US20100027347A1

US20100027347A1 - Three-Terminal Single Poly NMOS Non-Volatile Memory Cell

Info

Publication number: US20100027347A1
Application number: US12/582,646
Authority: US
Inventors: Evgeny Pikhay; Yakov Roizin
Original assignee: Tower Semiconductor Ltd
Current assignee: Tower Semiconductor Ltd
Priority date: 2008-02-25
Filing date: 2009-10-20
Publication date: 2010-02-04
Also published as: US7859043B2; US20100188901A1; US20090213660A1

Abstract

A three terminal non-volatile memory (NVM) cell for a CMOS IC is formed by a standard CMOS process flow. The NVM cell includes two transistors that share a common floating gate. The floating gate includes a first portion disposed over the channel region of the first (NMOS) transistor, a second portion disposed over the channel region of the second (NMOS or PMOS) transistor, and a third portion extending into an enlarged drain diffusion area away from the channel regions, whereby the gate-to-drain capacitance is higher than the gate-to-source capacitances. A pocket implant or CMOS standard LV N-LDD is formed under the second transistor to enhance CHE programming. Both HV LDD and LV LDD implants are introduced together enabling LDD implant merging under the floating gate extension. The floating gate is formed using substantially T-shaped, C-shaped, U-shaped, Y-shaped or O-shaped polysilicon structures. Various array addressing schemes are disclosed.

Description

RELATED APPLICATIONS

This application is a continuation of U.S. patent application for “Three-Terminal Single Poly NMOS Non-Volatile Memory Cell”, U.S. application Ser. No. 12/398,912, filed Mar. 5, 2009 which is a continuation in part of U.S. patent application for “Asymmetric Single Poly NMOS Non-Volatile Memory Cell”, U.S. application Ser. No. 12/037,051, filed Feb. 25, 2008. This application also claims priority of U.S. Provisional Patent application for “Three Terminal NVM Cell And Memory Array”, U.S. App. Ser. No. 61/115,773, filed Nov. 18, 2008.

FIELD OF THE INVENTION

The present invention relates to non-volatile memory (NVM) cells. More specifically, the present invention relates to NVM cell arrays that are “embedded” in (i.e., integrally formed with) complementary metal-oxide-semiconductor (CMOS) integrated circuits (ICs), and to methods for fabricating the NVM cells in an inexpensive manner using standard CMOS process flows.

BACKGROUND OF THE INVENTION

“CMOS” refers to both a particular style of digital circuitry design, and the family of processes used to implement that circuitry on IC “chips” or “die”. CMOS logic uses a combination of p-type and n-type metal-oxide-semiconductor field-effect transistors (MOSFETs) to implement logic gates and other circuits found in computers, telecommunication equipment, and signal processing equipment. Typical commercial CMOS ICs include millions (or hundreds of millions) of n-type and p-type MOSFETs.
Most CMOS IC manufacturers (aka, “fabs”) generate standardized process “flows” for generating CMOS ICs on monocrystalline silicon wafers. Each CMOS process flow includes a series of processing steps (e.g., material deposition, photolithographic exposure, and etching) that are required to produce a desired CMOS IC product. Standard CMOS process flows are typically developed to produce “normal” CMOS IC devices (i.e., CMOS IC devices that comprise mainly volatile n-type and p-type MOSFETS) using a minimum number of processing steps in order to minimize overall production costs. Significant effort is typically expended by each manufacturer to make their standard CMOS process flow as time and cost efficient as possible. Once a standard CMOS flow is optimized, it can typically be used to make a large number of CMOS IC designs by merely by providing a different set of photolithograpy masks for each IC design, and then repeating the standard CMOS process flow using the selected set of photolithograpy masks.
Although most standard CMOS process flows facilitate the inclusion of non-MOSFET circuit components into the CMOS IC products, a problem arises when a circuit design requires a circuit component that cannot be produced by the standard CMOS process flow. In this case, the CMOS process flow must be modified at great expense to include additional steps in order to produce the needed circuit component. It is therefore desirable to develop methods for producing the non-standard circuit component using the steps of the existing CMOS process flow.
Non-volatile memory (NVM) or “floating gate” cells represent one type of non-standard circuit component that is often needed in large scale CMOS ICs. In contrast to volatile (aka primary storage) memory built of typical n-type and p-type MOSFETs which require continuous power to retain stored information, NVM cells are able to retain a stored state even when power to an IC is turned off, thereby allowing the IC to “remember” important operating conditions and information upon restart. Several types of NVM cells have been developed that can be produced with minimal changes to a standard CMOS process flow. One NVM cell that has a small floating gate capacitively coupled to the drain area is disclosed in U.S. Pat. No. 6,678,190. This NVM cell is a programmable read only memory comprising two serially connected P-type metal-oxide semiconductor (MOS) transistors, wherein the floating gate coupling to the source and drain is symmetric, and wherein a control gate is omitted because a special gate bias is not necessary in the programming mode. A problem with the PMOS NVM cells of U.S. Pat. No. 6,678,190 is that only a small portion of Vd is transferred to the floating gate, so that a subsequent erase operation is practically impossible, thus limiting the application of corresponding cells to OTP (one-time program) memories only.
What is needed is an NVM cell that is small to medium in size, can be produced using a standard CMOS process flow having a single polysilicon layer, and exhibits high endurance (allows multiple cycling, i.e., program/erase operations).

