KR20090110806A - Floating gate memory device with interpoly charge trapping structure - Google Patents

Abstract

PURPOSE: A floating gate memory device with an interpoly charge trapping structure is provided to replace an interpoly dielectric layer by assembling a charge trapping device on a top part of a floating gate. CONSTITUTION: A memory cell(100) includes a semiconductor substrate(104), a multi stack, and a top conductive layer(101). A semiconductor substrate has a surface including a source region and a drain region separated by a channel region. The multi stack includes a first tunneling barrier structure(105), a floating gate, a second tunneling barrier structure, a charge trapping dielectric layer, and a top dielectric structure(107). The first tunneling barrier structure is arranged on the channel region on a surface of the substrate. The floating gate is arranged on the channel region of the first tunneling barrier structure. The second tunneling barrier structure is arranged on the floating gate.

Description

FLOATING GATE MEMORY DEVICE WITH INTERPOLY CHARGE TRAPPING STRUCTURE}

This application claims priority from US Provisional Patent Application No. 61 / 124,652 dated April 18, 2008.

The present invention is related to US Patent Application No. 11 / 756,559, “CHARGE TRAPPING DEVICES WITH FILED DISTRIBUTION LAYER OVER TUNNELING BARRIER,” filed May 31, 2007, which is incorporated herein by reference.

TECHNICAL FIELD The present invention generally relates to nonvolatile memory devices, and more particularly to flash memory and manufacturing methods of flash memory devices.

Flash memory technology includes a memory cell that stores charge between a channel and a gate of a field transistor. The stored charge affects the threshold of the transistor, and a change in the threshold due to the stored charge is sensed to represent data.

One well known type of charge storage memory cell is a floating gate memory cell. In a floating gate memory cell, a conductive material, such as a floating gate of polysilicon, is formed over a tunneling barrier structure, and an interpoly dielectric is formed over the floating gate to form a wordline of the memory cell. , Or it can be isolated from the control gate. The geometry of the floating gate is designed to establish a high coupling ratio for the voltage between the floating gate and the channel so that the voltage across the control gate leads to a stronger electric field across the tunneling barrier structure. For example, the floating gate is implemented using a T-shape, or a U-shape, which results in a larger surface area between the control gate and the floating gate than between the floating gate and the channel, and thus There will be greater capacitance between the controls. Although this technique has been widely successful, as the size of memory cells and the spacing between memory cells shrink, due to interference between neighboring floating gates, the performance of floating gate technology has begun to degrade.

Another type of memory cell based on storing charge between the channel and gate of a field effect transistor uses a dielectric charge trapping structure. In this type of memory cell, a dielectric charge trapping structure is formed over the tunneling barrier structure, wherein the tunneling barrier structure isolates the dielectric charge trapping structure from the channel. An upper dielectric layer is formed over the charge trapping structure to isolate it from the wordline, or gate. Representative devices include silicon-oxide-nitride-oxide-silicon (SONOS) cells.

In a memory cell using a dielectric charge trapping structure, the device may be planar because there is no coupling ratio design included in the design. As the minimum features for the fabrication process exceed about 45 nanometers, memory cells using dielectric charge trapping structures will overtake floating gate memory cells because of the planar structure and because there is little coupling between adjacent cells. It is foreseen.

SONOS-type memory cells suffer from performance degradation as the minimum feature size is less than about 45 nanometers. In particular, due to the fringing field, charges are unevenly injected along the channel width of the dielectric charge trapping structure, so that the region toward the center of the channel effectively cuts off the edge of the memory cell, compared to having a high threshold. The following area will effectively have a low threshold voltage. Low threshold regions along the edge can lead to performance degradation. US Patent Application No. 11 / 756,559, entitled "CHARGE TRAPPING DEVICES WITH FIELD DISTRIBUTION LAYER OVER TUNNELING BARRIER" (published US No. 2008-0116506), filed on May 31, 2007, having the present invention and the same applicant and the inventor. A combination of a floating gate structure, referred to as a field distribution layer, and a charge trapping layer between the gate and floating gate structure, is described.

Thus, even when there is a non-uniform charge concentration in the charge trapping structure along the channel width size, it is desirable to maintain a more constant threshold voltage along the channel width size.

The present invention relates to a nonvolatile memory device, and more particularly, to a nonvolatile memory device including a floating gate between a first tunneling barrier structure and a dielectric charge trapping structure, wherein the second tunneling barrier structure is a floating gate and a floating gate. In contact, the tunneling barrier is asymmetrical. To facilitate the migration of electrons in the floating gate to the dielectric charge trapping layer while preventing electron tunneling from the dielectric charge trapping layer through the floating gate to the substrate, for example by bandgap engineering and / Or by using different materials, or different thicknesses of materials, the second tunneling barrier structure can be engineered to have a different electron tunneling probability function than the first tunneling barrier structure. The floating gate distributes the electric field more uniformly across the channel, acting by a charge trapped in the dielectric charge trapping layer, even when there is a non-uniform charge concentration in the charge trapping structure along the channel width size. It results in a more consistent threshold low pressure below the conductive layer along the width size. The combination of floating gate and dielectric charge trapping structure enables the use of memory cell structures (eg, cells including planar floating gates) that reduce interference between neighboring devices in a dense array. In addition, the combination of the floating gate and the dielectric charge trapping structure together with the first and second tunneling barrier structures arranged such that a deep trap in the dielectric charge trapping layer retains most of the charge trapped by the combination. This results in improved data retention in high density flash memory.

Thus, the embodiment described herein includes a memory cell comprising a source region and a drain region isolated by a channel region. The first tunneling barrier structure is disposed above the channel region and has a thickness and a dielectric property to establish a tunneling probability function. The floating gate is disposed over the second tunneling barrier structure covering the channel region. The second tunneling barrier structure is disposed above the floating gate. The dielectric charge trapping structure is disposed over the second tunneling barrier structure, and the blocking dielectric structure is disposed over the charge trapping structure. The upper conductive layer disposed over the blocking dielectric structure acts as a gate. The second tunneling barrier structure tunnels the second tunneling barrier structure under bias conditions applied for programming and erasing the memory cell, according to an embodiment of the memory cell described herein, more efficient than the first tunneling barrier structure. It has thickness and dielectric properties that make it a conductor of current. That is, the second tunneling barrier structure has a higher tunneling probability than the first tunneling barrier structure under a given bias state. Also in some embodiments in the bias state applied during a read operation, the probability may be reversed such that under a given bias state the second tunneling barrier structure has a lower tunneling probability than the first tunneling barrier structure. In this way, the charge trapped in the memory cell is swept from the floating gate to the dielectric charge trapping layer during program or erase, where the charge is poor as the cell size and spacing between adjacent cells all decrease. It is more resistant to charge leakage which will result in retention.

Also described is an integrated circuit memory device comprising a cell implemented as mentioned above.

A method for manufacturing a memory cell of the present invention includes forming a first tunneling barrier structure on a surface of a semiconductor substrate, forming a floating gate layer on the tunneling barrier structure, and forming a floating gate layer on the surface of the floating gate layer. Forming a second tunneling barrier structure on the substrate, forming a charge trapping structure on the second tunneling barrier structure, forming an upper dielectric structure on the charge trapping structure, and forming an upper portion on the dielectric structure. Forming a conductive layer. As mentioned above, the second tunneling barrier structure is different from the first tunneling barrier structure. By implanting dopants into the semiconductor substrate, the source and drain regions can be formed such that the source and drain regions are isolated by the channel and the channel is located below the tunneling barrier structure.

In essence, a novel memory is described that is a “fusion” of floating gate (FG) and bandgap engineered BE-SONOS charge-trapping device. Unlike conventional flash memory structures, a charge-trapping device (BE-SONOS) can be assembled on top of the floating gate (FG) to replace the interpoly dielectric layer (IPD). Planar structures with ultra-thin poly FG (<5 nanometers) can be provided, which enables pitch scaling unlike conventional structures.

1 shows the basic structure of a prior art SONOS-type memory cell. The cell is formed over the semiconductor substrate 10 in which the first doped region 11 functions as a source terminal and the second doped region 12 functions as a drain terminal. A control gate 13 is formed over the charge trapping structure including the lower tunneling barrier structure 14, the dielectric charge trapping layer 15, and the upper dielectric 16. The channel of the memory cell is the region of the substrate 10 between the first doped region 11 and the second doped region 12. The magnitude L shown in FIG. 1 is commonly referred to as the channel length L because along the size of this channel a current flows between the source and the drain. The SONOS-type memory cell shown in FIG. 1 is often configured in a NAND array configuration, where a column of arrays comprises a set of memory cells arranged in series between ground and a global bitline contact. Include.

2 shows a cross-sectional view taken in parallel with the word line 13 in a NAND array configuration along the channel width size of a basic prior art SONOS-type memory cell. Source and drain terminals are disposed above and below the plane of the figure of FIG. Individual columns of memory cells are isolated by an isolation structure, such as shallow trench isolation (STI) structure 20. In this way, the columns of memory cells can be arranged in a dense array separated by the width of the STI structure 20, where the width of the STI structure is the minimum feature size of the technology used to manufacture the device ( F) level. Similarly, the channel width W may be at the level of the minimum feature size F for the NAND array configuration. In FIG. 2, it is shown that electrical force lines (eg, electrical force lines 21 and 22) passing through a dielectric layer (eg, control gate 13) between the channel and the wordline are located on the edge of the channel. Electric force lines 21 and 22 represent fringe fields that reduce the effect of charge trapping at the edge of charge trapping layer 15. In the embodiment shown in FIG. 2, the EOT (typically 20 nanometer level) of the combination of the lower tunneling barrier structure 14, the charge trapping layer 15, and the upper dielectric 16 is greater than the channel width W. Certainly small, the fringing field effect does not interfere with the operation of the device. The EOT is defined to be equal to the thickness of the dielectric layer scaled by the ratio of the dielectric constant of silicon dioxide to the dielectric constant of the material of the layer.

