阅读说明：本技术 基于双晶体管鳍式场效晶体管的***栅极非易失性浮动栅极闪存存储器及其制造方法 (Split gate non-volatile floating gate flash memory based on two transistor fin field effect transistors and method of manufacturing the same ) 是由 S·乔尔巴 C·德科贝尔特 Z·冯 J·金 X·刘 N·多 于 2019-01-23 设计创作，主要内容包括：本发明公开了一种非易失性存储器单元,该非易失性存储器单元形成在半导体衬底上,该半导体衬底具有上表面,该上表面具有向上延伸的鳍片,该向上延伸的鳍片具有相对的第一侧表面和第二侧表面。第一电极和第二电极与该鳍片的第一部分和第二部分电气接触。该鳍片的沟道区包括在该鳍片的该第一部分和该第二部分之间延伸的该第一侧表面和该第二侧表面的部分。浮动栅极沿着该沟道区的该第一部分的该第一侧表面延伸,其中该浮动栅极中没有一个部分沿着该第二侧表面延伸。字线栅极沿着该沟道区的第二部分的该第一侧表面和该第二侧表面延伸。控制栅极设置在该浮动栅极上方。擦除栅极具有第一部分和第二部分,该第一部分与该浮动栅极横向相邻设置,该第二部分竖直地设置在该浮动栅极上方。(A non-volatile memory cell is formed on a semiconductor substrate having an upper surface with an upwardly extending fin having opposing first and second side surfaces. First and second electrodes are in electrical contact with the first and second portions of the fin. The channel region of the fin includes portions of the first side surface and the second side surface extending between the first portion and the second portion of the fin. Floating gates extend along the first side surface of the first portion of the channel region, wherein none of the floating gates extend along the second side surface. Word line gates extend along the first and second side surfaces of the second portion of the channel region. A control gate is disposed over the floating gate. The erase gate has a first portion disposed laterally adjacent to the floating gate and a second portion disposed vertically above the floating gate.)

1. A non-volatile memory cell, comprising:

a semiconductor substrate having an upper surface with upwardly extending fins including first and second side surfaces opposite one another;

a first electrode in electrical contact with a first portion of the fin;

a second electrode in electrical contact with a second portion of the fin, wherein the first and second portions of the fin are spaced apart from each other such that a channel region of the fin includes portions of the first and second side surfaces and extends between the first and second portions of the fin;

a floating gate extending along a first portion of the channel region, wherein the floating gate extends along and is insulated from the first side surface, and wherein no portion of the floating gate extends along the second side surface;

a word line gate extending along a second portion of the channel region, wherein the word line gate extends along and is insulated from the first and second side surfaces;

A control gate disposed over and insulated from the floating gate;

an erase gate having a first portion disposed laterally adjacent to and insulated from the floating gate and a second portion disposed vertically above and insulated from the floating gate.

2. The non-volatile memory cell of claim 1, wherein no conductive gate is disposed along and insulated from a portion of the second side surface opposite a portion of the first side surface along which the floating gate extends.

3. The non-volatile memory cell of claim 1, wherein the word line gate comprises a metallic material, and wherein the word line gate is insulated from the first and second side surfaces by a high-K insulating material.

4. The non-volatile memory cell of claim 3, wherein the floating gate, the control gate, and the erase gate each comprise a polysilicon material.

5. The non-volatile memory cell of claim 4, wherein the first electrode and the second electrode each comprise a metallic material.

6. The non-volatile memory cell of claim 1, wherein the upper surface of the substrate comprises an upwardly extending second fin comprising a third side surface and a fourth side surface opposite one another, and wherein at least a portion of the floating gate is disposed between the fin and the second fin.

7. The non-volatile memory cell of claim 1, wherein the control gate and the erase gate are each disposed vertically above the fin.

8. The non-volatile memory cell of claim 1, wherein the first portion and the second portion of the fin each have a width that is greater than a width of the channel region of the fin.

9. The non-volatile memory cell of claim 1, wherein the first portion and the second portion of the fin each have a height that is greater than a height of the channel region of the fin.

10. The non-volatile memory cell of claim 1, wherein the first electrode extends along the first and second side surfaces of the first portion of the fin, and wherein the second electrode extends along the first and second side surfaces of the second portion of the fin.