SUMMARY OF THE INVENTION

The present invention is directed to medium-sized, three- or four-terminal (two-channel) NVM (logic) cells that are produced using standard single-poly CMOS process flows (i.e., do not require the use of masks other than those used in the standard CMOS process flow) and exhibit high endurance. Each NVM cell includes two NMOS transistors—a read transistor and an injection (program/erase) transistor that share a common drain region and a common floating gate. In the three-terminal embodiments, the read transistor and injection transistor have respective (first and second) source regions disposed adjacent to the shared drain region, and the floating gate has a first portion disposed over a first channel region located between the first source region and the drain region, a second portion disposed over a second channel region located between the second source region and the drain region, and a third (extension) portion that overlaps (i.e., extends “over”) the shared drain (diffusion) region at a location away from the first and second channel regions, and is coupled to the drain region such that the gate-to-drain capacitance between the floating gate and the drain region is substantially higher than the gate-to-source capacitances between the floating gate and either of the first and second source regions. In addition, the injection transistor is formed with a channel hot electron (CHE) enhancing pocket implant (e.g., Boron or BF₃), and the reading transistor is formed using standard CMOS implants (i.e., outside of the special pocket implant area). This two-channel arrangement facilitates desirable channel hot electron (CHE) or secondary channel hot electron (CHISEL) programming of the floating gate. In the first case this is achieved by applying a positive programming voltage to the drain region and coupling the injection transistor's (second) source region to ground (while the source of reading transistor is floating), whereby the drain voltage is transferred to the floating gate as a result of the high capacitive coupling, and CHE injection from the drain region into the floating gate is induced in the region of the injection transistor due to the CHE enhancing pocket implant. In the second embodiment (CHISEL), the programming process is enhanced by applying a negative bias to the transistor bulk. For this purpose, the whole device is placed in an isolated P-well, or positive voltage is applied to the source of the injection transistor with the bulk connected to ground. Erasing the floating gate involves floating the drain diffusion and the source of the read transistor and applying the positive voltage to the injection (second) source region, thereby causing Band-to-Band Tunneling (BBT) generation of holes and their injection into the floating gate. Tests performed on two-channel NVM cells produced in accordance with the present invention exhibit superior performance (suppressed read disturb) than comparable one-channel (two-terminal) NVM cells. During cell read-out operations a read voltage lower than the required programming voltage, e.g., 1.5-2V, is applied to the shared drain region, the injection (second) source region is disconnected (floating), and the read (first) source region is coupled directly to ground, and the cell drain-source current is compared with the reference current to test whether the cell is programmed or erased. Because the read transistor is located outside of the CHE enhancing pocket implant, CHE are not expressed in the read transistor during the read operation, thus read-disturb effects in the memory device are strongly suppressed. Therefore, three-terminal NVM cells produced in accordance with the present invention are ideal for incorporating into low cost CMOS integrated circuits (ICs) that require high endurance, high density (i.e., several Kbits to 1 Mbit) NVM cells.
According to one embodiment of the invention, a CMOS IC includes both high voltage (HV) MOSFETs and low voltage (LV) MOSFETs and at least one three-terminal NVM cell formed on a substrate using a standard CMOS process flow. The HV MOSFET and the LV MOSFET are fabricated using processing steps that are optimized for producing these different devices. For example, HV MOSFETs are produced using lower density LDD (HV-LDD) implants, thicker gate oxides (HV-OX) and wider polysilicon gate lines compared with LV devices. HV transistors are optimized for the higher (e.g., 5V or 3.3V) applied voltages. In contrast, LV MOSFETs are produced using higher density LDD (LV-LDD) implants, thinner gate oxides (LV-OX) and narrower polysilicon gate lines that are optimized for the lower (e.g., 1.8V) applied voltages expected on LV MOSFETs during operation. The various process steps used to fabricate HV and LV MOSFETs are included in standard CMOS process flows. In accordance with an aspect of the invention, these various HV and LV process steps are utilized to optimize the three-terminal NVM cells by using the LV gate design rules (LV DR) to define the width of the extension portion of the floating gate, and utilizing one or more of the LDD implants (e.g., Phosphorous and Arsenic) to form a continuous implant region that merges (i.e., extends entirely across) the N+ drain diffusion under the extension portion. The present inventors have found that superior two-channel NVM cells are produced by forming the floating gate on the thicker HV gate oxide, utilizing the HV-NLDD in read transistor and LV-NLDD in the injection transistor, and generating the implant region under the extension portion using the both LV-NLDD and HV-NLDD processing steps. In addition, the CHE enhancing pocket implant is formed under the injection transistor using P-type dopants (e.g., Boron or BF₃) that are utilized in the standard CMOS process flow to produce PMOS transistors. It is clear that the process of three-terminal (two channel) NVM cell fabrication does not require all the operations used in fabrication of LV devices in the core CMOS process flow, thus embodiments where only the necessary operations, such as an additional LVNLDD implant is used in a 5V only 0.18 um process flow are possible.
In accordance with another aspect of the present invention, various drain region and floating gate configurations are used to optimize cell function and minimize cell size. In one embodiment, the first (read), second (injection) portions of the floating gate are connected to the third (drain) by fourth and fifth polysilicon portions such that a T-shaped polysilicon structure is formed. In another embodiment, the floating gate is patterned in a C-shaped structure with both the read and injection transistors located at the same side of the drain region. In yet other embodiments, the first and second floating gate portions are formed in a C-shaped or box-like O-shaped pattern such that two extension portions extend from read/injection transistors disposed at opposite sides of the drain region.
In accordance with another aspect of the present invention, the three-terminal NVM cells are arranged in arrays that facilitate cell operations while minimizing occupied chip space. In one embodiment, each NVM cell is connected between two bit lines and one word line, and program/erase operations are performed by applying programming voltages to the word line and one of the bit lines, while read operations are performed by applying a voltage to the word line and reading the other bit line.
In accordance with another embodiment of the present invention, a four-terminal NVM cell includes a PMOS injection transistor and an NMOS read transistor. The advantage of the described embodiment is low program currents.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other features, aspects and advantages of the present invention will become better understood with regard to the following description, appended claims, and accompanying drawings, where:

FIG. 1 is a simplified circuit diagram showing a CMOS IC including LV and HV MOSFETs and a two-channel NMOS NVM cell according to an embodiment of the present invention;

FIGS. 2(A), 2(B) and 2(C) are simplified circuit diagrams showing connections to the two-channel NMOS NVM cell of FIG. 1 during program, erase and read operations, respectively, according to an embodiment of the present invention;

FIG. 3 is a flow diagram depicting a standard CMOS flow utilized to produce CMOS circuit of FIG. 1 according to another embodiment of the present invention;

FIG. 4 is simplified top plan (layout) view showing portions of a CMOS IC including a three terminal (two-channel) NMOS NVM cell according to a specific embodiment of the present invention;

FIG. 5 is a cross-sectional side view showing the CMOS IC taken along section line 4-4 of FIG. 4;

FIG. 6 is simplified top plan (layout) view showing a three terminal NMOS NVM cell according to another specific embodiment of the present invention;

FIG. 7 is simplified top plan (layout) view showing a three terminal NMOS NVM cell according to yet another specific embodiment of the present invention;

FIG. 8 is simplified top plan (layout) view showing a three-terminal NMOS NVM cell according to yet another specific embodiment of the present invention;

FIGS. 9(A) and 9(B) are simplified circuit diagrams showing arrays of three-terminal NMOS NVM cells connected to an addressing scheme according to a specific embodiment of the present invention;

FIG. 10 is simplified partial plan view showing arrays of three-terminal NMOS NVM cells connected to addressing schemes according to another specific embodiment of the present invention;

FIG. 11 is a simplified partial plan view showing an array of three-terminal (two-channel) NMOS NVM cells connected to an addressing scheme according to another specific embodiment of the present invention;

FIG. 12 is a simplified partial plan view showing an array of three-terminal NMOS NVM cells connected to an addressing scheme according to another specific embodiment of the present invention;

FIG. 13 is a simplified partial plan view showing an array of three-terminal NMOS NVM cells connected to an addressing scheme according to another specific embodiment of the present invention;

FIG. 14 is a graph including cell programming data generated for three-terminal NMOS NVM cells showing the benefit of using an injection transistor;

FIG. 15 is a graph showing read disturb data generated for three-terminal NMOS NVM cells of the present invention;

FIG. 16 is a simplified partial plan view showing an NVM cell according to another specific embodiment of the present invention;

FIG. 17 is a simplified circuit diagram showing the NVM cell of FIG. 16 in additional detail; and

FIG. 18 is a simplified partial plan view showing an array including the NVM cells of FIG. 16 connected to an addressing scheme according to another specific embodiment of the present invention.

