT0,0 switch element release-plate is at zero volts, the switch-plate is at V_SWvolts, and the NT switch is at zero volts. The NT′0,0 switch element release-plate is at V_SWvolts, the switch-plate is at zero volts, and the NT′ switch is at zero volts. The BL0 V_SWvoltage is applied to the switch-plate of non-volatile storage element NT0,0 and the release-plate of non-volatile storage element NT′0,0 by the controlled source of select device T0,0. The zero BL0b voltage is applied to the release-plate of non-volatile storage element NT0,0, and to the switch-plate of non-volatile storage element NT′0,0, by the controlled source of select device T′0,0. The difference in voltage between the NT0,0 switch-plate and NT switch is V_SWand generates an attracting electrostatic force. The voltage difference between the release-plate and NT switch is zero so there is no electrostatic force. The difference in voltage between NT′0,0 release-plate and NT′ switch is V_SWand generates an attracting electrostatic force. The voltage difference between the switch-plate and NT′ switch is zero so there is no electrostatic force. If V_SWexceeds the nanotube threshold voltage V_NT-TH, the nanotube structure switches to “ON” state or logic “1” state, that is, the nanotube NT switch and switch-plate of non-volatile storage element NT0,0 are electrically connected as illustrated inFIG. 53B, and the nanotube NT switch and release-plate dielectric of non-volatile storage element NT′0,0 are in contact as illustrated inFIG. 53B. The near-Ohmic connection between switch-plate7120 andNT switch7140 in position7140S I represents the “ON” state or “1” state. If the power source is removed, cell C0,0 remains in the “ON” state.
• For a write “0” operation to cell C1,0, select devices T1,0 and T′1,0 are activated when WL0 transitions from 0 to V_SW+V_FET-TH, after BL1 has transitioned to zero volts and after BL1b has transitioned to V_SWvolts. REF voltage is at zero volts. The NT1,0 switch element release-plate is at V_SWvolts, the switch-plate is at zero volts, and the NT switch is at zero volts. The NT′1,0 switch element release-plate is at zero volts, the switch-plate is at V_SWvolts, and the NT′ switch is at zero volts. The BL1 zero volts is applied to the switch-plate of non-volatile storage element NT1,0 and the release-plate of non-volatile storage element NT′1,0 by the controlled source of select device T1,0. The V_SWBL0b voltage is applied to the release-plate of non-volatile storage element NT1,0, and to the switch-plate of non-volatile storage element NT′1,0, by the controlled source of select device T′1,0. The difference in voltage between the NT1,0 switch-plate and NT switch is zero and generates no electrostatic force. The voltage difference between the release-plate and NT switch is V_SWso there is an attracting electrostatic force. The difference in voltage between NT′1,0 release-plate and NT′ switch is zero volts and generates no electrostatic force. The voltage difference between the switch-plate and NT′ switch is V_SWso there is an attracting electrostatic force. If V_SWexceeds the nanotube threshold voltage V_NT-TH, the nanotube structure switches to “OFF” state or logic “0” state, that is, the nanotube NT′ switch and switch-plate of non-volatile storage element NT′1,0 are electrically connected as illustrated inFIG. 53C, and the nanotube NT switch and release-plate dielectric of non-volatile storage element NT1,0 are in contact as illustrated inFIG. 53C. The near-Ohmic connection between switch-plate7120′ and NT′switch7140′ inposition7140′S1 represents the “OFF” state or “0” state. If the power source is removed, cell C1,0 remains in the “ON” state.
