Method for Fabricating Multilayer Insulator-Semiconductor Memory Apparatus

This invention relates to semiconductor apparatus, and more particularly to processes for fabricating semiconductor memory devices which contain a multilayer insulator contacting the semiconductor.

In the prior art of semiconductor memory apparatus, SI.sub.1 I.sub.2 M layered structures have been used as memory devices. Here, "S" denotes a semiconductor substrate or layer; "I.sub.1 " and "I.sub.2 " denote first and second insulator layers, respectively; and "M" denotes a metal electrode layer. For electrical erase of this SI.sub.1 I.sub.2 M structure, a positive voltage is applied to the metal electrode, so that the captured electrons (if any) are transported back to the metal by means of Fowler-Nordheim tunneling in the opposite direction from that during the write-in. In such a memory device, the presence versus absence of captured electrons at surface states at the I.sub.1 I.sub.2 interface of the insulator layers defines the memory state of the device.

Other types of SI.sub.1 I.sub.2 M structures in the prior art rely upon the phenomenon of tunneling of charge carriers between the I.sub.1 I.sub.2 interface and the semiconductor, rather than the metal electrode. Again, the presence versus absence of captured electrons at the I.sub.1 I.sub.2 interface state defines the memory state of the device.

The above-mentioned SI.sub.1 I.sub.2 M structures can be incorporated in integrated circuit arrays for mass memories, as known in the art. In such arrays, instead of measuring capacitance of a two-terminal device as previously described, each of the I.sub.1 I.sub.2 M portions of many such SI.sub.1 I.sub.2 M structures is advantageously fabricated as the gates of insulated gate field effect transistors (IGFET's), in which the gates are all integrated on a single semiconductor substrate. As also known in the art, these arrays can be addressed for selective write-in, readout, and erase by various selective crosspoint electrical circuit techniques, such as described for example in U.S. Pat. No. 3,665,423 issued to S. Nakanuma et al. on May 23, 1972.

As set forth in U.S. Pat. No. 3,877,054, issued to Boulin et al. on Apr. 8, 1975, the interface states in the region of the I.sub.1 I.sub.2 interface of SI.sub.1 I.sub.2 M memory devices can be controlled and stabilized by introducing dispersed impurity metal atoms such as tungsten, in a surface concentration between about 10.sup.14 and 2 .times. 10.sup.15 per square centimeter the interface region. By "dispersed" is meant that the impurities do not clump or cluster together, which would form a Fermi level characteristic of the impurity itself. These impurity atoms also serve to increase the capture (trapping) efficiency of electronic charge carriers (electrons or holes), particularly those charge carriers which can be transported from the semiconductor (or metal) to the I.sub.1 I.sub.2 interface region by the phenomenon of Fowler-Nordheim tunneling to the interface states.

In the aforementioned Boulin et al. U.S. Pat. No. 3,877,054, it was suggested to introduce tungsten impurities by evaporation techniques or by purposely including a mixture of some tungsten halide impurites with an aluminum halide advantageously only during an initial stage of chemical vapor deposition of aluminum oxide as the I.sub.2 layer. However, tungsten evaporation techniques entail the use of relatively large amounts of costly ultra-pure tungsten sources (when compared to the exceedingly small amounts of tungsten deposited in the I.sub.1 I.sub.2 interface), high evaporation temperatures (in excess of 3000.degree.C), which consumes large amounts of energy, and furthermore require careful and skilled operators to control the process. Moreover, the inclusion of tungsten halide as an impurity in the initial phase of chemical vapor deposition of aluminum oxide tends to produce an I.sub.2 layer of poor electrical quality, thereby degrading charge storage times in the final device. Therefore it would be desirable to have more economical and easily controllable techniques for introducing metallic impurities, such as tungsten, into an I.sub.1 I.sub.2 interface region of a SI.sub.1 I.sub.2 M memory device.

In order to fabricate an SI.sub.1 I.sub.2 M memory structure with a molecularly dispersed metallic impurity, such as tungsten, in an I.sub.1 I.sub.2 interface region, an oxide of the impurity is deposited on the then exposed I.sub.1 layer just prior to deposition of the I.sub.2 layer. For example, oxide(s) of tungsten, typically tungsten trioxide, is deposited on the exposed I.sub.1 layer by means of a reactive evaporation of tungsten with oxygen. In this way, somewhat higher concentrations of tungsten, (up to 4 .times. 10.sup.15 nuclei of tungsten per square centimeter) can be deposited in a workable memory device, thereby decreasing the required write-in voltages during operation.