DETAILED DESCRIPTION OF THE DRAWINGS

The present invention relates to an improvement in embedded CMOS NVM cells. The following description is presented to enable one of ordinary skill in the art to make and use the invention as provided in the context of a particular application and its requirements. The terms “coupled” and “connected”, when used alone herein, are defined as follows. The term “connected” is used to describe a direct connection between two circuit elements or structures, for example, by way of a conducting diffusion or metal line formed in accordance with normal integrated circuit fabrication techniques. In contrast, the term “coupled” is used to describe either a direct connection or an indirect connection between two circuit elements. For example, two coupled elements may be directly connected by way of a metal line, or indirectly connected by way of an intervening circuit element (e.g., a capacitor, resistor, inductor, or by way of the source/drain terminals of a transistor). In contrast to “coupled” (alone), the phrases “capacitive coupling” and “capacitively coupled” indicates the transfer of voltage by means of the capacitance between two nodes or circuit structures. In addition, the term “region” is defined herein to describe a volumetric (three-dimensional) area having substantially identical electrical properties and/or doping concentrations. Various modifications to the preferred embodiment will be apparent to those with skill in the art, and the general principles defined herein may be applied to other embodiments. Therefore, the present invention is not intended to be limited to the particular embodiments shown and described, but is to be accorded the widest scope consistent with the principles and novel features herein disclosed.
FIG. 1 is a simplified circuit diagram showing a CMOS IC 100 including an LV MOSFET 110, an HV MOSFET 130, and a three-terminal (two-channel) NMOS NVM cell 150 that are all fabricated on a monocrystalline silicon “chip” 101 using a standardized CMOS process, along with other circuit components that are omitted for brevity. HV MOSFET 130 is typically utilized to perform voltage conversion and other interface functions, and is therefore designed and fabricated for high external (system) operating voltages (e.g., 3.3V or 5V), and are typically located on exterior peripheral areas 103 of chip 101. In the 5V only designs HV MOSFET 130 may be located in the central area of the chip where LV MOSFET 110 are shown. NVM cell 150 is located in an area 104 that is located either in internal area 102, external area 103 or both, and serves to store control and/or data bits that are needed after power down.
Referring to the top left portion of FIG. 1, as is well known in the art, NMOS LV MOSFET 110 includes an N+ source region S₁₁₀and an N+ drain region D₁₁₀separated by a p-type channel region C₁₁₀. Formed over channel region C₁₁₀is a polysilicon gate structure 120-1 that is separated from an upper surface of substrate 101 by a low-voltage gate oxide layer LV-OX. NMOS LV MOSFET 110 also includes low voltage n-type LDD regions LDD₁₁₀having a first (relatively high) doping concentration that are connected to each of source region S₁₁₀and drain region D₁₁₀and extend into channel region C₁₁₀.
Referring to the top right portion of FIG. 1, as is also well known in the art, NMOS HV MOSFET 130 includes an N+ source region S₁₃₀and an N+ drain region D₁₃₀separated by a p-type channel region C₁₃₀. Formed over channel region C₁₃₀is a polysilicon gate structure 120-3 that is separated from an upper surface of substrate 101 by a high-voltage gate oxide layer HV-OX. Note that high-voltage gate oxide layer HV-OX is typically thicker than low-voltage oxide layer LV-OX of LV MOSFET 110. NMOS HV MOSFET 130 also includes high-voltage n-type LDD regions LDD₁₃₀having a second, relatively low doping concentration that are connected to each of source region S₁₃₀and drain region D₁₃₀and extend into channel region C₁₃₀.
The HV LDD₁₃₀consists of a single implant only, while NLDD₁₁₀of consists of two components—N-type and P-type. The second, P-type part of the LV NLDD implant is relatively strong and deep (to influence stronger P-wells of LV devices compared with HV transistors). A typical 40-50 keV BF2 6-8 10¹³/cm²implant serves as a subsurface anti-punchthrough protecting in LV transistors. In the injection channel of the tree-terminal NVM cell it forms an abrupt p-n region with enhanced electrical field, thus facilitating the hot carrier generation)
Referring to the lower portion of FIG. 1, two-channel NVM cell 150 includes a “read” (first) floating gate NMOS transistor 151 having a read (first) source region S₁₅₁, an injection (second) transistor 155 having an injection (second) source region S₁₅₅, and a shared drain region D₁₅₀that is separated from first source region S₁₅₅by a first p-type channel region C₁₅₁, and separated from injection source region S₁₅₅by a second p-type channel region C₁₅₅. Read source region S₁₅₁, injection source region and shared drain region D₁₅₀are diffused into corresponding regions of substrate 101 using known CMOS processing techniques. A high-voltage oxide layer HV-OX, which is substantially identical to that utilized by NMOS HV MOSFET 130 (discussed above) is formed over channel regions C₁₅₁and C₁₅₅and over drain region under the floating gate extension 123A in FIG. 4. A polysilicon floating gate 120-5 is shared by both read transistor 151 and injection transistor 155, and includes a first portion 121 that is at least partially disposed over read channel region C₁₅₁, a second portion 122 that is at least partially disposed over second channel region C₁₅₅, and a third portion 123 that is formed over drain region D₁₅₀. Third portion 123 of floating gate 120-5 is fabricated using techniques described below such that a gate-drain capacitance C_DSbetween floating gate 120-5 and drain region D₁₅₀is substantially (i.e., at least 4 times) higher than a first gate-source capacitance C_GS-1between floating gate 120-5 and first source region S₁₅₁, and is substantially (i.e., at least four times) higher than a second gate-source capacitance C_GS-2between floating gate 120-5 and second source region S₁₅₅. In particular, as indicated in FIG. 1, floating gate 120-5 includes a first portion 121 disposed over read channel region C₁₅₁(i.e., such that first portion 121 is disposed adjacent to a first section DS1 of drain region D₁₅₀), a second portion 122 disposed over injection channel region C₁₅₅(i.e., such that second portion 122 is disposed adjacent to a second section DS2 of drain region D₁₅₀, and a third portion 123 extending over a third section DS3 of drain region D₁₅₀. As described in the various specific embodiments below, floating gate 120-5 is an integral polysilicon structure that is etched from a single polysilicon layer (e.g., using a single photolithographic mask and etching step) such that, in at least one specific embodiment, first portion 121 is connected to third portion 123 either directly (i.e., third portion 123 is connected to and extends from first portion 121) or connected by an optional intermediate (fourth) portion 125, and such that second portion 122 is connected to third portion 123 either directly or by way of an optional intermediate (fifth) portion 125. As indicated by the dashed line connecting first portion 121 with second portion 122, in another alternative embodiment these portions may also be connected by an optional intervening sixth polysilicon structure 126. As such, floating gate 120-5 is formed such that when floating gate 120-5 is programmed, portions 121, 122 and 123 have the same potential (e.g., 4V), and when floating gate 120-5 is erased, portions 121, 122 and 123 also have the same potential (below ˜0.5V).
Referring to the bottom portion of NVM cell 150, NVM cell 150 also includes first LDD regions LDD₁₅₁that are formed on each side of read channel C₁₅₁(i.e., connected to each of source region S₁₅₁and first section DS1 of drain region D₁₅₀), second LDD regions LDD₁₅₅that are formed on each side of injection channel C₁₅₅(i.e., connected to each of source region S₁₅₅and second section DS2 of drain region D₁₅₀), and in accordance with an aspect of the present invention, an implant region 152 disposed under third section DS3 of drain region D₁₅₀. First LDD regions LDD₁₅₁and second LDD regions LDD₁₅₅comprise the same dopant and concentration as (i.e., are simultaneously formed with) LDD regions LDD₁₃₀of NMOS HV MOSFET 130 and region LDD₁₁₀of NMOS LV MOSFET, correspondingly (discussed above). In contrast, implant region 152 is formed using the dopants and concentrations of both NMOS LV MOSFET 110 and NMOS HV MOSFET 130 (i.e., dopant is diffused into implant region 152 during the formation of both LDD₁₁₀and LDD₁₃₀) in order to facilitate the desired gate-to-drain capacitance C_DS.
