Description METHOD OF MANAGING PRODUCT INFORMATION IN PRODUCTION LINE Technical Field
[1] The present invention relates to a method of collecting and managing information about products manufactured under process conditions that a manufacturer does not desire in a production line formed of a series of unit processes, and more particularly, to a method of managing information by which a yield lowering factor is effectively analyzed in manufacturing a product. Background Art
[2] The present rapid industrial development is achieved by the development of the manufacturing industry. The continuous growth of the manufacturing industry has been mostly achieved by innovation in a production method through an successive process and such innovation ever continues. One of factors enabling the innovation is an effort to lower defectiveness at the production site. To secure priority over competing products, the manufacturer exerts all possible efforts to find causes for the defectiveness in products and reflect improvements in production to increase productivity.
[3] In order to improve productivity and competitiveness in a raw price, regardless of all kinds of industries such as automobiles, vessels, electronic devices, and semiconductors, an accurate operation of a manufacturing process line in which products are manufactured is indispensable. The accurate operation of the manufacturing process line not only improves product quality but also guarantees competitiveness of a company in a harsh competitive society.
[4] Thus, it is natural to monitor whether process equipments in most production lines are accurately operated as a manufacturer desires and record the result of monitoring. In addition, the manufacturer samples some of products after each unit process is complete and evaluates whether the process is performed as the manufacturer desires.
[5] However, the evaluation work is usually performed by arbitrarily sampling some of products completing each unit process or sampling according to a preset rule. For example, a product is sampled out of every 100 products or for every several hours. However, it is difficult to recognize the defect factors of a product mass-produced in an up-to-date production line in the above method.
[6] The difficulty in recognizing defect factors is described below with examples of a
process of manufacturing a high-tech semiconductor device and a semiconductor production line.
[7] A semiconductor device is manufactured into a complete form that can be used by common users after undergoing several hundreds or thousands complicated and super- accurate fine processes in a production line. The process of manufacturing a semiconductor device can be classified into a preprocess and a post-process. The preprocess is from inputting a wafer on a semiconductor production line until an electrically operating chip is maintained in a wafer state. In the post-process, the chip in the wafer state is sliced for an end user to use and a product is packaged.
[8] Most of the preprocess is sequentially and repeatedly performed using several tens or hundreds of different manufacturing devices provided in a clean room that is referred to as a fab and is very clean. The preprocess includes a photo process, an etch process, an ioninplant process, and a thin film forming process. These preprocesses use exclusively manufactured equipments, are performed at a high temperature and high pressure environment, and use various kinds of chemicals. In the preprocess, patterns of devices for electrical operation are gradually adjusted on a wafer through a repeated process such as the photo process, the etch process, the ioninplant process, and the thin film forming process. After the preprocess is complete (fab-out), identical semiconductor chips are repeatedly formed on the wafer and the operation of these chips can be electrically checked.
[9] The post-process includes sub-processes of testing semiconductor chips on the wafer and sorting them into quality products and defective products (electrical die sorting, hereinafter, referred to as ΕDS'), slicing the respective chips (slicing), selecting the quality products among the sliced chips and attached the selected quality products on a die (die-bond), bonding pads formed inside the chip using pins and fold wires of a package (wire bond), completing a package with a plastic or ceramic material to provide a package body and pins(packaging), after bonding is complete, and producing complete products which are finally determined as quality products through a final test.
[10] For semiconductor products completed through a complicated manufacturing process, it is difficult to know in which manufacturing process, by which manufacturing equipment, and under which manufacturing condition a defect is caused which produces defective products. Thus, a semiconductor manufacturer exerts a great effort to achieve defect reduction by managing various factors affecting the yield of semiconductor devices.
[11] The defect factors can be classified into a defect in a raw material of a semiconductor, a process defect due to instability in a process, an equipment defect due to characteristic of the manufacturing equipments, and an environmental defect generated during a manufacturing process. Although most of the defect factors are reduced by the development of a semiconductor technology and the accumulation of experience in a semiconductor development technology, these defect factors cannot be sharply reduced.
[12] A test equipment having an electronic beam or a high resolution light source is used to find a fine defect in a material such as a very fine particle, dislocation, and stacking fault and an environmental defect by a fine particle.
