Defect identification system and method for repairing killer defects in semiconductor devices

Abstract

A method for improving semiconductor yield by in-line repair of defects during manufacturing comprises inspecting dies on a wafer after a selected layer is formed on the dies, identifying defects in each of the dies, classifying the identified defects as killer or non-critical, for each killer defect determining an action to correct the defect, repairing the defect and returning the wafer to a next process step. Also disclosed is a method for determining an efficient repair process by dividing the die into a grid and using analysis of the grid to find a least invasive repair.

Description

SPECIFIC DATA RELATED TO THE INVENTION

[0001] This application claims the benefit of U.S. Provisional Application No. 60/571,435, filed May 14, 2004.

TECHNICAL FIELD OF THE INVENTION

[0002] The present invention is directed, in general, to semiconductor fabrication and, more specifically, to an in-line defect identification system and method for repairing killer defects in semiconductor devices upon detection.

BACKGROUND OF THE INVENTION

[0003] In the realm of semiconductor fabrication, systems and methods for maximizing chip yield are critical to the success of a semiconductor manufacturing company. Higher yields allows companies to distribute the manufacturing costs over a greater quantity of products, thereby reducing the sales price or increasing the profit margin.

[0004] Optical, laser-based and SEM inspection tools are key pieces of equipment for yield maintenance and improvement. They are used to inspect wafers for defects at numerous points in the production process. Their data is used for three fundamental purposes: statistical process control (SPC), identification and quantification of the defects limiting yield for process improvement purposes, and yield modeling.

[0005] Inspection data is generally monitored using SPC since the number of defects is a good indicator of yields. A change in the distribution of defects can indicate a yield problem. Therefore, by monitoring inspection data, the yield problem may be detected in a timely manner. Despite the effectiveness of monitoring data, there is opportunity for improvement. Many types of defects may arise during manufacturing. These defects may be classified as either "killer" or "non-critical" defects. Killer defects cause a malfunction or failure of the semiconductor device, whereas non-critical defects do not substantially affect the performance of the semiconductor device. U.S. Pat. No. 6,047,083 describes one method of identifying defects in semiconductor products during their manufacture and for classifying such defects as "killer" or non-critical. However, once the defect has been identified as "killer", the only solution is to adjust the manufacturing process to attempt to prevent future defects. It is not believed that the industry has addressed the repair of killer defects in order to salvage what otherwise would be defective products.

[0006] The majority of the yield loss for integrated circuits is due to killer defects that are a) of sub-micron size, b) short elements of a single mask level (such as metal or gate-stack runners) together, or create opens at these same levels, and c) can be detected using in-line inspection tools such as an inspection SEM. This invention proposes a method of eliminating these defects in-line, thereby recovering large amounts of yield.

BRIEF DESCRIPTION OF THE DRAWINGS

[0007] FIGS. 1A and 1B schematic representations of a metal layer of a semiconductor device or die illustrating a shorting defect and its repair;

[0008] FIGS. 2A and 2B are schematic representations of an open conductor defect in a metal layer and its repair;

[0009] FIG. 3 is a flow chart representation of a process for defect analysis and repair; and

[0010] FIG. 4 illustrates a parsing process for die repair.

DETAILED DESCRIPTION OF THE INVENTION

[0011] The majority of the yield lost for integrated circuits is due to defects that are of sub-micron size, short elements of a single mask level (such as metal or gate stack runners) together or create opens at the same levels and can be detected using in-line inspection tools. The shorting mechanisms include metal to metal shorts in either copper or aluminum technologies, gate-stack to gate-stack shorts, gate-stack to window shorts and active region to active region shorts. In-line inspection tools are capable of detecting and locating most of these shorted conditions. For example, U.S. Pat. No. 6,047,083 describes a method and apparatus for pattern inspection that can be used to identify killer defects on semiconductor dies. Most killer defects create a single short between two adjacent elements and therefore only need a minor repair to become non-yield limiting. However, the present invention contemplates that killer defects that cause shorts between more than two elements could be repaired by multiple step corrective action.

[0012] Referring now to FIG. 1A, there is shown a plan view of a small segment of a semiconductor die and, more particularly, a representation of a metal layer on a die such as, for example, metal 3. The metal layer includes a plurality of electrical conductors 12, 14 and 16. A defect 18 which may be characterized as a killer defect short circuits conductor 12 to conductor 14. FIG. 2B shows the same die layer after repair of the killer defect by simply cutting through the defect 18 so that the electrical short between the conductor elements 12 and 14 has been eliminated.