According to another aspect of the present invention, injection transistor 155 has an additional hot electron (CHE) enhancing pocket implant 158, for example, using Boron or BF₃, implants that are utilized in the CMOS fabrication flow, however the dose and energy are different compared with those existing in the core CMOS process flow. In contrast, read transistor 151 (i.e., source region S₁₅, and first section DS1 of drain region D₁₅₀) is formed outside pocket implant 158 (i.e., using standard NMOS transistor formation. In one embodiment, pocket implant 158 has a dopant concentration in the range from 5×10¹²to 5×10¹³cm⁻².
Referring again to FIG. 1, in addition to the above-described transistor types, CMOS IC 100 includes a memory address/control circuit 160 and NVM program/erase circuitry 170 that are utilized in the manner described below to access LV MOSFET 110 and HV MOSFET 130, and used for programming floating gate 120-5 of NVM cell 150 by transferring a positive programming potential from drain region D₁₅₀to floating gate 120-5, and for erasing floating gate 120-5 by transferring an approximately zero volt potential from the drain region D₁₅₀to the floating gate 120-5. FIGS. 2(A), 2(B) and 2(C) shown NVM cell 150 during program, erase and read operations, respectively, according to a specific embodiment. Referring to FIG. 2(A), floating gate 120-5 is programmed by applying a programming voltage V_PROGto drain region D₁₅₀and connecting injection source region S₁₅₅to ground (read source S₁₅₁is disconnected (floating)), which generates channel hot electrons that enter floating gate 120-5 from drain D₁₅₀. As indicated in FIG. 2(B), floating gate 120-5 is erased by applying an erase voltage V_ERASEto injection source S₁₅₅(drain region D and read source S₁₅₁are disconnected (floating)), which initiates the injection of band-to-band tunneling (BBT) holes into floating gate 120-5 from injection source region S₁₅₅. As indicated in FIG. 2(C), read transistor 151 is used during read-out operations by applying a read voltage V_readwhich is lower than the program voltage to drain region D₁₅₀and coupling source read region S₁₅₁to ground, while the injection source S₁₅₅is floating, and measuring the resulting current I_readflowing from drain to the “read” source (injection source region S₁₅₅is disconnected). Note that CHE are not expressed in read transistor 151 due to the absence of the CHE enhancing implant discussed above, thus the read-disturb effects in NVM cell 150 are strongly suppressed compared with the one-channel (two-terminals) cell. It is clear to the skilled in the art that programming efficiency can be enhanced by applying bias to the body of the cell, which is equivalent to applying voltage to S₁₅₅(e.g. 1-3V) while keeping the potential difference between S₁₅₅and D₁₅₀constant. Such an operation enhances CHISEL and allows lower programming currents.
CMOS 100 is produced using a standardized CMOS process flow that is depicted in relevant part in FIG. 3. Fabrications steps that are not essential to describe the present invention, such as those utilized to produce p-type MOSFETs, are omitted for brevity. The CMOS process flow includes a process of forming P-well regions in substrate 101 (block 210) that define the boundaries of LV MOSFET 110, HV MOSFET 130 and NMOS cell 150. Referring to the left and central portions of FIG. 3, gate oxide layers are then formed on the upper surface of substrate 101 using separated processing steps. This gate oxide layer production includes forming a relatively thin gate oxide LV-OX (e.g., 30 Angstroms); step 220A) for LV MOSFET 110 and a relatively thick gate oxide HS-OX (e.g., 70 Angstroms; step 220B) for HV MOSFET 130. Next, a single polysilicon layer having a thickness of approximately 2000 Angstroms is formed and patterned using known techniques (block 230). Note that LV gate structure 120-1 of LV MOSFET 110 has a first (relatively narrow) gate line width (e.g., 0.14-0.18 μm (microns)) that is defined by the design rules (DR) of the standard CMOS process flow. In contrast, the HV gate structure 120-3 of HV MOSFET 130 has a second (relatively wide) gate line width (e.g., 0.35 μm) that is selected to facilitate high voltage operation. Following the polysilicon etch process, two separate LDD implants are performed using two masks and implant processes. LDD implants consist of combinations of donor and acceptor (pocket) implants. In particular, an LV-LDD implant (block 240A) is performed using a first mask to provide LDD regions LDD₁₁₀of LV MOSFET 110, and an HV-LDD implant (block 240B) is performed using a second mask to provide LDD regions LDD₁₃₀of HV MOSFET 130. Following the LV LDD and the HV LDD implant processes, sidewall spacers are formed on the polysilicon structures (block 260), and N+ source/drain region diffusions (block 270) are formed in accordance with known techniques. Finally, first pre-metal dielectric, contacts, metallization, including interlayer dielectrics and metal vias, and passivation are performed according to known techniques.
Referring to the right side of FIG. 3, in accordance with an aspect of the present invention, NVM cell 150 (FIG. 1) is entirely fabricated utilizing the existing LV and HV process steps of the standard CMOS process flow that are utilized to form LV MOSFET 110 and HV MOSFET 130. In particular, NVM cell 150 is produced using only the process steps of a standard CMOS process flow (i.e., no extra masks are required), which in turn facilitate the incorporation of NVM cell 150 into CMOS ICs without any additional manufacturing costs. As indicated on the right side of FIG. 3, source regions S₁₅₁and S₁₅₅and drain region D₁₅₀of NVM cell 150 are formed using the same p-well process step (block 210) as that used to form LV MOSFET 110 and HV MOSFET 130. In addition, NVM cell 150 is produced using the same gate oxide HV-OX step (block 220C) that is used to form HV MOSFET 130. The single polysilicon layer used to form gates 120-1 and 120-3 is also patterned to form NVM cell 150 (step 230), with modifications to the floating gate shape being incorporated into the poly mask that are described in detail below. In accordance with an aspect of the present invention, NVM cell 150 is then subjected to two LDD implant steps (blocks 240C1 and 240C2), with the HV LDD implant step utilized for HV MOSFET 130 being used to form LDD regions LDD₁₅₁(i.e., adjacent channel C₁₅₁; see FIG. 1), and to provide a first implant dose to second diffusion region 152 (block 240C1), and the LV-LDD implant step utilized for LV MOSFET 110 being used to form LDD regions LDD₁₅₅(i.e., adjacent channel C₁₅₅; see FIG. 1), and to provide a second implant dose to second diffusion region 152 (see FIG. 1; block 240C2). As set forth below, by forming NVM cell 150 using the thicker HV gate oxide and utilizing both the LV-LDD and HV-LDD implant processes in the manner described above, the present invention facilitates forming an implant region that extends entirely under third portion 123 of floating gate 120-5, whereby an optimal high drain field is provided that facilitates CHE programming. Following the LDD implants, in accordance with another aspect of the present invention, pocket implant 158 is formed in the area of injection transistor 155 by a boron implant process (block 250), (see FIG. 1) utilizing standard CMOS techniques. Floating gate 120-5 is then subject to the same sidewall spacer process (block 260), N+ source/drain diffusion process (block 270), metal via, and metallization processes (block 280) that are utilized in the fabrication of LV MOSFET 110 an HV MOSFET 130.
