US 20060245245 A1
A non-volatile memory device has a channel region between source/drain regions, a floating gate, a control gate, a first dielectric region between the channel region and the floating gate, and a second dielectric region between the floating gate and the control gate. The first dielectric region includes a high-K material. The non-volatile memory device is programmed and/or erased by transferring charge between the floating gate and the control gate via the second dielectric region.
1. A method of making a non-volatile storage device, comprising:
depositing a high-K material over a region of a semiconductor to be used as a channel region;
depositing a floating gate over said high-K material;
adding a dielectric region over said floating gate; and
adding a control gate over said dielectric material, said non-volatile storage device is erased by transferring charge from said control gate to said floating gate via said dielectric region.
2. A method according to
said dielectric region includes tunnel oxide.
3. A method according to
depositing a poly-silicon layer; and
depositing a low resistivity layer above said poly-silicon layer.
4. A method according to
depositing a tungsten nitride barrier layer; and
depositing a tungsten layer over said tungsten nitride layer.
5. A method according to
adding an epitaxially grown silicon region.
6. A method according to
said steps of depositing a high-K material, depositing a floating gate, depositing a second dielectric, and depositing a control gate include performing any one of chemical vapor deposition, physical vapor deposition, or atomic layer deposition.
7. A method according to
performing sidewall oxidation, said sidewall oxidation causes rounding of edges of said floating gate and said control gate.
8. A method according to
said non-volatile storage device is programmed by transferring charge from said floating gate to said control gate.
9. A method according to
said non-volatile storage device is a multi-state flash memory device.
10. A method according to
said non-volatile storage device is a multi-state NAND flash memory device.
11. A method according to
said depositing a high-K material includes depositing Aluminum Oxide.
12. A method according to
adding an oxide spacer next to said high-K material, said dielectric region, said floating gate and said control gate.
13. A method according to
adding a spacer next to said floating gate.
14. A method according to
adding a nitride spacer on a side of said high-K material.
15. A method according to
adding a metal fin on a side of said control gate.
16. A method of making NAND flash memory, comprising:
creating source/drain regions within a substrate;
depositing a high-K material over said substrate;
depositing a floating gate layer over said high-K material;
etching through said high-k material and said floating gate layer to form NAND strings;
adding a dielectric layer over said floating gate layer;
adding a control gate layer over said dielectric layer; and
etching through said control gate layer to separate non-volatile storage devices on said NAND strings, said non-volatile storage devices program by transferring electrons from their floating gates to their control gates via said dielectric layer and erase by transferring electrons from their control gates to their floating gates via said dielectric layer.
17. A method according to
said non-volatile storage devices program by tunneling electrons from their floating gates to their control gates via said dielectric layer and erase by tunneling electrons from their control gates to their floating gates via said dielectric layer.
18. A method according to
said non-volatile storage devices program by Fowler-Nordheim tunneling electrons from their floating gates to their control gates via said dielectric layer and erase by Fowler-Nordheim tunneling electrons from their control gates to their floating gates via said dielectric layer.
19. A method of making a non-volatile storage device, comprising:
depositing a high-K material over a region of a semiconductor to be used as a channel region;
depositing a floating gate over said high-K material;
adding a dielectric region over said floating gate; and
adding a control gate over said dielectric material, said non-volatile storage device transfers electrons from said control gate to said floating gate via said dielectric region.
20. A method according to
said adding said dielectric region includes depositing or growing oxide material;
said non-volatile storage device erases by transferring charge from said control gate to said floating gate via said dielectric region; and
said non-volatile storage device programs by transferring charge from said floating gate to said control gate via said dielectric region.
This application is a divisional of U.S. patent application Ser. No. 10/762,181, “Non-Volatile Memory Cell Using High-K Material and Inter-Gate Programming,” filed Jan. 21, 2004, inventors Nima Mokhlesi and Jeffrey W. Lutze, incorporated herein by reference in its entirety.
1. Field of the Invention
The present invention relates to non-volatile memory devices.
2. Description of the Related Art
Semiconductor memory devices have become more popular for use in various electronic devices. For example, non-volatile semiconductor memory is used in cellular telephones, digital cameras, personal digital assistants, mobile computing devices, non-mobile computing devices and other devices. Electrical Erasable Programmable Read Only Memory (EEPROM) and flash memory are among the most popular non-volatile semiconductor memories.
Typical EEPROMs and flash memories utilize a memory cell with a floating gate that is provided above a channel region in a semiconductor substrate. The floating gate is separated from the channel region by a dielectric region. For example, the channel region is positioned in a p-well between source and drain regions. A control gate is provided over and separated from the floating gate. The threshold voltage of the memory cell is controlled by the amount of charge that is retained on the floating gate. That is, the level of charge on the floating gate determines the minimum amount of voltage that must be applied to the control gate before the memory cell is turned on to permit conduction between its source and drain.
