NONVOLATILE SEMICONDUCTOR MEMORY DEVICE AND ITS FABRICATION METHOD
A memory cell includes a selective gate and a memory gate arranged on one side surface of the selective gate. The memory gate includes one part formed on one side surface of the selective gate and the other part electrically isolated from the selective gate and a p-well through an ONO layer formed below the memory gate. A sidewall-shaped silicon oxide is formed on side surfaces of the selective gate, and a sidewall-shaped silicon dioxide layer and a silicon dioxide layer are formed on side surfaces of the memory gate. The ONO layer formed below the memory gate is terminated below the silicon oxide, and prevents generation of a low breakdown voltage region in the silicon oxide near an end of the memory gate during deposition of the silicon dioxide layer.
Latest Patents:
- METHODS AND THREAPEUTIC COMBINATIONS FOR TREATING IDIOPATHIC INTRACRANIAL HYPERTENSION AND CLUSTER HEADACHES
- OXIDATION RESISTANT POLYMERS FOR USE AS ANION EXCHANGE MEMBRANES AND IONOMERS
- ANALOG PROGRAMMABLE RESISTIVE MEMORY
- Echinacea Plant Named 'BullEchipur 115'
- RESISTIVE MEMORY CELL WITH SWITCHING LAYER COMPRISING ONE OR MORE DOPANTS
The present application claims priority from Japanese patent application No. JP 2006-67088 filed on Mar. 13, 2006, the content of which is hereby incorporated by reference into this application.
TECHNICAL FIELD OF THE INVENTIONThe present invention relates to a semiconductor memory device and a fabrication technique for the semiconductor memory device, and particularly relates to a technique effective to be applied to a nonvolatile semiconductor memory device including memory cells each including a charge trapped layer constituted by a silicon nitride layer.
BACKGROUND OF THE INVENTIONAn electrically erasable and programmable nonvolatile read only memory is quite an important element in various LSI applied fields since stored information remains in each memory cell even if the memory is turned off.
There is described a so-called floating-gate nonvolatile memory and a nonvolatile memory using an insulating layer in S. Sze, “Physics of Semiconductor Devices, 2nd edition”, A Wiley-Interscience Publication (Non-Patent Document 1), pages 496-506. As disclosed in the Non-Patent Document 1, it is known that there is no need to separately form a conductive layer for accumulating charges in a nonvolatile memory in which charges are accumulated in a trap of an insulating layer or in an interface of a multilayer insulating layer, differently from a floating-gate nonvolatile memory in which charges are accumulated in a polycrystalline silicon layer. It is, therefore, possible to form memory cells with high consistency with CMOS-LSI process.
However, the nonvolatile semiconductor memory configured so that charges are accumulated in the insulating layer is required to include the insulating layer capable of maintaining sufficient charge holding characteristics even if injection and emission of charges are repeated. It is, therefore, difficult to realize such a nonvolatile semiconductor memory. There has been proposed, by contrast, a nonvolatile semiconductor memory that rewrites stored information by injecting charges having different signs in place of emitting charges. Operation performed by this nonvolatile semiconductor memory is described in “Symposium on VLSI Technology in 1997” (Non-Patent Document 2), pages 63-64. The nonvolatile semiconductor memory is characterized in that a polycrystalline silicon gate for causing each memory cell to operate and a gate for selecting one of the memory cells are formed separately. The same characteristic is also disclosed in U.S. Pat. No. 5,969,383 (Patent Document 1) and U.S. Pat. No. 6,477,084 (Patent Document 2).
A memory cell of the nonvolatile semiconductor memory disclosed in the Non-Patent Document 2 or the like is basically constituted by two transistors (a selective transistor and a memory transistor) each based on an n-channel MOSFET. The memory transistor is arranged next to the selective transistor to be connected to the selective transistor in a so-called ‘stacked in series’ manner.
In the memory cell shown in
Basic operations performed by the memory cell are four operations, i.e., (1) write, (2) erasing, (3) holding, and (4) read operations. It is to be noted that these four operations are denoted by typical notations and that the write and erasing operations are sometimes denoted differently. Furthermore, typical operations will be described; however, various other operations can be considered. Although the memory cell constituted by the n-channel MOSFETs will be described herein, a memory cell constituted by p-channel MOSFETs are identical to the former memory cell in principle.