[13] In the thin film forming process, an equipment for measuring a film quality is used to evaluate whether a thin film is properly formed. In the photo process, a defect factor in the photo process is analyzed by measuring electric and geometric characteristics of a minute line.
[14] The semiconductor manufacturer monitors and analyzes purity of chemicals in a gas or liquid state injected in each manufacturing process and the pressure and temp erature of the process environment. The semiconductor manufacturer exerts every effort to find the defect factors affecting the yield of a semiconductor device using various expensive analyzing and measuring equipments and methods and reduce generation of the defect factors and improve the yield. However, as the size of the semiconductor device decreases, the equipment requires higher performance.
[15] According to the conventional technology, to reduce the defects in the entire semiconductor manufacturing process, devices are randomly or regularly sampled to measure various values when each process is complete, so that whether the process is completed as the manufacturer desires.
[16] For example, in the thin film forming process, devices are sampled after the thin film forming process is complete or the thickness of a thin film or electrical or mechanical film quality is measured by using a monitoring device input into the thin film forming process, so that whether the manufactured thin film fits a standard the manufacturer desires is determined.
[17] However, this conventional method has a limit in finding numerous factors generating defects in the semiconductor manufacturing process that is more divided and complicated and figuring out a generation mechanism thereof. The limit is applied to not only the semiconductor device but also to all types of industrial products. Thus, a method of analyzing detect factors generated in a certain process and feed-
forwarding the analyzed defect factors to the subsequent process so that defects in the respective processes are analyzed by being conelated with one another, is needed. In addition, a new information management system to analyze and manage how the defect factors generated in each process affect the overall product yield by collecting and analyzing product production information in each process, is needed. Disclosure
[18] To solve the above andor other problems, the present invention provides a method of managing defect information by which various defect factors generated during manufacturing of products are effectively analyzed.
[19] The present invention provides a method of effectively linking and analyzing defects to figure out how a defect generated in a particular process among various steps of manufacturing processes affects other process or a final product.
[20] The present invention provides a method of analyzing defect information to improve the quality of a product and productivity so that a manufacturer can manufacture more economic products.
[21] According to an aspect of the present invention, a method of managing product information in a production line by which information on an ID and characteristics of a product in a production line formed of a plurality of consecutive unit processes comprises recording process conditions needed for production of products in each process, determining one of the products processed in a state in which given process conditions are not satisfied as a preliminary defective sample, determiriing one of the products processed in a state in which the given process conditions are satisfied and processed under conditions closest to those of the preliminary defective sample as a preliminary normal sample, measuring characteristics of the preliminary defective sample andor the preliminary normal sample, and storing a result of the measurement.
[22] According to another aspect of the present invention, a method of managing product information in a production line by which information on an ID and characteristics of a product in a production line formed of a plurality of consecutive unit processes comprises recording process conditions needed for production of products in each process, determining one of the products processed in a state in which given process conditions are not satisfied as a preliminary defective sample, determining one of the products processed in a state in which the given process conditions are satisfied and processed under conditions closest to those of the preliminary defective sample as a preliminary normal sample, measuring characteristics of the preliminary defective sample andor the preliminary normal sample, automatically measuring characteristics
of the preliminary defective sample and the preliminary normal sample in a subsequent process, and storing a result of the measurement.
[23] In the step of recording process conditions needed for production of products in each process, all process conditions applied to manufacturing equipments for each manufacturing process are recorded real-time. Snce each manufacturing equipment has various recording apparatuses to monitor whether the manufacturing process is smoothly performed, the conditions given for the process are recorded real-time by using the recording apparatuses.
[24] In the step of determining one of the products processed in a state in which given process conditions are not satisfied as a preliminary defective sample, by using the recording apparatuses of each manufacturing equipment, whether there is a product processed out of the process conditions the manufacturer sets is checked. Ideally, all products must be manufactured under the given process conditions. Practically, however, a change occurs somewhat from the process conditions that the manufacturer sets. When the range of the change is relatively great, the change is out of the desired process conditions. The product processed under the changed process conditions is preliminary determined as a sample having a defect and carefully observed by the manufacturer at the later time.