11. The non-volatile memory cell of claim 1, wherein the floating gate has a rectangular vertical cross-section.

12. The non-volatile memory cell of claim 1, wherein the floating gate has a U-shaped vertical cross-section.

13. The non-volatile memory cell of claim 12, wherein the control gate comprises a lower portion that extends into the U-shaped vertical cross-section of the floating gate.

14. A method of forming a non-volatile memory cell, the method comprising:

forming a trench into an upper surface of a semiconductor substrate such that the upper surface includes an upwardly extending fin including first and second side surfaces opposite one another;

forming a first electrode in electrical contact with a first portion of the fin;

forming a second electrode in electrical contact with a second portion of the fin, wherein the first and second portions of the fin are spaced apart from each other such that a channel region of the fin includes portions of the first and second side surfaces and extends between the first and second portions of the fin;

Forming a floating gate extending along a first portion of the channel region, wherein the floating gate extends along and is insulated from the first side surface, and wherein no portion of the floating gate extends along the second side surface;

forming a word line gate extending along a second portion of the channel region, wherein the word line gate extends along and is insulated from the first and second side surfaces;

forming a control gate disposed over and insulated from the floating gate;

an erase gate is formed having a first portion disposed laterally adjacent to and insulated from the floating gate and a second portion disposed vertically above and insulated from the floating gate.

15. The method of claim 14, wherein no conductive gate is disposed along and insulated from a portion of the second side surface opposite a portion of the first side surface along which the floating gate extends.

16. The method of claim 14, wherein the word line gate comprises a metallic material, and wherein the word line gate is insulated from the first and second side surfaces by a high-K insulating material.

17. The method of claim 14, wherein the floating gate, the control gate, and the erase gate each comprise a polysilicon material.

18. The method of claim 17, wherein the first electrode and the second electrode each comprise a metallic material.

19. The method of claim 14, wherein the upper surface of the substrate comprises upwardly extending second fins comprising third and fourth side surfaces opposite one another, and wherein at least a portion of the floating gate is disposed between the fins and the second fins.

20. The method of claim 14, wherein the control gate and the erase gate are each disposed vertically over the fin.

21. The method of claim 14, wherein the first portion and the second portion of the fin each have a width and a height that are greater than a width and a height, respectively, of the channel region of the fin.

22. The method of claim 14, wherein the first electrode extends along the first and second side surfaces of the first portion of the fin, and wherein the second electrode extends along the first and second side surfaces of the second portion of the fin.

23. The method of claim 14, wherein the floating gates have a rectangular vertical cross-section.

24. The method of claim 14, wherein the floating gate has a U-shaped vertical cross-section.

25. The method of claim 24, wherein the control gate includes a lower portion that extends into the U-shaped vertical cross section of the floating gate.

Technical Field

The invention relates to a non-volatile flash memory cell array.

Background

Non-volatile memory devices are well known in the art. For example, a split gate memory cell is disclosed in U.S. patent 5,029,130 (which is incorporated herein by reference for all purposes). The memory cell has a floating gate and a control gate disposed over and controlling the conductivity of a channel region of a substrate extending between source and drain regions. Various combinations of voltages are applied to the control gate, source, and drain to program the memory cell (by injecting electrons into the floating gate), erase the memory cell (by removing electrons from the floating gate), and read the memory cell (by measuring or detecting the conductivity of the channel region under the floating gate to determine the programmed state of the floating gate).

The configuration and number of gates in the nonvolatile memory cells can vary. For example, U.S. patent 7,315,056 (which is incorporated herein by reference for all purposes) discloses a memory cell that additionally includes a program/erase gate over a source region. U.S. Pat. No. 7,868,375, which is incorporated herein by reference for all purposes, discloses a memory cell additionally including an erase gate over a source region and a coupling gate over a floating gate. See also U.S. Pat. nos. 6,747,310, 7,868,375, 9,276,005, and 9,276,006 (which are also incorporated herein by reference for all purposes).