(1) In a write operation, a positive potential is applied to the diffused layer 56 on the memory gate 52 side, and the same ground potential as that applied to the substrate 50 is applied to the diffused layer 55 on the selective gate 51 side. By applying a gate overdrive voltage higher than that applied to the substrate 50 to the memory gate 52, a channel under the memory gate 52 is turned into an ON-state. By setting a potential of the selective gate 51 to a value higher than a threshold voltage by 0.1V to 0.2V, a channel under the selective gate 51 is turned into an ON-state. At this time, a highest electric field is generated near a boundary between the two gates 51 and 52, so that many hot electrons are generated and injected into the memory gate 52. A state of generating carriers by electric field acceleration and impact ionization is denoted by symbol A. An electron is denoted by a white circle and a hole is denoted by a hatched circle.
This phenomenon is known and referred to as “source side injection or SSI”, which is described in A. T. Wu et al., “IEEE International Electron Device Meeting”, Technical Digest, pages 584-587, 1986 (Non-Patent Document 3). In the Non-Patent Document 3, a floating-gate memory cell has been described. However, a memory cell in which an insulating layer is used as charge trapped layer is identical to the floating-gate memory cell in injection mechanism.
The hot electron injection by the above-stated method is characterized in that the hot electron injection concentrates on a selective gate 51-side end of the memory gate 52 because of concentration of the electric field near the boundary between the selective gate 51 and the memory gate 52. Furthermore, while in the floating gate-type memory cell, the charge trapped layer is constituted by a conductive layer, in an insulating layer-type memory cell, charges are accumulated in the insulating layer (ONO layer 54). Therefore, electrons are held in an extremely narrow region in the insulating layer-type memory cell.
(2) In an erasing operation, a negative potential is applied to the memory gate 52 and the positive potential is applied to the diffused layer 56 on the memory gate 52 side so as to cause strong inversion on an end of the diffused layer 56 on which the memory gate 52 overlaps with the diffused layer 56. An interband tunnel phenomenon is thereby generated and holes are generated (denoted by symbol B). The interband tunnel effect is disclosed in, for example, T. Y. Chan et al., “IEEE International Electron Device Meeting”, Technical Digest, pages 718-721, 1987 (Non-Patent Document 4).
In this memory cell, the generated holes are accelerated in channel direction, attracted by a bias applied to the memory gate 52, and injected into the ONO layer 54, whereby the erasing operation is performed. A state of generating a secondary electron-hole pair resulting from the generated hole is denoted by symbol C. The carries are also injected into the ONO layer 54. Namely, a threshold voltage of the memory gate 52 that has risen by electron charges can be reduced by charges of the injected holes.
(3) In a holding operation, the charges are held as the charges of the carriers injected into the ONO layer 54. Since movement of the carriers in the ONO layer 54 is quite small and slow, the charges can be satisfactorily held even if a voltage is not applied to the memory gate 52.
(4) In a read operation, by applying a positive potential to the selective gate 51-side diffused layer 55 and the selective gate 51, the channel below the selective gate 51 is turned into the ON-state. The held charge information is read as a current by applying potential appropriate for discriminating a threshold voltage difference in the memory gate 52 between the write and erasing states (that is, an intermediate potential between the threshold voltage in a write state and that in an erasing state).
To fabricate the memory cell shown in
According to studies of the inventors of the present invention, however, the conventional method of forming the memory gate on one side surface of the selective gate using the spacer process has the following disadvantages. The conventional disadvantages will be described with reference to
First, as shown in
As shown in
Next, as shown in
After removing the photoresist layer 57, the unnecessary ONO layer 54 left on an upper surface, one side surface and the like of the selective gate 51 is etched and removed as shown in
As a result, as shown in
As already stated, in the write operation performed by this memory cell, the high voltage (Vmg) is applied to the memory gate 52 and a low source voltage (Vs) is applied to the diffused layer 56, so that a strong longitudinal electric field is generated near the end of the memory gate 52. Due to this, if a low breakdown voltage region is present in the silicon dioxide layer 60 in an area of the strong longitudinal electric field, a short-circuit occurs between the memory gate 52 and the substrate 50 (diffused layer 56).