Some EEPROM and flash memory devices have a floating gate that is used to store two ranges of charges and, therefore, the memory cell can be programmed/erased between two states (e.g. a binary memory cell). A multi-bit or multi-state flash memory cell is implemented by identifying multiple, distinct threshold voltage ranges within a device. Each distinct threshold voltage range corresponds to predetermined values for the set of data bits. The specific relationship between the data programmed into the memory cell and the threshold voltage levels of the cell depends upon the data encoding scheme adopted for the cells. For example, U.S. Pat. No. 6,222,762 and U.S. patent application Ser. No. 10/461,244, “Tracking Cells For A Memory System,” filed on Jun. 13, 2003, both of which are incorporated herein by reference in their entirety, describe various data encoding schemes for multi-state flash memory cells. To achieve proper data storage for a multi-state cell, the multiple ranges of threshold voltage levels should be separated from each other by sufficient margin so that the level of the memory cell can be read, programmed or erased in an unambiguous manner.
When programming typical prior art EEPROM or flash memory devices, a program voltage is applied to the control gate and the bit line is grounded. Electrons from the channel are injected into the floating gate. When electrons accumulate in the floating gate, the floating gate becomes negatively charged and the threshold voltage of the memory cell as seen from the control gate is raised.
Typically, the program voltage Vpgm applied to the control gate is applied as a series of pulses. The magnitude of the pulses is increased with each successive pulse by a predetermined step size (e.g. 0.2 v). In the periods between the pulses, verify operations are carried out. That is, the programming level of each cell of a group of cells being programmed in parallel is read between each programming pulse to determine whether it is equal to or greater than each individual cell's targeted verify level to which it is being programmed. One means of verifying the programming is to test conduction at a specific compare point. The cells that are verified to be sufficiently programmed are locked out, for example, by raising the bit line voltage from 0 to Vdd to stop the programming process for those cells. The above described programming technique, and others described herein, can be used in combination with various self boosting techniques, for example, as described in U.S. patent application Ser. No. 10/379,608, titled “Self Boosting Technique,” filed on Mar. 5, 2003, incorporated herein by reference in its entirety. Additionally, an efficient verify technique can be used, such as described in U.S. patent application Ser. No. 10/314,055, “Smart Verify for Multi-State Memories,” filed Dec. 5, 2002, incorporated herein by reference in its entirety.
Typical prior art memory cells are erased by raising the p-well to an erase voltage (e.g. 20 volts) and grounding the control gate. The source and drain are floating. Electrons are transferred from the floating gate to the p-well region and the threshold voltage is lowered.
There is a trend to make smaller and smaller non-volatile memory devices. As devices become smaller, it is anticipated that the cost per bit of a memory system will be reduced. As the channel size is reduced, the capacitive coupling between the channel and the floating gate needs to be increased in order to maintain the gate's influence over the channel. One way to achieve this is to reduce the effective thickness of the dielectric region between the channel and the floating gate. Thinner effective gate oxide thicknesses will maintain the dominance of the gate to channel capacitance over other parasitic capacitances to the channel such as those of the drain, source and substrate. Otherwise, the source, drain, and/or substrate (i.e. P-well region for N-channel devices fabricated in a triple well) regions will have too much influence over the channel. However, if the thickness of the channel dielectric region becomes too small, the electric field from a charged floating gate can cause electrons to leak from the floating gate across the channel dielectric region and into the channel, source, or drain. In some cases, if the dielectric region is not thick enough, direct tunneling occurs when no tunneling is desired. Thus, there is a need to shrink device size of non-volatile memory devices, without suffering from the effects of thin dielectric regions.
The present invention, roughly described, pertains to non-volatile memory devices, including EEPROMS, flash memory and other types of non-volatile memory. One embodiment of the non-volatile memory device includes a channel region between source/drain regions, a floating gate, a control gate, a first dielectric region between the channel region and the floating gate, and a second dielectric region between the floating gate and the control gate. The first dielectric region includes a high-K material (and, maybe, other materials). When operating one embodiment of the above described non-volatile memory device, the non-volatile memory device is programmed and/or erased by transferring charge between the floating gate and the control gate via the second dielectric region (i.e. the inter-gate dielectric region). In one example implementation, the non-volatile memory device is programmed and/or erased by tunneling between the floating gate and the control gate via the second dielectric region.
In one embodiment of the present invention, the non-volatile storage device is a flash memory device (e.g. binary flash memory device or multi-state flash memory device). In other embodiments, the device is a different type of non-volatile memory device.
One or more of the non-volatile memory devices can be used in a system that includes a control circuit for operating the non-volatile memory devices. For example, a control circuit can include (individually or in combination) a controller, a state machine, decoders, drivers, sense amplifiers, other logic, subsets of the above and/or combinations of the above.
These and other objects and advantages of the present invention will appear more clearly from the following description in which the preferred embodiment of the invention has been set forth in conjunction with the drawings.