It is known that symmetric SONOS-type memory cells may suffer from performance degradation as the minimum feature size becomes less than about 45 nanometers. For example, FIG. 3 shows a structure similar to that of FIG. 2, where the channel width W is in the EOT of the combination of the lower dielectric 54 and the charge trapping layer 55 and the upper dielectric 56. It shrinks down to compare. In this embodiment, the memory cell includes a polysilicon wordline 57, where the columns of cells are isolated by the STI structure 60. In this embodiment, the electric field lines 61 and 62 representing the fringing field can sufficiently affect the effect of the charge trapping layer 55. In particular, the non-uniform injection of charge into the charge trapping layer along the channel width due to the fringing field allows the region along the center of the channel to have a high threshold while the region along the edge of the channel is effectively This can result in low thresholds.

4 illustrates an asymmetrical charge trapping distribution in a charge trapping layer along a channel along a channel width size of a prior art memory cell. As shown in FIG. 4, the concentration of charge in the charge trapping layer on the left side of the channel is low compared to the concentration near the center of the channel. Also, on the right side of the channel, the charge concentration of the charge trapping layer is low compared to the concentration near the center of the channel. FIG. 5 shows that the uneven charge trapping distribution shown in FIG. 4 results in a distribution of effective threshold voltages of the memory cell along the channel width magnitude. Thus, a memory cell programmed to a high threshold state may still have a low threshold region along the edge of the cell. FIG. 6 shows the drain current (Id) versus gate voltage (Vg) I-V characteristics through the channel of a SONOS-type cell undergoing an asymmetrical charge distribution caused by the fringing effect. Trace 50 on the left shows desirable I-V characteristics for an unprogrammed “fresh” cell. As programming proceeds and the charge trapped in the charge trapping layer increases, traces 51, 52, and 53 indicate, in particular, that the I-V characteristic is degraded in the subthreshold region. Since the charge cannot be trapped at the edge of the charge trapping structure, the current below the threshold is pinned, which is indicated by the dotted oval in the figure.

FIG. 7 shows a cross-sectional view taken along the channel length size L of a dielectric charge trapping memory cell 100 including a conductive layer 101 over a tunneling barrier structure 105, in accordance with one embodiment of the present invention. . In the embodiment shown in FIG. 7, memory cell 100 includes a substrate 104 having a doped region 102 and a doped region 103, each functioning as a source and a drain, wherein doped region 102 is employed. And doped region 103 are separated by a channel. In the embodiment shown in FIG. 7, a tunneling barrier structure 105 is disposed over a channel located on the surface of the substrate 104. In this embodiment, the tunneling barrier structure 105 is a single dielectric layer. In the embodiment shown in FIG. 7, the memory cell 100 includes a conductive layer 101 disposed over the tunneling barrier structure 105, a charge trapping structure 106 disposed over the conductive layer 101, and And an upper dielectric structure 107 disposed over the charge trapping structure 106 and an upper conductive layer 108 disposed over the upper dielectric structure 107. In certain embodiments, the tunneling barrier structure 105 may comprise silicon dioxide, or silicon oxynitride. In a particular embodiment, the tunneling barrier structure 105 includes silicon dioxide having a thickness on the order of 4-6 nanometers. In certain embodiments, charge trapping structure 106 includes silicon nitride, nano-particle embedded dielectrics, or “high-k” metal oxides such as Al ₂ O ₃ , Hf ₂ O _3, and the like. It includes other substances. In certain embodiments, charge trapping structure 106 includes silicon nitride having a thickness of 5 to 7 nanometers. In certain embodiments, upper dielectric structure 107 includes silicon dioxide, or other dielectric material (eg, “high-k” metal oxides such as Al ₂ O ₃ , Hf ₂ O _3, etc.). In an embodiment, the upper dielectric structure 107 includes silicon dioxide having a thickness on the order of 5 to 9 nanometers. In addition, in certain embodiments, the conductive layer 101 is p-type polysilicon and n- Type polysilicon and other doped semiconductor materials, or metals, such as aluminum, copper, or tungsten, In an exemplary embodiment, the conductive layer 101 is about 2-6 nanometers. And the doped polysilicon having a thickness, so that the interference caused by the electric field between the conductive layers of neighboring cells is small, so that the conductive layer 101 is thin, so as not to interfere with the performance of the memory cell. Above dispersion It may be thick enough for stable formation of the provided layer In certain embodiments, the top conductive layer 108 may be p-type polysilicon, n-type polysilicon, other doped semiconductor materials, or metals (aluminum). Or copper, or tungsten) The materials selected in this exemplary embodiment are representative materials that are readily manufactured, and various other materials and combinations thereof may also be used for the memory cell layers and structures. have.

FIG. 8 illustrates a cross-sectional view of an array of memory cells similar to that shown in FIG. 7 taken along a channel width size in accordance with one embodiment of the present invention. In the embodiment shown in FIG. 8, memory cells are isolated by trench isolation structure 110. In the embodiment shown in FIG. 8, the channel width W is comparable to the tunneling barrier structure 105 and the charge trapping structure 106 of the memory cell and the effective oxide thickness (EOT) of the upper dielectric structure 107. Size. Since the conductive layer 101 is a conductive layer rather than a dielectric, it does not affect the EOT of the memory cell. One characteristic of the conductive layer 101 is that it is not subject to design for inducing large coupling ratios, such as floating gate memory cells. Rather, the ratio of the area of the conductive layer 101 disposed over the channel to the area of the channel is the ratio of the area of the upper conductive layer 108 disposed over the channel to the area of the conductive layer 101 disposed over the channel. It may be equal to the ratio sufficiently. In this way, the electric field located above the conductive layer 101 is sufficiently equal to the electric field below the conductive layer 101. Even if some electrons are trapped in the conductive layer 101, the large electric field applied during programming will immediately wipe out most or all of the electrons into the charge trapping structure 106.

As shown in FIG. 8, the fringing fields 111, 112, and 113 are terminated by the constant potential of the conductive layer 101. Thus, the fringing field effect in the charge trapping structure 106 is reduced. In addition, as shown in FIG. 8, as long as non-uniform dispersion of charge occurs in the memory cell, the electrostatic potential of the conductive layer 101 uniformly distributes the electric field across the tunneling barrier structure 105, and in the channel The effective dispersion of the threshold voltage of will make the channel width more uniform.

FIG. 9 shows an example of charge concentration across the channel width size of a dielectric charge trapping memory cell with a conductive layer, as shown in FIG. 8. In this example, as shown in FIG. 8, the dispersion of charge concentration for the memory cell is similar to that of a conventional SONOS-type memory cell as mentioned above. FIG. 10 shows that the conductive layer 101 affects the dispersion of the threshold voltage VT in the channel width magnitude. As shown in FIG. 10, the potential of the conductive layer 101 leads to a symmetrical dispersion of the threshold voltage in the channel width magnitude. Thus, even in the asymmetrical dispersion of charge in the charge trapping structure 106, the performance of the memory cell does not substantially degrade.

FIG. 11 illustrates IV characteristics of the drain current (Id) versus gate voltage (Vg) of a dielectric charge trapping memory cell including a conductive layer undergoing asymmetrical charge dispersion in a charge trapping structure, in accordance with one embodiment of the present invention. Show the simulation. Trace 80 on the left shows the desired I-V characteristics for an unprogrammed “fresh” cell. As programming proceeds, and as the charge trapped in the charge trapping structure increases, traces 81 and 82 show that the I-V characteristics do not degrade. As the threshold voltage increases, the behavior of the sub-threshold current (value shift below the threshold) remains constant.

12 is a cross-sectional view taken along the wordline 200 of a conventional floating gate device. For example, a channel of a memory cell is formed on the semiconductor body 201 extending in a line perpendicular to the page in the NAND string. Each line in the semiconductor body is isolated by dielectric trench 202 using techniques such as shallow trench isolation (STI). The tunneling barrier structure 203 is formed on the semiconductor body 201. Polysilicon floating gate 204 is formed on tunneling barrier structure 203. An interpoly dielectric, in this example for example silicon oxide layer 205, silicon nitride layer 206 and silicon oxide layer 207 is formed over floating gate polysilicon 204. Interpoly dielectrics 205-207 are designed to block charge leakage between wordline 200 and floating gate 204. In addition, to provide a large surface area for coupling between floating gate 204 and wordline 200, the floating gate 204 should be relatively thick (typically 100 nanometers or more in current technology). This large surface area increases the gate coupling ratio of the floating gate device, which transfers a larger voltage from wordline 200 to floating gate 204 during programming and erasing. However, the thick floating gate device results in severe interference between floating gates in adjacent lines. In the figure, the electrons appear to be scattered around the surface of the floating gate on the left side of the feature. Similar electrons in the floating gate 204 on the right side create an undesirable electric field between the floating gates and de-trapping to oxide traps or surrounding defects in, for example, STIs. will lead to charge leakage through -trapping). This problem of interference between adjacent cells is a significant concern for the research and implementation of SONOS-type dielectric charge trapping devices in which charge is trapped in a deep trap and less interference occurs between adjacent cells resulting in charge leakage. I have been motivated.

In addition, as fabrication techniques for memory cells enable 30 nanometer devices, the number of electrons stored in a given cell becomes very small. For example, less than 100 electrons will be used to establish the memory state of a small memory cell. As the number of electrons establishing a memory state becomes smaller, other forms of interference and charge leakage between adjacent cells become increasingly important in the design of memory cell structures.