• An NRAM read operation does not change (destroy) the information in the activated cells, as it does in a DRAM, for example. Therefore the read operation in the NRAM is characterized as a non-destructive readout (or NDRO) and does not require a write-back after the read operation has been completed. For a read operation of cell C0,0, BL0 and BL0b are driven high to V_DDand allowed to float. WL0 is driven high to V_DD+V_FET-THand select devices T0,0 and T′0,0 turn on. REF0 is at zero volt. If cell C0,0 stores an “ON” state (“1” state) as illustrated inFIG. 53B, BL0b remains unchanged, and BL0 discharges to grounded REF line through a conductive path that includes select device T0,0 and non-volatile storage element NT0,0, the BL0 voltage drops, and the “ON” state or “1” state is detected by a sense amplifier/latch circuit (shown further below) that records the voltage drop of BL0 relative to BL0b by switching the latch to a logic “1” state. BL0 is connected by the select device T0,0 conductive channel of resistance R_FETto the switch-plate of NT0,0. The switch-plate of NT0,0 is in contact with the NT switch with a contact resistance R_SWand the NT switch contacts reference line REF0 with contact resistance R_C. The total resistance in the discharge path is R_FET+R_SW+R_C. Other resistance values in the discharge path, including the resistance of the NT switch, are much small and may be neglected.
• For a read operation of cell C1,0, BL1 and BL1b are driven high to V_DDand allowed to float. WL0 is driven high to V_DD+V_THand select devices T1,0 and T′1,0 turn on. REF1 is at zero volts. If cell C1,0 stores an OFF state (“0” state) as illustrated inFIG. 53C, BL1 remains unchanged, and BL1b discharges to grounded REF line through a conductive path that includes select device T′1,0 and non-volatile storage element NT′1,0, the BL1b voltage drops, and the OFF state or “0” state is detected by a sense amplifier/latch circuit (shown further below) that records the voltage drop of BL1b relative to BL1 by switching the latch to a logic “0” state. BL1b is connected by the select device T′1,0 conductive channel of resistance R_FETto the switch-plate of NT′1,0. The switch-plate of NT′1,0 is in contact with the NT′ switch with a contact resistance R_SWand the NT′ switch contacts reference line REF0 with contact resistance R_C. The total resistance in the discharge path is R_FET+R_SW+R_C. Other resistance values in the discharge path, including the resistance of the NT switch, are much small and may be neglected.
• FIG. 55 illustrates theoperational waveforms7800 ofmemory array7700 shown inFIG. 54 during read “1”, read “0”, write “1”, and write “0” operations for selected cells, while not disturbing unselected cells (no change to unselected cell stored logic states).Waveforms7800 illustrate voltages and timings to write logic state “1” in cell C0,0, write a logic state “0” in cell C1,0, read cell C0,0 which is in the “1” state, and read cell C1,0 which is in the “0” state.Waveforms7800 also illustrate voltages and timings to prevent disturbing the stored logic states (logic “1” state and logic “0” state) along selected word line WL0 in this example. Word line WL0 turns on transistors T2,0 and T′2,0 of cell C2,0 after bit lines BL2 and BL2b have been set to zero volts. No voltage difference exists between NT and NT′ switches and corresponding switch-plates and release-plates because REF is also at zero volts, and the stored state of cell C2,0 is not disturbed. All other unselected cells along active word line WL0 are also not disturbed. Word line WL1 is not selected and is held at zero volts, therefore all select transistors along word line WL1 are in the OFF state and do not connect bit lines BL and BL′ to corresponding source terminals. Therefore, cells C0,1, C1,1, C2,1, and any other cells along word line WL1 are not disturbed. Cells inmemory array7700 tolerate unlimited read and write operations at each memory cell location with no stored state disturbs, and hold information in a non-volatile mode (without applied power).