Finfet-type structures have been proposed because the problem of shrinking the lithographic dimensions and thus the channel width affects all semiconductor devices. In a finfet-type structure, a fin-shaped member of semiconductor substrate material connects a source region to a drain region. The fin member has a top surface and two opposing side surfaces. Then, current from the source region to the drain region may flow along the top surface and both side surfaces. Thus, by "folding" the channel region into two side surfaces, the surface width of the channel region can be increased, thus increasing current without sacrificing more semiconductor real estate, thereby reducing the "footprint" of the channel region. Non-volatile memory cells using such finfet transistors have been disclosed. Some examples of prior art finfet non-volatile memory structures include us patents 7,423,310, 7,410,913, 8,461,640, and 9,634,018. However, these prior art finfet structures have disclosed using floating gates as stacked gate devices, or using trapping materials, or using SRO (silicon rich oxide) or using nano-crystalline silicon to store charge, or using other memory cell configurations that are too simple for memory cells with more than 2 gates or too complex for the number of gates in question.

The inventors have discovered a number of problems when scaling down the size of memory cells. Ultra-thin polysilicon or amorphous silicon film deposition and doping techniques are complex and often suffer from under-and non-uniform doping and structural non-uniformity. Ballistic electron transport in ultra-thin polysilicon floating gates causes programming problems (difficulty in trapping hot electrons in ultra-thin floating gates). The integration of the control gate on top of the floating gate results in a thick polysilicon stack that presents serious process integration challenges for advanced CMOS technology (CMP planarization step and subsequent advanced photolithography steps used in high-K metal gate process flows). The capacitive coupling between adjacent floating gates increases significantly with horizontal scaling. This leads to strong cross talk effects and requires complex management by design (the read current of a cell becomes dependent on the charge state of the neighboring cell). Scaling of flat floating gate memory cells is limited by the reduction in read current associated with transistor width scaling. Lower read currents impose penalties on access times and require complex design techniques to meet high access time specifications. The flat floating gate architecture does not allow for effective control of sub-threshold leakage of the floating gate and select transistor at advanced technology nodes, resulting in high background leakage from unselected cells sharing the same bit line as the selected cell.

Disclosure of Invention

The above problem is solved by a non-volatile memory cell comprising: a semiconductor substrate having an upper surface with upwardly extending fins including first and second side surfaces opposite one another; a first electrode in electrical contact with a first portion of the fin; a second electrode in electrical contact with a second portion of the fin, wherein the first and second portions of the fin are spaced apart from each other such that a channel region of the fin includes portions of the first and second side surfaces and extends between the first and second portions of the fin; a floating gate extending along a first portion of the channel region, wherein the floating gate extends along and is insulated from the first side surface, and wherein no portion of the floating gate extends along the second side surface; a word line gate extending along a second portion of the channel region, wherein the word line gate extends along and is insulated from the first and second side surfaces; a control gate disposed over and insulated from the floating gate; and an erase gate having a first portion disposed laterally adjacent to and insulated from the floating gate and a second portion disposed vertically above and insulated from the floating gate.

A method of forming a non-volatile memory cell, the method comprising: forming a trench into an upper surface of a semiconductor substrate such that the upper surface includes an upwardly extending fin including first and second side surfaces opposite one another; forming a first electrode in electrical contact with a first portion of the fin; forming a second electrode in electrical contact with a second portion of the fin, wherein the first and second portions of the fin are spaced apart from each other such that a channel region of the fin includes portions of the first and second side surfaces and extends between the first and second portions of the fin; and forming a floating gate extending along a first portion of the channel region, wherein the floating gate extends along and is insulated from the first side surface, wherein none of the floating gates extends partially along the second side surface; forming a word line gate extending along a second portion of the channel region, wherein the word line gate extends along and is insulated from the first and second side surfaces; forming a control gate disposed over and insulated from the floating gate; and forming an erase gate having a first portion disposed laterally adjacent to and insulated from the floating gate and a second portion disposed vertically above and insulated from the floating gate.

Other objects and features of the present invention will become apparent by a review of the specification, claims and appended figures.

Drawings

Fig. 1A-19A are side cross-sectional views (along the row direction) showing steps of forming a split-gate non-volatile memory cell of the present invention.