To eliminate the recess 59, there is proposed wet-etching the ONO layer 54, thermally oxidizing the substrate 50, and thereby making the ONO layer 54 below the memory gate 52 thicker. However, since a thickness of the ONO layer 54 of a nonvolatile memory in the generation of 0.13 μm to 0.18 μm is equal to or larger than 20 nanometers, it is difficult to eliminate the recess 59 by thermal oxidation.
SUMMARY OF THE INVENTIONIt is an object of the present invention to provide a technique capable of improving the reliability of a nonvolatile semiconductor memory device including a memory cell that includes a charge trapped layer constituted by a silicon nitride layer.
The above and other objects and novel features of the present invention will be readily apparent from the description of the specification and the accompanying drawings.
An outline of typical elements of the invention disclosed in this application is described briefly as follows.
According to an aspect of the present invention, there is provided a method of fabricating a semiconductor device comprising steps of:
(a) forming a first conductive layer on a principal surface of a semiconductor substrate through a gate insulting layer, and then forming a selective gate by patterning the first conductive layer;
(b) forming an ONO layer on the semiconductor substrate including an upper surface and both side surfaces of the selective gate;
(c) forming a second conductive layer on the ONO layer, and then forming a memory gate electrically isolated from the selective gate and the semiconductor substrate through the ONO layer on each of the both side surfaces of the selective gate by anisotropically etching the second conductive layer, the memory gate being in a form of a sidewall;
(d) forming a first insulating layer on the semiconductor substrate, and then forming the first insulating layer in the form of the sidewall on an other side surface of the memory gate formed on each of the both side surfaces of the selective gate by anisotropically etching the first insulating layer;
(e) leaving the memory gate and the first insulating layer on one of the side surfaces of the selective gate and removing the memory gate and the first insulating layer formed on the other side surface of the selective gate by etching using a photoresist layer as a mask,
(f) after the step (e), leaving the ONO layer having an L-shaped cross section between one of the side surfaces of the selective gate and one side surface of the memory gate and below the memory gate by wet-etching the ONO layer; and
(g) after the step (f), forming a second insulating layer on the semiconductor substrate, and forming the second insulating layer in the form of the sidewall on an other side surface of the memory gate, and the second insulating layer in the form of the sidewall on the other side surface of the selective gate by anisotropically etching the second insulating layer.
According to the one aspect of the present invention, it is advantageously possible to improve the reliability of the semiconductor device.
Embodiments of the present invention will be described hereinafter in detail with reference to the drawings. The same constituent elements are basically denoted by the same reference symbols, respectively in all of the drawings for explaining the embodiments, and they will not be repeatedly described in the specification.
First EmbodimentThe memory cells MC1 and MC2 of the MONOS nonvolatile memory are formed in a p-type well 2 on a semiconductor substrate (hereinafter, simply “substrate”) 1 made of p-type monocrystalline silicon. The p-type well 2 is electrically isolated from the substrate 1 through a well-isolation n-type buried layer 4, and a desired voltage is applied to the p-type well 2.
Each of the memory cells MC1 and MC2 is configured to have a split gate structure including a selective gate 5 and a memory gate 6. The selective gate 6 is made of an n-type polycrystalline silicon layer and formed on a gate dioxide layer 5 made of a silicon dioxide layer. The memory gate 8 is made of an n-type polycrystalline silicon layer and arranged on one side surface of the selective gate 6. The memory gate 8 includes one part formed on one side surface of the selective gate 6 and the other part electrically isolated from the selective gate 6 and the p-type well 2 through an ONO layer 7 formed below the memory gate 8 and having an L-shaped cross section. The ONO layer 7 is constituted by two silicon dioxide layers and a silicon nitride layer (charge trapped layer) formed between the silicon dioxide layers. In a data write operation, hot electrons generated in a channel region are injected into the silicon nitride layer that forms part of the ONO layer 7 and captured in a trap of the silicon nitride layer.
An n+-semiconductor region 13 is formed in each of the p-type well 2 near the selective gate 6 and that near the memory gate 8. The n+-semiconductor region 13 functions as a source or a drain of a transistor that constitutes each of the memory cells MC1 and MC2. Furthermore, an n−-type semiconductor region 11 lower in impurity concentration than the n+-type semiconductor region 13 is formed in the p-type well 2 in a region adjacent to the n+-type semiconductor region 13. The n−-type semiconductor region 11 functions as an extension region for relaxing a high electric field on an end of the source and/or drain (n+-type semiconductor region 13).