• Non-volatile NT-on-sourceNRAM memory array7700 with bit lines parallel to reference lines is shown inFIG. 54 contains6 bits, a subset of a 2^N×2^Marray7700, and is a subset of non-volatileNRAM memory system7810 illustrated asmemory array7815 inFIG. 56A.NRAM memory system7810 may be configured to operate like an industry standard asynchronous SRAM or synchronous SRAM because nanotubenon-volatile storage cells7000 shown inFIG. 53A, inmemory array7700, may be read in a non-destructive readout (NDRO) mode and therefore do not require a write-back operation after reading, and also may be written (programmed) at CMOS voltage levels (5, 3.3, and 2.5 volts, for example) and at nanosecond and sub-nanosecond switching speeds. NRAM read and write times, and cycle times, are determined by array line capacitance, and are not limited by nanotube switching speed. Accordingly,NRAM memory system7810 may be designed with industry standard SRAM timings such as chip-enable, write-enable, output-enable, etc., or may introduce new timings, for example. Non-volatileNRAM memory system7810 may be designed to introduce advantageous enhanced modes such as a sleep mode with zero current (zero power—power supply set to zero volts), information preservation when power is shut off or lost, enabling rapid system recovery and system startup, for example.NRAM memory system7810 circuits are designed to provide thememory array7700waveforms7800 shown inFIG. 55.
• NRAM memory system7810 accepts timinginputs7812, acceptsaddress inputs7825, and acceptsdata7867 from a computer, or providesdata7867 to a computer using a bidirectional bus sharing input/output (I/O) terminals. Alternatively, inputs and outputs may use separate (unshared) terminals (not shown). Address input (I/P)buffer7830 receives address locations (bits) from a computer system, for example, and latches the addresses. Address I/P buffer7830 provides word address bits toword decoder7840 viaaddress bus7837; address I/P buffer7830 provides bit addresses to bit decoder7850 viaaddress bus7852; and address bus transitions provided bybus7835 are detected by function generating, address transition detecting (ATD), timing waveform generator, controller (controller)7820.Controller7820 provides timing waveforms onbus7839 toword decoder7840.Word decoder7840 selects the word address location withinarray7815.Word address decoder7840 is used to decode word lines WL and drives word line (WL) using industry standard circuit configurations resulting inNRAM memory system7810 on-chip WL waveforms for both write-one, write-zero, read-one, and read-zero operations as illustrated bywaveforms7800′ shown inFIG. 57.FIG. 57NRAM memory system7810waveforms7800′ correspond tomemory array7700waveforms7800 shown inFIG. 55.
• Bit address decoder7850 is used to decode bit lines BL.Controller7820 provides timing waveforms onbus7854 tobit decoder7850.BL decoder7850 uses inputs frombus7854 andbus7857 to generate data multiplexer select bits onbus7859. The output ofBL decoder7850 onbus7859 is used to select control data multiplexers using combined data multiplexers & sense amplifiers/latches (MUXs & SAs)7860.Controller7820 provides function and timing inputs onbus7857 to MUXs &SAs7860, resulting inNRAM memory system7810 on-chip BL waveforms for both write-one, write-zero, read-one, and read-zero operations as illustrated bywaveforms7800′ shown inFIG. 57 corresponding tomemory array7700waveforms7800 shown inFIG. 55. MUXs &SAs7860 are used to write data provided by read/write buffer7865 viabus7864 inarray7815, and to read data fromarray7815 and provide the data to read/write buffer7865 viabus7864 as illustrated inwaveforms7800′.
• Sense amplifier/latch7900 is illustrated inFIG. 56B.Flip flop7910, comprising two back-to-back inverters is used to amplify and latch data inputs fromarray7815 or from read/write buffer7865.Transistor7920 connectsflip flop7910 to ground when activated by a positive voltage supplied bycontrol voltage V_TIMING7980, which is provided bycontroller7820.Gating transistor7930 connects a bit line BL tonode7965 offlip flop7910 when activated by a positive voltage.Gating transistor7940 connects a bit line BLb to flipflop node7975 when activated by a positive voltage.Transistor7960 connects voltage V_DDto flipflop7910node7965,transistor7970 connects voltage V_DDto flipflop7910node7975, andtransistor7950 ensures that small voltage differences are eliminated whentransistors7960 and7970 are activated.Transistors7950,7960, and7970 are activated (turned on) when gate voltage is low (zero, for example).