FIGS. 4 and 5 are simplified plan and cross-sectional views showing an exemplary CMOS IC 100A fabricated on substrate 101A in accordance with a specific embodiment of the present invention. CMOS IC 100A including an LV MOSFET 110A, an HV MOSFET 130A, and a “T” shaped two-channel NMOS NVM cell 150A. Referring to the left side of FIGS. 4 and 5, LV MOSFET 110A includes an N+ source region S_110Aand an N+ drain region D_110Aseparated by a p-type channel region C_110A, a polysilicon gate structure 120-1A formed on LV gate oxide layer LV-OX, and includes n-type LDD regions LDD_110Athat are connected to each of source region S_110Aand drain region D_110Aand extend into channel region C_110A. Referring to the right side of FIGS. 4 and 5, HV MOSFET 130A includes an N+ source region S_130Aand an N+ drain region D_130Aseparated by a p-type channel region C_130A, a polysilicon gate structure 120-3A formed on HV gate oxide layer HV-OX, and includes n-type LDD regions LDD_130A. Referring to the central portions of FIGS. 4 and 5, three terminal NVM cell 150A includes a read NMOS transistor 151 having a source region S_151A, an injection NMOS transistor 155 having a source region S_155A, and an enlarged drain region D_150Aseparated from source regions S_151Aand S_155Aby channel regions C_151Aand C_155A, respectively, and a shared polysilicon floating gate 120-5A that is formed on a gate oxide HV-OX. Referring to FIG. 5, NVM cell 150 also includes LDD regions LDD_151Aand LDD_155A(formed using the HV-LDD implant and LV-LDD implants correspondingly or using special pocket implants in the second case (LDD_155A)) that are connected to each of source regions S_151Aand S_155Aand drain region D_150A, respectively. In FIG. 4, the square boxes with internal “X” shapes designated contact structures used to provide electrical connection to the source/drain and gate structures of CMOS IC 100A.
In accordance with an aspect of the present embodiment, as shown in FIG. 4, floating gate 120-5A comprises a substantially T-shaped polysilicon structure including a first portion 121A disposed over channel region C_151A, a second portion 122A disposed over channel region C_155A, and a third portion 123A extending parallel to first portion 121A and extending over a central area of enlarged drain region D_150A, a fourth portion 124A that is connected between respective ends of first portion 121A and third portion 123A, and a fifth portion 125A that is connected between respective ends of second portion 122A and third portion 123A. Note that fourth portion 124A and fifth portion 125A are disposed over areas located outside of the boundaries of channel regions C_151Aand C_155Aand drain region D_150A. First portion 121A has a gate line width defined by the high voltage design rule (HV-DR) of the standardized CMOS process flow (i.e., first portion 121A has the same width as gate 120-3A of HV MOSFET 130A). Second portion 122A has a width INJ-DR (between LV-DR and HV-DR) in order to enhance the CHE programming but not to cause channel leakage (e.g., in the range from 0.2 um to 0.35 um). Third portion 123A has a width defined by the minimum design rule (MIN-DR) width of the standardized CMOS process flow. In one specific embodiment, using a 0.18 micron CMOS process flow, the width of third portion 123A is 0.14 microns. The relatively large width of first portion 121A is required to support the read-out potentials that are generated between source region S_151Aand drain region D_150A(across channel region C_151A) without read disturb. In contrast, third portion 123A is not subject to the operating constraints and potentials of first portion 121A, and therefore is made narrower than first portion 121A in order to save space and increase capacitance to the floating gate (due to merging of NLDD implant under its narrow polycrystalline silicon (polysilicon) structure).
In accordance with another aspect of the present embodiment, NVM cell 150 includes both LV-LDD and HV-LDD implants during the respective LV-LDD and HV-LDD implant process steps. That is, as indicated in FIG. 4, the mask used during the HV-LDD process defines a first opening 310 ₁₃₀that is positioned over the source/drain/channel regions of HV MOSFET 130, and a second opening 310 ₁₅₀that is positioned over source/drain/channel regions of read transistor 151A, and is also utilized to diffuse material into implant region 152A. Similarly, the mask use during the LV-LDD process defines a first opening 320 ₁₁₀that is positioned over the source/drain/channel regions of LV MOSFET 110, and is also utilized to diffuse material into implant region 152A. As shown in FIG. 4, during the HV-LDD implant process, the mask openings respectively produce HV-LDD implant regions HV-LDD_130Aunder gate 120-3A, HV-LDD_151Aunder gate 121A, and an HV LDD implant under third portion 123A (i.e., in region 152A). In addition, during the LV-LDD implant process, the mask openings produce LV-LDD implant regions LV-LDD_110Aunder gate 120-1A, a LV-LDD implant in region 152A under second portion 123A, and implant region LDD_155Aunder gate portion and 122A of floating gate 120-5A. As such, implanted region 152A has the sum of both HV-LDD and LV-LDD doping concentrations of LDD regions LDD_110Aand LDD_130A, shown in FIG. 5. The present inventors have found that superior three-terminal (two-channel) NVM cells are produced by forming floating gate 120-5 on the relatively thick HV gate oxide such that second portion 123A has the minimal width set by the LV-DR, using HV-LDD implants in the source/drain/channel region below first portion 121A, and generating implant region 152A under third portion 123A using both the LV-LDD and HV-LDD implants. This combination of processes are believed to form implant region 152A such that it merges under (i.e., extends entirely under) third portion 123A, as depicted in FIG. 4, whereby implant region 152A extends between opposing edges of N+ drain region D_150Athat are located below opposing side edges of third portion 123A.
In accordance with another aspect of the present embodiment, NVM cell 150A is programmed using channel hot electron (CHE) injection and erased using band-to-band tunneling (BBT) holes. During programming, a positive programming voltage (e.g., 5V-6V is applied to drain region D_150Aand injection source region S_155Ais connected to ground (0V), and read source region S_151Ais floating, thereby causing CHE injection from drain region D₁₅₀into the floating gate 120-5A. In one embodiment, the programming process is enhanced by applying a body bias voltage (e.g., from −1 to −5V to the P-well in which injection transistor 155 is formed). In one specific embodiment, the body bias from −3V to −5V voltage is applied to the isolated P-well containing channel region C_155Awith a delay period of 1 to 100 μS (100 microseconds) after asserting the programming voltages 4.5V to drain region D_150Aand coupling source region S_155Ato ground. Erase operations are performed by keeping the drain region D_150Aand read source region S_151Afloating and applying the positive erase voltage to source region S_155A, thereby causing BBT generation of holes in the region S_155A, their acceleration in the field of source junction and tunneling into the floating gate 120-5.
FIG. 6 is simplified top plan view showing an three-terminal “C” or “horseshoe” shaped NMOS NVM cell 150B according to another specific embodiment of the present invention. NVM cell 150B is formed as part of a CMOS IC, similar to that described above, which includes both LV MOSFETs and HV MOSFETs, which are omitted from the discussion below for brevity. Similar to NVM cell 150A (discussed above), NVM cell 150B includes a read NMOS transistor 151B having a source region S_151B, an injection NMOS transistor 155B having a source region S_155B, and an enlarged drain region D_150Bseparated by a channel region C_151Bfrom source region S_151Band by a channel region C_155Bfrom source region S_155B, and a polysilicon floating gate 120-5B. Floating gate 120-5B is similar to floating gate 120-5A (discussed above) in that it is formed on an Hv oxide (not shown), and includes a first portion 121B disposed over channel region C_151Band adjacent to a first section DS1B of drain region D_150B, a second portion 122B disposed over channel region C_155Band adjacent to a second section DS2B of drain region D_150B, and third portions 123B disposed over an enlarged third section DS3B of drain region D_150B. Note that floating gate 120-5B differs from floating gate 120-5A (discussed above) in that two portions 123B extend from first portion 121B and second portion 122B over drain region D_150B, respectively, and first and second portions 121B and 122B are directly connected by a (sixth) portion 126B. NVM cell 150B is subjected to the LDD implant processes in the manner described above to facilitate the formation of drain implant 152B and CHE enhancing pocket implant 158B.