FIG. 13 is a cross sectional view taken along the wordline 20 of a prior art floating gate device, illustrating the design trend for improving the floating gate device. In the structure shown in FIG. 13, channels of memory cells are formed on the semiconductor body 211. Each line of the semiconductor body is isolated by a dielectric trench 212. The tunneling barrier structure 213 is formed on the semiconductor body 211. Floating gate 214 is formed on tunneling barrier structure 213. A planar or nearly planar interpoly dielectric, for example, silicon oxide layer 215, silicon nitride layer 216 and silicon oxide layer 217 in this embodiment is a floating gate along wordline 210. It extends over the upper surface of the structure. One problem with the planar, or nearly planar, devices shown in FIG. 13 is that they have a very low gate coupling ratio (GCR). To improve the gate coupling ratio, designers propose to use a non-traditional interpoly dielectric structure to improve the capacitance of the floating gate to the wordline interface. For example, interpoly dielectric structures (eg, 215-217) can be replaced with high-k dielectric materials. For example, aluminum oxide, or other materials have been proposed. Alternative high-k interpoly dielectric structures may include silicon oxide as the bottom buffer layer, or include a multilayer stack that includes silicon oxide as both the bottom buffer layer and the top buffer layer for high-k dielectric materials such as aluminum oxide. Can be.

The problem with planar floating gate devices as shown in FIG. 13 is that due to the presence of large electric fields during programming, charge is easily injected into the interpoly dielectric. Thus the interpoly dielectric easily traps charge. However, because it is very difficult to remove the charge on the left side of the interpoly dielectric, this makes the device difficult to erase and results in impractical results for many applications of flash memory devices.

14 is a cross-sectional view of a charge trapping / floating gate memory device that overcomes many of the problems presented by prior art flash memory designs. The cross section is taken along the word line 310. In the structure shown in FIG. 14, channels of memory cells are formed over the semiconductor body 311. For example, using implanted dopants, or inversion regions, the source and drain terminals are disposed on opposite sides of the word line. Each line in the semiconductor body is isolated by a dielectric trench 312. A first tunneling barrier structure 313 is formed on the semiconductor body 311. A floating gate 314 is formed over the first tunneling barrier structure 313, where a dielectric fill provides a planar or nearly planar structure that exposes the top surface of the floating gate 314. On top of the floating gate 314, a charge trapping structure is formed. The charge trapping structure includes a second tunneling barrier structure 315, a charge trapping layer 319, and a blocking dielectric layer 320. The second tunneling barrier structure 315 in the illustrated example is a layer of silicon dioxide 316, preferably less than about 2 nanometers thick, and a silicon nitride 317, preferably less than about 3 nanometers thick. And a multi-layer stack of bandgap barrier engineered materials comprising a layer and a layer of silicon oxide that is preferably less than about 3.5 nanometers thick. The word line 310 is formed over the blocking dielectric layer 320. In this way, the floating gate device (basically region 321) is capped by the charge trapping device (basically region 322).

The structure features a first tunneling barrier structure 313 that provides a larger tunneling barrier than that provided by the second tunneling barrier structure 315 in a bias state applied for program or erase. Thus, the first tunneling barrier structure 313 and the second tunneling barrier structure 315 together pass through the floating gate 314 from the charge trapping layer 319 under a bias condition applied for reading, thereby forming a semiconductor body. Under the positive gate bias condition applied to increase the threshold voltage while preventing electron tunneling to 311, electrons in the floating gate 314 move from the body, through the floating gate, to the charge trapping layer 319. It provides a means to. In this way, during programming, electrons are tunneled through the first tunneling barrier structure 313 to the floating gate, and because of the lower barrier height during programming, through the second tunneling barrier structure 315, a charge trapping layer ( 319), where the electrons are trapped in a relatively deep trap of dielectric material. By managing the relative barrier effects of the first and second tunneling structures 313, 315, the cell is designed such that in practice only a few electrons are stored in the floating gate 314.

As can be seen, using bandgap engineering (BE), the interpoly dielectric of a floating gate memory device is replaced with a dielectric charge trapping structure, resulting in a tunneling barrier with greater efficiency than the tunneling barrier structure between the substrate and the floating gate. Can be provided. Representative barrier engineering approaches include so-called crested barrier layers and layers with “U-shaped” barriers such as the above mentioned ONO structure. The charge trapping layer has a very high trapping efficiency and it is desirable that the desired data retention is provided by deep traps. Silicon nitride layers thicker than about 5 nanometers are representative examples of providing this property. In another embodiment, nano-particle embedded dielectric, or other material including “high-k” metal oxides such as Al ₂ O ₃ , Hf ₂ O _3, and the like. It can replace this silicon nitride.

The blocking dielectric layer 320 should have a low degree of leakage such as provided by silicon oxide and silicon oxynitride. At the boundary with the word line (control gate), the upper dielectric should have a high barrier height that suppresses gate injection. In addition, the use of a high-k layer (eg, aluminum oxide on top of a silicon oxide buffer layer) in the upper portion of the blocking dielectric layer 320 over a buffer oxide layer (eg, aluminum oxide on top of a silicon oxide buffer layer) is blocked. By reducing the electric field in layer 320, gate injection is further suppressed. The floating gate layer may be relatively thin, for example, 20 nanometers or less. Typical embodiments use n + doped polysilicon. Undoped polysilicon and p + doped polysilicon may also be used. It is desirable that the lower tunnel oxide layer, which serves as a tunneling barrier structure between the substrate and the floating gate, has a very low degree of leakage. Silicon dioxide thus becomes a preferred lower tunneling barrier structure material having a thickness of 5 to 7 nanometers.

In an exemplary embodiment, the wordline 310 serving as the gate of the memory cell includes p + polysilicon (work function is about 5.1 eV). N + polysilicon may also be used. Other embodiments include metals, or metal compounds, or combinations thereof, such as platinum, tantalum nitride, metal silicides, aluminum, or other metal or metal compound gate materials, such as Ti, TiN, relative to the gate. , Ta, Ru, Ir, RuO ₂ , IrO ₂ , W, WN, etc.) are used. In some applications, it is desirable to use materials having a work function of at least 4 eV, preferably 4.5 eV. Various high work function materials suitable for use as gate terminals are described in US Pat. No. 6,912,163. Typically such materials are deposited using sputtering and physical vapor deposition techniques and can be patterned using reactive ion etching.

In the embodiment of FIG. 14, the first tunneling barrier structure 313 is an ISSG comprising selective nitriding, for example by post deposition NO anneal or by addition of NO to the atmosphere during deposition. silicon oxide formed using in-situ steam generation. The thickness of the first tunneling barrier structure 313 of silicon dioxide is 70 GPa or less, about 40 GPa or more, and in a representative embodiment, about 50 GPa. Using a conventional floating gate polysilicon process, floating gate 314 is formed with a reduced thickness.

In the embodiment shown in FIG. 14, the second tunneling barrier structure 315 located above the floating gate 314 comprises a composite of material, such as by post deposition NO anneal, or deposition. First layer 316 (hole tunneling) of silicon dioxide located on the top surface of floating gate 314 formed using in-situ steam generation (ISSG) including selective nitriding by the addition of NO to the atmosphere during Layer, referred to as a hole tunneling layer. The thickness of the first layer 316 of silicon dioxide is 20 kPa or less, preferably 15 kPa or less. Exemplary embodiments have a thickness of 10 ms, or 12 ms.

Silicon nitride layer 317, referred to as a band offset layer, is a low-pressure chemical vapor deposition (LPCVD) using dichlorosilane (DCS) and NH ₃ precursor, for example, at 680 ° C. Is positioned over the first layer 316 of silicon oxide formed using &lt; RTI ID = 0.0 &gt; In another process, the band offset layer includes silicon oxynitride made using a similar process using an N 2 O precursor. The thickness of the layer 317 of silicon nitride is 30 kPa or less, preferably 25 kPa or less.

A layer 318 of silicon dioxide, referred to as an isolation layer, overlies a layer 317 of silicon nitride formed using, for example, LPCVD hot oxide HTO deposition. The thickness of the second layer 318 of silicon dioxide is 35 kPa or less, preferably 25 kPa or less. The valence band energy level at the first location is such that an electric field sufficient to induce hole tunneling through the thin region of the boundary between the semiconductor body and the first location is at the valence band energy level behind the first location. It is sufficient to elevate to the level that effectively removes the hole tunneling barrier in the engineered tunneling barrier structure behind the location. With this structure having an inverted “U-shaped” valence band, only a smaller electric field induced in the absence of an electric field or for other operations (eg, reading data from a cell, or neighboring cell programming) When present, hole tunneling assisted by an electric field is enabled at high speed, while effectively preventing charge leakage through the engineered tunneling barrier structure.

Thus, in a representative device, the engineered tunneling barrier structure 315 may comprise an ultrathin silicon oxide layer O1 (eg, 18 kV or less), an ultrathin silicon nitride layer N1 (eg, 30 kV or less), and an ultrathin silicon oxide layer O ₂ (E.g., 35 mA or less), which leads to an increase in the valence band energy level of about 2.6 eV at an offset of 15 ms or less from the boundary with the semiconductor body. The O ₂ layer is N1 from the charge trapping layer at a second offset (eg, about 30 kPa to 45 kPa from the boundary), by regions of lower valence band energy level (higher hole tunneling barrier) and higher conduction band energy level. Isolate the layer. An electric field sufficient to induce hole tunneling raises the valence band energy level after the second location to a level that effectively eliminates the hole tunneling barrier, since the second location is located longer from the interface. The O2 layer thus does not substantially interfere with electric field assisted hole tunneling, or electric field assisted electron tunneling, while improving the ability of the engineered tunneling barrier structure to block leakage during low electric fields.

Details of the structure of the engineered tunneling barrier layer will now be described in more detail with reference to FIGS. 14A and 14B.

In this embodiment, the charge trapping layer 319 comprises silicon nitride having a thickness of about 50 microns or more, for example 70 microns in some embodiments, formed using LPCVD, for example. Other trapping materials and structures can be used, such as silicon oxynitride (Si _x O _y N _z ), silicon-rich nitride, silicon-rich oxide, trapping layers with nano-particles, and the like. . Various charge trapping materials are described in the aforementioned US Patent Application Publication No. 2006/0261401 A1 “Novel Low Power Non-Volatile Memory and Gate Stack” (Bhattacharyya, published November 23, 2006).