In a specific embodiment of the invention, a silicon dioxide (I.sub.1) layer is grown on a monocrystalline silicon semiconductor substrate. The surface of the silicon dioxide is exposed to a stream of tungsten trioxide molecules which are formed by the reaction of the surface of a heated solid tungsten body (about 1050.degree.C to 1700.degree.C) with oxygen gas at a low pressure (about 10.sup.-.sup.5 to 10.sup.-.sup.3 torr) in a vacuum chamber. The tungsten body can be in the form of a hollow circular cylinder into one end of which the oxygen gas flows and out of the other end of which the tungsten trioxide flows onto the exposed silicon dioxide surface. Alternatively, the tungsten body can take the form of a heated tungsten filament located inside a plurality of hollow concentric tungsten cylinders acting as radiation shields, and oxygen is introduced into one end of the innermost cylinder while tungsten trioxide flows out of the other end of the cylinder onto the exposed silicon dioxide surface. Thus, tungsten trioxide is deposited onto the exposed silicon dioxide surface by a chemically reactive evaporation process. The amount of deposited tungsten trioxide is controlled by the tungsten body temperature, oxygen pressure, and time of exposure of the silicon dioxide surface to the tungsten trioxide flow, as well as by the distance of separation between the silicon dioxide surface and the tungsten body. In this way, the total number of deposited tungsten nuclei can range somewhat higher than in the case of nonreactive evaporation of tungsten itself, rather than its oxides .

This invention, together with its features, advantages and objects can be better understood from the following detailed description in conjunction with the drawings in which:

FIG. 1 is a cross section diagram of apparatus for carrying out the invention in accordance with specific embodiments; and

FIG. 2 is a cross section diagram of apparatus for carrying out the invention in accordance with another specific embodiment.

As shown in FIG. 1, a vacuum chamber 10 is evacuated through an outlet pipe controlled by an outlet valve 11. Oxygen at low pressure is introduced into the chamber 10 through an inlet pipe 12 under the control of an inlet valve 13. The inner region of a hollow cylindrical tungsten tube 17 accepts the flow of oxygen from the unsealed outlet end of the pipe 12. This tungsten tube is typically pre-formed by a standard sodium-free chemical vapor deposition process and is heated by means of an A.C. electrical source 14 applied to a pair of tantalum straps 15 and 16 press-fitted to contact the outside wall of the tube 17 near opposite ends thereof. Typically the tungsten tube 17 is about 18 centimeters long, with an outside diameter of about 0.6 centimeters and a wall thickness of about 0.05 centimeters. At a distance typically of about 40 centimeters from the nearer end of the tube 17 is located a target 20 comprising a silicon semiconductor substrate layer 21 upon a major surface of which has been grown a silicon dioxide layer 22 typically by a conventional dry oxidation process. As a result of chemical reaction of the tungsten of the inner surface of the tube 17 with the oxygen stream emanating from the pipe 12, a stream 23 of oxides of tungsten flows onto the then exposed surface of the silicon dioxide layer 22. The target 20 is located typically about 40 centimeters from the nearer end of the tube 17.

Advantageously, the temperature of the tube 17 and the pressure of the oxygen flowing in this tube are advantageously adjusted to produce a significant component of tungsten trioxide in the stream 23. For this purpose, the chamber 10 is first evacuted through the valve 11 (while valve 13 is closed) to a pressure of typically about 10.sup.-.sup.7 torr, and then oxygen is bled through the partially open valve 13 into the chamber to produce a pressure of oxygen therein of advantageously about 5 .times. 10.sup.-.sup.4 torr or less, typically about 5 .times. 10.sup.-.sup.5 torr. The A.C. source 14 is pre-adjusted to maintain the tungsten tube 17 at a temperature of in the range of about 1050.degree. to 1700.degree.C, typically about 1150.degree.C, in order to oxidize the tungsten and sublime the tungsten oxide at the inner surface of the tube 17 in response to the flowing oxygen. The stream 23 thus includes tungsten trioxide molecules which condense and are deposited on the then exposed major surface of the (unheated) silicon dioxide layer 22. The process of tungsten oxide deposition is terminated when the number of deposited tungsten nuclei reaches the desired amount, in the range of between about 10.sup.14 and 4 .times. 10.sup.15 per square centimeter, advantageously between about 4 .times. 10.sup.14 and about 2 .times. 10.sup.15 per cm.sup.2, for example about 7 .times. 10.sup.14 per square centimeter. The deposition time required depends upon the various parameters as the pressure of oxygen, the temperature of the tungsten tube 17, and the separation of the target from the tungsten tube. Typically, this deposition rate on a single target 20 is of the order of 5 .times. 10.sup.12 tungsten nuclei per square centimeter per second, so that exposure time is of the order of 100 seconds. The target 20 can advantageously be located on a rotating carousel containing many such similar targets which are sequentially moved into and out of the paths of the oxide stream 23. A shield is typically inserted in front of each target during the time periods when the deposition of tungsten oxide is not desired.