Fig. 1B-19B are side cross-sectional views illustrating the logic region of the substrate during steps of forming the split-gate non-volatile memory cell of the present invention.

Fig. 5C-19C are side cross-sectional views (along the column direction) showing steps of forming the split-gate non-volatile memory cell of the present invention.

Fig. 5D is a plan view showing the direction of the views of fig. 5A and 5C.

Fig. 15D-19D are side cross-sectional views (along the column direction) showing steps of forming the split-gate non-volatile memory cell of the present invention.

Fig. 20A and 20B are partial perspective views of a split gate non-volatile memory cell of the present invention.

Fig. 21A-25A are side cross-sectional views (along the row direction) showing steps in forming an alternative embodiment of a split-gate non-volatile memory cell of the present invention.

Fig. 21B-25B are side cross-sectional views illustrating the logic region of the substrate during steps in forming an alternative embodiment of the split-gate non-volatile memory cell of the present invention.

Fig. 21C-25C are side cross-sectional views (along the column direction) showing steps to form an alternative embodiment of a split-gate non-volatile memory cell of the present invention.

Figure 25D is a side cross-sectional view (along the column direction) showing the final step in forming an alternative embodiment of the split-gate non-volatile memory cell of the present invention.

Fig. 26A is a side sectional view (along the column direction) showing the memory cell unit of the first embodiment.

Fig. 26B is a side sectional view (in the column direction) showing a memory cell unit of the second embodiment.

Fig. 27A is a top view showing a memory cell unit of the first embodiment.

Fig. 27B is a top view showing a memory cell unit of the second embodiment.

Detailed Description

The embodiments described below effectively address scaling issues specific to conventional split gate memories. In particular, the split-gate memory cell of the present invention includes two transistors that are compatible with the mainstream finfet CMOS fabrication flow. The two transistors are connected in series. Each transistor is formed on two adjacent silicon fins. The first transistor (referred to as a wordline or select transistor) has a finfet architecture in which an HKMG gate electrode wraps around one of the silicon fins. A second transistor, referred to as a floating gate transistor, has a polysilicon floating gate interposed between two silicon fins. The floating gate transistor operates in a fully depleted SOI-like mode in which the ultra-thin channel is electrically controlled by the floating gate on only one side of the silicon fin that serves as the channel. The floating gate transistor has a split gate architecture (separate control gate and erase gate), allowing for improved endurance and reliability. This architecture allows for efficient control of cell leakage (from both selected and unselected cells) while maintaining reasonable floating gate physical dimensions and addressing main memory cell scalability challenges (high-K metal gate integration, read current scaling, floating gate cross talk and isolation, and selected and unselected cell leakage control). There are two disclosed embodiments. The first embodiment is implemented using a box-shaped floating gate. The second implementation is implemented using U-shaped floating gates, allowing improved control gate coupling to the floating gates to improve programming efficiency.

The formation of the first embodiment is illustrated in fig. 1A-19A, 5C-19C, 5D, and 15D-19D (which depict the formation of memory cells in a memory region of a substrate) and 1B-19B (which depict the formation of logic devices in a peripheral region (also referred to as a logic region) of the same substrate). The process begins by forming a silicon dioxide (oxide) layer 12 on both the memory region portion and the logic region portion of a silicon semiconductor substrate 10. A silicon nitride (nitride) layer 14 is formed on the oxide layer 12. A hard mask material 16 is formed on the nitride layer 12. A photoresist 18 is formed over the hard mask material 16. The photoresist is then patterned, which includes a photolithography process that includes selectively exposing portions of the photoresist, and selectively removing portions of the photoresist to expose selective portions of the underlying material (i.e., the strips of hard mask material 16 in this case). The resulting structure is shown in fig. 1A and 1B.

An etch is performed to remove the exposed portions of the hard mask material 16, leaving vertical stripes of hard mask material 16 (after photoresist removal) as shown in fig. 2A and 2B. Oxide spacers 20 are formed along the sides of the hard mask material strips 16 by performing an oxide deposition followed by an anisotropic oxide etch, which leaves the spacers 20 on the vertical sidewalls of the hard mask strips 16. A photoresist is formed over the structure and patterned to leave photoresist strips covering alternating spacers 20 in the memory area (e.g., right-hand spacers along each strip 16) and pairs of spacers 20 in the logic area. An oxide etch is then used to remove those oxide spacers 20 left by the photoresist exposure. As shown in fig. 3A and 3B, after the photoresist removal, etching is performed to remove the hard mask stripes 16.