The blocking dielectric layer 320 in this embodiment is silicon dioxide and may be formed by, for example, wet transition of nitride by a wet furnace oxidation process. Other embodiments may be implemented using high temperature oxide (HTO), or LPCVD SiO ₂ . The thickness of the blocking dielectric layer may be at least about 50 mm 3, for example 90 mm in some embodiments.

FIG. 14A shows the energy levels of conduction bands and valence bands in a dielectric tunneling structure including a stack of layers 316-318 of FIG. 14 under a low electric field showing “U-shaped” conduction bands and “inverse U-shaped” valence bands. Is a diagram of which shows the probability function for tunneling under low bias conditions that may be encountered during a read operation. From the right, the bandgap for the semiconductor body is shown as region 30, the bandgap for the offset layer is shown as region 32, the valence and conduction bands for the isolated layer appear in region 33, and charge trapping. The valence and conduction bands for the layer are shown in region 34. Electrons confined within the charge trapping region 34, represented by a circle with a minus sign (-), cannot tunnel through the conduction band in the channel, because in all three regions 31, 32, 33, This is because the conduction bands remain high relative to the trap's energy level. The probability of electron tunneling correlates with the area under the “U-shaped” conduction band of the tunneling barrier structure and above the horizontal line at the energy level of the trap for the channel. Thus, electron tunneling rarely occurs in a low electric field state, and the illustrated barrier structure is one embodiment, and the illustrated barrier structure is combined with the tunneling barrier structure between the floating gate and the body to be applied for reading. It is possible to effectively prevent electron tunneling from the dielectric charge trapping layer through the floating gate to the semiconductor body. Similarly, tunneling holes in the valence band of the channel in region 30 to the charge trapping layer (region 34) is due to the overall thickness of the regions 31, 32 and 33 and the high hole tunneling barrier height at the channel boundary. , Blocked. The probability of hole tunneling is correlated with the area above the “inverse U-shaped” valence band and below the horizontal line at the energy level of the channel for the charge trapping layer. Thus, hole tunneling is rarely present in low electric field conditions.

In an exemplary embodiment in which the hole tunneling layer comprises silicon dioxide, a hole tunneling barrier height of about 4.5 eV prevents hole tunneling. The valence band of silicon nitride remains below 1.9 eV, the work function of the valence band in the channel. Thus, substantially the valence bands in all three regions 31, 32, 33 of the tunneling barrier structure remain below the valence band in the channel region 30. Thus, the tunneling barrier structure described herein has a relatively large hole tunneling barrier height at a thin layer (region 31) at the boundary with the semiconductor body, and valence band energy level at the first location that is less than 2 nanometers from the channel surface. It is characterized by the band offset characteristic, such as the increase (37). The mand offset characteristic also includes a reduction 38 of valence band energy levels at the second location away from the channel, as a result of providing a thin layer of material (region 33) having a relatively high tunneling barrier height, This leads to an inverted U-shaped valence band form. Likewise, the conduction band derived by selecting the same material has a U-shape.

FIG. 14B shows a band diagram for a dielectric tunneling structure under an electric field condition of about −12 MV / cm in the tunneling region 31 to induce hole tunneling (in FIG. 14B, the O1 layer has a thickness of about 15 μs). ). Under the electric field, valence bands form an uphill from the channel surface. Thus, at offset intervals from the channel surface, the band energy level of the valence band of the tunneling barrier structure is increased, and in the figure, rises above the band energy of the valence band of the channel region. Thus, as the area between the level of valence bands in the channel and the inverted U-shaped valence bands of the tunneling stack (darkened areas in Fig. 14B) is reduced, the hole tunneling probability increases. The band offset blocks the offset layer in the region 32 and the isolation layer in the region 33 from the tunneling barrier structure during the high electric field, while allowing a large hole tunneling current under a relatively small electric field (eg, E <14 MV / cm). Effectively removes functionality.

Isolation layer (region 33) isolates the offset layer (region 32) from the charge trapping layer (region 34). This increases the effective blocking ability during low electric fields, for both electrons and holes, thereby improving charge retention.

The offset layer (region 32) of this embodiment is thin enough with negligible charge trapping efficiency. The offset layer is also dielectric, not conductive. Thus, for embodiments using silicon nitride, the offset layer should be 30 angstroms thick or less, more preferably about 25 kPa or less.

In one embodiment using silicon dioxide, the hole tunneling region 31 should be 20 mu m or less in thickness, more preferably 15 mu m or less in thickness. For example, in a preferred embodiment, the hole tunneling region 31 is about 13 GPa, or 10 GPa thick silicon dioxide, exposed to the nitriding process as mentioned above, resulting in ultrathin silicon oxynitride To derive

Tunneling barrier structure 315 located above floating gate 314 may be implemented in an embodiment of the present invention that utilizes a composite of silicon oxide, silicon oxynitride, and silicon nitride, the inverse U- of which the composite is required. As long as the valence band is derived, there is no elaborate movement between layers, and there is a change in the energy level of the valence band at offset intervals from the channel surface required for efficient hole tunneling. Combinations of other materials may also be used to provide band offset techniques.

The description of the dielectric tunneling barrier structure 315 focuses on “hole tunneling” rather than electron tunneling because the technology solves the problem associated with the need to rely on hole tunneling in SONOS type memory. For example, a tunneling barrier structure composed of silicon dioxide thin enough to support practical speed hole tunneling would be too thin to block leakage by electron tunneling. However, the effect of engineering improves the performance of electron tunneling. Thus, using barrier engineering, programming by electron tunneling and cancellation by hole tunneling are sufficiently improved. The barrier structure in combination with the tunneling barrier structure between the floating gate and the body, shown in one embodiment, tunnels electrons from the body through the floating gate to the dielectric charge trapping layer under a positive gate bias condition applied for programming. Can be effectively enabled.

15 is a cross-sectional view of the charge trapping / floating gate structure, as in the structure of FIG. 14, the charge trapping structure isolates the dielectric charge trapping structure for adjacent cells in the wordline direction, resulting in a very dense array between adjacent cells. The patterning is done to eliminate the possibility of charge transfer at. 15 is taken along the word line 330. In the structure shown in FIG. 15, channels of memory cells are formed on the semiconductor body 331. Each line of semiconductor body is isolated from each other by dielectric trench 332. A first tunneling barrier 333 is formed on the semiconductor body 331, where the dielectric fill provides a planar or nearly planar structure exposing the top surface of the floating gate 334. On top of the floating gate 334, a charge trapping structure is formed. The charge trapping structure includes a second tunneling barrier 335, a charge trapping layer 339, and a blocking dielectric layer 340. The second tunneling barrier 335 in the illustrated example comprises a multilayer stack of barrier-engineered material, which stack preferably comprises a layer 336 of silicon dioxide having a thickness of less than about 2 nanometers, Preferably a layer 337 of silicon nitride having a thickness of less than about 3 nanometers, and preferably a layer 338 of silicon oxide having a thickness of less than about 3.5 nanometers. Wordline 330 is formed over blocking dielectric layer 340. The dielectric stack forming the charge trapping structure is patterned along both the wordline direction and a direction perpendicular to the wordline direction, providing an isolated charge trapping island above each floating gate 334. Can be. In the embodiment of FIG. 15, by isolating the charge trapping structure in both the wordline direction and the bitline direction, reduction of the lateral movement of any possible charge in the dielectric charge trapping layer during high-temperature storage may be assisted. Can be.

In both FIGS. 14 and 15 embodiments, the charge trapping / floating gate structure provides higher reliability and more desirable erase performance, while at the same time having a planar structure in a similar manner as a planar floating gate device. The interpoly dielectric of a standard floating gate device is designed to be a non-trapping interpoly dielectric. The interpoly dielectric of the standard floating gate device in the structures of FIGS. 14 and 15 is replaced with a charge trapping device that is tuned to operate to store the majority of the charge in the memory cell that establishes a memory state.

The preferred charge trapping structure is based on the bandgap engineered SONOS device described in US Patent Application Publication No. 2007/0268753, BE-SONOS, which provides a very efficient tunneling barrier structure, which is relatively thick It provides a weaker barrier to tunneling during programming and erase bias states than the first tunneling barrier structures 313 and 333 of silicon dioxide do. Since most of the injected charge, in the form of holes or electrons, is swept from the floating gate to the deep traps of the dielectric charge trapping layers 319 and 339, even at high thresholds, the floating gate remains almost unchanged. That is, it remains in a neutral state.

In an exemplary embodiment, the first tunneling barrier structure (313 in FIG. 14) includes a silicon dioxide layer 5-7 nanometers thick. This is relatively thick compared to the effective thickness that contributes to the barrier height during the program and erase bias states of the bandgap engineered tunneling barrier layer (or other embodiments) of the second tunneling barrier structure (315 of FIG. 14). However, in a typical floating gate device, it is common for the tunneling dielectric to have a thickness of at least 7 nanometers because of the increased probability of leakage due to charge storage in the floating gate.

In one example, for a BE-SONOS-type charge trapping structure over a floating gate, the silicon dioxide layer 316 has a thickness of about 13 microseconds, the silicon nitride layer 317 has a thickness of about 20 microseconds, and the silicon oxide layer ( 318 has a thickness of about 25 microseconds, charge trapping dielectric layer 319 may be about 50 microseconds thick silicon nitride, and blocking dielectric layer 320 may be about 50 microseconds thick silicon dioxide. However, the thickness of the charge trapping dielectric layer 319 may be 70 kΩ, or more. In addition, the thickness of the blocking dielectric layer 320 of the silicon dioxide embodiment may be 70 GPa or more. The overall thickness of the stack is one parameter that determines the operating voltage. Thus, the larger the overall thickness, the greater the operating voltage is needed.