FIG. 7 is a simplified top plan view showing a three-terminal (two-channel) NMOS NVM cell 150C according to another specific embodiment of the present invention. Similar to NVM cell 150B (discussed above), NVM cell 150C includes read NMOS transistor 151C having a source region S_151C, an injection transistor 155C having a source region S_155C, and an enlarged drain region D_150Cseparated by a channel regions C_151Cand C_155C, and a “C” shaped polysilicon floating gate 120-5C formed on an HV oxide (not shown). However in this embodiment, read transistor source region S_151Cand injection transistor source region S_155Care disposed on opposite sides of drain region D_150Csuch that channel regions C_151Cand C_155Care separated by drain region D_150Cand such that an intervening (sixth) portion 126C of floating gate 120-5C extends over a portion of drain region D_150Cbetween first portion 121C and second portion 122C.
NVM cells 150B and 150C (shown in FIGS. 6 and 7) are used, for example, when there is no confidence in implant merging under the coupling extension polysilicon. This is typical in special processes used, e.g., for fabrication of Power Management IC. Corresponding products are often “5 V only” to save the masks intended for 1.8v logics in the core CMOS process flow. In this case, only ˜100 A GOX is used (both in the NVM area and in the logic part of the microcircuit).
FIG. 8 is simplified top plan view showing a three-terminal (two-channel) NMOS NVM cell 150D according to yet another specific embodiment of the present invention. NVM cell 150D is similar to NVM cell 150C (discussed above) in that NVM cell 150D includes a read NMOS transistor 151D having a source region S_151D, and an injection transistor 155D having a source S_155Dthat are disposed on opposite sides of an enlarged drain region D_150D, and a floating gate 120-5D that includes first and second portions 121D and 122D that are connected by an extension portion 126D. However, NVM cell 150D differs from NVM cell 150C in that (third) portion 123 extends entirely across drain region D_150D, such that a substantially “O” (ring-like or circular) shape is formed by polysilicon gate structure 120-5D. The ring-like structure of polysilicon gate 120-5D make NVM cell 150D the most compact implementation of read-disturb-free NVM cell based on the proposed principle.
FIGS. 9(A) and 9(B) are simplified circuit diagrams showing various arrays of three-terminal NMOS NVM cells, which are constructed in the manner described above with reference to the specific embodiments of FIGS. 3-8, as arranged and connected to associated addressing schemes. These addressing schemes facilitate cell operations while minimizing occupied chip space, and are optimized in the manner described below to facilitate programming and erasing operations.
FIG. 9(A) is a simplified circuit diagram showing a portion of a CMOS IC 100E including three-terminal NMOS NVM cells 150-x,y arranged in an array including three rows and two columns, where “x,y” represents the row number and column number, respectively. The array includes word lines W1 and W2 and bit lines B11, B12, B21, B22, B31 and B32 that are connected to associated NVM cells 150-x,y, and are also connected to addressing/control circuitry (e.g., circuits 160 and 170; see FIG. 1) in order to perform program/erase and read operations on the NVM cells. NVM cells 150-x,y are implemented using any of the specific three-terminal NVM cells described herein. Word lines W1 and W2 are disposed parallel to the columns of three-terminal NVM cells, and each word line is connected to the drain region of each NVM cell of an associated column. For example, word line W1 is connected to drain region of NVM cell 150-1,1, and also to the drain regions of NVM cell 150-2,1 and 150-3,1. Conversely, bit lines B11 to B32 are disposed parallel to the rows of NVM cells, and each bit line is connected to a read source region or injection source region of each NVM cell of an associated row. For example, bit line B11 is connected to the read source regions of NVM cells 150-1,1 and 150-1,2, and bit line B12 is connected to the injection source regions of NVM cells 150-1,1 and 150-1,2. With this arrangement, the program/erase/read circuitry (not shown) addresses a selected NVM cell (e.g., NVM cell 150-1,1) during program/erase operations by selectively applying predetermined programming voltages to the word line (e.g., W1) and the bit line (e.g., B12) that is associated with the selected NVM cell, and during read operations by applying a predetermined read voltage to the word line (e.g., W1), by applying 0 voltage to the source (e.g., B11) that is associated with the selected NVM cell, and measuring the drain-source current. Note that the remaining word lines and bit lines are disconnected (floating) during the program/erase and read operations directed to the selected NVM cell.
FIG. 9(B) is a simplified circuit diagram showing a portion of a CMOS IC 100F according to an alternative embodiment, where word lines W41, W42 and W43 are disposed parallel to the rows of three-terminal NVM cells, and each word line is connected to the drain region of each NVM cell of an associated row. For example, word line W41 is connected to drain region of NVM cell 150-1,1, and also to the drain region of NVM cell 150-1,2. Bit lines B41 to B44 are disposed parallel to the columns of NVM cells, and each bit line is connected to a read source region or injection source region of each NVM cell of an associated column.
FIG. 10 is a partial simplified plan view showing an IC circuits 100G including NVM cells 150A-1,1 to 150A-4,2 arranged in four rows and two columns. NVM cell 150A-1,1 is consistent with NVM cell 150A, described above with reference to FIGS. 4 and 5. Consistent with the arrangement described above with reference to FIG. 9(A), each column of NVM cells is coupled to a common word line (e.g., NVM cells 150A-1,1, 150A-2,1, 150A-3,1 and 150A-4,1 are connected to word line W51, and NVM cells 150A-1,2, 150A-2,2, 150A-3,2 and 150A-4,2 are connected to word line W52), and each NVM cells in each row of the array of CMOS circuit 100G share two adjacent bit lines (e.g., the read transistor source regions of cells 150A-1,3 and 150A-1,2 are connected to bit line B51, and the injection transistor source regions of cells 150A-1,1 and 150A-1,2 are connected to bit line B52). Bit lines B53 and B54 are similarly connected to the sources of the second row, bit lines B55 and B56 are similarly connected to the sources of the third row, and bit lines B57 and B58 are similarly connected to the sources of the fourth row. To minimize space, according to the embodiment shown in FIG. 10, drain regions of each adjacent pair of NVM cells in each column are combined (connected together) and connected to the associated word line by a common connection point. For example, cells 150A-1,1 and 150A-2,1 are inverted and share a common drain D_150G-1that is connected to word line W51. For similar reasons, at least two cells in each column share a common drain diffusion and CHE enhancing pocket implant, which are formed in accordance with the description provided above. For example, NVM cells 150A-1,1 and 150A-2,1 share elongated drain diffusion 152G and pocket implant 158G, which also extend under cells 150A-3,1 and 150A-4,1.
FIG. 11 shows a portion of a CMOS IC 100H including an array of NVM cells arranged according to another embodiment, wherein each NVM cell is consistent with the NVM cell 150B described above with reference to FIG. 6, and the drain regions of NVM cells in each row are connected together and to a shared word line, and source regions of NVM cells in each column are connected to shared bit lines in the manner described above with reference to FIG. 9(B). For example, the drain regions of NVM cells 150B-1,1 and 150B-1,2 are connected to word line W61, where the source regions of NVM cell 150B-1,1 are respectively connected to bit lines B61 and B62 and the source regions of NVM cell 150B-1,2 is connected to bit lines B63 and B64. Similar to the arrangement described above with reference to FIG. 10, each adjacent pair of cells in each column share a common pocket implant region, but drain and drain implant regions are shared along the rows in the present embodiment. For example, adjacent column cells 150B-1,1 and 150B-2,1 share a common pocket implant 158H, but a drain D_150H-1and a drain implant 152H are shared by adjacent row cells 150B-1,2 and 150B-2,2. Similar diffusions are shared by cells 150B-2,1, 150B-2,2, 150C-3,1, 150C-3,2, 150C-4,1 and 150C-4,2 of the remaining rows and columns.