After the deposition of the tungsten oxide, the I.sub.2 layer is then formed on the exposed I.sub.1 (silicon dioxide) layer now containing the deposited tungsten nuclei. For example, aluminum oxide as the I.sub.2 layer is deposited in the same chamber 10 by conventional aluminum halide chemical vapor deposition at an elevated temperature of typically about 900.degree.C to a thickness in the range of about 300 to 700 Angstroms, typically about 500 Angstroms. Alternatively, the I.sub.2 layer can be of other relatively high dielectric constant insulators compared to the I.sub.1 layer, such as silicon nitride typically also from about 300 to 700 Angstroms thick. The I.sub.1 layer 22 is, as described in the aforementioned Boulin et al. U.S. Pat. No. 3,877,054, advantageously between about 50 and 200 Angstroms thick, typically about 100 Angstroms of silicon dioxide grown by dry thermal oxidation of a major surface of the silicon substrate layer 21. At some sacrifice of charge storage times in the final memory device, this I.sub.1 layer can be somewhat thinner, as little as about 20 Angstroms thick.

After fabrication of the I.sub.2 on the I.sub.1 layer, a metal electrode is applied to the exposed surface of the I.sub.2 layer, in accordance with the poly-silicon gate technology, for example. Many memory devices can be fabricated on a single silicon substrate, together with access circuitry; and source and drain regions for forming insulated gate field effect transistors can be introduced together with further metallization in accordance with conventional integrated circuit techniques. Thereby, electrical access readout of the memory states of the devices with gain is achieved, as known in the art.

FIG. 2 illustrates the reactive evaporation of tungsten from a tungsten filament 29 rather than the tube 17. Many of the elements in FIG. 2 can be the same as in FIG. 1 and accordingly those elements have been denoted by the same reference numerals. Concentric hollow cylindrical radiation shields 26, 27 and 28 advantageously surround the filament 29, to prevent heat loss due to radiation. These shields are typically made of chemically vapor deposited tungsten, each of the order of 0.05 centimeters in wall thickness. The filament itself is of multiple stranded drawn tungsten, which should be sodium free, typically about 10 centimeters long with a filament thickness of about 0.08 centimeters, and has about 14 turns of inside diameter about 0.8 centimeters. The innermost shield 28 is typically about 10 centimeters long, about 0.05 centimeters in wall thickness, and has an inside (cylinder) diameter of about 1.5 centimeters. The tungsten filament is connected to copper leads from the A.C. source by means of metallic bolts and screws.

At a typical filament temperature of about 1140.degree.C and a typical oxygen pressure of about 3 .times. 10.sup.-.sup.5 torr, the tungsten trioxide deposition rate is about 8 .times. 10.sup.12 atoms per square centimeter per second when the substrate target is located at about 25 centimeters from the filament. Thus, a target exposure time of about 1.5 minutes to the stream of tungsten oxide is useful for the deposition of about 7 .times. 10.sup.14 tungsten nuclei per square centimeter in the form of tungsten oxide. As in the case of the tungsten tube 16, the oxygen pressure and the tungsten filament temperature can be varied over about the same limits. Whereas the tungsten tube 17 as the tungsten source has the advantage of ruggedness, the tungsten filament 29 has the advantage of ready commercial availability.

While this invention has been described in terms of specific embodiments, various modifications can be made without departing from the scope of the invention. For example, solid tungsten trioxide itself can be used as the source of the tungsten oxide stream, by heating it to a temperature of about 1130.degree.C for example. Commercially pure tungsten trioxide, free of sodium oxide, is not readily available, however. Accordingly, since sodium oxide sublimes at 1275.degree.C the source temperatures of the impure commercial tungsten trioxide should advantageously be kept lower than 1130.degree.C for a sodium-free tungsten oxide deposit. With pure (sodium-free) tungsten trioxide, a source temperature of 1140.degree.C is feasible.

Alternatively, also, the control over the amount of tungsten deposit can be achieved by pre-oxidizing a known amount of tungsten of a sodium-free tungsten filament, at a low temperature where tungsten oxide volatility is negligible, equal to the amount to be deposited. For example, a tungsten filament can be oxidized for about 15 minutes at about 800.degree.C in 760 torr of oxygen, and placed in a vacuum chamber in proximity to the target and the tungsten filament is heated to a somewhat higher temperature of about 1130.degree.C for sufficient time (15 minutes) in order to sublimate the entire pre-oxidized portion of the tungsten filament. Alternatively, much more than the desired amount of tungsten oxide to be deposited on the target can be pre-oxidized into the tungsten filament and the amount of deposited tungsten oxide into the target can be controlled by control over the source temperature and time of deposition as well as distance from source to target.

This invention can be practiced with a variety of methods for depositing tungsten oxides on the exposed silicon dioxide layer, in order to form the desired dispersed tungsten impurity nuclei. Moreover, other impurities such as oxides of platinum, iridium, tantalum or niobium may be used instead of tungsten; and other semiconductor substrates and insulators may also be used in conjunction therewith.