Silicon nitride antifuse for use in diode-antifuse memory arrays

Abstract

Silicon nitride antifuses can be advantageously used in memory arrays employing diode-antifuse cells. Silicon nitride antifuses can be ruptured faster and at a lower breakdown field than antifuses formed of other materials, such as silicon dioxide. Examples are given of monolithic three dimensional memory arrays using silicon nitride antifuses with memory cells disposed in rail-stacks and pillars, and including PN and Schottky diodes. Pairing a silicon nitride antifuse with a low-density, high-resistivity conductor gives even better device performance.

Description

This application is related to co-pending U.S. application Ser. No. 10/611,245 by S. Brad Herner, entitled “Low-Density, High-Resistivity Titanium Nitride Layer for use as a Contact for Low-Leakage Dielectric Layers,” filed on even date herewith, which application is hereby incorporated by reference in its entirety.

BACKGROUND OF THE INVENTION

Antifuses are known elements in semiconductor devices. An antifuse separates a conductor or semiconductor from another conductor or semiconductor, and is characterized by having two states. Initially electrically insulating, when an antifuse is subjected to a high current, it ruptures and becomes conductive, and remains so.

These two distinct states make an antifuse useful in a nonvolatile memory cell. A cell containing, or isolated from a conductor by, an antifuse in an intact, insulating state may be considered to correspond to a zero or a one, while the same cell with a ruptured, conductive antifuse corresponds to the opposite value. This state remains whether power is applied to the device or not.

Antifuses can be made of a variety of materials. A common choice has been intrinsic silicon, as in Kimura et al., U.S. Pat. No. 5,311,039. Silicon dioxide and other oxides have been used as well, as in Rioult, U.S. Pat. No. 3,787,822.

Silicon nitride has been a less frequent choice. Silicon nitride antifuses, however, become particularly advantageous when used in certain structures or paired with certain materials.

SUMMARY OF THE INVENTION

The present invention is defined by the following claims, and nothing in this section should be taken as a limitation on those claims. In general, the invention is directed to the use of silicon nitride antifuses in diode-antifuse memory arrays.

A first aspect of the invention provides for a memory array comprising a plurality of memory cells, each memory cell comprising a first diode portion; a second diode portion vertically above the first diode portion; and a silicon nitride antifuse in contact with the first or the second diode portion.

Another aspect of the invention provides for a memory array comprising a plurality of memory cells, each memory cell comprising a first diode portion; a second diode portion wherein the first or the second diode portion comprises in-situ doped polysilicon; and a dielectric-rupture antifuse comprising silicon nitride in contact with the second diode portion.

An embodiment of the invention provides for a monolithic three dimensional memory array comprising first conductors extending in a first direction at a first height above a substrate; second conductors extending in a second direction at a second height above the substrate, wherein the second direction is different from the first direction; and antifuses comprising silicon nitride.

Still another aspect of the invention provides for a monolithic three dimensional memory array comprising memory cells, each memory cell comprising a first diode portion; a second diode portion; and an antifuse comprising silicon nitride.

Each of the aspects of the invention and preferred embodiments can be used alone or in combination with one another.

The preferred embodiments will now be described with reference to the attached drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a chart showing time to rupture for antifuses formed of various materials adjacent to various contacts.

FIG. 2 is a cross-section of a memory cell employing a silicon nitride antifuse in contact with a low-density, high-resistivity layer formed according to an aspect of the present invention.

FIG. 3 is a chart showing leakage current for antifuses formed of various materials adjacent to various contacts.

FIG. 4 is a chart showing programmed current for antifuses formed of various materials adjacent to various contacts.

FIG. 5 is a chart showing programmed current for silicon-nitride antifuse devices programmed with various programming voltages.

DETAILED DESCRIPTION OF THE INVENTION

There are many types of nonvolatile memory cells, some of which incorporate, for example, transistors, tunnel junctions, etc. Among the types of nonvolatile memory cells are diode-antifuse memory cells.

A diode allows current to flow more easily in one direction than the other. Two types of diodes commonly used in diode-antifuse memories are PN diodes and Schottky diodes. In a PN diode, p-type semiconductor material is adjacent to n-type semiconductor material. Silicon is the most commonly used semiconductor material. A Schottky diode is formed when a semiconductor material is adjacent to a metal.

In some configurations, an antifuse can separate diode portions. The diode portions become a diode when the antifuse is ruptured. Diode portions separated by an intact antifuse can be termed an incipient diode. An antifuse separating diode portions is used in the '999 application, the '188 application, and in the '359 application; the present application claims priority from all three.

