Abstract: A storage device structure (10) has two bits of storage per control gate (34) and uses source side injection (SSI) to provide lower programming current. A control gate (34) overlies a drain electrode formed by a doped region (22) that is positioned in a semiconductor substrate (12). Two select gates (49 and 50) are implemented with conductive sidewall spacers adjacent to and lateral to the control gate (34). A source doped region (60) is positioned in the semiconductor substrate (12) adjacent to one of the select gates for providing a source of electrons to be injected into a storage layer (42) underlying the control gate. Lower programming results from the SSI method of programming and a compact memory cell size exists.

Abstract: In some embodiments, non-volatile memory (NVM) devices are formed on a silicon-on-insulator (SOI) substrate (12) by forming elevated sources and drains (56) in contact with extensions (46) within the top silicon layer (18) of the SOI substrate (12). Buried conductive regions (42) are formed within the top silicon layer (18) below the extensions (46) to mitigate floating body effects that occur when using SOI substrates. In other embodiments, NVM devices are formed using elevated sources and drains (56), extensions (46) and the buried conductive regions (42) in bulk semiconductor substrates. In any embodiment, logic devices may be formed in conjunction with NVM devices, wherein the logic and NVM devices have elevated sources and drains (56), extensions (46) and the buried conductive regions (42).

Abstract: A non-volatile memory (NVM) has a silicon germanium (SiGe) drain that is progressively more heavily doped toward the surface of the substrate. The substrate is preferably silicon and the drain is formed by first forming a cavity in the substrate in the drain location. SiGe is epitaxially grown in the cavity with an increasing doping level. Thus, the PN junction between the substrate and the drain is lightly doped on both the P and N side. The drain progressively becomes more heavily doped until the maximum desired doping level is reached, and the remaining portion of the SiGe drain is doped at this maximum desired level. As a further enhancement, the perimeter of the SiGe in the substrate is the same conductivity type as that of the substrate and channel. Thus a portion of the channel is in the SiGe.

Abstract: A floating gate memory cell has a floating gate in which there are two adjacent floating gate layers. The top layer is made to have a contour while leaving the lower layer substantially unchanged. An interlevel dielectric and a control gate follow the contour of the floating gate to increase capacitance between the control gate and the floating gate. The two layers of the floating gate can be polysilicon in which the top layer has the contour formed therein by use of a sacrificial layer. The sacrificial layer is formed over the bottom polysilicon layer and etched. The top polysilicon layer is formed over the sacrificial layer. Subsequent processing of the top polysilicon layer exposes the remaining portion of the sacrificial layer so it can be removed; leaving the contour in the top polysilicon layer for the interlevel dielectric and the control gate.

Abstract: A plurality of memory cell devices is formed by using an intermediate dual polysilicon-nitride control electrode stack overlying nanoclusters. The stack includes a first-formed polysilicon-nitride layer and a second-formed polysilicon-containing layer. The second-formed polysilicon-containing layer is removed from areas containing the plurality of memory cells. In one form the second-formed polysilicon-containing layer also contains a nitride portion which is also removed, thereby leaving the first-formed polysilicon-nitride layer for the memory cell devices. In another form the second-formed ploysilicon-containing layer does not contain nitride and a nitride portion of the first-formed polysilicon-nitride layer is also removed. In the latter form a subsequent nitride layer is formed over the remaining polysilicon layer. In both forms a top portion of the device is protected from oxidation, thereby preserving size and quality of underlying nanoclusters.

Abstract: Forming a non-volatile memory device includes providing a semiconductor substrate, forming a masking layer having a first plurality of openings overlying the semiconductor substrate, forming diffusion regions in the semiconductor substrate at locations determined by the masking layer, forming a dielectric within the first plurality of openings, removing the masking layer to form a second plurality of openings, forming sacrificial spacers along edges of the second plurality of openings and adjacent to the dielectric, forming a separating dielectric to separate the sacrificial spacers within each of the second plurality of openings, forming a sacrificial protection layer overlying the separating dielectric, removing the sacrificial spacers, removing the sacrificial protection layer, forming at least two memory storage regions within each of the second plurality of openings, and forming a common control electrode overlying the at least two memory storage regions.

Abstract: To increase the gate coupling ratio of a semiconductor device 10, discrete elements 22, such as nanocrystals, are deposited over a floating gate 16. In one embodiment, the discrete elements 22 are pre-formed in a vapor phase and are attached to the semiconductor device 10 by electrostatic force. In one embodiment, the discrete elements 22 are pre-formed in a different chamber than that where they are attached. In another embodiment, the same chamber is used for the entire deposition process. An optional, interfacial layer 17 may be formed between the floating gate 16 and the discrete elements 22.