A sidewall-shaped silicon dioxide layer 12 is formed on an opposite side surface of the selective gate 6 to that on which the ONO film 7 is formed. Sidewall-shaped silicon dioxide layers 9 and 12 are formed on an opposite side surface of the memory gate 8 to that on which the ONO film 7 is formed. A part formed on the p-type well 2 of the ONO film 7 having the L-shaped cross section is terminated below the silicon dioxide layer 9.
A Co (cobalt) silicide layer 14 is formed on surfaces of the selective gate 6, the memory gate 8, and the n+-type semiconductor region 13. The Co silicide layer 14 is formed to reduce resistances of the selective gate 6, the memory gate 8, and the n+-type semiconductor region 13, respectively.
A bit line BL is formed above the memory cells MC1 and MC2 configured as stated above through a silicon nitride layer 20 and a silicon dioxide layer 21. The bit line BL is electrically connected to one of the source and the drain (the n+-type semiconductor region 13 shared between the two memory cells MC1 and MC2) through a plug 19 in a contact hole 18 formed in the silicon nitride layer 20 and the silicon dioxide layer 21. The bit line BL is made of a metal film mainly consisting of Al (aluminum alloy), and the plug 19 is made of a metal film mainly consisting of W (tungsten).
Since a memory array using the memory cells MC1 and MC2 are identical in configuration to the memory array shown in
Referring next to
First, as shown in
As shown in
As shown in
As shown in
As shown in
As shown in
After removing the photoresist layer 10, the unnecessary ONO film 7 left on the upper surface, one side surface and the like of the selective gate 6 is etched and removed as shown in
If the wet etching is performed on the ONO layer 7, the ONO layer 7 is retreated (side-etched) in a downward direction of the memory gate 8 below the silicon dioxide layer 9 formed on the side surface of the memory gate 8 as shown in an enlarged view of
As shown in
As shown in
As shown in
Next, the silicon nitride layer 20 and the silicon dioxide layer 21 are deposited on the substrate 1 by the CVD, and the silicon nitride layer 20 and the silicon dioxide layer 21 are then etched to form the contact hole 18. After forming the plug 19 in the contact hole 18, the bit line BL is formed on the silicon dioxide layer 21, thereby completing the memory cell MC1 or MC2 shown in
The semiconductor device thus fabricated is configured so that the end of the ONO film 7 protrudes outside of the end of the memory gate 8 as shown in
Moreover, even if the ONO layer is to be removed using isotropic etching for forming the silicon dioxide layer 9 that functions as a mask on the side surface of the memory gate, it is possible to leave the ONO layer below the memory gate.
In the above-stated fabrication method, when impurities are implanted into the substrate 1 to form the n−-type semiconductor region 11 (see
In the first embodiment, the impurity ions are implanted into the substrate 1 to form the n−-type semiconductor region 11 as shown in
In a second embodiment, by contrast, impurity ions are implanted into the substrate 1 to form the n−-type semiconductor region 11 as shown in
As shown in
In a semiconductor device fabricated by a fabrication method according to the second embodiment, the n+-type semiconductor region 13 (source and/or drain) is formed after removal of the silicon dioxide layer 9, whereby the n+-type semiconductor region 13 on the memory gate 8 side can be formed to be close to the memory gate 8, as compared with the first embodiment. Furthermore, ends of the ONO layer 7 are located outside of the respective side surfaces of the memory gate 8, so that no low breakdown region is generated when the silicon dioxide layer 12 is deposited. It is thereby possible to prevent the short-circuit between the memory gate 8 and the n+-type semiconductor region 13 caused by the dielectric breakdown even if a high voltage is applied between the memory gate 8 and the n+-type semiconductor region 13.