• In operation, V_TIMINGvoltage is at zero volts whensense amplifier7900 is not selected.NFET transistors7920,7930, and7940 are in the “OFF” (non-conducting) state, because gate voltages are at zero volts.PFET transistors7950,7960, and7970 are in the “ON” (conducting) state because gate voltages are at zero volts. V_DDmay be 5, 3.3, or 2.5 volts, for example, relative to ground.Flip flop7910nodes7965 and7975 are at V_DD. If sense amplifier/latch7900 is selected, V_TIMINGtransitions to V_DD,NFET transistors7920,7930, and7940 turn “ON”,PFET transistors7950,7960, and7970 are turned “OFF”, andflip flop7910 is connected to bit line BL and to bit line BLb. As illustrated by waveforms BL0, BL0b, BL1, and BL1b ofwaveforms7800′, bit line BL and BLb are pre-charged prior to activating a corresponding word line (WL0 in this example). Ifcell7000 of memory array7700 (memory system array7815) stores a “1”, then bit line BL and BLb inFIG. 56B correspond to BL0 and BLb, respectively, inFIG. 54. BL is discharged bycell7000, voltage droops below V_DD, BLb is not discharged, and sense amplifier/latch7900 detects a “1” state. Ifcell7000 of memory array7700 (memory system array7815) stores a “0”, then bit line BL and BLb inFIG. 20B corresponds to BL1 and BL1b, respectively, inFIG. 54. BLb is discharged bycell7000, voltage droops below V_DD, BL is not discharged, and sense amplifier/latch7900 detect a “0” state. The time from sense amplifier select to signal detection by sense amplifier/latch7900 is referred to as signal development time. Sense amplifier/latch7900 typically requires 75 to 100 mV difference voltage in order to switch. It should be noted thatcell7000 requires a nanotube “OFF” resistance to “ON” resistance ratio of greater than 10 to 1 for successful operation. A typical bit line BL has a capacitance value of 250 fF, for example. A typical nanotube storage device (switch) or dimensions 0.2 by 0.2 um typically has 8 nanotube filaments across the suspended region, for example, as illustrated further below. For a combined contact and switch resistance of 50,000 ohms per filament, as illustrated further below, the nanotube “ON” resistance ofcell7000 is 6,250 ohms. For a bit line of 250 fF, the time constant RC=1.6 ns. The sense amplifier signal development time is less than RC, and for this example, is between 1 and 1.5 nanoseconds.
• Non-volatileNRAM memory system7810 operation may be designed for high speed cache operation at 5 ns or less access and cycle time, for example. Non-volatileNRAM memory system7810 may be designed for low power operation at 60 or 70 ns access and cycle time operation, for example. For low power operation, address I/P buffer7830 operation requires 8 ns;controller7820 operation requires 16 ns;bit decoder7850 plus MUXs &SA7860 operation requires 12 ns (word decoder7840 operation requires less than 12 ns);array7815 delay is 8 ns; sensing7900 operation requires 8 ns; and read/write buffer7865 requires 12 ns, for example. The access time and cycle time of non-volatileNRAM memory system7810 is 64 ns. The access time and cycle time may be equal because the NDRO mode of operation of nanotube storage devices (switches) does not require a write-back operation after access (read).
• Method of Making Two FED Device NT-on-Source Memory System and Circuits
• Two FED480 (FIG. 2D) controllable source devices are interconnected to form a non-volatile two transistor, two nanotube (2T/2NT) NRAM memory cell that is also referred to as a two device NT-on-source memory cell. The 2T/2NT NT-on-source NRAM memory array is fabricated using the same method steps used to fabricate 1T/1NT NT-on-source memory structure3225 shown inFIG. 30M′.
• FIG. 22 describes thebasic method3000 of manufacturing preferred embodiments of the invention. In general, preferred methodsfirst form3002, a base structure including field effect devices similar to a MOSFET, having drain, source, gate nodes, and conductive studs on source and drain diffusions for connecting to additional layers above the MOSFET device that are used to connect to the nanotube switch fabricated above the MOSFET device layer, bit lines, and other structures.Base structure8102 withsurface8104 illustrated inFIG. 58A is similar tobase structure3102′ withsurface3104′ shown inFIG. 30M′ with transistors, except source diffusions have been elongated to accommodate connection to a NT-on-sourcenanotube switch structure8233 and acell interconnect structure8235. Thecell interconnect structure8235 contacts sourcediffusion region8124 and is formed in the same way asdrain contact structure8118 and8118A, and is used for internal (local) cell wiring as is explained further below.