[0013] FIG. 2A illustrates another form of killer defect in a semiconductor die layer in which the electrical conductor 20 has an open or space so that continuity through the conductor element is lost. As shown in FIG. 2B, the open killer defect indicated at 22 can be repaired by creating an electrically conducting bridge between the spaced-apart ends of the conductor element 20.

[0014] The process of cutting through a killer defect such as defect 18 in FIG. 1A can be performed using an in-line focused ion beam or FIB tool to cut a trench through the shorting conductor. However, one of the concerns with using the FIB tool is that such use may result in gallium contamination. However, if gallium contamination is an issue, a gallium barrier or layer that getters gallium could be incorporated into the process flow for forming the semiconductor layer after transistor formation thereby allowing FIB cuts to be used in interconnect layers. Another possibility for cutting through such killer defects is to use laser assisted microchemical machining technology. Revise, Inc. markets a product under the designation 9850 Silicon Etcher that could be used not only for cutting through short circuiting defects such as defect 18 but also could be used to create the repair indicated at 24 in FIG. 2B. It is also possible that micromachining using MEMS and nano technology could be adapted to affect cutting through of these micron sized defects.

[0015] The actual repair process involves a number of steps that are illustrated in block diagram form in FIG. 3. In a first step, block 30, the integrated circuit or die layer needs to be inspected to identify dies which have defects. Such inspection may be carried out using optical examination, laser based or even scanning electron microscope (SEM) inspection tools. The aforementioned U.S. Pat. No. 6,047,083 describes one form of defect identification using SEM inspection tools. Once the defects have been located, further examination of each individual defect is necessary in order to be able to classify the defect as either a killer or a non-critical defect. Typically, a SEM inspection tool would be used to image the area identified as having defects so each individual defect could be manually examined. Block 32 indicates the step of classifying and filtering out defects that are non-killer or non-critical. While the SEM inspection tool is believed to be currently the best tool choice for the classification task, other alternative tools such as optical review tools may be suitable for this application. It is also possible that the classification process could be converted into an automated process such as is described in the aforementioned U.S. Pat. No. 6,047,083. Killer defects may be subdivided into two subclasses, shorts and opens. To be a short, the defect must bridge two elements in the circuit and must be conductive material. The composition may be surmised by the appearance of the defect and/or by composition analysis techniques such as energy dispersive x-ray spectroscopy EDX. To be an open, a conductive element must be broken into two separate parts by the defect and the defect must be non-conductive.

[0016] Once the potentially killer defects have been identified, it is then necessary to filter out large defects that are not practical to repair, block 34. Generally small defects that affect one or two elements will be easier to repair than a defect that affects three or more elements. First, only a single cut or bridge is needed and second, the defect is less likely to bridge to another level. Defects that bridge to another level are not likely repairable. Other factors in filtering out killer defects may include the success rate in repairing similar defects and the type of defect. Once the defects have been classified and filtered, the next step in the process is to determine what material has to be removed from the die layer, block 36, in order to repair the die. After determining what material has to be removed or what material has to be added in the case of an open conductor, the next step is to either remove the material, block 38, or deposit new material and then to return the wafer to the next process step, block 40, to complete manufacturing of the semiconductor devices.

[0017] The classification and filtering of blocks 32 and 34 can be done manually using optical or SEM examination of the die layer. Alternately, the automated identification and classification procedure described in U.S. Pat. No. 6,047,083 could be applied. However, once the devices have been sorted into those that can be repaired, the next step is to determine what material needs to be removed or which conductors need to have material added in order to repair open spaces. A brute force method would be to remove all material from a device layer that differs from the intended pattern. Such a method would be economically unfeasible both from a cost and time standpoint. Accordingly, applicants propose an improved procedure which analyzes the area of the defect and determines the simplest way to correct the defect without having to completely remove all of the defective area.