One or more etches are performed to remove those portions of the nitride 14, oxide 12, and upper portions of the substrate 10 that are not under the oxide spacers 20, which results in the formation of trenches 24 that extend into the substrate, and thin fin structures 26 of the substrate 10 between adjacent trenches 24. As shown in fig. 4A and 4B, the fins 26 extend in the vertical/column direction, and are parallel to each other in both the memory area and the logic area. An insulating material 28 (e.g., oxide) is formed over the structure (including filling the trenches 24 with oxide 28), followed by oxide planarization to remove any portion of the oxide 28 on top of the nitride 14. A photoresist 30 is formed over the structure and patterned to remove portions of the photoresist 30 over the alternately filled trenches in the memory area as shown in fig. 5A and 5C-5D, and to leave the entire logic area covered as shown in fig. 5B. An oxide etch is performed to remove those portions of the oxide 28 left exposed by the photoresist 30, removing most of the oxide 28 from the alternating trenches 24 in the memory area. After the photoresist 30 is removed, as shown in fig. 6A-6C, a thick layer of polysilicon is then formed over the structure, filling the alternating trenches 24 in the memory region with polysilicon 32. A polysilicon etch is used to remove the top portion of the polysilicon 32 and recess the polysilicon in the trench 24 below the top of the oxide 28. An oxide deposition and planarization (e.g., CMP) is performed to form an oxide over the polysilicon, leaving blocks 32 of polysilicon between alternating fins 26 in the memory region as shown in fig. 7A-7C.

The nitride 14 is removed from the memory and logic regions by a nitride etch. An insulating layer (e.g., ONO with oxide-nitride-oxide sublayers) 34 is formed over the structure. A photoresist 36 is formed over the structure and patterned to expose those portions of the logic and memory regions laterally adjacent to the polysilicon block 32. As shown in fig. 8A-8C, an oxide etch is then used to remove some of the oxide 28 in each trench 24 between the ONO layer 34 and fin 26 in the logic region, and to remove the ONO layer 34 and oxide 28 laterally adjacent one side of the poly block 32. Polysilicon 38 is deposited over the structure. A photoresist 40 is formed over the structure and patterned to expose a portion of the memory area partially over the polysilicon block 32. As shown in fig. 9A-9C, etching is used to remove the exposed portions of the ONO layer 34.

After the photoresist is removed, an oxide layer 42 is formed over the structure. Nitride deposition and etching are used to form nitride spacers 44 along the vertical portions of the oxide layer 42 at the sidewalls of the polysilicon layer 38. The exposed portions of oxide layer 42 are removed using an oxide etch. Oxide deposition is then used to form an oxide layer 46 over the structure, including over the exposed portions of the polysilicon block 32. The oxide 46 will act as a tunnel oxide for the erase operation. As shown in fig. 10A-10C, a polysilicon layer 48 is formed over the structure. Polysilicon etch back, and oxide etch and planarization are performed to remove the upper portions of polysilicon layer 48 and oxide 46 and planarize the upper surface. As shown in fig. 11A-11C, a series of depositions are performed to form a silicon carbon nitride (SiCN) layer 50, an oxide layer 52, a SiCN layer 54, and a hard mask layer 56 on the structure.

A photoresist is formed over the structure and patterned, leaving only photoresist strips in the memory areas. An etch is performed to remove the exposed portions of the hard mask 56, leaving strips of hard mask 56 extending in the row/horizontal direction. After photoresist removal, an oxide deposition and etch are performed to form oxide spacers 58 against the remaining hard mask material strips. A photoresist 60 is formed over the structure and patterned to cover only one of the spacers (for each memory cell) positioned over the poly block 38. As shown in fig. 12A-12C, an oxide etch is used to remove all of the spacers 58 except for the spacers of each memory cell under the photoresist 60. After photoresist removal, photoresist 61 is formed on the structure and patterned so that it remains only over a portion of polysilicon block 32 and over a portion of polysilicon block 48 in the memory region. A SiCN etch is performed to remove the exposed portions of the SiCN layer 54. An oxide etch is performed to remove the oxide layer 52 and the exposed portions of the oxide spacers 58. In the logic region, the oxide layer 58 and the SiCN layer 54 are removed. The resulting structure is shown in fig. 13A to 13C.