• Oncebase structure8102 is defined, then methods of fabricating 2T/2NT NT-on-source memory arrays is the same as methods of fabricating 1T/1NT NT-on-source memory arrays already described.Preferred methods3004 shown in FIGS.23,23′, and23″ and associated figures;methods3036 shown inFIG. 26 and associated figures;methods3006 shown inFIGS. 27 and 27′ and associated figures; andmethods3008 shown inFIGS. 28 and 28′ and associated figures. Conductors, semiconductors, insulators, and nanotubes are formed in the same sequence and are in the same relative position in the structure. Length, width, thickness dimensions may be different and the choice of conductor material may be different reflecting differences in design choices. Also, interconnections may be different because of cell differences. The function of electrodes are the same, however, interconnections may be different.FIGS. 58A and 58B cross sections illustrated further below correspond toFIG. 30M′ of 1T/1NT NT-on-source cross section.
• FIG. 58A illustrates cross section A-A′ ofarray8000 taken at A-A′ of the plan view ofarray8000 illustrated inFIG. 58D, and showsFET device region8237 in the FET length direction,elongated source8124 to accommodatenanotube switch structure8233 andcell interconnect region8235.Bit line8138 contacts drain8126 throughcontact8140, conductingstuds8118A and8118, andcontact8123. WhenFET device region8237 FET channel is formed insubstrate8128 belowFET gate8120,bit line8138 is electrically connected to elongatedsource diffusion8124, which connects to switch-plate8106 throughcontact8121, conductingstud8222, andcontact8101, and to release-plate extension8205R throughcontact8340, conductingstuds8300 and8300A, andcontact8320, as illustrated in54A.Nanotube switch structure8233 corresponds to nanotubeswitch structure3133 inFIG. 30M′ with switch-plate8106,dielectric layer8108 betweennanotube8114 layer andswitch plate8106, combinedconductors8119 and8117 forming a picture frameregion contacting nanotube8114 layer,insulator8203 insulates the underside of release-plate8205. Nanotube reference (picture-frame)region extension8119R contacts and is a part ofreference array line8400 shown inFIG. 58D. Structures are embedded indielectric layer8116, SiO₂for example, except for gap regions above and below nanotube layers in the nanotube switching region.FIG. 58B illustrates cross section B-B′ ofarray8000 taken at B-B′ of the plan view ofarray8000 illustrated inFIG. 58D, and showsFET device region8237′ in the FET length direction,elongated source8124′ to accommodatenanotube switch structure8233′ andcell interconnect region8235′.Bit line8138′ contacts drain8126′ throughcontact8140′,conductive studs8118A′ and8118′, andcontact8123′. WhenFET device region8237′ FET channel is formed insubstrate8128 belowFET gate8120,bit line8138′ is electrically connected to elongatedsource diffusion8124′, which connect to switch-plate8106′ throughcontact8121′, conductingstud8222′, andcontact8101′, and to release-plate extension8205R′ throughcontact8340′, conductive studs8300′ and8300A′, andcontact8320′, as illustrated inFIG. 58B.Nanotube switch structure8233′ corresponds to nanotubeswitch structure8233 andstructure3133 inFIG. 30M′. Nanotube reference (picture-frame)region extension8119R′ contacts and is a part ofreference array line8400 shown inFIG. 58D.FIG. 58C illustrates cross section C-C′ ofarray8000 taken at C-C′ of plan view ofarray8000 illustrated inFIG. 58D, and showsnanotube switch structure8233 with switch-plate8106 connected to sourcediffusion8124 as further described with respect toFIG. 58A. Release-plate8205extension8205R connects release-plate8205 to sourcediffusion8124′ throughcontact8320′, conductingstuds8300A′ and8300′, andcontact8340′, all withincell8500 boundaries. Thus,source8124 diffusion is electrically connected to switch-plate8106 ofnanotube switch structure8233, andsource8124′ diffusion is electrically connected to release-plate8205 ofnanotube switch structure8233 as illustrated inFIG. 