An antifuse can be in series with a diode, separating it from a conductor. This approach is employed in Johnson et al., U.S. Pat. No. 6,034,882, “Vertically Stacked Field Programmable Nonvolatile Memory and Method of Fabrication,” hereinafter the '882 patent; and in Herner et al., U.S. patent application Ser. No. 10/326,470 “An Improved Method for Making High Density Nonvolatile Memory,” filed Dec. 19, 2002, hereinafter the '470 application; both of which are hereby incorporated by reference in their entirety.

Antifuses have been formed of a variety of materials, including intrinsic silicon and various oxides, especially silicon dioxide. Silicon nitride has been used less frequently.

The instant invention takes advantage of silicon nitride antifuses, which rupture more quickly than silicon dioxide antifuses, allowing memory cells to be programmed faster, as shown in FIG. 1. The chart of FIG. 1 is probability plot, in which each data point is the measurement of one device on a wafer. The probability plot allows the distribution for many devices to be shown, giving statistical significance to the data. Curves A-E show time to rupture for cells formed in the same memory cell, in which an antifuse separates a diode from a conductor. Curves A and B are silicon dioxide antifuses, while curves C, D, and E are silicon nitride antifuses.

Silicon nitride antifuses can be ruptured at a lower breakdown field, thus requiring less power. Further, nitrides are not reduced as easily as oxides, enabling more materials to be used as the contacts. Titanium, for example, will reduce silicon dioxide but not silicon nitride.

Diode-antifuse memory arrays with the antifuse between the diode portions (sometimes called antifuses-at-junction) have been taught with the diode portions of the memory cell formed as a pillar, with one diode portion formed over the other, as in the '188 application. In memory arrays comprising such memory cells, any silicon dioxide antifuse could be replaced with a silicon nitride antifuse.

Diode-antifuse memory arrays with the antifuse formed in series with the diode have been taught with the diode or diode portions of the memory cell formed as a pillar, with one diode portion formed over the other, as in the '882 patent or the '470 application. In the pillar memories with an antifuse in series with a diode, for example formed on top of the diode, any silicon dioxide antifuse could be replaced with a silicon nitride antifuse.

Memory arrays including memory cells in which a silicon nitride antifuse is formed between diode portions, and in which the diode portions form parts of rail-stacks, are taught in the '999 application and the '359 application. In such memory arrays, generally a plurality of first rail-stacks formed at a first height above a substrate extend in a first direction, with dielectric fill between them. First diode portions are located in the first rail-stacks. Either a single, continuous blanket of antifuse material is formed covering the first rail-stacks and intervening dielectric fill, or the antifuse material is formed only directly on top of the first rail-stacks.

A plurality of second rail-stacks is then formed on top, at a second height above a substrate extending in a second direction, the second direction different from the first direction. Second diode portions are located in the second rail-stacks; the second diode portions are above the first diode portions. The antifuse material forms antifuses wherever first rail-stacks intersect second rail-stacks. Both first rail-stacks and second rail-stacks comprise conductors.

It should be noted that pillar memories comprise first conductors formed at a first height above a substrate extending in a first direction, and second conductors formed at a second height above a substrate extending in a second direction, the second direction different from the first direction. A pillar, including diode portions and an antifuse which may be silicon nitride, is formed wherever a vertical projection of a first conductor intersects a second conductor.

Thus both rail and pillar monolithic three dimensional memory arrays comprise first conductors formed at a first height above a substrate extending in a first direction; second conductors formed at a second height above the substrate extending in a second direction, the second direction different from the first direction; and antifuses at a third height between the first height and the second height. The antifuses may comprise silicon nitride. There may be multiple vertically stacked memory levels, for example eight levels.

The incorporated and priority documents teach, inter alia, memory cells comprising a first diode portion and a second diode portion, the second diode portion vertically above the first. In all such cells an antifuse, which may comprise silicon nitride, is in contact with the first or the second diode portion.

In cells in which the antifuse is in series with this diode, the antifuse is below the first or above the second diode portion. In antifuse-at-junction cells, the antifuse is between the diode portions. The incorporated and priority documents teach, inter alia, first diode portions comprising a semiconductor of a first conductivity type (n-type or p-type) and second diode portions comprising a semiconductor of a second conductivity type, the second conductivity type opposite the first conductivity type. The semiconductor of either type is preferably in situ doped polycrystalline silicon, also called polysilicon.

