Method and Structure for Improving Patterning Design for Processing

This invention relates to interactive feedback to a router, and in particular, to interactive feedback of necessary process physics after completion of physical circuit layout to provide a reliable, manufacturable product.

The metallization process in semiconductor technology is well-known in the art. This process begins in the masking area where contacts, i.e. small holes, are etched through all of the surface layers down to the active regions of the chip. Following contact masking, a thin layer of conducting material is deposited by sputtering, CVD techniques, or vacuum evaporation over the entire wafer. The unwanted portions of this layer are removed by patterning, i.e. photomasking and etch procedures. This step leaves the surface covered with thin lines of conductor.

Patterning errors can cause warped or misaligned patterns that result in undesirable electrical characteristics. Hence, the patterning process is critical to ensuring a satisfactory product. Alignment and exposure continue to be at the heart of patterning.

Two important parameters must be considered when photomasking. The first parameter is the aligner's ability to produce a particular size image, i.e. its resolution capability. The second parameter is the aligner's ability to place images in correct relationship to one another, i.e. its registration capability.

During the photomasking process, an optical phenomena, diffraction, occurs which causes printing to vary from one section of the circuit to another. Diffraction is due to the bending effects of a wave of energy as it passes the opaque edge of a mask. Improvement of aligners has been achieved by using shorter wavelengths which lessen the diffraction effect. However, undesirable resolution and registration still occur even with shorter wavelengths.

For example, if a circuit is core limited, i.e., having lots of gates connected with metal lines, then the printing of lines in the circuit remains homogeneous. However, if the circuit has one section which is heavily populated with lines and another section which is less heavily populated with lines (hereinafter called lonely lines), then the width of the lines and the roughness of the edges of the lines vary from one section to another because of this diffraction phenomenon. As circuits become smaller, the undesirability of varying line size dramatically increases. For example, the difference in line size in one micron technology is probably about 0.05 microns which is approximately a 2 percent difference in size. But at one-half micron technology, the approximate size of the effect is still the same, but the difference is now approximating 10 percent, which is unacceptable.

Etching in semiconductor processing may also entail inherent limitations due to a circuit's physical layout. An ideal anisotropic etch leaves vertical walls in the resist and metal layers. However, because the etching chemical dissolves the top of the wall for a longer time than the bottom, the resulting hole is wider at the top than at the bottom. Hence, the etch is isotropic. This etch undesirably undercuts the metal layer beneath the resist which may result in resist lifting or narrow lines. Dry etching processes, such as reactive ion etching, have decreased undercutting, but have not completely solved this problem.

Dry etch techniques rely in part on material from the masking layer (usually photoresist) to achieve anisotropic profiles. This has the undesirable side effect of making the etch anisotropicly sensitive to masking pattern density. Hence, lonely lines in an isolated pattern will etch more isotropic than a heavily populated pattern. Both patterns may exist on the same chip design.

Another problem to be addressed, effectively the mirror image of the problem above, is the issue of microloading where the etching rate of the material is dependent upon the amount of material to be etched. Hence, more surface area to be exposed will take a longer period of time to etch. It logically follows that a part of a chip having different and more densely populated area lines will take a shorter time to etch than an area of less density populated lines.

Another factor to consider during patterning is electromigration. Typically, in circuit layout design, lines are designed to be at one predetermined width, irrespective of their future use. This layout design may create electromigration problems, especially in lines which must carry a heavy load. Long, very thin metal lines, typically formed of aluminum, carrying high currents are particularly prone to electromigration. The high current sets up an electric field in the lead and generates heat. As current and frequency increase, the electromigration resistance goes down. During the electromigration, the aluminum in the lead becomes mobile and begins to diffuse to either end of the lead. Under extreme conditions, the lead itself is severed. In the past, a worst case current density was assumed and all metal lines were made wide enough to carry that current. This is undesirable as line widths become smaller and more functions are put on a single chip.

Another phenomena occurring during patterning is inherent stress due to layering. Because various layers of material are printed on the circuit, all of which may have different coefficients of expansion/contraction and degrees of hardness, an intrinsic stress builds up between these layers. This stress may result in the linear expansion of the softer materials, i.e. generally metals, causing metal voiding even with no voltage. Therefore, stress due to layering may also produce an electrical disconnection.

Therefore, a need arises for a method of improving patterning design for processing.

In accordance with the present invention, dummy lines are placed on the circuit with working lines until a predetermined amount of metallization is present on the circuit. These dummy lines are positioned adjacent lonely lines by at least a predetermined distance.

In one embodiment of the present invention, the metallization amount and placement of dummy lines is implemented with tile use of a computer utilizing a computer software program.

FIG. 1 illustrates dummy, lines interposed between working lines to counter lithographic or etching proximity effects.

FIG. 2 shows one example of a stress relief point on a metal line.