58C, andFIG. 58D. A corresponding interconnection means is used to electrically connectsource8124′ to switch-plate8106′ ofnanotube switch structure8233′, and also to electrically connectsource8124 to releaseplate8205′ ofnanotube switch structure8233′ as illustrated inFIG. 58D.FIG. 58D illustrates a plan view of non-volatile 2T/2NT NT-on-source array8000 including two interconnected NT-on-source FED480 structures having twotransistor regions8237 and8237′ and twonanotube switch structures8233 and8233′; twocell interconnect regions8235 and8235′ including release-plate interconnect extensions8205R and8205R′, and nanotube reference (picture-frame)region extensions8119R and8119R′ contacting arrayreference line REF8400; array word line8120Aforms gates8120 and8120′ of the FET select devices;bit line BL8138 contactingdrain8126 throughcontact8140 and underlying stud and contact shown inFIG. 58A;bit line BLb8138′ contactingdrain8126′ throughcontact8140′ and underlying stud and contact shown inFIGS. 58B; in terms of minimum technology feature size, 2T/2NT NT-on-source cell8500 area is approximately 45 F². If sub-minimum technology features are used in the NT switch structure (not shown), theminimum cell8500 area in terms of minimum technology feature size is 30 F². Nanotube-on-source array8000 structures illustrated inFIGS. 58A, 58B,58C, and58D correspond to 2T/2NT nanotube-on-source array7700 schematic representations illustrated inFIG. 54.Bit line3138 structures correspond to any of bit lines BL0 to BL2 schematic representations;bit line8138′ structures correspond to any of bit lines BL0b to BLb2 schematic representations;common reference line8400 structures correspond to common reference lines REF schematic representations; word line3120A structures correspond to any of word lines WL0 and WL1 schematic representations;nanotube switch structures3233 and3233′ correspond to any of NT0,0 to NT2,1 and NT′0,0 to NT′2,1 schematic representations, respectively;FET3237 and3237′ structures correspond to any of FETs T0,0 to T2,1 and T′0,0 to T′2,1 schematic representations, respectively; andexemplary cell8500 corresponds to any of cells C0,0 to cell C2,1 schematic representations.
• Methods to increase the adhesion energies through the use of ionic, covalent or other forces may be used to alter the interactions with the electrode surfaces. These methods can be used to extend the range of stability within these junctions.
• Nanotubes can be functionalized with planar conjugated hydrocarbons such as pyrenes which may then aid in enhancing the internal adhesion between nanotubes within the ribbons. The surface of the substrate used can be derivatized/functionalized to create a more hydrophobic or hydrophilic environment to promote better adhesion of nanotubes. The nature of the substrate allows control over the level of dispersion of the nanotubes to generate monolayer nanotube fabric.
• Preferred nanofabrics have a plurality of nanotubes in contact so as to form a non-woven fabric. Gaps in the fabric, i.e., between nanotubes either laterally or vertically, may exist. The fabric preferably has a sufficient amount of nanotubes in contact so that at least one electrically conductive, semi-conductive or mixed conductive and semi-conductive pathway exists from a given point within a ribbon or article to another point within the ribbon or article (even after patterning of the nanofabric).
• Though certain embodiments prefer single-walled nanotubes in the nanofabrics, multi-walled nanotubes may also be used. In addition, certain embodiments prefer nanofabrics that are primarily a monolayer with sporadic bilayers and trilayers, but other embodiments benefit from thicker fabrics with multiple layers.
• It will be further appreciated that the scope of the present invention is not limited to the above-described embodiments but rather is defined by the appended claims, and that these claims will encompass modifications and improvements to what has been described.