Abstract: A non-volatile memory (30) comprises nanocrystal memory cells (50, 51, 53). The program and erase threshold voltage of the memory cell transistors (50, 51, 53) increase as a function of the number of program/erase operations. During a read operation, a reference transistor (46) provides a reference current for comparing with a cell current. The reference transistor (46) is made from a process similar to that used to make the memory cell transistors (50, 51, 53), except that the reference transistor (46) does not include nanocrystals. By using a similar process to make both the reference transistor (46) and the memory cell transistors (50, 51, 53), a threshold voltage of the reference transistor (46) will track the threshold voltage shift of the memory cell transistor (50, 51, 53). A read control circuit (42) is provided to bias the gate of the reference transistor (46).

Abstract: Nanoclusters are blanket deposited on an integrated circuit and then removed from regions where the nanoclusters are not desired. A sacrificial layer is formed in those regions where the nanoclusters are not desired prior to the blanket deposition. The nanoclusters and the sacrificial layer are then removed. In one form, the sacrificial layer includes a deposited nitride containing or oxide containing layer. Alternatively, the sacrificial layer includes at least one of a pad oxide or a pad nitride layer previously used to form isolation regions in the substrate. Nanocluster devices and non-nanocluster devices may then be integrated onto the same integrated circuit. The use of a sacrificial layer protects underlying layers thereby preventing the degradation of performance of the subsequently formed non-nanocluster devices.

Abstract: A non-volatile memory (30) comprises nanocrystal memory cells (50, 51, 53). The program and erase threshold voltage of the memory cell transistors (50, 51, 53) increase as a function of the number of program/erase operations. During a read operation, a reference transistor (46) provides a reference current for comparing with a cell current. The reference transistor (46) is made from a process similar to that used to make the memory cell transistors (50, 51, 53), except that the reference transistor (46) does not include nanocrystals. By using a similar process to make both the reference transistor (46) and the memory cell transistors (50, 51, 53), a threshold voltage of the reference transistor (46) will track the threshold voltage shift of the memory cell transistor (50, 51, 53). A read control circuit (42) is provided to bias the gate of the reference transistor (46).

Abstract: Each cell of a memory is programmed by first using a source bias that is typically effective for programming the cells. If a cell is not successfully programmed in the first attempt, that is typically because a number of cells on the same column as that of the cell that did not successfully program have a relatively low threshold voltage, a low enough threshold voltage that these memory cells are biased, even with grounded gates, to be conductive. The vast majority of the cells do not have this problem, but it is common for there to be a few memory cells that do have this low threshold voltage characteristic. To overcome this, a different source bias is applied during subsequent programming attempts. Thus, the vast majority of the cells are programmed at the faster programming condition, and only the few that need it are programmed using the slower approach.

Abstract: A non volatile memory (100) includes an array (102) of transistors (30) having discrete charge storage elements (40). The transistors are programmed by using a two step programming method (60) where a first step (68) is hot carrier injection (HCl) programming with low gate voltages. A second step (78) is selectively utilized on some memory cells to modify the injected charge distribution to enhance the separation of charge distribution between each memory bit within the transistor memory cell. The second step of programming is implemented without adding significant additional time to the programming operation. In one example, the first step injects electrons and the second step injects holes. The resulting distribution of the two steps removes electron charge in the central region of the storage medium.

Abstract: A semiconductor device (10) has a highly doped layer (26) having a first conductivity type uniformly implanted into the semiconductor substrate (20). An oxide-nitride-oxide structure (36, 38, 40) is formed over the semiconductor substrate (20). A halo region (46) having the first conductivity type is implanted at an angle in only a drain side of the oxide-nitride-oxide structure and extends under the oxide-nitride-oxide structure a predetermined distance from an edge of the oxide-nitride-oxide structure. A source (52) and drain (54) having a second conductivity type are implanted into the substrate (20). The resulting non-volatile memory cell provides a low natural threshold voltage to minimize threshold voltage drift during a read cycle. In addition, the use of the halo region (46) on the drain side allows a higher programming speed, and the highly doped layer (26) allows the use of a short channel device.