The word line 310 is typically of polysilicon structure. In a preferred embodiment, higher work function materials, such as P + doped polysilicon, are preferred to suppress gate injection during the erase operation. Higher work function materials may be used at the boundary between the polysilicon wordline and the charge trapping structure, or at the location of the polysilicon wordline. Such materials include TaN, WN, Pt.

The results of the simulation of the program and erase performance of the memory cell described with reference to FIG. 14 are shown in FIGS. 16-19A and 19B. In the simulated cell, the second tunneling barrier layer included a 13 micron thick silicon dioxide layer 316, a 20 micron thick silicon nitride layer 317, and a 25 micron thick silicon oxide layer 318. The charge trapping dielectric layer 319 was 50 nm thick silicon nitride and the blocking dielectric layer 320 was 50 mm thick silicon dioxide. The first tunneling barrier structure was 50 microns thick silicon dioxide. The floating gate is polysilicon in the layer that can be as thin as 20 microseconds, or less. Thicker polysilicon in the range of 100 to 1000 microns provides the benefits described herein. However, in order to produce very dense arrays, it is preferred that the polysilicon layer has a thickness of 100 kPa or less.

FIG. 16 is a graph showing the change in threshold voltage over time under programming bias from gate to substrate of a cell to induce Fowler-Nordheim tunneling, where the bias voltage is + 21V on trace 400 and the bias voltage is + 20V on trace 401, bias voltage is + 19V on trace 402, and bias voltage is + 18V on trace 403. Thus, the memory cell can be programmed at the appropriate time and with the appropriate programming bias voltage.

FIG. 17 shows the calculated bars of electrons (Q _trap ) trapped inside the dielectric charge trapping layer over time, under a programming bias of + 21V, as trace 404 and electrons (Q _FG) trapped inside the floating gate. Is a graph showing the calculated bar of. The simulation shows that the charge trapped in the dielectric charge trapping layer is much more than the charge trapped in the floating gate. This occurs because the tunneling efficiency of the bandgap engineered tunneling barrier layer between the floating gate and the dielectric charge trapping layer is much better than the tunneling efficiency of the tunneling barrier layer between the substrate and the floating gate.

FIG. 18 is a graph showing the change in threshold voltage over time under an erase bias from the gate to the substrate of a cell to induce Fowler-Nordheim tunneling, where the bias voltage is -21 V on trace 410 and the bias voltage Is -20V on trace 411, the bias voltage is -19V on trace 412, and the bias voltage is -18V on trace 413. Therefore, the memory cell can be erased at an appropriate time and with an appropriate erase bias voltage. The programming operation is proportional to the programming potential, where the slope of incremental step pulse programming (ISPP) is approximately one.

FIG. 19 shows a calculated bar of electrons (Q _trap ) trapped inside the dielectric charge trapping layer over time, under an erase bias of −21 V, as trace 415, and of electrons Q _FG trapped inside the floating gate. It is a graph showing the calculated bar as the trace 414. The simulation shows that the charge trapped in the dielectric charge trapping layer is quickly removed and hole trapping in the dielectric charge trapping layer is more desirable than hole trapping in the floating gate. This occurs because the tunneling efficiency of the bandgap engineered tunneling barrier layer between the floating gate and the dielectric charge trapping layer is more desirable than the tunneling efficiency of the tunneling barrier structure between the substrate and the floating gate. The simulation shows the erase saturation condition as the gate injection from the polysilicon wordline occurs after a longer erase time.

The simulation shows that the charge trapped in the dielectric charge trapping layer is quickly removed and that hole trapping in the dielectric charge trapping layer is much more desirable than hole trapping in the floating gate.

As shown by the simulation represented by FIGS. 17-19, the charge trapping / floating gate memory cells described herein provide a new operating condition different from prior art devices. The floating gate establishes an equal potential region over the channel and controls the threshold distribution across the channel. Although the trapped charge may be non-uniform in the dielectric charge trapping layer, the channel is still controlled by a floating gate as an equi-potential conductor. Thus, the device can have nearly ideal parallel movement in the current / voltage curve during program and erase operations, controlled by local charge trapping at the edge of the device, or by configuration of the edge of a shallow trench isolation (STI) structure. It is due to the fact that it is not. Therefore, it is possible to be immune to the edge problem problem of the charge trapping dielectric based memory cell of the prior art.

As shown by the simulation shown by FIGS. 17-19, most of the injected charge is stored inside the dielectric charge trapping layer rather than inside the floating gate. The charge is thus stored in a deep trap in the dielectric charge trapping layer, providing desirable data retention and providing desirable immunity to substrate induced leakage current (SILC). In addition, since the polysilicon layer used to form the floating gate stores only a small amount of charge, the dielectric tunneling barrier layer between the floating gate and the substrate is reduced in thickness (eg, typically required for floating gate flash memory). 7 nanometers or less). As mentioned above, the simulated cell used a tunneling barrier structure about 5 nanometers thick between the floating gate and the substrate.

The charge trapping floating gate memory device thus combines the desirable features of both floating gate and charge trapping device technology. In addition, the structure can be used to solve the scaling and reliability problems that all of these prior art types of flash memory devices have.

20-24 illustrate one embodiment of a process flow for fabricating a NAND flash memory array using charge trapping floating gate memory cells. A first stage according to one embodiment of a manufacturing process is shown in FIG. 20, where a stack of material is formed on a semiconductor substrate 500. First, in this embodiment a tunneling barrier structure 501 is formed on the surface of the substrate 500 that includes a layer of silicon dioxide 5-7 nanometers thick. Thereafter, a polysilicon layer 502 is formed on the tunneling barrier structure 501. In an exemplary embodiment, the polysilicon layer has a thickness of 100 nanometers or less, and has a thickness of 10 nanometers or less, thereby achieving significant scaling of the device. In the embodiment shown in FIG. 20, a hard mask layer 503 is then formed over the polysilicon layer 502. In certain embodiments where the hard mask layer 503 comprises silicon nitride, the silicon nitride may be about 100 nanometers thick.

21 shows the next stage of the manufacturing process according to one embodiment of the process. At this stage, a photolithography process or other pattern forming process may be used to form the location of the trench isolation structure. After that, the trench is etched according to the pattern. The etching is made to the substrate 500 through the hard mask layer 503, the floating gate poly layer 502, and the tunneling barrier structure 501 to form a substrate trench for isolating columns of memory cells. Can be. The gap is then filled with silicon dioxide, or other dielectric material, using, for example, a high density plasma (HDP) chemical vapor deposition technique, on top of the hard mask layer between the trenches. Positioning hat-shaped structures 513 and 514 are derived, and trench isolation structures 510, 511 and 512 extending into the substrate 500 are derived. In certain embodiments, trench isolation structures extend about 200 nanometers into the substrate.

22 shows the next stage of the manufacturing process. The embodiment shown in FIG. 22 removes any excess oxide from the trench isolation deposition process, for example using chemical mechanical polishing (CMP), and strips off the hard mask layer, for example. Removing oxide from the top of floating gate poly layer 502 using a wet hydrogen fluoride solution dip etch.

Figure 23 shows the next stage of the manufacturing process. At this stage, the formation of the layers 536-540 of the dielectric used to form the charge trapping structure and the formation of the wordline polysilicon layer 530 are completed. In the illustrated embodiment, preferably a layer of silicon dioxide 536 having a thickness of 2 nanometers or less and a layer of silicon nitride having a thickness of 3 nanometers or less (preferably less than 2.5 nanometers thick) 537 and, preferably, a layer 538 of silicon oxide having a thickness of less than or equal to 3.5 nanometers (more preferably less than or equal to 3 nanometers) may be formed to provide a second tunneling barrier structure. A layer 539 of silicon nitride about 5-7 nanometers thick may be formed, providing a charge trapping layer. A blocking dielectric layer 540 is then formed that includes silicon dioxide having a thickness of about 5-9 nanometers. A wordline polysilicon layer 530 is formed over the blocking dielectric layer 540.

The layer 530 is then cleaned in preparation for the lithography step, or other patterning step, to form the word lines of the array. The pattern for the word line may be etched through the floating gate polysilicon layer 502 to provide separate floating gates 502-1 and 502-2, resulting in a row of memory cells.

24 simply shows a portion of the final structure having a first word line 615-1 and a second word line 615-2 arranged along a row of memory cells. Thereafter, a dopant is implanted between the word lines, a source and drain region is formed on opposite sides of the word line, and dielectric fill is deposited between the word lines to isolate the floating gate in the cell. By further providing, source and drain terminals are formed, and metal layer pattern processing or the like is performed, thereby completing the device. In FIG. 24, the structure is simplified to represent the floating gate element 601 and the dielectric charge trapping element 602 as one single unit, and these portions can be made using a combination of different materials and thicknesses. . See, for example, FIGS. 27-28.

The nearly planar structure of the final device makes it possible to scale the pitch of the memory cell as the minimum feature size reduced for the fabrication process. In addition, the combination of floating gate and dielectric charge trapping elements, or the substitution of an interpore dielectric structure with a charge trapping dielectric element, is unlike any prior art device. In the described embodiment, the charge trapping element is configured such that the tunneling efficiency from the floating gate to the charge trapping layer is much better than the tunneling efficiency between the substrate and the floating gate. Thus, bandgap engineered tunneling barrier structures are suitable for use in such structures. In addition, by providing a memory cell in which most of the charge is stored inside the dielectric charge trapping layer, desirable retention and reliability for the device can be obtained. In addition, the channel of the device is directly controlled by an equi-potential floating gate. This provides the same DC characteristics as a conventional MOSFET.

The channel region of the cell implemented as shown in FIG. 24 has a length between the source region and the drain region defined by the width of the word line 615-1 that is reduced by diffusion of the dopant during the ion implantation process. The width W of the channel is defined by the spacing between the STI structures 511 and 512. The length and width of the channel as defined herein establishes the area of the active region of the channel, which area is less than or equal to the spacing between the STI structures multiplied by the width of the wordline 615-1.