After the photoresist is removed, a SiCN etch is performed to remove the exposed portions of the SiCN layer 50 and the SiCN layer 54 in the memory region. A polysilicon etch is then used to remove the exposed portions of polysilicon blocks 38 and 48 in the memory area, leaving polysilicon blocks 38a and 38b of virgin polysilicon 38, and reducing the lateral dimensions of polysilicon block 48, as shown in fig. 14A-14C. A photoresist is formed over the structure and patterned to remove it from a portion of the memory area adjacent to the poly block 48. An oxide etch is performed to recess the exposed portions of oxide 28 (on one side of poly block 48). After the photoresist is removed, the exposed silicon fin at the edge of the memory cell in the memory cell region may optionally be subjected to epitaxial growth to enlarge the top portion 26a of the silicon fin. The oxide layer 52 and SiCN layer 50 are removed from the memory region (using the photoresist over the logic region). A conformal oxide layer 62 is formed over the structure. A nitride layer 64 is then formed over the structure as shown in fig. 15A-15D. An oxide 66 (e.g., ILD oxide) is formed over the structure and CMP planarized using nitride 64 as a polish stop. A photoresist 68 is formed over the structure and patterned to expose the area above poly block 38a and the logic area. As shown in fig. 16A-16D, a nitride etch is used to remove portions of nitride layer 64 over the tops of polysilicon blocks 38a in the memory area, as well as portions of nitride layer 64 in the logic area.

After photoresist removal, polysilicon etch and oxide etch are used to remove polysilicon block 38a and surrounding oxide in the memory region, and polysilicon layer 38 in the logic region. A high-K metal gate layer HKMG 70 is then formed over the structure, filling the void left by the removed polysilicon block 38 a. HKMG layer 70 includes an insulating layer 72 of high-K material (i.e., having a dielectric constant K greater than that of an oxide such as HfO2, ZrO2, TiO2, Ta2O5, or other suitable material, etc.) underlying a conductive metal layer 74. The forming may be accomplished using atomic layer chemical vapor deposition and etch back. A high-K metal gate layer will also be formed in the logic region (i.e., the process flow for forming the HKMG is the same for both the memory region and the logic region). The resulting structure is shown in fig. 17A to 17D.

Nitride is formed over the structure and then CMP planarized to cover the top of HKMG 70. ILD oxide 66 is removed from the memory regions by patterned photoresist and oxide etch. Metal contacts 78 (one on the side of HKMG 70 and one on the side of polysilicon block 48) are formed on the structure, and then CMP is performed using nitride 76 as an etch stop. The metal contact 78 may include Ti 80, TiN82, and W84 sublayers annealed prior to the W CMP step. A nitride 86 is formed over the structure. The resulting structure is shown in fig. 18A to 18D. An oxide 88 is formed over the structure. A patterned photoresist is used to form metal contacts 90 that extend through the oxide 88 and nitride 86 to make contact with various features. The final structure is shown in fig. 19A to 19D.

Fig. 20A and 20B show perspective views of the final structure of a single memory cell having a floating gate transistor and a word line transistor formed along one of the substrate fins 26. The memory cell includes spaced apart bit line contacts 78a and source line contacts 78b, which are electrodes that are in electrical contact with the enlarged top portion 26a of the fin 26. Each electrode 78a and 78b is disposed on opposite sides and on top of fin 26. Word line gate 70, floating gate 32, control gate 38b, and erase gate 48 are located between electrodes 78a and 78 b. Word line gates 70 extend along and are insulated from both opposing sides and the top of fin 26. Floating gate 32 is disposed on only one side of fin 26. A control gate is disposed over fin 26 and over floating gate 32. Erase gate 48 is disposed over fin 26 and partially over floating gate 32 and includes a notch 48a facing the upper corner of floating gate 32 to improve erase efficiency. Fin 26 includes a channel region that extends from electrode 78a (which is a drain or bit line contact) to electrode 78b (which is a source contact). The conductivity of a portion of the channel region surrounded by the word line gate is controlled by the word line gate and the conductivity of another portion of the channel region adjacent to the floating gate is controlled by the floating gate.