FIGS. 10A-F depict the non-volatile memory device of
Between N+ diffusion regions 24 is the channel 16. Above channel 16 is dielectric area 30. Above dielectric area 30 is floating gate 32. The floating gate, under low voltage operating conditions associated with read or bypass operations, is electrically insulated/isolated from channel 16 by dielectric area 30. Above floating gate 32 is dielectric area 34. Above dielectric area 34 is a poly-silicon layer of control gate 36. Above poly-silicon layer 36 is a conductive barrier layer 138 made of Tungsten Nitride (WN). Above barrier layer 138 is a low resistivity metal gate layer 40 made of Tungtsen. WN layer 38 is used to reduce the inter-diffusion of Tungsten into the poly-silicon layer of control gate 36, and also of silicon into Tungsten layer 40. Note that, in one embodiment, control gate 36 consists of layers 36, 38, and 40 as they combine to form one electrode. In other embodiments, a single metal layer, or multiple metal layers without using a poly control gate sub-layer 36 can be used. Dielectric 30, floating gate 32, dielectric 34, poly-silicon layer of control gate 36, WN layer 38 of control gate, and Tungsten metal layer 40 of control gate comprise a stack. An array of memory cells will have many such stacks.
Various sizes and materials can be used when implementing the memory cell of
Use of high-K dielectric materials between the crystalline silicon channel, and a poly gate typically creates two interfacial layers above and below the high-K material itself. These interfacial layers are composed of SiO2, or Silicon Oxy-nitride (SiON), with some fraction of metal atoms that may have diffused from the high-K material itself. These interfacial layers are usually formed naturally and not intentionally, and in many applications these interfacial layers are undesirable, as their dielectric constant tends to be substantially lower than the dielectric constant of the high-K material. In the present application, because the high-K dielectric is substantially thicker than that used for gate dielectrics of advanced MOS logic transistors, an interfacial layer that is 1 nm thick or even thicker may not only be tolerable, but also a welcome feature. This will especially be the case if the lower K interfacial layer provides higher mobility for channel electrons, and/or higher immunity to leakage currents because of the higher energy barrier (bottom of the conduction band offset) that the interfacial layer may offer. Higher energy barriers reduce the possibility of electron injection into the high-K dielectric by both direct tunneling, and Fowler-Nordheim (FN) tunneling. Silicon nitride or other inter-diffusion barrier insulators and oxygen diffusion barrier insulators may also be deposited or grown at the interface of silicon and high-K material in order to impede inter-diffusion of various atoms across material boundaries and/or impede further growth of interfacial silicon oxide layers. Toward these ends, in some embodiments, layers of silicon oxide and/or silicon nitride may be intentionally grown and/or deposited to form part of the interfacial layers above and/or below the high-K dielectric(s).
Floating gate 32 is 20 nm and is typically made from poly-silicon that is degenerately doped with n-type dopants; however, other conducting materials, such as metals, can also be used. Dielectric 34 is 10 nm and is made of SiO2; however, other dielectric materials can also be used. Control gate sub layer 36 is 20 nm and is made from poly-silicon; however, other materials can also be used. The WN conducting diffusion barrier layer 38 is 4 nm thick. Tungsten metal control gate layer 40 is 40 nm thick. Other sizes for the above described components can also be implemented. Additionally, other suitable materials, such as replacing W/WN with Cobalt Silicide, can also be used. The floating gate and the control gate can also be composed of one or more layers of poly-silicon, Tungsten, Titanium, or other metals or semiconductors.
As mentioned above, dielectric 30 includes a high-K material. A “high-K material” is a dielectric material with a dielectric constant K greater than the dielectric constant of silicon dioxide. The dielectric constant K of silicon dioxide is in the range 3.9 to 4.2. For the same actual thickness, a high-K material will provide more capacitance per unit area than silicon dioxide (used for typical dielectric regions). In the background discussion above, it was stated that as channel size becomes smaller, the thickness of the dielectric region between the channel and the floating gate should be reduced. What is learned is that it is the effective thickness that must be reduced because it is the effective thickness that determines the control of the floating gate over the channel. Effective thickness is determined as follows:
A high-K material will have an effective thickness that is lower than its actual thickness. Therefore, a high-K material can be used with a smaller channel size. The smaller effective thickness accommodates the smaller channel size, allowing the gate to maintain the appropriate influence over the channel. The larger actual thickness of a high-K material helps prevent the leakage discussed above.
In one embodiment, the programming and erasing is performed by transferring charge between floating gate 32 and control gate 36, across dielectric 34. This is advantagous because the programming mechanism (e.g. tunneling) is now not so burdened with strong coupling. Rather, the strong steering function is placed between the floating gate and the channel, matching the strong channel coupling dictate for scaled channels. Thus, the memory cell of
Some advantages which may be realized with some embodiments of the above described memory cell includes the ability to properly scale the device; wear associated with program/erase can be confined to the inter-gate region (away from the channel), which can increase endurance; lower program/erase voltages and/or higher reliability by using thicker dielectrics; and the elimination of the need to aggressively scale tunnel oxide of traditional NAND (or flash memories with other architectures such as NOR). A designer of a memory cell according to the present invention should be mindful of GIDL and a lower control gate coupling ratio (less Qfg, stronger magnification of channel noise and larger manifestations of cell-to-cell variations).