In the structure shown, the floating gate element 601 is sufficiently planar and has approximately the same area on the top and bottom surfaces. The area of the floating gate element 601 is defined by an etching process that sets the spacing between the STI structures and the width of the wordline. Thus, the area of the upper and lower surfaces of the floating gate element 601 are substantially equal to each other, and substantially equal to the product of the width of the wordline and the spacing between the STI structures. Likewise, the area of the top contact for the cell in this embodiment is defined by an etching process that sets the spacing between the STI structures and the width of the wordline. Thus the area of the bottom surface of the wordline serving as the top contact for the cell is substantially equal to the area of the top surface of the conductive layer, which is defined by the product of the width of the wordline and the spacing between the STI structures.

As shown from the illustration of FIG. 24, the channel region of a typical memory cell fabricated using a process for forming critical features of 45 nanometers or less has a length between source and drain of 45 nanometers or less, It has a width of 45 nanometers or less orthogonal to the length.

A representative embodiment fabricated using a process to form key features of 30 nanometers or less has a length between source and drain of 30 nanometers or less, has a width of 30 nanometers or less orthogonal to the length, It has an effective oxide thickness of about 20 nanometers or less and the channel region is orthogonal to the length and has a width that is 1.5 times or less of the effective oxide thickness of the multilayer stack.

In an exemplary embodiment, the channel width W of the memory cell is less than 45 nanometers. The effective oxide thickness (EOT) (the actual thickness scaled as a function of the dielectric constant of silicon dioxide divided by the dielectric constant of the material) of the charge trapping element 602 of the memory cell may be on the order of 15-25 nanometers. In a memory cell having such a structure, the channel width may be about 1.5 times or less of the effective oxide thickness of the charge trapping element 602 of the memory cell calculated as the EOT of the combination of the tunneling barrier structure and the charge trapping structure and the upper dielectric layer. And more preferably about the same as the effective oxide thickness of the memory cell. Embodiments with channel widths of 20 nanometers or less and smaller than the EOT of a memory cell are implemented using photoresist trimming techniques, or phase shift masking, or other sublithography patterning techniques. Can be.

In a particular embodiment, the memory cells may be composed of NAND arrays having channel widths of 45 nanometers or less, preferably at the EOT level of the multilayer stack.

In addition, the memory cells described herein may be implemented in other array structures. For example, an array structure using the charge trapping / floating gate memory cells described herein can be implemented in NOR and AND configurations. In addition, memory arrays using charge trapping / floating gate memory cells as described herein are described, for example, in US Patent Publication No. 2008/0175053, filed July 24, 2008, filed with the same applicant. 12 / 056,489, filed March 27, 2008, which may be implemented using a thin film transistor (TFT) technique and a silicon-on-insulator technique. Cited by reference.

The fabrication process is very similar to that applied for current floating gate flash memory technology, but has been modified to provide a thin polysilicon layer for floating gates and near-plane interpoly dielectric charge trapping structures. Therefore, it can be easily applied to various array structures. In addition, floating gate / charge trapping memory devices may be implemented in both p-channel and n-channel technologies.

25 is a simplified diagram of an integrated circuit having an array of charge trapping floating gate (CTFG) memory cells. Integrated circuit 1950 includes a memory array 1900 implemented on a semiconductor substrate using a non-volatile CTFG memory cell, as described herein. The memory cells of the array 1900 may be interconnected in parallel, in series, or in a virtual ground array (VGA). A row decoder 1901 is connected to a plurality of word lines 1902 arranged along a row of the memory array 1900. The memory cells described herein may be composed of NAND arrays, NOR arrays, or other types of array structures. A column decoder 1903 is connected to a plurality of bit lines 1904 arranged along a column of the memory array 1900. Via bus 1905, an address is provided to column decoder 1903 and row decoder 1901. The sense amplifier and data-influx structure of block 1906 are connected to column decoder 1903 via data bus 1907. Data is entered from the input / output port on the integrated circuit 1950, or from another data source inside or outside the integrated circuit 1950, via a data-in line 1911, block 1906. It is supplied in the data-inflow structure of. Data is located from the sense amplifier of block 1906, through the data-outflow line 1915, to an input / output port on the integrated circuit 1950, or any other location within or outside the integrated circuit 1950. It is supplied to the data destination. The bias arrangement state machine 1909 can be used to determine the bias arrangement supply voltage 1908, such as, for example, an erase verify voltage and a program verify voltage. It controls the application and controls the arrangement for programming, erasing and reading memory cells. The array can be combined on integrated circuits with other modules such as processors, other memory arrays, programmable logic, dedicated logic, and the like.

FIG. 26 illustrates an effective manner in which the fabricated memory cells described herein can be integrated with CMOS devices used in peripheral circuits. In FIG. 26, memory cells such as those in FIG. 23 are shown with the same reference numerals. The peripheral MOSFET is shown on the right. As can be seen, the peripheral MOSFET has a channel body 550 that can be formed and patterned simultaneously with the channel / bitline structure of the memory cell. Similarly, the peripheral MOSFET has a gate dielectric layer 551 that can be placed at the same time as the tunneling barrier structure 501 is formed in the memory array region. For process integration, a stack 536-540 of layers used to form the charge trapping structure above the floating gate 502-1 is formed in a blanket process covering all of the peripheral region of the array. In the peripheral region, lithographic processes, or other pattern processing processes, may be used to form contact openings, such as openings 600 through the stack of layers 536-540. When polysilicon is deposited on the wordline 530 in the array, polysilicon fills the contact opening to form the first polysilicon layer used to form the floating gate in the memory cell, and the wordline in the memory cell. Form and connect to the second polysilicon used to form the transistor gate structure and interconnect in the peripheral MOSFET. The fabrication process thus requires only one additional pattern processing step to form the contact opening 600 for completely integrating the memory cell with the peripheral circuitry. The final MOSFET can be used as a local bitline select transistor in the memory array, connecting the local bitline to a global metal bitline.

Peripheral CMOS circuits often require multiple gate oxide thicknesses and implementations. Thick gate oxide and thin gate oxide are all formed in a process by a technique of first forming a thick oxide on a device, applying a pattern to a designated area where a thick oxide is required, and removing a thick oxide outside the pattern. Can be implemented. After removing the thick oxide, a thin oxide such as tunneling barrier structure 501 and thin gate oxide 551 for the gate dielectric layer may be grown. The thickness of the thick oxide hardly changes during the growth of the thin oxide.

FIG. 27 illustrates another charge trapping floating gate memory cell, where the charge trapping dielectric element 602 is modified as compared to the charge trapping dielectric stack 322 of FIG. 14, in this embodiment the buffer layer 640. And a barrier dielectric layer comprising a stack comprising a high-k capping layer 641. The reference numerals used in FIG. 27 match those used in FIG. 14 when appropriate. The high-k herein refers to dielectric constants greater than 7 and is found in materials such as Al ₂ O ₃ , HfO, ZrO ₂ , La ₂ O ₃ , AlSiO, HfSiO and ZrSiO.

By wet conversion of nitride by a wet furnace oxidation process, a buffer layer 640 of silicon dioxide may be formed. Other embodiments may be implemented using high temperature oxide (HTO), or LPCVD SiO ₂ . Aluminum oxide capping dielectric layer 641 may be made by atomic vapor deposition, including post rapid thermal anneal at about 900 ° C. for 60 seconds to strengthen the film.

Using these processes, a layer of silicon oxide (which can be formed almost without defects), a high-k capping layer, and a high conduction band offset material such as aluminum oxide are combined to provide desirable retention properties and very low erase saturation. A blocking dielectric layer may be provided that provides a voltage. Thus the EOT can be reduced and the operating voltage can be lowered.

Also for a combination of silicon oxide (k ₁ = 3.9) and aluminum oxide (k ₂ = about 8), the ratio of the thickness of the top layer 641 to the thickness of the bottom layer 640 of the blocking dielectric layer may be no greater than two. have. In general, the top layer 641 may have a thickness that is less than the product of the ratio of the dielectric constant (8 / 3.9) and the thickness of the bottom layer 640. Thus, the blocking dielectric layer described herein is in contact with the charge trapping dielectric layer and has a first layer 640 having a dielectric constant k ₁ and a second layer 641 having a dielectric constant k ₂ higher than k ₁ of the first layer. Wherein the second layer has a thickness smaller than the thickness of the first layer multiplied by k ₂ / k ₁ . Since the barrier height of Al ₂ O ₃ is approximately equal to SiO ₂ , the electron barrier height, or conduction band offset, of the aluminum oxide including the N + polysilicon gate is about 3.1 eV. Generally, according to this embodiment, the second layer has a dielectric constant k ₂ higher than k ₁ of the first layer, and the second layer has a thickness smaller than the thickness of the first layer multiplied by k ₂ / k ₁ . Have The structure of FIG. 27 provides for lowering the gate injection current during the negative bias Fowler-Nordheim erase process while maintaining desirable retention.

Another alternative is shown in FIG. 28 where the stack of FIG. 14 is replaced with a charge trapping dielectric stack 652 so that both the tunneling barrier structure 315 and the blocking dielectric layer 320 are replaced. The second tunneling barrier structure 650 is implemented using a single layer tunneling oxide as used in conventional MONOS devices, or using another tunneling barrier structure. In this embodiment, the tunneling barrier structure 650 is implemented using silicon dioxide having a thickness of, for example, 3 nanometers or less, as mentioned above, thereby allowing for tunneling over the first tunneling barrier structure 313. It should provide a smaller barrier. In addition, the blocking dielectric layer 651 is implemented by a high-k material, for example aluminum oxide. The wordline layer 653 may also be implemented using TaN, or other high work function materials, to highlight alternative materials for the wordline 310 of FIG. A final structure, commonly known as a MANOS / TANOS structure, is provided in place of the interpoly dielectric on the floating gate.