FIG. 3 illustrates a flow chart of the computerized process steps to be performed in accordance with the present invention to determine placement of dummy lines and ensure availability of carbon molecules for anisotropic etching.

FIG. 4 shows one example of placement of dummy lines after partial removal of a local line mesh.

FIG. 5 shows the method of removing redundant local lines performed in pass 2 of the computer program.

FIG. 6 illustrates two forms of dummy lines.

In accordance with the present invention, dummy lines are printed on the circuit to insure standard sizing of all working lines. The number and positioning of dummy lines is determined by the spacing between lonely lines in both the horizontal and vertical directions. In other words, if the spacing exceeds a predetermined threshold distance in both the horizontal and vertical directions, then dummy lines are inserted. In one embodiment of the present invention, specifically addressing lithographic effects the threshold distances in the horizontal and vertical directions is typically approximately less than 2.mu.. This threshold value will vary when considering the undercutting problems separately. For example, as the percentage of photoresist on the surface area of the chip increases, the threshold distance increases to greater than 100.mu.. Clearly, the threshold distance in the horizontal and vertical directions may differ.

Referring to FIG. 1, the distance between working line 2 and working line 3 in the horizontal direction exceeds the threshold distance thresh -x, or the distance between working lines 5-7 and lines 8-10 in the vertical direction exceeds the threshold distance thresh -y. Therefore, dummy lines 11-16 are printed in the area defined by the dashed box 17 to ensure uniform printing of all working lines.

These dummy lines are left floating, i.e. not connected to the working lines in the circuit, to avoid possible shorting. The size of dummy lines 11-16 typically remains constant. Hence, a larger defined area between working lines requires the insertion of more dummy lines.

To facilitate easy identification, the dummy lines may be formed differently than working line. FIG. 6 shows two possible methods. In FIG. 6, working line 70 is bonded by two dummy lines 71 and 72. Dummy line 71 is typically used for countering lithographic effects while line 72 is generally implemented to solve undercutting problems. Solving for lithographic effects involve very small tolerances in distances, therefore having one smooth side in line 71 is important to solving this problem. Although dummy lines may have many forms, forming the dummy lines into a closed loop may produce an undesirable antenna effect.

Interactive feedback which measures the amount of photoresist on the metal layer significantly reduces the problem of undercutting while etching. Photoresist comprises, among other ingredients, a light-sensitive, i.e. energy-sensitive, polymer. A polymer is a heavy molecule containing carbon, hydrogen, and oxygen formed into a particular repetitive pattern. When multiple lines are etched adjacent one another, i.e. within the threshold distance mentioned previously, the process provides carbon molecules from the polymer which form a coating on the sidewall to prevent undercutting. Hence, the etching will be substantially anisotropic.

However, when the line to be etched is not adjacent to another, no carbon is available to form the necessary coating. Therefore the etch becomes ragged, i.e. more isotropic, which may result in circuit breakdown due to undercutting.

In accordance with the present invention and current processing techniques, at least approximately 17 percent of the chip must be covered with photoresist in order to provide sufficient amounts of carbon to form the coating which subsequently insures an anisotropic etch. Note the amount of resist is, in effect, the amount of metallization on the chip. If the amount of photoresist (or metallization) is at 17 percent, then placement of dummy lines is unnecessary regardless of less populated areas. Therefore, the determination of photoresist coverage precedes the printing of dummy lines.

To solve the electromigration problem, the current each line will carry is determined from the electrical design before processing. In accordance with the present invention, line width will vary depending upon the kind of duty cycle and current density that will be on each line. In one embodiment, lines carrying larger currents would be widened to greater than 2.mu. while lines carrying less current could be narrower. A second or complimentary method could be to route a parallel line(s) for high current paths, reducing the current density carried by any one metal clue.

Because stress due to layering is a function of the amount of metal and the ability of a harder material i.e. a dielectric, to exert a force over that metal, the user enhances the probability of electrical disconnection by running a very long line along the chip. In accordance with the present invention, a metal line is provided with stress relief points. In one embodiment, the stress relief point is a 90-degree jog on line 20 as seen in FIG. 2. A 90-degree jog is used because it is easy to digitize. A 45 or 60-degree angle, although also solving the layering stress problem would require a smaller resolution and therefore take longer to digitize the output. Stress relief may also be accomplished by vertically jogging from one metal layer to another through a via.

In one embodiment, the above-mentioned considerations are implemented in a router system using a computer programmed with the software found in Appendix A. The computer software found in Appendix A specifically addresses the undercutting problem mentioned above. Clearly, the invention may also be implemented by user manipulation of the router software to direct the router to include dummy lines where appropriate. This software program is written in C Language and is usable at any work station with a Modular Development Environment Layout Development System (MDE-LDS), release 10, which is commercially available from LSI Logic Corporation, located in Milpitas, Calif. The definition of the options (and default values) used in the software program, and various portions of the Specification hereinafter, is set forth below in Table 1.