In one embodiment, the memory cell of
Dielectric 130, floating gate 132, dielectric 134, poly-silicon 136, WN layer 138, Tungsten layer 140, and hard mask Silicon Nitride (Si3N4) layer 142 form a stack. The memory cell of
In some embodiments, between oxides spacers of neighboring stacks are epitaxially grown silicon regions 144 (e.g., positioned over N+ diffusion regions 124). The utilization of such epitaxially grown raised source/drain regions obviates the implanted source/drain regions underneath them, increasing the effective channel length of the device in keeping with the dictates of proper scaling of MOS devices. This reduces punch through and improves the sub-threshold swing of the NAND devices. The issues arising from offset source/drain diffusion regions that degrade the endurance characteristics of standard NAND device should not arise here as the tunneling and the associated charge trapping have been moved from the channel dielectric to the inter-gate dielectric. The epitaxially grown silicon regions 144 also provides extra capacitance between the floating gate and the channel/source/drain, reducing the high voltage requirements for program and erase operations. It is desired to have the floating gate more capacatively coupled to the channel than the control gate. In some implementations, there is a goal to maximize the voltage drop across dielectric 134 and have less of a voltage drop across dielectric 130. By using the high-K material in dielectric 130 in combination with the epitaxially grown silicon regions 144, the coupling between floating gate 132 and channel 166 is increased. Yet another benefit derived from epitaxially grown source/drain regions is their capability to reduce the capacitive coupling between neighbor floating gates on neighbor word lines by shielding these floating gates from one another. This effect is a major problem resulting in eroded threshold sensing margins. This effect was first published in May 2002 issue of IEEE Electron Device Letters, Vol. 23, No. 5, page 264 by Jae-Duk Lee, et. al. in an article titled: “Effects of Floating Gate Interference on NAND Flash Memory Cell Operation”. Also see U.S. Pat. Nos. 5,867,429 and 5,930,167, which patents are incorporated herein by reference in their entirety.
The article titled “A Novel Gate-Offset NAND Cell (GOC-NAND) Technology Suitable for High-Density and Low-Voltage-Operation Flash Memories” by Shinji Satoh, et. al. published in the Technical Digest of 1999 IEDM (section 11, number 2, page 275) discusses the issue of parasitic cells formed in the off-set region of GOC-NAND devices that impact the cycling endurance of cells through trap-up occurring in the oxide residing above these parasitic cells. While this is a serious issue plaguing the conventional implementation of GOC-NAND, the gate-offset embodiments of the present invention should not suffer from this issue because the tunneling action should be confined to the inter-gate dielectric.
High-K channel dielectric 230, floating gate 232, inter-gate dielectric 234, lower control gate 236, WN barrier layer 238 and Tungsten layer 240 form a stack. The memory cell of
Dielectric 230A, floating gate 232A, inter-gate dielectric 234A, lower control gate 236A, WN barrier layer 238A and Tungsten layer 240A form a stack that is trapezoidal in shape (tapered toward the top), which helps the dielectric 230A provide more coupling to the floating gate as compared to the control gate coupling to the floating gate.
The memory cell of
Above epitaxially grown silicon regions 144 and between the stacks is a SiO2 filler layer 252A. Above SiO2 filler layer 252A and also between the stacks is a booster fin 250A. In one embodiment, booster fin 250A is made of a metal, for example, Tungsten.
A booster fin is a variation of a booster plate. Booster plates are made of metal layers that usually wrap around word line stacks and provide isolation for floating gate to floating gate capacitive interference effects. They can be manufactured in a connected form, covering the entire memory array, or be broken up into distinct electrodes with each individual electrode covering a single plane of memory, covering a single erase block, or covering a few erase blocks. Additional relevant background information can be found in U.S. Pat. No. 5,877,980; U.S. Pat. No. 6,093,605; U.S. Pat. No. 6,246,607; U.S. Pat. No. 5,990,514; U.S. Pat. No. 6,044,017; U.S. Pat. No. 5,936,887; Choi et al., “A Novel Booster Plate Technology in High Density NAND Flash Memories for Voltage Scaling-Down and Zero Program Distrubance”, IEEE Symposium on VLSI Technology Digest of Technical Papers, 1996, pp. 238-239; Kim et al., “Fast Parallel Programming of Multi-Level NAND Flash Memory Cells Using the Booster-Line Technology”, Symposium on VLSI Technology Digest of Technical Papers, 1997, pp. 65-66; Choi et al., “A Triple Polysilicon Stacked Flash Memory Cell With Wordline Self-Boosting Programming”, IEEE, 1997, PP. 283-286; and Satoh et al., “A Novel Channel Boost Capacitance (CBC) Cell Technology with Low Program Disturbance Suitable for Fast Programming 4 Gbit NAND Flash Memories”, IEEE Symposium on VLSI Technology Digest of Technical Papers, 1998, pp. 108-109; all of which are incorporated herein by reference. One embodiment of the
Booster fins are similar to booster plates, except that they only consist of fins that are placed between stacks within the memory array, and the fins can be electrically connected to each other in the shunt areas of the array. Shunt areas consist of breaks in the memory array that run in the direction of the bit lines and occur at a frequency of once every few hundred bit lines. A shunt area separates two neighboring bit lines from one another. While booster plates cover the top of all word lines, booster fins do not cover the top of word lines. One embodiment would allocate a single, isolated booster fin or plate to each erase block.
