Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Mechanical measurement system
Utility Patent
1997-06-27
2001-01-02
Shah, Kamini (Department: 2857)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Mechanical measurement system
Utility Patent
active
06169960
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to the field of semiconductor processing and, more particularly, to a method for determining the damage potential of defects that occur during the manufacturing of semiconductor devices.
2. Description of Relevant Art
The manufacturing of integrated circuits involves the formation of devices upon semiconductor wafers, typically silicon wafers. The manufacturing of the devices involves a sequence of physical and/or chemical processes for the formation of structures and layers upon or proximal to the upper surface of the wafer. These processes include the introduction of external species into the substrate(either by diffusion or implantation), patterning using photolithography techniques, removing material using wet or dry etching, polishing using mechanical and/or chemical means, chemical and/or physical vapor deposition or thermal growing of films, as well as other physical and chemical processes.
Each wafer is subdivided into several sites, called die, their number depending on the size of each die and the size of the wafer. The different die or “chips” can be microprocessors, static random access memories, dynamic random access memories, flash memories, or other forms of integrated circuits. In the past 25 years, the diameter of silicon wafers has steadily increased from less than 1 in to 8 in (200 mm) diameter wafers which are currently used by many semiconductor manufacturing companies.
Throughout the history of semiconductor wafer processing, a challenge has been to maximize the yield of functional die on a given wafer. Many steps in the manufacturing process can cause a die to become nonfunctional and thus depress the yield. A die is considered nonfunctional even if the device is functioning but certain critical parameters are not within the design specifications. The word functional herein after refers to functioning die with parameters which are within the design specifications. Defects that can occur in these manufacturing steps are the major cause of depressed yields. Defects can be caused by foreign matter, patterning problems, or by a faulty process.
Small particles are a common cause of defects. Small particles found in the air can attach themselves to the wafer and interfere with the manufacturing process. Most modern fabrication laboratories try to minimize small particles in the air by maintaining a clean environment using special filters and high circulation of air. The density of small particles in the air in modem manufacturing “clean rooms” can be as low as 1 particle per cubic foot.
Faulty processing equipment can also be a source of foreign matter. For example, a rubber o-ring can break into small particles which can be deposited upon the surface of the wafer. Deposition tools may collect deposits upon the equipment's sidewalls which can later break away and again be deposited upon the surface of the wafer. Human hair and even saliva from operating technicians may also cause defects to the wafers. Gas-phase nucleation is another source of process induced particles.
Photolithography is the most popular technique used in semiconductor manufacturing for defining structures upon wafers. Typically, a layer of the material to be patterned is first deposited. A layer of photoresist is deposited upon that layer and the photoresist is then exposed using a mask and a form of radiation that typically breaks down the photoresist (positive photoresist). Negative photoresist, which requires radiation to form bonds and thus become resistant to being developed, is also available. Subsequently, the exposed photoresist may be removed with the appropriate developer. Many of the defects on a wafer occur during this patterning process. If the mask is not in perfect alignment with the wafer, the pattern is misaligned and, depending on the severity of the defect and the misalignment, the particular die may become nonfunctional. Particles on the back side of the wafer or top of the chuck cause small areas of the wafer to be out of focus causing “hot spots” or local areas poorly resolved patterns. Bubbles in the developer can cause some areas not to be developed away.
Foreign matter that may exist on the mask may also cause defects since it may cause a portion of the photoresist not to be exposed to the radiation source. A non-exposed portion of the photoresist will not be etched away and thus create additional features on the wafer with the possibility of causing damage.
A first step in eliminating or reducing defects is their detection and classification. There are two major categories of tools currently used in the industry for the detection of defects. The first category of tools uses laser scanning of the wafer. A laser beam is scanned across the surface of the wafer as the wafer is scanned. The presence of particles is expected to change the scattering of the beam. Such tools are relatively fast in scanning wafers and finding defects. These tools are usually not capable of detecting planar defects or previous level defects.
A second category of defect detection and classification tools is image comparison tools. Image comparison tools typically form an image of a particular die by shining white light on the wafer and then detecting the image with a microscope and a charge-coupled device camera. Using special algorithms, the image of each die is then compared to the image of two neighboring die. A defect is assumed when corresponding pixels of the images differ in contrast by more than a certain threshold. Image comparison tools are typically slower in processing wafers due to the computation time required for all the image comparisons. However, image comparison tools are good at finding planar defects and previous layer defects.
After the defects are found, an operator or an automated defect classification tool will review and classify the defects. The defects are also sorted according to which die they were found on. As many as 1000 defects can be found on a particular die. Typically, not all of those defects will cause a die to fail. Operators and engineers need to focus their resources on determining the origin of the most harmful. There is no need to waste resources on defects that are too small or the wrong type to cause any significant die failure. In the past, the damage potential of each type of defect was determined by the percentage of functional die with only one of a given type of defect. However, this simple model does not take into account the size of the defect. In addition, the model requires the existence of many die with only one type of defect. It would thus be desirable to have a model that takes into account the size of the defect and a technique for determining the damage potential of each type of defect even if multiple defects exist on a given die.
SUMMARY OF THE INVENTION
A model is first formed to predict the probability (single probability) that a single defect on a die will not render that die nonfunctional. This model depends on a single, undetermined parameter, which is specific to the type of that particular defect, as well as on the size of the defect. In a preferred embodiment, only one undetermined parameter is preferably used to avoid statistical error which can occur from too many free parameters in a model. The expected probability in this single defect model should have the general behavior of initially decreasing linearly with the size of the defect and then taper off and approach 0 as the size keeps increasing. An initial threshold defect size may also exist. That is, in some cases it may be reasonable to assume that the probability remains at 1 up to a certain threshold size before linearly decreasing. Assuming that the effect of different defects is statistically independent, the probability that a die will be functional when multiple defects exist on the die (die probability) is then simply given by the product of all the single probabilities.
After the end of the manufacturing process, a wafer map is produced by electrically testing all the die on a
Advanced Micro Devices , Inc.
Conley Rose & Tayon
Daffer Kevin L.
Shah Kamini
LandOfFree
Method for determining the damage potential of the different... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for determining the damage potential of the different..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for determining the damage potential of the different... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2446444