Third EmbodimentAs shown in
In a third embodiment, one of the memory gates 8 formed on the respective side surfaces of the selective gate 6 is covered with a photoresist layer 22, and the other memory gate 8 is etched and removed, thereby leaving the memory gate 8 on one side surface of the selective gate as shown in
As shown in
As shown in
In the first embodiment, when the memory gate 8 on one side surface of the selective gate 6 is etched and removed (see
In the first embodiment, after the memory gates 8 are formed on the respective both side surfaces of the selective gate 6 as shown in
In a fourth embodiment, by contrast, after forming memory gates 8 on the respective both side surfaces of the selective gate 6 as shown in
As shown in
In the first embodiment, when the memory gate 8 on one side surface of the selective gate 6 is etched and removed (see
If the polycrystalline 25 containing impurities is formed on each of the side surfaces of the selective gate 6 constituted by the n-type polycrystalline silicon layer, when a voltage is applied to the selective gate 6, the voltage is also applied to the polycrystalline silicon layer 25. Namely, the polycrystalline silicon layer 25 substantially functions as a part of the selective gate 6. Therefore, a high potential difference is generated between the polycrystalline silicon layer 25 and the n+-type semiconductor region 13, and a strong voltage is applied to the low breakdown voltage region.
It is, therefore, preferable to constitute the polycrystalline silicon layer 25 by undoped polycrystalline silicon into which no impurities are implanted. In this case, the voltage applied to the selective gate 6 is not applied to the polycrystalline silicon layer 25, so that strong voltage is never applied to the low breakdown voltage region. Moreover, prior to the step of depositing the polycrystalline silicon layer 25 on the substrate 1 (see
In the first to fourth embodiments, the MONOS nonvolatile memory having the split gate structure including the selective gate 5 and the memory gate 6 has been described. However, the present invention is also applicable to a MONOS nonvolatile memory including a single memory gate.
To form the memory cell thus configured, after the ONO layer 30 is formed on the substrate 1, an n-type polycrystalline silicon layer deposited on the ONO layer 30 is patterned to form the memory gate 31. Next, the ONO layer 30 in regions other than the region below the memory gate 31 is etched and removed. However, if the ONO layer 30 is removed by dry etching, the substrate 1 is damaged by the etching. The ONO layer 30 is, therefore, removed by wet etching that does not damage the substrate 1. If so, as shown in
The present invention has been specifically described with reference to the embodiments so far. However, it goes without saying that the present invention is not limited to the embodiments but that various changes and modifications can be made of the present invention without departure of the scope of the invention.
According to the invention disclosed in the specification of the present application, the ends of the ONO layer below the memory gate protrude outside of the memory gate. Due to this, no low breakdown voltage region is generated in a second insulating layer near each end of the memory gate. Therefore, it is possible to realize the semiconductor device capable of preventing the short-circuit between the memory gate and the semiconductor substrate caused by the dielectric breakdown even if a high potential difference is generated between the memory gate and the semiconductor substrate while the memory cell operates.
Moreover, even if isotropic etching is performed after the memory gate is formed, to form the mask outside the memory gate further, the semiconductor device can be fabricated without removing the ONO layer below the memory gate.
The present invention is effective to be used for a nonvolatile semiconductor memory device including a memory cell that includes a charge trapped layer constituted by a silicon nitride layer.
Claims
1. A semiconductor device comprising a split-gate memory cell, the memory cell including:
- a selective gate formed on a principal surface of a semiconductor substrate through a gate insulating layer;
- a memory gate formed on one side surface of the selective gate, the memory gate being in a form of a sidewall; and
- an ONO layer with a generally L-shaped cross section, includes one part formed between one side surface of the selective gate and one side surface of the memory gate, and an other part formed below the memory gate,
- wherein a second insulating layer is formed on an other side surface of the memory gate through a first insulating layer, the first insulating layer being in the form of the sidewall, the second insulating layer being in the form of the sidewall,
- the second insulating layer in the form of the sidewall is formed on an other side surface of the selective gate, and
- one end of the ONO layer formed on the semiconductor substrate is terminated below the first insulating layer.
2. The semiconductor memory device according to claim 1,
- wherein hot electrons generated in the semiconductor substrate are injected into the ONO layer by applying a first voltage to the semiconductor substrate near the memory gate and a second voltage higher than the first voltage to the memory gate during a write operation.
3. The semiconductor memory device according to claim 2,
- wherein an erasing operation is performed by injecting holes into the ONO layer into which the hot electrons have been injected.
4-10. (canceled)
Type: Application
Filed: Jan 17, 2007
Publication Date: Sep 13, 2007
Applicant:
Inventors: Digh Hisamoto (Kokubunji), Kan Yasui (Kodaira), Shinichiro Kimura (Kunitachi), Daisuke Okada (Kunitachi)
Application Number: 11/653,832
International Classification: H01L 29/788 (20060101);