In some embodiments, the individual booster fins or blocks are driven by an NMOS device to drive them to positive voltages and a PMOS device to drive them to high negative voltages. In some embodiments a fixed negative voltage of, for example, −5V is applied to booster fins or plates during read and verify operations with the objective of bringing some of the otherwise negative range of cell threshold voltages into the positive range which then become measurable by control gates which can only take positive voltage values. In some other embodiments the booster fins or plates will have the same voltage as the selected word lines for read operations. The advantage of these embodiments is that the control gate to floating gate coupling ratio for read and verify operations is enhanced by booster plates or fins to floating gate coupling ratio. The effects of threshold voltage variations due to dopant fluctuation or geometric variations, and 1/f noise or random telegraph signal (RTS) noise that are a result of trapping and de-trapping of charges into interface and deeper trap sites are magnified by the inverse of the control gate coupling ratio when the cell's threshold voltage is measured from the control gate. In this sense a high control gate coupling ratio is desirable. However, a low control gate coupling ratio is desirable because it allows inter-gate program and erase operations to be accomplished at substantially lower voltages. Therefore, for program and erase operations, it maybe advantagous to apply as high a voltage as may be possible in the opposite direction or polarity as the word lines. For example, in order to program, 15V may be applied to the word line while the P-well, and the channel are at or near zero volts. The floating gate may be at a voltage in the range 3V to 6V depending on how much charge is on it. A grounded booster plate or fin will couple down the floating gate and make it easier to program. An added advantage is that a booster plate or fin that is at a lower voltage than the floating gate will tend to inhibit edge dominated tunneling and, thus, provide a more uniform tunneling behavior without having to utilize high temperature side wall oxidation in order to round the floating gate corners.
In the embodiment of
Note that the embodiments of
Also note that the memory cells of 1, 3, 4 and 4A include one floating gate per memory cell. In other embodiments, more than one floating gate per memory cell can be used.
The memory cells of
The memory cells described in
In the present devices the substrate is tightly coupled to the floating gate via the high dielectric constant material and the control gate is relatively weakly coupled to the floating gate so that reversing the polarity of the definition of erase and program is convenient. That is, when the substrate is raised to a high potential, the floating gate is also raised to a relatively high potential, and many electrons are transferred to the floating gate by tunneling from a grounded control gate, resulting in the collection of cells having a high threshold as viewed from the control gate. Programming, or setting a variable threshold to represent the data state, is accomplished by selectively removing some electrons by raising the control gate in a controlled fashion and terminating the electron removal on a cell by cell basis. This results in selectively reducing the threshold voltage as seen from the control gate, in direct contrast to the prior art devices. This will be described more completely below in conjunction with
In one example, the drain and the p-well will receive 0 volts while the control gate receives a set of programming pulses with increasing magnitudes, such as depicted in
One means for verifying is to apply a pulse at the word line corresponding to the target threshold value and determine whether the memory cell turns on. If so, the memory cell has reached its target threshold voltage value. For arrays of flash memory cells, many cells are verified in parallel. For some embodiments of multi-state flash memory cells, after every individual program pulse the memory cells will experience a set of verification steps to determine which state the memory cell is within. For example, a multi-state memory cell capable of storing data in eight states may need to perform verify operations for seven compare points. Thus, seven verify pulses are applied in order to perform seven verify operations between two consecutive programming pulses are. Based on the seven verify operations, the system can determine the state of the memory cells. Performing seven verify operations after each programming pulse slows down the programming process. One means for reducing the time burden of verifying is to use a more efficient verify process, for example, as disclosed in U.S. patent application Ser. No. 10/314,055, “Smart Verify for Multi-State Memories,” filed Dec. 5, 2002, incorporated herein by reference in its entirety.
In one embodiment of a two state memory cell according to the teachings of
Compacting the wide erase distribution 302 into a narrower distribution 304 is referred to as soft programming. In standard NAND memories sufficient tightening of a wide erase distribution 302 by soft programming is achieved in a massively parallel operation where all the word lines in one erase block are simultaneously raised to a suitable soft programming starting voltage for a first soft programming pulse, and the soft programming pulses are stair cases in the same manner as regular programming. A single verify operation is performed after each soft programming pulse with all the word lines grounded, the roles of source and drain is reversed by applying VDD voltage to the source of NAND strings, and sensing the bit line voltage. As long as the bit line voltage rises above a first erase verify voltage (EV1) of for example, 1V, the soft programming operation will continue on that bit line. This rise of the bit line voltage indicates that the threshold voltage of none of the cells on the corresponding NAND string has risen to a high enough value of typically −0.8V to shut off the current in the string. During a soft programming verify operation, when an individual bit line voltage does no longer rise above EV1, that corresponding NAND string is locked out of subsequent soft programming pulses through the usual boosting techniques used for program inhibit. A final verify operation using grounded word lines, and a second sensing trip point EV2 of, for example 0.7V, is used to make sure no more than a tolerable number of strings contain one or more cells with threshold voltages above, for example −0.5V. Applying the same read voltage to all the word lines of a NAND string results in gaining the following information: 1) if the string is “ON” then all cells in the string have a threshold voltage below the voltage applied to all word lines, and 2) if the string is “OFF” then at least one cell has a threshold voltage greater than the applied word line voltage.