[0018] FIG. 4 shows a diagram of the layout of FIG. 1A in which the area containing the defect has been broken down into a grid structure forming a plurality of squares. Preferably, the grid is the same resolution as the grid used in the original design file for the die circuit being examined. The grid structure can be defined by a series of XY coordinants with the X coordinants being identified by i.sub.1-i.sub.n. The Y coordinants are then defined by j.sub.1-j.sub.n. For each column l.sub.1-l.sub.n, the number of squares that would have to be removed in order to clear the connection between the adjacent conductors is calculated. For example, in FIG. 4, the defect can be removed by clearing the squares 42,44 and 46 in column l.sub.4. Each of the columns is evaluated and the number of squares that would have to be cleared in order to remove the defect is totaled in order to identify the column which would require the least number of squares to be cleared. The same analysis is then applied along the Y axis to identify the number of rows that must be cleared in order to remove the defect. Thereafter, the minimals in the rows are compared to the minimals in the columns to determine which of the rows or columns would require the least amount of change in order to remove the defect. Selection of a row or column, for example, a horizontal or vertical cut, is then based upon which cut would require the least material to be removed. In the example of FIG. 4, it is apparent that the least material is removed by selecting the column i.sub.4. This scenario assumes that removal of a single row or column of squares provides sufficient resistance to effectively fix the defect and allow the device to work sufficiently. Possible it may be necessary to provide multiple square-widths of isolation between runners. This algorithm may be modified accordingly to define a minimum acceptable cut width.

[0019] Next for opens, the depth of the cut must be determined. For full stack extras, the cut should be slightly more than the stack height. For W puddles, the cut could be a proportion of the diameter of the W puddle. For shorts, the thickness of the bridge would be based on the conductivity of the material used and the composition and line width of the defective runner.

[0020] As discussed above, once the squares in the grid have been identified for removal of material and the repair thickness determined, the material removal process can be implemented by using FIB, laser assisted microchemical machining or some form of micromaching using MEMS or nano technology. Further, in the case of an open conductor, the open conductor can be repaired using a laser assisted chemical deposition technique. It should be noted that in the case of an open conductor, the analysis of the best way to correct the defect may not be as complicated since the open occurs in a conductor and the direction of the conductor will define the direction and location of the repair.

Claims

What is claimed is:

1. A method of repairing killer defects in a semiconductor die prior to completion of semiconductor processing comprising: identifying a die having a killer defect; determining a location and action for a minimally invasive repair; implementing the repair action at the repair location; and continuing processing of the semiconductor die.

2. The method of claim 1 wherein the step of determining includes: defining a grid structure overlaying an area of the die containing the defect; analyzing the grid structure to locate squares in the grid structure containing portions of the defect; and determining a minimum number of squares to modify in order to correct the defect.

3. The method of claim 1 wherein the killer defect comprises a shorting connection between conductors and the step of repairing comprises cutting the shorting connection.

4. The method of claim 1 wherein the killer defect comprises an open conductor space and the step of repairing comprises deposition of conductor material in the open conductor space.

5. The method of claim 3 wherein the step of cutting comprises one of focused ion beam etching, laser etching and microchemical machining.

6. The method of claim 4 wherein the step of deposition comprises laser assisted chemical deposition.

7. The method of claim 2 wherein the step of determining a minimum number of squares comprises: defining the grid structure by columns and rows; computing the number of squares in each column that would require clearing to remove the defect; identifying a column having a minimum number of squares requiring clearing; repeating the steps of computing and identifying squares for each row to locate a row having a minimum number of squares requiring clearing; comparing the minimum number of squares in the column to the minimum number of squares in the row to effect clearing; and selecting the one of the column and the row having a minimum number of squares requiring clearing.

8. The method of claim 7 and including the further steps of determining a minimum width for clearing the defect and clearing other squares to achieve the minimum width.

9. A method for improving semiconductor yield by in-line repair of defects during manufacturing comprising: inspecting dies on a wafer after a selected layer is formed on the dies; identifying defects in each of the dies; classifying the identified defects as killer or non-critical; for each killer defect determining an action to correct the defect; repairing the defect; and returning the wafer to a next process step.

10. The method of claim 9 wherein the killer defect comprises a shorting connection between conductors and the step of repair comprises cutting the connection.

11. The method of claim 10 wherein the step of cutting comprises one of focused ion beam etching, laser etching and microchemical machining.

12. The method of claim 9 wherein the defect comprises an open space in a conductor and the step of repair comprises laser assisted microchemical deposition.