29-38 show the results of a test of the FG-BE-SONOS device, where the first tunneling barrier structure is silicon oxide (over 54 microns thick) formed by in situ steam generation (ISSG) and the floating gate is undoped Polysilicon (40 microns thick or more), the second tunneling barrier structure is a bandgap engineered structure comprising O 1 / N 1 / O 2 having a thickness of about 13 microns, 20 microns and 25 microns, respectively, and the charge trapping layer is about 50 microns thick silicon. Nitride, and the blocking dielectric layer is about 40 GPa thick silicon oxide. Self-aligned STI structures were fabricated to isolate the floating gate poly in both the STI and WL directions. The device under test was a large area device that allowed the monitoring of internal properties without ambient effects while experimenting with the viability of the structure.

29 shows test results under a + FN programming bias in the range of + 15V to + 22V. The initial threshold voltage V _t of the device was about 2.2V. The threshold is saturated at a level of about 6V on the sample under test. As can be seen, the cell can be programmed to a level greater than 4V very quickly within 10ms or less, under a bias state of + 18V or less. Faster programming speeds can be achieved with higher electric fields.

30 shows test results under negative FN erase bias of −17 V to −20 V. FIG. The erased state may be about 2V below the initial threshold. In this device, erase saturation occurs at a threshold voltage of about 0V. The device can be erased to a threshold voltage of 1V or less, within 100ms or less, under a negative voltage bias condition having a magnitude of 18 or less. With higher electric fields, faster erase rates can be obtained.

FIG. 31 shows the result of + FN incremental step pulse programming (ISPP) for the device under test, wherein the start program pulse size is 17-20 V, using 100 Hz pulses per increment. In this example, the ISSP slope is about 0.7. In general, the device under test requires a relatively high voltage, on the order of 25V, to achieve a threshold voltage of 6V or higher. In addition, ISPP programming is initiated at a relatively higher program potential of 16V or more, compared to standard bandgap engineered SONOS where ISPP programming is initiated at a 12V level potential. The relatively higher voltage operation takes into account the EOT of the BE-SONOS structure, which is about 13 nanometers, the EOT which is about 5 nanometers of the tunneling barrier structure below the floating gate, and the effective oxide thickness of the floating gate in the depletion mode. It results from the larger effective oxide thickness of the structure.

32 and 33 show drain current versus gate voltage (IV) curves for a first cell having a length of about 0.2 μm and a width of about 0.07 μ and a second cell having a length of about 0.2 μ and a width of about 0.15 μ Each is shown. In the test, the drain and body were grounded while the source was connected to a potential of about 0.5V. In the figure, the IV curve at the initial threshold voltage cannot be read. In practice, however, the IV curve at the initial threshold voltage is parallel to all remaining lines, starting at a threshold of about 2.2V. As represented by the data in these figures, the IV curves move in parallel during the program and erase operations.

34 shows the shift of the sub-threshold for the memory cell under test. As can be seen, there is a very slight variation in the value shift (SS) below the threshold over a large range of threshold voltages.

35 shows the variation in transconductance g _m over the threshold voltage. The sub-threshold value shift and transconductance are relatively smaller than those seen in a standard BE-SONOS cell. This effect is the result of the fact that the floating gate of the charge trapping floating gate cell provides an equipotential plane, which offsets the edge effect that occurs in standard dielectric charge trapping cells.

FIG. 36 illustrates a stepped gate voltage and stepped drain voltage starting with a gate voltage of 8V and a drain voltage of 4V, with a pulse of 1 microsecond and a step of 0.1V. Is used to show the performance of the cell under test, under channel hot electron programming bias. After programming, reverse read and forward read are applied to the cell.

FIG. 37 shows the gate voltage versus drain current on a logarithmic scale for an erased state made using -FN tunneling and a programmed state made using CHE programming from the drain side. FIG. 38 shows the gate voltage versus drain current on a linear scale of 0-5x10-5 amps. 36-38 show that the forward and reverse reads lead to essentially the same current. This means that a thin floating gate (about 40 microns thick in this example) shields the channel from any non-uniform implantation that would occur due to the dielectric charge trapping layer. CHE programming is also provided for programming up to threshold voltages above 7V at very high speeds, allowing memory windows above 4V. CHE programming can be much more efficient than Fowler Nordheim programming.

39 and 40 show the results of testing a charge trapping floating gate device in which the first and second tunneling barrier structures are the same layer of silicon dioxide about 54 microns thick. The floating gate layer, the charge trapping layer and the blocking layer were the same as tested using the bandgap engineered structure of the second tunneling barrier structure. 39 shows a threshold voltage versus programming voltage graph for an ISPP programming task on a device. As can be seen, the device may not be programmed efficiently. 40 shows the threshold voltage versus erase time for an erase voltage of -20V. Again, this figure shows that the cell may not be erased. Based on this test, as mentioned above, in order to facilitate charge tunneling into the charge trapping layer while operating to prevent leakage, while providing the desired durability, the second tunneling barrier structure is different from the first tunneling barrier structure. It can be concluded that it should have different tunneling behavior.

In sum, as described herein, a flash memory device comprising a thin floating gate memory cell may be provided, wherein in the thin floating gate memory cell, an interpoly dielectric is planar arranged to operate as a charge trapping device, or Replaced by a nearly planar dielectric stack, the tunneling efficiency between the floating gate and the charge trapping device is greater than the tunneling efficiency between the channel and the floating gate. Most of the charge stored inside the memory cell is trapped inside the dielectric charge trapping device. However, channel operation is controlled by trapped charge and the equipotential floating gate structure between the channels, providing DC performance similar to that of MOSFETs or traditional floating gate cells. The memory cell can be applied to many types of array structures, such as NAND, NOR, and virtual ground AND-type structures. The device can be implemented using both n-channel and p-channel techniques. In addition, such a configuration allows for a structure that avoids the edge effect caused by the trench isolation structure adjacent to the charge trapping element, provides desirable data retention and immunity to tunnel oxide defects, and this configuration also provides It can be made using fabrication steps that are easily integrated into floating gate flash memory technology.

Although the present invention has been described with reference to the preferred embodiments and examples described above in detail, these examples are intended to be illustrative and not restrictive. The assembly steps and structures described herein do not cover the complete process flow for the fabrication of the entire integrated circuit. The invention can be realized in connection with various integrated circuit assembly techniques known in the art or developed in the future.

1 shows the basic structure of a prior art SONOS-type memory cell.

2 shows a cross-sectional view of a basic prior art SONOS-type memory cell taken along a channel width size parallel to the wordline in a NAND array configuration.

FIG. 3 shows a structure similar to FIG. 2 with a reduced channel width compared to the effective oxide thickness of the combination of the bottom dielectric, charge trapping layer and top dielectric.

4 illustrates an asymmetric charge trapping derivation along the channel width size of a prior art memory cell.

FIG. 5 illustrates the asymmetric charge trapping distribution shown in FIG. 4 resulting in a distribution of effective threshold voltages of the memory cells along the channel width magnitude.

FIG. 6 shows the drain current vs. gate voltage I-V through the channel of an undesired SONOS-type memory cell due to an asymmetric charge trapping distribution.

FIG. 7 shows a cross-sectional view taken along the channel length of a dielectric charge trapping memory cell including a conductive layer located over a tunneling barrier structure.

FIG. 8 illustrates a cross-sectional view taken along the channel width size of an array of memory cells as shown in FIG. 7.

FIG. 9 shows one example of charge concentration across a channel width size of a dielectric charge trapping memory cell including a conductive layer as shown in FIG. 8.

FIG. 10 shows a constant potential in the conductive layer and illustrates the effect that constant potential can have on the distribution of the threshold voltage in the channel width magnitude.

FIG. 11 shows a simulation of the drain current versus gate voltage I-V characteristics of a dielectric charge trapping memory cell including a conductive layer having an undesirable asymmetrical charge distribution.

12 shows a cross-sectional view taken as the word line size fEk of a floating gate memory cell of the prior art.

13 illustrates a cross-sectional view taken along the wordline size of a prior art floating gate memory cell, including a planar cell structure.

14 shows a cross-sectional view taken along the wordline in the charge trapping floating gate memory cell of the present invention.

FIG. 14A is a band diagram for a bandgap engineered tunneling barrier structure including a band offset technique at low electric field, showing a relatively low tunneling probability.

14B is a band diagram for a bandgap engineered tunneling barrier structure that includes a band offset technique at high electric fields, showing a relatively high tunneling probability.

FIG. 15 illustrates a cross-sectional view taken on the wordline of a charge trapping floating gate memory cell of the present invention, including a charge trapping layer isolated from the floating gate by a second tunneling barrier structure.

FIG. 16 is a graph showing the results of a simulation of a programming operation for a charge trapping floating gate memory cell disclosed herein.

FIG. 17 is a graph showing the results of simulation of the concentration of confined charges for programming operations for the charge trapping floating gate memory cell of the present invention.

18 is a graph showing the results of a simulation of an erase operation for the charge trapping floating gate memory cell of the present invention.

19 is a graph showing the results of a simulation of the concentration of confined charges for an erase operation for the charge trapping floating gate memory cell of the present invention.

20 illustrates a stage of one of the methods for fabricating a memory array comprising a first tunneling barrier structure and a floating gate polysilicon layer formed on a semiconductor substrate.

21 illustrates one stage of a method for fabricating a memory array that includes etching an isolation trench and depositing a dielectric material in the isolation trench inside the substrate.

FIG. 22 illustrates a stage of one of the methods for manufacturing a memory array that includes removing excess oxide from an isolated trench deposition step, stripping off the hard mask layer, and removing oxide from the floating gate polysilicon layer.

FIG. 23 illustrates a method of fabricating a memory array comprising forming a stack of materials including a second tunneling barrier structure, a charge trapping layer, a blocking dielectric layer, and an additional top layer conductive material for wordline formation. One stage is shown.