Since during soft programming verify operation the goal is finding the first cell on each string whose threshold voltage becomes smaller than a designated value, the massive multiple word line verify parallelism that is utilized in conventional NAND will no longer work for some embodiments of the present invention. One approach for soft programming can be the following. Apply, for example 4V, to every word line during verify operations, and lock out each string when it is detected to be “ON”. Each string will be detected as being “ON” only when every cell in the string has been programmed to a threshold voltage below 4V. With this approach, the hope is that the distribution of threshold voltages within each group cells belonging to the same string is tight enough that when the threshold voltage of the slowest cell to program becomes less than 4V, the fastest to program cell will not have a threshold voltage that is less than 3V. This has to be the case for millions of strings. A final verify operation that has to proceed word line by word line is performed to make sure no more than an acceptable number of cells per each page has a threshold voltage below 3V. This last operation will not have the same parallelism as the conventional NAND. In the rare event that this approach fails, the block has to be erased again, and soft programming has to be performed one word line at a time, and in the same manner as regular programming. Another approach for increasing the soft programming speed is to use a coarser soft programming step size, which will result in a wider soft programmed distribution.
The memory cells of
It should be noted that in flash memory chips, the convention has been to use the same floating gate oxide that is used between the floating gate and the channel for the gate oxide of low, and some medium voltage transistors in order to save extra process steps. Therefore the conventional tunnel oxide with a thickness that is usually greater than 8 nm has been limiting the performance, sub-threshold slope, and on-current drive of the low and some medium voltage transistors. This has resulted in slower program, and read characteristics. One advantage of the present invention is to provide a peripheral transistor gate oxide that is electrically and effectively much thinner than the conventional tunnel oxide, and is physically thicker than the conventional tunnel oxide. In other words, the peripheral circuitry will benefit from replacing the conventional tunnel oxide gate with high-K material(s) in alignment with the general trend of the semiconductor industry towards high-K materials.
Step 402 of
Step 408 of
In step 416 Chemical Mechanical Polishing (CMP), or another suitable process, is used to polish the material flat until reaching the floating gate poly-silicon. The floating gate is polished to 20 nm (10-100 nm in other embodiments). In step 418, the inter-poly tunnel dielectric (e.g. dielectric 34) is grown or deposited using ALD, CVD, PVD, Jet Vapor Deposition (JVD) or another suitable process.
In one embodiment, the inter-poly tunnel oxide layer can be created in the manner disclosed by “Resonant Fowler-Nordheim Tunneling through Layered Tunnel Barriers and its Possible Applications,” Alexander Korotkov and Konstantin Likharev, 1999 IEEE, 0-7803-5413-3/99 (hereinafter “Likharev I”); “Riding the Crest of a New Wave in Memory, NOVORAM: A new Concept for Fast, Bit-Addressable Nonvolatile Memory Based on Crested Barriers,” Konstantin and Likharev, Circuits and Devices, July 2000, p. 17 (hereinafter “Likharev II”); or U.S. Pat. No. 6,121,654 granted Sep. 19, 2000 titled: “Memory device having a crested tunnel barrier” all of which are incorporated herein by reference in their entirety. The oxide layer bottom of the conduction band energy diagram can be rounded near the mid-depth region of the tunnel dielectric, instead of forming a sharp triangle as in
In step 440 of
On top of the Tungsten layer, a hard mask of Si3N4 is deposited using, for example, CVD in step 446. In step 448, photolithography is used to create patterns of perpendicular strips to the NAND chain, in order to etch the multi-gate stack and form word lines (i.e. control gates) that are isolated from one another. In step 450, etching is performed using plasma etching, ion milling, ion etching that is purely physical etching, or another suitable process to etch the various layers and form the individual word lines. In one embodiment, the etching is performed until the high-k material is reached. The process attempts to leave as much high-K material as possible, but tries to etch completely through the floating gate material. In another embodiment, the process will etch all the way to the substrate.
In step 452, sidewall oxidation, sidewall oxide deposition, or a combination of the two is performed. For side wall oxidation, the device is placed in a furnace at a high temperature and some fractional percentage of ambient oxygen gas, so that the exposed surfaces oxidize, which provides a protection layer. Sidewall oxidation can also be used to round the edges of the floating gate and the control gate. An alternative to high temperature (e.g. over 1000 degrees Celsius) oxide growth is low temperature (e.g. 400 degrees Celsius) oxide growth in high density Krypton plasma. More information about sidewall oxidation can be found in “New Paradigm of Silicon Technology,” Ohmi, Kotani, Hirayama and Morimoto, Proceedings of the IEEE, Vol. 89, No. 3, March 2001; “Low-Tempetrature Growth of High Silicon Oxide Films by Oxygen Radical Generated in High Density Krypton Plasma,” Hirayama, Sekine, Saito and Ohmi, Dept. of Electronic Engineering, Tohoku University, Japan, 1999 IEEE; and “Highly Reliable Ultrathin Silicon Oxide Film Formation at Low Temperature by Oxygen Radical Generated in High-Density Krypton Plasma,” Sekine, Saito, Hirayama and Ohmi, Tohoku University, Japan, 2001 IEEE; all three of which are incorporated herein by reference in their entirety. Another way to deposit low temperature tunnel oxide may be by using Krypton Plasma, in conjunction with atomic layer deposition of Silicon Oxide or silicon Oxi-Nitride.