Claims (19)

1. A method of making a non-volatile field effect device, the method comprising:

forming a semiconductor field effect device comprising a gate, source, and drain;

forming a gate terminal, a source terminal, and a drain terminal;

forming a control terminal; and

forming a nanotube switching element between one of the gate, source, and drain and its corresponding terminal, the nanotube switching element being switchable, in response to electrical stimulus on the control terminal and at least one of the gate, source, and drain terminals, between a first non-volatile state that enables current flow between the source and drain and a second non-volatile state that disables current flow between the source and drain.

2. The method ofclaim 1, further comprising providing a direct electrical connection between the others of the gate, source and drain and their corresponding terminal,

3. The method ofclaim 1, wherein forming the nanotube switching element comprises forming a non-woven fabric of nanotubes.

4. The method ofclaim 3, wherein forming the nanotube switching element further comprises defining a pattern in the non-woven fabric of nanotubes and removing nanotubes in accordance with the defined pattern.

5. The method ofclaim 3, further comprising forming a sacrificial layer, forming the non-woven fabric of nanotubes over the sacrificial layer, and then removing the sacrificial layer to suspend a portion of the non-woven fabric of nanotubes.

6. The method ofclaim 1, comprising forming the nanotube switching element in direct electrical communication with the one of the gate, source, and drain.

7. The method ofclaim 6, comprising forming the nanotube switching element over the one of the gate, source, and drain.

8. The method ofclaim 1, comprising forming the nanotube switching element in direct electrical communication with the terminal corresponding to the one of the gate, source, and drain.

9. The method ofclaim 1, comprising forming the nanotube switching element and control terminal in spaced relation to one another.

10. The method ofclaim 9, further comprising depositing a dielectric material on a surface of the control terminal facing the nanotube switching element.

11. The method ofclaim 1, comprising forming the nanotube switching element on a substantially horizontal surface.

12. The method ofclaim 1, further comprising forming a field-modulatable channel adjacent the gate and between the source and drain, and capable of current flow in response to electrical stimulus at the gate.

13. A method of making a non-volatile field effect device, the method comprising:

forming a gate, a source, and a drain, with a conductive channel between the source and the drain; and

forming a nanotube switch having a corresponding control terminal and being positioned to control electrical conduction through said conductive channel.

14. The method ofclaim 13, wherein the nanotube switch is positioned to be switchable between a first nonvolatile logical state in which electrical conduction through said conductive channel is enabled, and a second nonvolatile logical state in which electrical conduction through said conductive channel is disabled.

15. The method ofclaim 14, wherein the first nonvolatile logical state comprises the nanotube switch contacting one of the gate, the source, and the drain, and wherein the second nonvolatile logical state comprises the nanotube switch contacting the control terminal.

16. The method ofclaim 14, wherein the first nonvolatile logical state comprises the nanotube switch contacting a terminal corresponding to one of the gate, the source, and the drain, and wherein the second nonvolatile logical state comprises the nanotube switch contacting the control terminal.

17. The method ofclaim 13, wherein forming the nanotube switch comprises forming a non-woven fabric of nanotubes.

18. The method ofclaim 17, wherein forming the nanotube switch further comprises defining a pattern in the non-woven fabric of nanotubes and removing nanotubes in accordance with the defined pattern.

19. The method ofclaim 13, comprising forming the nanotube switch and control terminal in spaced relation to one another.

Images