FIG. 12 shows a portion of a CMOS IC 100J including an array of NVM cells arranged according to another embodiment, wherein each NVM cell is consistent with the NVM cell 150C described above with reference to FIG. 7, and the drain regions of NVM cells in each row are connected together and to a shared word line, and source regions of NVM cells in each column are connected to shared bit lines in the manner described above with reference to FIG. 9(B). For example, the drain regions of NVM cells 150C-1,1 and 150C-1,2 are connected to word line W71, where the source regions of NVM cell 150C-1,1 are respectively connected to bit lines B71 and B72 and the source regions of NVM cell 150C-1,2 is connected to bit lines B73 and B74. All cells in each row share a common pocket implant region, a drain and a drain implant region. For example, adjacent row cells 150C-1,1 and 150C-1,2 share a common pocket implant 158J, a shared drain D_150J-1, and a shared drain implant 152J. Similar diffusions are shared by cells 150C-2,1 and 150C-2,2 of the second row, and cells 150C-3,1 and 150C-3,2 of the second row.
In yet another embodiment shown in FIG. 13, a CMOS IC 100K includes an array of NVM cells wherein each NVM cell is consistent with the NVM cell 150D described above with reference to FIG. 8, and the drain regions of NVM cells in each row are connected together and to a shared word line, and source regions of NVM cells in each column are connected to shared bit lines in the manner described above with reference to FIG. 9(B). For example, the drain regions of NVM cells 150D-1,1 and 150D-1,2 are connected to word line W81, where the source regions of NVM cell 150D-1,1 are respectively connected to bit lines B81 and B82 and the source regions of NVM cell 150D-1,2 is connected to bit lines B83 and B84. All cells in each row share a common pocket implant region, a drain and a drain implant region. For example, adjacent row cells 150D-1,1 and 150D-1,2 share a common pocket implant 158K, a shared drain D_150K-1, and a shared drain implant 152K. Similar diffusions are shared by cells 150D-2,1 and 150D-2,2 of the second row, and cells 150D-3,1 and 150D-3,2 of the second row.
FIGS. 14 and 15 include graphs showing experimentally generated data on prototype two-channel NVM cells (on silicon) at the single cell and miniArray level.
FIG. 14 is a graph showing cell programming test data generated for three-terminal NMOS NVM cells, and indicates that when programming is performed by way of the injection source regions of the cells, programming time is reduced and accumulated charge is increased in comparison to programming by way of the read transistor. It is seen that with 4.5V at drain, only several milliseconds are necessary to obtain a ˜2V memory window by programming through the W=0.35 um (min DR) injection transistor. Several seconds are needed to program the NVM cell to the same level through the reading transistor with the same Poly CD.
FIG. 15 is a graph showing “Time-to-failure” Takeda plot (lifetime vs. reciprocal drain voltage) data for Two-Terminal and Three-Terminal cells. The failure is defined as a ˜400 mV FG potential increase. The graph demonstrates the extrapolated time necessary for the 400 mV FG potential increase at different drain voltages in the read-out operation. For two-terminal cells, a 10 day continuous read at a typical 2V voltage results in failure. In the proposed Three-Terminal NVM cell the read disturb is not an issue (3000 years of continuous operation till the read disturb results in failure).
Although the present invention has been described with respect to certain specific embodiments, it will be clear to those skilled in the art that the inventive features of the present invention are applicable to other embodiments as well, all of which are intended to fall within the scope of the present invention. For example, in alternative embodiments a number of different programming and erasing methods may be used, e.g., erasing with ultraviolet (UV) light, programming using an avalanche breakdown in the drain region, etc. In addition, a pocket implant or another special drain engineering scheme may be performed in the transistor source/drain regions to enhance lateral electrical fields and thus increase the efficiency of programming and erase, while it is absent in the extension area. An additional Poly topology over the drain connected Active area may be used to increase coupling to the floating gate.
Moreover, although the present invention is described with reference to NMOS NVM cells, those skilled in the art will recognize that the concepts described above may be utilized to produce NVM cells having similar characteristics but employing PMOS transistors placed in N-wells, e.g., a PMOS NVM injection transistor programmed using hot hole induced electron injection, NMOS readout transistor and array organization without select transistors. An NVM cell 150L utilizing a PMOS transistor for programming and an NMOS transistor for readout is shown in FIGS. 16 and 17, and a corresponding array incorporating NVM cell 150L is shown in FIG. 18.
Referring to FIGS. 16 and 17, memory cell 150L consists of injection channel PMOS transistor 155L and a readout NMOS transistor 151L that share a “Y” shaped floating gate 120-5L. Readout NMOS transistor 151L includes a source region S_151Land a drain region D_151Lthat are separated by a first portion 121L of floating gate 120-5L. Injection transistor 155L is formed in an N-Well 158L and includes a source region S_155Land a drain region D_155Lthat are separated by a second portion 122L, where drain region D_155Lincludes two sections D_155L-1and D_155L-2that are bisected by a narrow third portion 123L of gate 120-5L, which acts as an extension for increasing the drain coupling. A channel C_155Lof injection transistor 155L is implanted with LV PLDD implant. The area of extension (i.e., the region including sections D_155L-1and D_155L-2) is implanted with both HV PLDD and LV PLDD to stimulate implant merging under the undersized Poly of the extension with the width of 0.14 um. Channel region C_151Lof readout transistor 151L is implanted with LV NLDD implant.
Programming of cell 150L is performed by channel hot holes induced electron injection (CHHEI). For this purpose, voltage of 5-6V is applied to N-well 158L, and drain region D_155Lof injection transistor 155L is connected to ground. Source region S_155Lis at the same potential as that of N-Well 158L, and the source and drain of readout transistor 151L are coupled to ground. Due to capacitive coupling, a voltage of −1V to −2V with respect to N-Well 158L is transferred to floating gate 120-5L. Injection transistor 155L is thus slightly open (Vt of PMOS is of the order of −0.8V), and there is a flow of holes in channel C_155L. The holes create electron-hole pairs in drain region D_155Land secondary electrons are injected into floating gate 120-5L (the vertical field in the drain region facilitates injection). To erase the cell, zero voltage is applied to N-well 158L, while negative voltage (from 0 to −6V) is applied to P+ diffusions of the injection transistor (i.e., drain region D_155Land source region S_155L). Due to strong capacitive coupling, negative potential is transferred to floating gate 120-5L. This enhances the BBT of holes and their injection into floating gate 120-5L in drain region D_155Lof readout transistor 151L. The advantage of the described embodiment is low program currents. This is similar to CHISEL case described for other embodiments, but the currents are lower (in the range 1 to 10 uA/cell for the same programming times). Vertical field assists electron injection thus enhancing its efficiency.
In the read-out operation all terminals of injection transistor 151L are maintained at 2-2.5 V, while drain region D_151Lof read-out transistor 151L is maintained at ˜1V (source S_151Lis coupled to ground). This ensures absence of the read disturb (low drain voltages in read and longer channel C_151L, e.g. L=0.5-1 um, thus no channel hot electron effects are pronounced though drain field is enhanced to ensure efficient erase (in addition to low Vd in read)). The price for lower programming currents and absence of read disturb is up to 25% increase of the cell area compared with previous embodiments (due to “N-Well to S/D P+ spacing” DR). The embodiment allows similar to the previous cases cross-point array organization (shown in FIG. 18). The drains of the memory cells are connected in two groups. A memory cell 1501 for programming is selected by choosing an “Injection Source” column line SC1, SC2 or SC3 and an “Injection drain” row line SR1 or SR2. The same principle is used in the read-out and erase. The adjacent sources are bound and connected to ground during read and left floating during erase. Drains are selected out of the two groups.