To achieve uniform tunneling a processing step may be employed in order to make the inter-gate tunnel dielectric thicker at the edges where the field lines may be more concentrated than near the middle. Oxidation may be a suitable way of achieving this end.
In step 454, an implant process is performed to create the N+ source/drain regions by Arsenic implantation. In one embodiment, a halo implant is also used. In step 456, an anneal process is performed. In one embodiment, a low temperature anneal process is performed to prevent damage to the high-K material. In some embodiments, a high-K material can be used that has a high thermal budget (e.g., able to endure high temperatures without degrading). In step 458, the process includes isotropically depositing and aniotropically etching sidewall material to form sidewall spacers.
There are many alternatives to the above described structures and processes within the spirit of the present invention. Textured gate (asperities) inter-gate tunneling is also possible, as well as Silicon-rich oxides, and graded band dielectrics. As in the existing NAND embodiment, an alternative is to fabricate the memory cells from PMOS devices with opposite polarity bias conditions for the various operations as compared to the existing NMOS implementation.
The low control gate coupling ratio will reduce the amount of floating gate charge needed to cause one volt of threshold shift as measured from the control gate as compared to existing NAND devices with its relatively high control gate coupling ratio. The benefit of this is lower programming/erase voltage levels, as compared to existing NAND. Alternatively, this advantage can be used to increase dielectric thicknesses, maintaining same program/erase voltages as in use today, but increasing overall cell reliability. Negative consequences of this are that effects of cell noise and electron charge gain or loss become amplified by the inverse of the control gate coupling ratio. These become manifest as larger shifts in threshold voltage for smaller values of control gate coupling ratio. In this respect, it is desirable not to have too small a control gate coupling ratio. A very small control gate coupling ratio will also limit the range of the amount of readable excess charge on the floating gate.
One embodiment would have a high temperature tolerant channel dielectric, such as Hafnium Silicate or Aluminum Oxide. A relatively thin poly-silicon floating gate, a suitable inter-gate dielectric, and a word line consisting of poly-silicon, covered by Tungsten Nitride, followed by Tungsten, constitute an embodiment that does not have to resort to a Damascene process. However, if poly-crystallization of amorphous-as-deposited silicon floating gate is to be avoided, then a low thermal budget process may have to be adopted that may include the Damascene process. An amorphous floating gate may offer a better quality tunneling oxide grown or deposited there upon.
Silicon Nitride has been proposed as a tunneling material for flash memories. A Damascene process can be employed to implant and anneal the source/drain junction of the memory array before the stack gates or some layers of the stack are deposited. Some materials such as Hafnium Oxide tend to crystallize at moderately high processing temperatures, which can lead to leakage currents at grain boundaries. To avoid crystallization a Damascene process avoiding such high temperature exposure post high-K dielectric deposition can be adopted.
The data stored in the memory cells are read out by the column control circuit 504 and are output to external I/O lines via data input/output buffer 512. Program data to be stored in the memory cells are input to the data input/output buffer 512 via the external I/O lines, and transferred to the column control circuit 504. The external I/O lines are connected to controller 518.
Command data for controlling the flash memory device is input to controller 518. The command data informs the flash memory of what operation is requested. The input command is transferred to state machine 516, which controls column control circuit 504, row control circuit 506, c-source control 510, p-well control circuit 508 and data input/output buffer 512. State machine 516 can also output status data of the flash memory such as READY/BUSY or PASS/FAIL.
Controller 518 is connected or connectable with a host system such as a personal computer, a digital camera, personal digital assistant, etc. Controller 518 communicates with the host in order to receive commands from the host, receive data from the host, provide data to the host and provide status information to the host. Controller 518 converts commands from the host into command signals that can be interpreted and executed by command circuits 514, which is in communication with state machine 516. Controller 518 typically contains buffer memory for the user data being written to or read from the memory array.
One exemplar memory system comprises one integrated circuit that includes controller 518, and one or more integrated circuit chips that each contain a memory array and associated control, input/output and state machine circuits. The trend is to integrate the memory arrays and controller circuits of a system together on one or more integrated circuit chips. The memory system may be embedded as part of the host system, or may be included in a memory card (or other package) that is removably inserted into the host systems. Such a removable card may include the entire memory system (e.g. including the controller) or just the memory chip(s) and associated peripheral circuits (with the Controller being embedded in the host). Thus, the controller can be embedded in the host or included within a removable memory system.