[25] In the step of determining a product processed under the conditions closest to those of the preliminary defective sample as a preliminary normal sample after the preliminary defective sample is determined, comparing the product processed under the conditions closest to those of the preliminary defective sample with the preliminary defective sample is one of the best methods. Although the preliminary normal sample is the same as the preliminary defective sample in most of the conditions, since only a limited number of the process conditions of the numerous process conditions are different, it is preferable to compare and analyze the preliminary defective sample and the preliminary normal sample. By the comparison and analysis, the manufacturer can easily recognize how the abnormal process conditions affect the characteristics of a product and production yield.
[26] In the step of measuring characteristics of the preliminary defective sample andor the preliminary normal sample, the preliminary defective sample and the preliminary normal sample are compared and analyzed to obtain information that the manufacturer wants. The information can be valuably used in the subsequent process. For example, the samples determined as the preliminary defective sample and the preliminary normal sample are monitored in the subsequent process. The item of the characteristics
of a sample to be measured may vary according to the nature and type of the sample or can be set in advance by the manufacturer.
[27] An example of measuring the characteristics of a product is provided below with a case of a semiconductor device.
[28] The measurement of the characteristics of the semiconductor device includes measuring not only the ID of the device but also electrical, physical, and mechanical characteristics of the semiconductor device after the process is complete, and extracting and storing values of the results.
[29] The ID of the semiconductor device may have various meanings. For example, in a semiconductor manufacturing process, the ID in the photo process signifies a photo shot unit. Snce several chips can be included in a shot according to the size of a semiconductor chip, a single shot is a unit sample in the photo process performed in units of shots.
[30] In a chemical vapor deposition (CVD) method, since a thin film forming process is performed in units of wafers, a single wafer becomes a unit sample. In a process performed in units of lots where tens of wafers makes one lot, a single lot becomes a unit sample. The unit of samples may vary according to the type or characteristic of each product such as chemicals, electronic products, or mechanical products.
[31] The preliminary defective sample signifies a sample processed in an environment that is out of the preset process conditions and is highly likely to be finally determined as a defective product. The preliminary normal sample signifies a sample processed in an environment that is under the preset process conditions and is highly likely to be finally determined as a normal product. Thus, to analyze a defect generating factor when the preliminary defective sample is finally determined as a defective product, the preliminary defective sample and the preliminary normal sample processed under the conditions closest to those of the preliminary defective sample are sampled together and compared.
[32] Thus, according to the method of processing and managing defect information according to the present invention, the semiconductor manufacturer can compare a defect or preliminary defect of a device that may be generated in each unit process with a normal device to easily analyze a defect factor. Also, the same device in question can be carefully monitored in the subsequent process based on the result of sampling in the preceding unit process. Thus, it is possible to analyze and expect how one defect factor generated in each process can affect the other process. Description of Drawings
[33] FIG. 1 is a block diagram illustrating a typical production line formed of a consecutive process;
[34] FIG. 2 is a graph showing a change in temperature in a thin film forming process of a semiconductor product to explain an embodiment of the present invention;
[35] FIG. 3 is a flow chart for explaining a method of sampling a device having a defect factor;
[36] FIG. 4 is a wafer map showing a method of sampling a semiconductor device when the semiconductor device has a defect in its material or environment; and
[37] FIG. 5 is a wafer map for explaining a method of sampling a defect factor in units of shots in a photo-process of a semiconductor production process. Mode for Invention
[38] The present invention is described below in detail with reference to the accompanying drawings. In the drawings, the same reference numerals denote the same elements.
[39] FIG. 2 is a graph showing a change in temperature which may occur in an actual process and a desired process temperature range, assuming that a process of forming a thin film on a semiconductor to explaining an embodiment of the present invention. In most thin film forming processes, an appropriate pressure and temperature atmosphere is established to form desired type and thickness of a thin film and chemicals are injected in a thin film processing equipment to induce a chemical reaction with a semiconductor wafer. When the desired temperature range is between 800 + 1 ° C during the thin film forming process, as shown in FIG. 2, most devices are processed within the temperature range. However, there may be a case in which the thin film forming process is performed out of the above temperature range for various reasons. In FIG. 2, a device d2 is processed out of the set temperature range while devices dl and d3 are processed within the normal temperature range. In this case, the device d2 is sampled. The sampling in this case includes not only storing intrinsic number such as an ID indicating the device d2 but also measuring the thickness and electrical and mechanical features of the thin film of the device d2 after the process is complete and storing the measured values in an electrical method.