Memory arrays comprising Schottky diode portions in which a silicon nitride antifuse is interposed between a first diode portion and a second diode portion, in which the first diode portion comprises metal or metalloid and the second diode portion comprises semiconductor material, and one diode portion is located vertically above the other, are taught in the '999 application, (in which the cells are formed in rail-stacks) and the '188 application, (in which the cells are formed in pillars.)

Ideally, before rupture, no current flows across an antifuse. In reality, there is always some unwanted current, called leakage current. The leakage current associated with a silicon nitride antifuse is generally higher than, for example, that of a silicon dioxide antifuse of the same thickness. If the contact to a silicon nitride antifuse is a low-density, high-resistivity conductor, such as those described in Hemer, application number 10/611,245, leakage current across the silicon nitride antifuse is significantly reduced.

The low-density, high-resistivity conductors of Hemer, application Ser. No. 10/611,245, can be titanium nitride, tungsten nitride, tantalum nitride, titanium tungsten, tungsten, or aluminum; preferably titanium nitride. This low-density, high-resistivity titanium nitride has a resistivity greater than about 300 microOhm-cms, preferably greater than about 500 microOhm-cms, more preferably greater than about 1000 microOhm-cms. Its density is less than about 4.25 grams/cm³, preferably less than about 4.00 grams/cm³, preferably about 3.98 grams/cm³. For the other materials named, the density is less than about 75 percent of the theoretical density for the metal or metalloid.

Accordingly, in a preferred embodiment, a low-density, high-resistivity conductor, preferably titanium nitride, is paired with a silicon nitride antifuse by positioning the silicon nitride antifuse adjacent to a metal or metalloid. Such a contact can be formed in a Schottky diode in which the antifuse is interposed between the metal and the semiconductor, as described earlier. The metal or metalloid contact adjacent to the silicon nitride antifuse can be formed of low-density, high-resistivity conductor.

In another embodiment, a low-density, high-resistivity conductor is paired with a silicon nitride antifuse which separates a PN diode from the conductor, as in Herner, application Ser. No. 10/611,245. These devices can be vertically stacked in multiple levels to form monolithic three dimensional memory arrays.

A detailed example of a memory array using a silicon nitride antifuse, illustrated in FIG. 2, will be given. The '470 application teaches a PN diode formed in a pillar, with a first diode portion of one conductivity type, and a second diode portion of the opposite conductivity type formed on top of it. An antifuse 104 on top of the pillar PN diode 102 separates it from an overlying conductor 110. The antifuse layer 104 can be formed of silicon nitride instead of silicon dioxide. This is preferably done by forming the PN diode 102 according to the teachings of the '470 application, and filling and planarizing to expose the top of PN diode 102. Next, silicon nitride is thermally grown, for example by a rapid thermal annealing at 750 to 800 degrees Celsius, flowing 4 liters/minute of NH₃ for 60 seconds at atmospheric pressure, with a temperature ramp rate of 50 degrees C. per second, forming a layer of silicon nitride preferably about 23 angstroms thick. The silicon nitride layer could be deposited instead. This layer of silicon nitride serves as antifuse 104. The low-density, high-resistivity titanium nitride layer 106 is formed next, with tungsten layer 108 above it.

When low-density, high-resistivity titanium nitride is used adjacent to the silicon nitride antifuse, the pre-rupture leakage current across the antifuse is less than when conventionally-formed titanium nitride is used, as shown in FIG. 3.

The layer of titanium nitride in this example is adjacent to a silicon nitride antifuse, which is intended, before rupture, to allow minimal leakage current. The antifuse has a roughly circular shape with a diameter of about 0.15 μm. The cross-sectional area for which the leakage is measured, then, is:π×((0.15 μm)/2)²=0.0177 μm²

In this case the median leakage current with a voltage lower than the rupture voltage applied across the antifuse was significantly less when the silicon nitride antifuse was paired with low-density, high-resistivity titanium nitride, shown in FIG. 3 on curves G (formed at 750 degrees C.) and H (formed at 800 degrees C.), than when the silicon nitride antifuse was in contact with conventional, dense titanium nitride, shown on curve F. The antifuses in curves I and J were made of silicon dioxide, which has still less leakage.

After antifuse rupture, higher current is preferred for device performance, and the forward current after rupture is as large or larger than when standard titanium nitride is used, as shown in FIG. 4. Conventional, denser titanium nitride was used with a silicon nitride antifuse on curve K, while low-density, high-resistivity titanium nitride was used with a silicon nitride antifuse on curve L (grown at 750 deg. C.) and M (grown at 800 deg. C.) Curves N and O show silicon dioxide antifuses used with low-density, high-resistivity titanium nitride.