In some implementations, some of the components of
In one embodiment of the present invention, NAND type flash memory cells are used. The NAND cells are arranged with multiple transistors in series between two select gates. The transistors in series and the select gates are referred to as a NAND string. The discussion herein is not limited to any particular number of memory cells in a NAND string or NAND chain. Furthermore, the present invention is not limited to NAND flash memory cells. In other embodiments flash memory cells other than NAND cells (e.g. NOR cells or other cells) can be used to implement the present invention. In yet other embodiments, non-volatile memory cells other than flash memory cells can be used to implement the present invention.
Relevant examples of NAND type flash memories and their operation are provided in the following U.S. Patents/Patent Applications, all of which are incorporated herein by reference in their entirety: U.S. Pat. No. 5,570,315; U.S. Pat. No. 5,774,397; U.S. Pat. No. 6,046,935; U.S. Pat. No. 5,386,422; U.S. Pat. No. 6,456,528 and U.S. Pat. Application. Ser. No. 09/893,277 (Publication No. US2003/0002348). Information about programming NAND flash memory, including self boosting techniques, can be found in U.S. patent application Ser. No. 10/379,608, titled “Self Boosting Technique,” filed on Mar. 5, 2003; and in U.S. patent application Ser. No. 10/629,068, titled “Detecting Over Programmed Memory,” filed on Jul. 29, 2003, both applications are incorporated herein by reference in their entirety. Other types of flash memory devices can also be used with the present invention. For example, the following patents describe NOR type flash memories and are incorporated herein by reference in their entirety: U.S. Pat. Nos. 5,095,344; 5,172,338; 5,890,192 and 6,151,248. Another example of a flash memory type is found in U.S. Pat. No. 6,151,248, incorporated herein by reference in its entirety.
During read and programming operations, 4,256 memory cells are simultaneously selected. The memory cells selected have the same word line and the same kind of bit line (e.g. even bit lines or odd bit lines). Therefore, 532 bytes of data can be read or programmed simultaneously. In one embodiment, these 532 bytes of data that are simultaneously read or programmed form a logical page. Therefore, one block can store at least eight logical pages (four word lines, each with odd and even pages). When each memory cell stores two bits of data (e.g. a multi-level cell), one block stores 16 logical pages. Other sized blocks and pages can also be used with the present invention. Additionally, architectures other than that of
In the read and verify operations, the select gates (SGD and SGS) and the unselected word lines (e.g., WL0, WL1 and WL3) are raised to a read pass voltage (e.g. 4.5 volts) to make the transistors operate as pass gates. The selected word line (e.g. WL2) is connected to a voltage, a level of which is specified for each read and verify operation in order to determine whether a threshold voltage of the concerned memory cell has reached such level. For example, in a read operation for a two level memory cell, the selected word line WL2 may be grounded, so that it is detected whether the threshold voltage is higher than 0V. In a verify operation for a two level memory cell, the selected word line WL2 is connected to 2.4V, for example, so that it is verified whether the threshold voltage has reached at least 2.4V. For a multi-state memory cell, a read operation to distinguish between whether the memory cell is in a state corresponding to threshold distribution 306 or a state corresponding to threshold distribution 308 may include placing a voltage on the word line corresponding to a compare point between threshold distribution 306 and threshold distribution 308 (e.g., the mid point between threshold distribution 306 and threshold distribution 308). The source and p-well are at zero volts. The selected bit lines (BLe) are pre-charged to a level of, for example, 0.7V. If the threshold voltage is higher than the read or verify level on the word line, the potential level of the concerned bit line (BLe) maintains the high level because of the non-conductive memory cell. On the other hand, if the threshold voltage is lower than the read or verify level, the potential level of the concerned bit line (BLe) decreases to a low level by the end of sensing integration time, for example less than 0.3V, because of the conductive memory cell. The state of the memory cell is, thereby, detected by a sense amplifier that is connected to the bit line.
The erase, read and verify operations described above are performed according to techniques known in the art. Thus, many of the details explained can be varied by one skilled in the art. Other read and verify techniques known in the art can also be used.
In step 708, the second read operation is performed. A second read compare point (e.g. Vr2), equivalent to a threshold voltage between state 2 (e.g. threshold voltage distribution 308 of
In step 710, the third read operation is performed. A third read compare point (e.g. 0V), equivalent to a threshold voltage between state 3 and state 2 is applied to the selected word line, and the sense amplifier on each bit line makes a binary decision as to whether the cell at the intersection of the selected word line and the corresponding bit line is on or off. An “off” bit line will indicate that the corresponding cell is either in state 0, in state 1, or in state 2. An “on” bit line will indicate that the corresponding memory cell is in state 3. The information obtained during the three sequential steps explained above is stored in latches. A decoder is used to combine the results of the three read operations in order to find the state of each cell. For example, state 1 would be a result of the following three read results: on in step 706, off in step 708, and off in step 710. The above sequence of the read operations can be reversed, corresponding to the verify waveform sequence depicted in
The foregoing detailed description of the invention has been presented for purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise form disclosed. Many modifications and variations are possible in light of the above teaching. The described embodiments were chosen in order to best explain the principles of the invention and its practical application to thereby enable others skilled in the art to best utilize the invention in various embodiments and with various modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the claims appended hereto.