It should be understood that although the figures herein show a single memory cell, multiple memory cells are formed end-to-end along fin 26, forming one column of memory cells, and that the memory cells of other columns are formed along other fins that extend parallel to the illustrated fin 26. Each memory cell is formed over or adjacent to two fins formed in a silicon substrate and has a non-uniform doping profile in both the vertical and horizontal directions. The isolation trenches between the fins are filled with silicon oxide. This finfet architecture allows for an increase in the effective read current density per cell area due to the extension of the channel width in the vertical direction. The floating gate is preferably n + type polysilicon (or amorphous silicon) that is disposed between and adjacent to a portion of the fins and is isolated from the fins by oxide. The floating gate extends in a vertical direction over the top of the fin. The floating gate has a box-like shape with a rectangular vertical cross-sectional shape (i.e., a cross-section viewed from the side, see fig. 19A and 19C). The floating gate and the adjacent portion of the fin form a first transistor of the cell, referred to as the floating gate transistor. The floating gate transistors operate like fully depleted silicon-on-insulator transistors. This configuration allows for an effective reduction in the parasitic coupling of the floating gate to neighboring cells while maintaining reasonable floating gate dimensions. A portion of the floating gate is disposed below the top of the fin, which may limit the height of the gate stack, facilitating integration with the HKMG CMOS process flow.

The control gate is preferably n + type polysilicon (or amorphous silicon), covers a portion of the floating gate, and is isolated from the floating gate and the top of the fin by a dielectric (a silicon oxide-nitride-oxide stack or the like). The erase gate is preferably n + type polysilicon (or amorphous silicon) and overlies the remainder of the floating gate and is isolated from the floating gate by a tunnel oxide. On one side, the erase gate is isolated from the control gate by silicon oxide or by silicon oxide/nitride/oxide spacers. On the other side, the erase gate extends over the floating gate, forming a wrap-around corner shape (i.e., notch 48a) for erase operation by the efficient corner-improved fowler-nordheim tunneling mechanism. The word line gate is preferably a metal (including tungsten and/or a work function adjusting metal) covering a second portion of the finfet channel region on both side surfaces of the fin. The metal word line gate and the underlying fin portion form a word line transistor. The finfet architecture of the wordline transistors allows for improved control of sub-threshold leakage current from unselected cells sharing the same column, and improved high temperature read performance and associated memory partitioning.

Source electrode 78b is comprised of a W/TiN/TiSi2 stack (or the like) on top of the enlarged epi shape 26a grown over the finfet transistor and adjacent to the floating gate transistor. The source electrodes 78b of the cells share the source electrodes of other cells in the same row, forming a common source line extending in the row direction. The bitline electrode 78a is composed of a W/TiN/TiSi2 stack (or the like) on top of the enlarged epi shape 26a grown over the finfet transistor and adjacent to the wordline transistor. The bit line electrodes 78a are connected to bit line electrodes of adjacent cells sharing the same column, and are connected together to form a common bit line 92 extending in the column direction. The wordline transistors and the floating gate transistors are surrounded by silicon nitride spacers which isolate them from the source line electrode regions and the bitline electrode regions. Fin 26 is preferably n + doped in those regions adjacent to the bit line electrode, source line electrode, under the silicon nitride spacer, and under the erase gate in regions not adjacent to the floating gate. Fin 26 is p-type doped in the region adjacent to the floating gate and in the region under the word line gate. The vertical P-type doping profile is highly non-uniform, with relatively low doping at the top of the fin and very high doping at the bottom of the fin. The top of the highly doped profile should be above the bottom of the floating gate to avoid forming a parasitic leakage path in the bottom portion of the finfet channel. The desired doping profile may be achieved by one or more implantations performed at different stages of the above-described manufacturing process flow. The injection may be shared with the logic area or implemented only on the memory area.