In the above table, OB stands for obscured options, and # stands for number of microns multiply 1000 if not specified. For example, if #=200000=>200.mu..

FIG. 3 illustrates a flow chart of the process steps performed by the computer to determine placement of dummy lines and to ensure availability of carbon molecules for anisotropic etching. To achieve the goal of having a predetermined amount of photoresist on wafer during etching, it may be necessary to traverse the flow chart loop several times. In the first pass through the loop, the metal utilization being analyzed is based on the metallization used in implementing the electrical circuit. As will be explained more full below, the metallization area being measured in the first step of subsequent passes includes dummy lines added during the preceding pass(es).

In step 40, the computer determines the metal utilization on the chip. The metal utilization is functionally equivalent to the area of photoresist on the surface of the chip. After this determination is made, the computer proceeds to step, 41.

Step 41 questions whether the metal utilization is lower than the minimal requirement. In one embodiment of the present invention, the software specifies this minimal requirement as being at least approximately 17 percent of the surface area of the chip.

If the metal utilization is lower than the minimal requirement, the computer proceeds to step 42 which is initializing the mesh. In step 42, initialization includes creating a mesh of dummy lines, also called local wires, in a particular pattern. Referring to FIG. 4, the mesh M in this embodiment is comprised of local wires 100-180, which run parallel to each other. Typically, although not shown in this embodiment, the local wires are positioned in the predominant direction of the working lines, also called physical wires, used for the layer of metallization being examined. Mesh M is superimposed over the already laid out physical wires which are identified in FIG. 4 by reference characters 101, 102, 103 and 104. In one embodiment, local wires 100-180 have a width of approximately 8.mu. with a distance between adjacent edges of the local wires being approximately 8-9.mu.. As will be appreciated from FIG. 4, that drawing, as well as other drawings are not necessarily to scale.

In pass 1 which is found in step 43, all overlap between the physical wires and the local wires is removed. This ensures that no shorting will occur between the local wires and physical wires. At this time, the distance between a local wire and physical wire is made to be at least 37 bloat 1", i.e. a minimum distance between the physical wire and local wire. In one embodiment, "bloat 1" is approximately 15.mu.. For example, as illustrated in dashed line block 50, the blackened portions at local lines 160-180 will be eliminated from the mesh. Furthermore, local wire segments are removed if their length becomes less than the predetermined threshold distance which in one embodiment is 100.mu..

During pass 2 found in step 44, local wires are removed if they are found to be "redundant". Removing redundancy is accomplished as follows: local wires are marked from left and right of physical wires if their distance is less a predetermined distance D. Distance D is generally defined as the threshold distance minus ("bloat 1", the minimum distance between the physical and local wire, plus the width of the local wire.) Distance D is roughly slightly larger than the threshold distance divided by two. For example, if a segment 58 of local wire 59, as shown in FIG. 5, is marked from both the left and the right of physical wires 60 and 61 by a distance D, respectively, then segment 58 of local wire 59 is removed. Once again, if the remaining local wire segment is less than the threshold value, it is also removed. As illustrated in FIG. 5, portion 63 which extends above dashed line 62 is less than the threshold value, and will be removed.

During pass 3 found in step 45, portions of local lines inside a predetermined periphery of the physical wires are marked and will be maintained, while portions not marked will be removed. For example, as seen in FIG. 4, dummy lines, or portions thereof, outside a predetermined periphery (indicated by dashed line boxes 50-53) of each physical line (100-103) are removed to minimize the amount of dummy line metallization added to the chip. If the boundary contacts a local line, as box 50 does with line 150, the local line is not removed. Thus, the cross-hatched portions of the mesh of dummy lines 10-18 are removed. The dummy lines to remain are indicated by the double-thick hatching.

During the final step 46, all remaining portions of local lines are generated in a distinguishable pattern to facilitate identification of these local line portions. After step 46 is completed, the computer returns to the initial step 40 for determining metal utilization which is now the sum of the physical wires and remaining portions of the local lines. If the metal utilization is found to still be lower than the minimal requirement, then the computer repeats steps 42-46 until the metal utilization is not lower than the minimal requirement. Before repeating steps 42-46 on the second traverse through these steps, the local line portions which were created during the first traverse of steps 42-46 are treated as physical wires and the creation of the mesh for the second traverse is conducted in the same manner as previously done. Local lines created in the second and any subsequent loops are similarly treated. When this requirement is met, the computer ends the program at step 47, and the resulting layout is then ready for fabrication.

The preceding description is meant to be illustrative only and not limiting. The above describes the metal etch undercutting problem. For lithographic, electromigration, and layering stress problems the methods would be similar but not identical. Those skilled in the art will be able to apply other methods in accordance with this invention based on the appended claims.