[40] According to the technical concept of the present invention, since the device d2 that is processed out of the normal temperature range has a possible defect presented at the later time, it is classified as a preliminary defective sample. The device d3 that is processed under the normal temperature range and is a preliminary normal sample is also sampled for comparison with the preliminary defective sample (hereinafter,
refened to as the couple sampling).
[41] Snce the above-described couple sampling is to analyze the effect of the change in temperature on a defect, it is most preferable that the preliminary defective sample and the preliminary normal sample are processed under the closest conditions except for the condition of temperature. In the example shown in FIG. 2, the device d3 is sampled because the device d3 is processed under the process conditions that are normal and simultaneously closest to the process atmosphere in which the device d2 is processed.
[42] When a plurality of devices are processed out of the normal temperature range and the number of devices is not more than a predetermined number, all devices are couple-sampled. In this case, the number of preliminary normal samples is the same as that of the preliminary defective samples.
[43] Although in FIG. 2 temperature is used as a parameter, the couple sampling can be performed using concentration of a chemical injected during a process, pressure, or time as a parameter. To analyze how the other parameters in addition to temperature affect yield later, it is most preferable to perform couple sampling using all input process conditions as parameters in the above-described manner.
[44] In the above-described embodiment, although the device close to the preliminary defect device in terms of time is taken as a preliminary normal sample, the technical concept of the couple sampling is not limited to the factor of time as shown below.
[45] In the photo process of the semiconductor manufacturing process, the process is performed in the units of being exposed to a light source at one time, that is, a shot unit. Thus, the preliminary defective sample can be classified in the units of shots. The conditions such as the strength of the light source, expose time, and the wavelength of light are parameters. A device processed when the parameters are out of a desired range can be classified as the preliminary defective sample. The preliminary normal sample in the photo process is sampled from devices processed when the process parameters are within the desired range. A device geometrically close to the preliminary defective sample can be sampled as the preliminary normal sample.
[46] According to the technical concept of the present invention, the device classified as the preliminary defective sample in each unit process can be carefully monitored in the subsequent process based on the result of sampling pursuant to the instruction by the manufacturer.
[47] The device d2 of FIG. 2 determined as a preliminary defective sample or the device d3 of FIG. 2, for example, determined as a preliminary normal sample in a certain process can be classified as a preliminary defective sample in the subsequent
process by another defect factor. The new preliminary defective sample is assigned a new ID different from that of the preliminary defective sample in the preceding process and couple-sampled in the same manner described above to be used in a later defect analysis process. Thus, by carefully observing the devices determined as a preliminary defective sample by a new defect factor, the semiconductor manufacturer is able to analyze a conelation between a defect factor in the preceding process and a defect in the subsequent process affected by the defect factor in the preceding process. Also, a conelation between the devices having defect factors in the respective unit processes and finally determined defective products can be analyzed.
[48] The present invention is not limited to a case in which devices processed out of process conditions that a manufacturer desires are merely sampled, analyzed, and evaluated. According to the present invention, the manufacturer arbitrarily generates a defect factor by changing the process conditions to analyze the effect by the defect factor on yield. Also, the present invention can be used to sample devices by a unit or time interval designated by the manufacturer in a particular process or equipment and observing the sampled devices. By couple-sampling and analyzing the preliminary defective sample and the preliminary normal sample manufactured with the arbitrary defect generating factor, a conelation between a defect factor in a certain process and yield affected by the defect factor can be recognized.
[49] The present invention is not limited to the thin film forming process in manufacturing a semiconductor device as shown in FIG. 2. For example, in the photo process, the strength of a light source, a scope of the allowable wavelength of light, or a degree of time exposed to the light source is a parameter and the unit of sampling devices is a shot. The measurement and analysis after the photo process is performed by analyzing and evaluating a item that is well know as critical dimension (hereinafter, refened to as 'CD').