The memory cell architecture allows for reasonable physical size of the floating gate, simplifies processing, and addresses problems associated with ballistic transport and ultra-thin polysilicon deposition processes. Furthermore, embedding at least some of the floating gates between fins solves the associated isolation and cross-talk issues while optimizing the gate stack topology for high-K metal gate integration and providing a way for further cell size scaling.

The formation of the second embodiment is illustrated in fig. 21A-25A, 21B-25B, 21C-25C, and 25D. The process starts with the same structure of fig. 5A to 5C. An oxide etch is performed to remove those portions of the oxide 28 left exposed by the photoresist 30, removing most of the oxide 28 from the alternating trenches 24 in the memory area. After the photoresist 30 is removed, rather than forming a thick layer of polysilicon 32 filling the alternating trenches as disclosed above with reference to fig. 6A-6C, a thin conformal layer of in-situ n-type doped polysilicon 94 is formed over the structure, lining the walls of the alternating trenches 24 in the memory region with polysilicon 94 as shown in fig. 21A-21C. Oxide deposition followed by CMP planarization is used to fill the remainder of trench 24 with oxide 96 and planarize the structure, leaving polysilicon layer 94 with a U-shaped cross-sectional shape. The top of the U-shaped polysilicon layer 94 is then recessed using a polysilicon etch, as shown in fig. 22A-22C. An oxide is formed over the structure and CMP is used to lower and planarize the upper surface of the structure (using nitride 14 as a CMP stop). Nitride 14 is removed using a nitride etch. As shown in fig. 23A to 23C, an oxide layer is formed over the structure, and then CMP or etch back is performed using the polysilicon 94 as a stopper.

A photoresist is formed over the structure and patterned to leave exposed oxide 96 inside the U-shaped polysilicon 94. An oxide etch is then used to remove the oxide 96 inside the U-shaped polysilicon 94. As shown in fig. 24A-24C, after the photoresist is removed, an ONO insulating layer 34 is then formed over the structure as described above. Processing continues using the processing steps set forth above with reference to fig. 8A-8C-19A-19C and 15D-19D, resulting in the final structure shown in fig. 25A-25D. The final memory cell in this second embodiment has substantially the same structure as the final memory cell in the first embodiment, except that the floating gate has a U-shaped cross-sectional shape instead of a box shape, and the control gate 38b has a lower portion that extends down into the U-shaped interior of the floating gate 94 to improve the capacitive coupling between the floating gate 94 and the control gate 38 b. Fig. 26A and 26B show the difference in cross-sectional shape of the first embodiment and the second embodiment, respectively. Fig. 27A and 27B are top views of the first and second embodiments, respectively.

It is to be understood that the present invention is not limited to the embodiments described above and shown herein, but encompasses any and all variations within the scope of any claims supported thereby. For example, reference to the invention herein is not intended to limit the scope of any claims or claim terms, but rather only to one or more features that may be covered by one or more claims. The above-described examples of materials, processes, and values are illustrative only and should not be construed as limiting any claim. For example, the floating gate may be formed of amorphous silicon instead of polysilicon. In addition, not all method steps need be performed in the exact order illustrated. Finally, a single layer of material may be formed as multiple layers of such or similar materials, or vice versa.

It should be noted that as used herein, the terms "above … …" and "above … …" both inclusively include "directly on … …" (with no intervening material, element, or space disposed therebetween) and "indirectly on … …" (with intervening material, element, or space disposed therebetween). Similarly, the term "adjacent" includes "directly adjacent" (no intermediate material, element, or space disposed therebetween) and "indirectly adjacent" (intermediate material, element, or space disposed therebetween), "mounted to" includes "directly mounted to" (no intermediate material, element, or space disposed therebetween) and "indirectly mounted to" (intermediate material, element, or space disposed therebetween), and "electrically connected to" includes "directly electrically connected to" (no intermediate material or element therebetween that electrically connects the elements together) and "indirectly electrically connected to" (intermediate material or element therebetween that electrically connects the elements together). For example, forming an element "over a substrate" can include forming the element directly on the substrate with no intervening materials/elements therebetween, as well as forming the element indirectly on the substrate with one or more intervening materials/elements therebetween.