Abstract: A non-volatile memory (30) comprises nanocrystal memory cells (50, 51, 53). The program and erase threshold voltage of the memory cell transistors (50, 51, 53) increase as a function of the number of program/erase operations. During a read operation, a reference transistor (46) provides a reference current for comparing with a cell current. The reference transistor (46) is made from a process similar to that used to make the memory cell transistors (50, 51, 53), except that the reference transistor (46) does not include nanocrystals. By using a similar process to make both the reference transistor (46) and the memory cell transistors (50, 51, 53), a threshold voltage of the reference transistor (46) will track the threshold voltage shift of the memory cell transistor (50, 51, 53). A read control circuit (42) is provided to bias the gate of the reference transistor (46).

Abstract: A non-volatile memory (30) comprises nanocrystal memory cells (50, 51, 53). The program and erase threshold voltage of the memory cell transistors (50, 51, 53) increase as a function of the number of program/erase operations. During a read operation, a reference transistor (46) provides a reference current for comparing with a cell current. The reference transistor (46) is made from a process similar to that used to make the memory cell transistors (50, 51, 53), except that the reference transistor (46) does not include nanocrystals. By using a similar process to make both the reference transistor (46) and the memory cell transistors (50, 51, 53), a threshold voltage of the reference transistor (46) will track the threshold voltage shift of the memory cell transistor (50, 51, 53). A read control circuit (42) is provided to bias the gate of the reference transistor (46).

Abstract: Each cell of a memory is programmed by first using a source bias that is typically effective for programming the cells. If a cell is not successfully programmed in the first attempt, that is typically because a number of cells on the same column as that of the cell that did not successfully program have a relatively low threshold voltage, a low enough threshold voltage that these memory cells are biased, even with grounded gates, to be conductive. The vast majority of the cells do not have this problem, but it is common for there to be a few memory cells that do have this low threshold voltage characteristic. To overcome this, a different source bias is applied during subsequent programming attempts. Thus, the vast majority of the cells are programmed at the faster programming condition, and only the few that need it are programmed using the slower approach.

Abstract: A semiconductor device (30) comprises an underlying insulating layer (34), an overlying insulating layer (42) and a charge storage layer (36) between the insulating layers (34, 42). The charge storage layer (36) and the overlying insulating layer (42) form an interface, where at least a majority of charge in the charge storage layer (36) is stored. This can be accomplished by forming a charge storage layer (36) with different materials such as silicon and silicon germanium layers or n-type and p-type material layers, in one embodiment. In another embodiment, the charge storage layer (36) comprises a dopant that is graded. By storing at least a majority of the charge at the interface between the charge storage layer (36) and the overlying insulating layer (42), the leakage of charge through the underlying insulating layer is decreased allowing for a thinner underlying insulating layer (34) to be used.

Abstract: An erase of a non-volatile memory (NVM) is achieved by first using oxide tunneling followed by hot hole injection (HHI). The subsequent use of HHI completes the erase that the tunneling cannot complete due to saturation. By first using tunneling, preferably Fowler-Nordheim tunneling (FNT), the damage to a bottom dielectric that normally occurs by HHI is significantly reduced. The damage due to HHI is significantly greater at the beginning of the erase when the electric field is greater. By reducing the damage due to HHI, the bottom dielectric can be smaller than that normally used for HHI so that high voltages are not required. Accordingly, the transistors in the periphery do not need to be so oversized as is normally required for HHI and thus saving area and power.

Abstract: A non-volatile multiple bit memory (10, 50) has electrically isolated storage elements (17, 21, 78, 80) that overlie a channel region having a central area (24, 94) with high impurity concentration. A planar gate (30, 84) overlies the storage elements. The high impurity concentration may be formed by a centrally located region (24) or by two peripheral regions (70, 72) having lower impurity concentration than the central portion of the channel. During a read or program operation, the channel area of high impurity concentration effectively controls a channel depletion region to enhance reading or programming of stored data bits. During a hot carrier program operation, the channel area of high impurity concentration enhances the programming efficiency by decreasing leakage currents in a memory array.

Abstract: A semiconductor device (10) has a highly doped layer (26) having a first conductivity type uniformly implanted into the semiconductor substrate (20). An oxide-nitride-oxide structure (36, 38, 40) is formed over the semiconductor substrate (20). A halo region (46) having the first conductivity type is implanted at an angle in only a drain side of the oxide-nitride-oxide structure and extends under the oxide-nitride-oxide structure a predetermined distance from an edge of the oxide-nitride-oxide structure. A source (52) and drain (54) having a second conductivity type are implanted into the substrate (20). The resulting non-volatile memory cell provides a low natural threshold voltage to minimize threshold voltage drift during a read cycle. In addition, the use of the halo region (46) on the drain side allows a higher programming speed, and the highly doped layer (26) allows the use of a short channel device.