As described in the '470 application, in preferred embodiments the top of the diode is heavily doped polysilicon, which provides a good ohmic contact after antifuse rupture. Some of this layer is lost during the planarization performed before growth of the antifuse. If too much heavily doped polysilicon is lost, forward current is decreased, which can hinder device performance. It will be observed that on curves N and O in FIG. 4, which have silicon dioxide antifuses, that several devices have unacceptably low forward current, while fewer devices with silicon nitride antifuses display this poor performance. Thus silicon nitride antifuses are more tolerant of varied doping conditions at the top of the diode.

While devices using silicon dioxide antifuses have an even bigger difference between pre-rupture leakage current and post-rupture forward current than do those using silicon nitride antifuses, the pre- and post-rupture difference for silicon nitride antifuses in contact with low-density, high-resistivity conductive layer is more than adequate for most uses.

As mentioned earlier, antifuses formed of silicon nitride can be ruptured at a lower breakdown field, thus requiring less power. In the device described in the '470 application, for example, in top antifuse embodiments when the diameter of the diode is about 0.15 μm, a voltage of about 10 volts is required to reliably rupture antifuses formed of silicon dioxide. When silicon nitride antifuses are used in the same device at the same diode diameter, the antifuse can be ruptured at lower voltages, for example 6-8 volts. FIG. 5 illustrates forward current after programming with 2 volts applied for the device of the '470 application when the device was programmed at 6 volts (curve P), 8 volts (curve Q), and 10 volts (curve R.) The antifuse for all three curves was silicon nitride grown at 750 deg. C. As will be seen, the forward current is virtually indistinguishable whether 10 volts or 6 volts was used to program the device. This allows memory cells to be scaled to smaller devices, as lower voltages allow smaller programming transistors to be used.

Accordingly, the inventors of the instant application have found it advantageous to trade off the larger programmed versus unprogrammed current differential of silicon dioxide antifuses in favor of the faster and lower-power rupture of silicon nitride antifuses in many applications.

In addition to the earlier incorporated patents and applications, monolithic three dimensional memories are described in Johnson, U.S. Pat. No. 6,525,953, “Vertically Stacked Field Programmable Nonvolatile Memory and Method of Fabrication”; Knall et al., U.S. Pat. No. 6,420,215, “Three Dimensional Memory Array and Method of Fabrication”; and Vyvoda, U.S. patent application Ser. No. 10/440,882, “Rail Schottky Device and Method of Making,” filed May 19, 2003, all assigned to the assignee of the present invention and herein incorporated by reference. The memory cells in these memory arrays are diodes and incipient diodes.

In some monolithic three-dimensional memory arrays taught in these incorporated patents and applications, the memory cells comprise diode portions and antifuses. The diode portions may be found in either rails or pillars. A silicon nitride antifuse can be advantageously paired with the low-density high-resistivity material (titanium nitride, tungsten nitride, tantalum nitride, titanium tungsten, tungsten, or aluminum) described herein.

The foregoing detailed description has described only a few of the many forms that this invention can take. For this reason, this detailed description is intended by way of illustration, and not by way of limitation. It is only the following claims, including all equivalents, which are intended to define the scope of this invention.

Claims

1. A memory array comprising a plurality of memory cells, each memory cell comprising: a first diode portion;a second diode portion; anda dielectric-rupture antifuse comprising silicon nitride in contact with the second diode portion,wherein:the first diode portion comprises in-situ doped polysilicon, andthe second diode portion comprises a layer of titanium nitride in contact with the antifuse, the titanium nitride layer comprising at least a portion having a thickness greater than about 20 angstroms, wherein, for any portion of the titanium nitride layer with a thickness greater than about 20 angstroms, the titanium nitride has a density less than 4.0 grams/cm3.

2. The memory array of claim 1, wherein the first diode portion or the second diode portion forms part of a rail-stack.

3. The memory array of claim 1, wherein the first diode portion or the second diode portion forms part of a pillar.

4. A monolithic three dimensional memory array comprising: first conductors extending in a first direction at a first height above a substrate;second conductors extending in a second direction at a second height above the substrate, wherein the second direction is different from the first direction; andantifuses comprising silicon nitride,wherein:the antifuses are in contact with a first metallic material, andthe first metallic material is titanium nitride comprising at least a portion having a thickness of greater than about 20 angstroms, and for any thickness of titanium nitride greater than about 20 angstroms, the density of the titanium nitride is less than about 4.0 grams/cm3 and the resistivity of the titanium nitride is greater than about 300 microOhm-cms.

5. The monolithic three dimensional memory array of claim 4 wherein the antifuses are at a third height between the first height and the second height.