24 is an illustration of one stage of a method for assembling a memory array including a memory cell of the present invention that includes a combination of a floating gate structure and a dielectric charge trapping structure.

25 is a simplified diagram in accordance with one embodiment of an integrated circuit that includes an array of charge trapping floating gate (CTFG) memory cells and control circuitry.

Figure 26 illustrates a fabrication stage integrating a process for fabricating a memory cell of the present invention comprising a transistor in a peripheral circuit.

27 shows an alternative charge trapping floating gate structure using a high-k capped buffer oxide blocking dielectric layer.

28 illustrates an alternative charge trapping floating gate structure using an aluminum oxide blocking layer and a workfunction gate material.

29 is a graph showing results of testing a positive voltage Fowler-Nordheim (FN) programming operation using the charge trapping floating gate memory device of the present invention.

30 is a graph showing the results of testing a negative voltage Fowler Nordheim (FN) erase operation using the charge trapping floating gate memory device of the present invention.

31 is a graph showing the results of testing an incremental step pulse programming (ISPP) operation using the charge trapping floating gate memory device of the present invention.

32 is a graph of gate voltage versus drain current generated by testing a charge trapping floating gate memory device having a first length and width under a shift in threshold voltage.

FIG. 33 is a graph of gate voltage versus drain current generated by testing a charge trapping floating gate memory device having a second length and width under a shift in threshold voltage.

34 is a graph showing results of testing threshold shifts for the charge trapping floating gate memory device of the present invention.

35 is a graph showing the results of testing the transconductance for the charge trapping floating gate memory device of the present invention.

36 is a graph of CHE programming time versus threshold voltage of a charge trapping floating gate memory device of the present invention, wherein the threshold voltage is sensed for reverse and forward read operations.

FIG. 37 is a graph of logarithmic magnification of gate voltage versus drain current for the charge trapping floating gate memory device of the present invention in both erase and programming states.

38 is a graph of the linear magnification of gate voltage versus drain current for the charge trapping floating gate memory device of the present invention in both erase and programming states.

FIG. 39 is a graph of programming voltage versus threshold voltage for a tested charge trapping floating gate memory device that includes a tunneling barrier structure between the floating gate and a charge trapping layer that is identical to the tunneling barrier structure between the semiconductor body and the floating gate.

40 is a graph of erase voltage vs. threshold voltage for a tested charge trapping floating gate memory device having a tunneling barrier structure between the floating gate and a charge trapping layer that is identical to the tunneling barrier structure between the semiconductor body and the floating gate.

Claims (24)

In a memory cell, the memory cell A semiconductor substrate having a surface comprising a source region and a drain region isolated by a channel region; A first tunneling barrier structure disposed over the channel region on the surface of the substrate, a floating gate disposed over the first tunneling barrier structure and over the channel region, and over the floating gate A second tunneling barrier structure disposed thereon, a charge trapping dielectric layer disposed over the second tunneling barrier structure and over the channel region, and a top dielectric disposed over the charge trapping dielectric layer a multilayer stack overlying the channel, including a structure; A top conductive layer disposed over the top dielectric structure and over the channel region Wherein the second tunneling barrier structure has a different electron tunneling probability function than the first tunneling barrier structure under a bias condition applied to program and erase the memory cell. 10. The memory cell of claim 1, wherein the first tunneling barrier structure comprises a layer of silicon oxide 4-7 nanometers thick. The method of claim 1, wherein the second tunneling barrier structure comprises a combination of materials, the low valence energy level near the channel surface and the increase in valence energy level near the first distance from the channel surface. And to establish a reduction in valence band energy at a second distance greater than two nanometers from the channel surface. 2. The second tunneling barrier structure of claim 1, wherein the second tunneling barrier structure comprises a first silicon oxide layer adjacent to a charge conductor or semiconductor layer and having a thickness of 18 kV (angstroms) or less and a thickness of 30 kPa or less located over the first silicon oxide layer. And a silicon nitride layer having a silicon oxide layer on the silicon nitride layer, the silicon oxide layer having a thickness of less than 35 GPa. The memory cell of claim 1, wherein the second tunneling barrier structure is a barrier engineered tunneling barrier structure. 10. The memory cell of claim 1, wherein the floating gate comprises a semiconductor layer less than 10 nanometers thick. 2. The structure of claim 1, wherein the upper dielectric structure comprises a first layer in contact with a charge trapping dielectric layer and having a dielectric constant k ₁ and a second layer in contact with a gate, wherein the second layer is the first layer. And a dielectric constant k ₂ higher than k ₁ , wherein the second layer has a thickness less than the thickness of the first layer multiplied by k ₂ / k ₁ . 8. The method of claim 7, wherein the first layer of blocking dielectric layer comprises silicon oxide, or silicon oxynitride, wherein the second layer of blocking dielectric layer comprises aluminum oxide, and the charge trapping layer is silicon nitride and silicon A memory cell comprising at least one of oxynitrides. 2. The memory cell of claim 1, wherein the dielectric constant k ₂ of the second layer of the blocking dielectric layer is at least seven. 2. The memory cell of claim 1, wherein the charge trapping dielectric layer comprises silicon nitride. The method of claim 1, wherein the memory cell A control circuit connected to the memory cell Wherein the control circuit comprises a program mode and an erase mode, wherein in the program mode, between the channel and the floating gate, and between the floating gate and the charge trapping structure, to increase negative charge trapped therein. A bias condition is applied to the memory cell to induce tunneling, and in the erase mode, to induce tunneling between the charge trapping structure and the floating gate, and between the floating gate and the channel, in order to reduce the negative charge trapped. And a bias state for said memory cell. In a memory cell, the memory cell A semiconductor body having a single surface, Floating gate, a charge trapping dielectric layer overlying the floating gate, a top dielectric structure disposed over the charge trapping layer, and applied for reading Under positive bias conditions, a positive gate bias is applied to increase the threshold voltage from the charge trapping layer through the floating gate to prevent electron tunneling to the semiconductor body. a multilayer stack positioned on the semiconductor body including means for moving electrons from the body through the floating gate to the charge trapping layer under a bias condition; An upper conductive layer disposed over the upper dielectric structure and over the channel region Memory cell comprising a. The method of claim 12, wherein the memory cell is A control circuit connected to the memory cell Wherein the control circuit comprises a program mode and an erase mode, wherein in the program mode, between the channel and the floating gate, and between the floating gate and the charge trapping structure, to increase negative charge trapped therein. A bias condition is applied to the memory cell to induce tunneling, and in the erase mode, to induce tunneling between the charge trapping structure and the floating gate, and between the floating gate and the channel, in order to reduce the negative charge trapped. And a bias state for said memory cell. In a memory cell, the memory cell A semiconductor substrate having a surface comprising a source region and a drain region isolated by a channel region; A gate dielectric comprising 4-7 nanometer thick silicon oxide disposed on the surface of the substrate and over the channel region; A semiconductor floating gate over the gate dielectric and over the channel region; A tunneling barrier structure positioned on the floating gate, the tunneling barrier structure comprising: a first silicon oxide layer adjacent to the floating gate and having a thickness of 18 kΩ or less, and having a thickness of 30 kΩ or less located on the first silicon oxide layer; A tunneling barrier structure comprising a silicon nitride layer and a silicon oxide layer over the silicon nitride layer and having a thickness of 30 microns; A layer of 4 to 7 nanometers thick silicon nitride overlying the tunneling barrier structure, A blocking dielectric structure disposed over the charge trapping layer, A top conductive layer disposed over the top dielectric structure Memory cell comprising a. 15. The method of claim 14, wherein the blocking dielectric structure comprises a _first layer in contact with the charge trapping dielectric layer and having a dielectric constant k ₁ , and a second layer in contact with a gate, wherein the second layer is the second layer. And a dielectric constant k ₂ higher than one layer of k ₁ , wherein the second layer has a thickness smaller than the thickness of the first layer multiplied by k ₂ / k ₁ . 15. The memory cell of claim 14, wherein the floating gate comprises a semiconductor layer having a thickness of 10 nanometers or less. A method for manufacturing an integrated circuit, the method comprising Forming a gate dielectric layer on the semiconductor substrate; Forming a patterned polysilicon layer over the gate dielectric layer, the pattern comprising a first pattern in a memory region on the substrate and a second pattern in a peripheral region on the substrate Steps, Forming a multilayer dielectric stack overlying the patterned polysilicon layer, wherein the multilayer dielectric stack comprises a tunneling barrier structure in contact with the patterned polysilicon layer; A charge trapping dielectric layer overlying the tunneling barrier layer and an upper dielectric structure disposed over the charge trapping dielectric layer; Forming a contact opening through the multilayer dielectric stack to expose the patterned polysilicon layer at a selected location in the peripheral region; Forming a patterned conductor over the multilayer dielectric stack and contacting the patterned polysilicon layer through the contact opening at the selected location; Forming a source and a drain region in a substrate adjacent to the patterned conductor Method for manufacturing an integrated circuit comprising a. 18. The method of claim 17, wherein the tunneling barrier structure has a higher tunneling efficiency for charge tunneling than the gate dielectric layer structure under bias conditions applied for program and erase operations. . 18. The method of claim 17 wherein the gate dielectric layer comprises silicon oxide having a thickness of 4 to 7 nanometers. 18. The method of claim 17, wherein forming the tunneling barrier structure comprises forming a bandgap engineered tunneling barrier structure comprising a plurality of dielectric layers. . 18. The method of claim 17 wherein the patterned conductor layer comprises polysilicon. 18. The method of claim 17, wherein etching the patterned polysilicon layer in the memory region to provide a floating gate. The method for manufacturing an integrated circuit further comprising. 18. The method of claim 17, wherein the upper dielectric structure comprises silicon dioxide having a thickness of 5 to 9 nanometers. 18. The method of claim 17 wherein the top conductive layer comprises polysilicon.

Images