Claims

1. A complimentary metal-oxide-silicon (CMOS) integrated circuit (IC) comprising:

a substrate; and

a multi-terminal non-volatile memory (NVM) cell including:

a first source region and a second source region, each including a first dopant diffused into the substrate such that the first source region is electrically isolated from the second source region;

a drain region including a second dopant diffused into the substrate and disposed such that the drain region is separated from the first source region by a first channel region and separated from the second source region by a second channel region; and

a polycrystalline silicon (polysilicon) floating gate including a first portion that is at least partially disposed over the first channel region, a second portion that is at least partially disposed over the second channel region, and a third portion that is at least partially disposed over the drain region;

wherein the floating gate, the first and second source regions and the drain region are formed such that a gate-drain capacitance between said floating gate and said drain region is substantially higher than both a first gate-source capacitance between said floating gate and said first source region, and a second gate-source capacitance between said floating gate and said second source region.

2. The CMOS IC of claim 1,

wherein the second source region, the second channel region, and a section of the drain region are disposed in a pocket implant disposed in said substrate and comprising means for enhancing channel hot electrons, and

wherein the first source region, the first channel region, and at least one section of the drain region are disposed outside of said pocket implant.

3. The CMOS IC of claim 2, wherein said pocket implant comprises at least one of boron and boron-triflouride (BF₃).

4. The CMOS IC of claim 1, wherein the floating gate comprises:

an elongated first portion disposed over the first channel region such that the first portion is disposed adjacent to a first section of the drain region,

an elongated second portion disposed over the second channel region such that the second portion is disposed adjacent to a second section of the drain region, and

at least one elongated third portion extending over a third section of the drain region, the third section being spaced from the first section and the second section.

5. The CMOS IC of claim 4, wherein the floating gate further comprises a fourth portion connected between the elongated first portion and the elongated third portion and a fifth portion connected between the elongated second portion and the elongated third portion such that the first, second, third, fourth and fifth portions comprise one of a substantially T-shaped polysilicon structure.

6. The CMOS IC of claim 4, wherein the floating gate further comprises a sixth portion connected between the elongated first portion and the elongated second portion, and the at least one elongated third portion extends from at least one of said first and second portions such that the first, second, third, and sixth portions comprise one of a substantially C-shaped polysilicon structure and a substantially O-shaped polysilicon structure and a substantially U-shaped polysilicon structure.

7. The CMOS IC of claim 1, further comprising:

a low voltage (LV) MOSFET including a first polysilicon gate having predefined first width that is formed on a first gate oxide having a first oxide thickness, and first lightly doped drain (LDD) regions having a first doping concentration; and

a high voltage (HV) MOSFET including a second polysilicon gate having predefined second width that is formed on a second gate oxide, the second gate oxide having a second oxide thickness that is greater than the first oxide thickness, the HV MOSFET also including second LDD regions having a second doping concentration,

wherein said floating gate of said multi-terminal NVM cell is formed on a third gate oxide that has the second oxide thickness, and

wherein said multi-terminal NVM cell further comprises an implanted region disposed below the third section of the floating gate, said implanted region having a third doping concentration including a sum of said first doping concentration and said second doping concentration.

8. The CMOS IC of claim 7, wherein said implanted region extends entirely under said third section of said floating gate such that said implanted region extends between opposing edges of said drain region located below opposing side edges of said third portion.

9. The CMOS IC of claim 1, further comprising means for generating channel hot electrons (CHE) in the drain region, and means for generating band-to-band tunneling (BBT) holes in the second source region.

10. The CMOS IC of claim 9, wherein said means for programming the floating gate further comprises means for generating a first bias voltage in said channel region such that secondary electrons are generated in the channel region and injected into the floating gate.

11. The CMOS IC of claim 10, wherein said means for programming further comprises, following a delay period of 1 to 100 microseconds after the first bias voltage is applied to the second channel region, applying a first programming voltage to said drain region and a second programming voltage to said second source region.

12. The CMOS IC of claim 1 further comprising:

an array including a plurality of said multi-terminal NVM cells arranged in rows and columns, a word line disposed parallel to an associated column of said multi-terminal NVM cells, a first bit line disposed parallel to an associated row of said multi-terminal NVM cells, and a second bit line disposed parallel to said associated row of said multi-terminal NVM cells, wherein said word line is connected to the drain region each NVM cell of said associated column of NVM cells, said first bit line is connected to the first source region of each said NVM cell of said associated row of NVM cells, and said second bit line is connected to the second source region of each said NVM cell of said associated row of NVM cells, and

means for programming and erasing a selected NVM cell by selectively applying predetermined programming voltages to said word line and said first and second bit lines.

13. The CMOS IC of claim 12, wherein the drain regions of each adjacent pair of NVM cells in said associated column are connected together and to said associated word line.

14. The CMOS IC of claim 13, further comprising an elongated pocket implant region disposed such that the second source regions and second channel regions of each adjacent pair of NVM cells in said each associated column are disposed in said elongated pocket implant region.

15. The CMOS IC of claim 13, further comprising an elongated implanted region connecting the drain regions of said each adjacent pair of NVM cells in said each associated column.

16. The CMOS IC of claim 1 further comprising:

an array including a plurality of said multi-terminal NVM cells arranged in rows and columns, a word line disposed parallel to an associated row of said multi-terminal NvM cells, a first bit line disposed parallel to an associated column of said multi-terminal NVM cells, and a second bit line disposed parallel to said associated column of said multi-terminal NVM cells, wherein said word line is connected to the drain region of each NVM cell of said associated row of NVM cells, said first bit line is connected to the first source region of each said NVM cell of said associated column of NVM cells, and said second bit line is connected to the second source region of each said NVM cell of said associated column of NVM cells, and

means for performing one of programming, erasing and reading a selected NVM cell by selectively applying an associated one of a predetermined programming voltage, a predetermined erasing voltage and a predetermined reading voltage to said word line and one of said first and second bit lines.

17. The CMOS IC of claim 16, wherein the drain regions of all NVM cells in said associated row are connected together and to said associated word line.

18. The CMOS IC of claim 17, further comprising an elongated pocket implant region disposed such that the second source regions and second channel regions of each adjacent pair of NvM cells in said each associated row are disposed in said elongated pocket implant region.

19. The CMOS IC of claim 17, further comprising an elongated implanted region connecting the drain regions of said each adjacent pair of NVM cells in said each associated column.

20. A complimentary metal-oxide-silicon (CMOS) integrated circuit (IC) comprising:

a substrate; and

a multi-terminal non-volatile memory (NVM) cell including:

a first transistor including a first source region and a first drain section separated by a first channel region;

a second transistor including a second source region and a second drain section separated by a second channel region; and

a polycrystalline silicon (polysilicon) floating gate including a first portion that is at least partially disposed over the first channel region, a second portion that is at least partially disposed over the second channel region, and a third portion that is at least partially disposed over a third drain section,

wherein the first, second and third drain sections comprise a contiguous drain region including a first dopant diffused into the substrate, and