[50] FIG. 3 is a flow chart for explaining a method of couple-sampling in a unit process according to an embodiment of the present invention. Referring to FIG. 3, each of processes starts by inputting parameters indicating process conditions and injecting materials needed for the processes (S10). One the process conditions are input, the process conditions are maintained as they are, except for a special case, so that an assembly process in manufacturing a product is maintained. When the process is performed, whether the process is performed in conformation with the parameters indicating the process conditions is monitored real-time (S20). A device that is processed not to conform with given process conditions is preliminarily determined as
a device having a defect (SO). The ID of the device determined as a preliminary defective sample is recorded and stored (S40). When the process is performed to conform with the given process conditions, whether a process time allowed by the manufacturer ends is checked (S50). When the process time is complete, since all products in the present process are processed under normal process conditions, a preliminary normal sample is selected according to a rule preset by the manufacturer (S60). The ID of the preliminary normal sample in S60 is recorded and stored (S70). The characteristics of the preliminary normal sample are checked and measured (S80) and the result thereof is stored (SI 10) so that the unit process is terminated.
[51] In the meantime, the ID of the preliminary defective sample that does not conform with the given process conditions is recorded and stored (S20 through S40) until the allowed process time ends (S90). As the process is performed in a state of conforming with the given process conditions, products processed under conditions closest to those of the preliminary defective sample is searched for and determined as a preliminary normal sample (S43). The ID of the preliminary normal sample is recorded and stored (S45).
[52] As described above, the couple sampling that samples the preliminary defective sample and the preliminary normal sample as a couple may include two samples that are geometrically close to each other or processed close to each other in terms of time.
[53] The couple sampling in the present invention is not only to compare and analyze a defect factor caused by a particular process condition causes a change in the preliminary defective sample and the preliminary normal sample, but also to manage products manufactured from an assembly production line in an overall viewpoint.
[54] In most cases, a plurality of process condition parameters are given in a single unit process. For example, for the thin film forming process shown in FIG. 2, the process condition parameters can be temperature, time, and pressure. Thus, in the real-time monitoring step in which whether the process conditions are satisfied is monitored (S20), the same number of monitoring apparatuses as the number of the given process condition parameters are needed. Snce the monitoring apparatuses are equipments performing the process or those attached to the equipments as shown in FIG. 1, the equipments are used therefor. The above-described real-time monitoring step (S20) is not only performed independently and simultaneously for each process condition parameter or at each process by using these monitoring apparatuses.
[55] When the process is performed with the given process condition parameters that are all satisfied, there are no preliminary defective samples. In this case, one of the
preliminary normal samples is selected (S43) and the ID of the selected device is sampled (S45). When the process is complete (S90), the characteristic of the device is measured (SI 00) and the result of the measurement is stored (SI 10). Even when the process is performed with the given process condition parameters that are satisfied, the result of the measurement in S80 may be within a regulation value, or a specification range, that the manufacturer desires, or out of the specification range. Even a device having a value out of the specification range may be finally turned out to be a normal product. However, since the device needs to be carefully monitored in the subsequent process, the device is classified as an item needing careful observation in the subsequent process and automatically sampled independently or in a couple in the subsequent process to measure the characteristic thereof.
[56] As shown in the above example, as a defect occurring in a semiconductor, there are fine particles on a wafer or a defect in a material which are generated from the raw material or the manufacturing environment. In particular, fine defects having a size not more than 100 nm which are difficult to find out and thus diligently searched for by the semiconductor manufacturer are detected in an optical method. The defects generated by these factors can also be presented as defects in a final product.
[57] FIG. 4 illustrates a wafer map to explain a method of sampling a device having a defect factor. Although a point is marked on a wafer to discriminate a device having a fine particle or a material defect factor, no point is marked on an actual wafer.
[58] Snce the devices having the above defects need to be carefully observed in the subsequent manufacturing process, these devices are preferably couple-sampled. The preliminary defective samples are appropriately sampled according to the position on the wafer. In the example shown in FIG. 4, when a defect is found at a position X0Y1, a device neighboring in a column direction can be sampled or a device in a row direction can be sampled. Devices to be couple-sampled according to the concept of the present invention are marked by a dotted line in FIG. 4.
[59] If a defect is not found on the wafer being observed, an item to be carefully observed in the subsequent process may not exist.
[60] The wafer map shown in FIG. 4 can be used not only for a material or environmental defect, but also in all process in which a defect is generated. For reference, in the photo process in which the process is performed in units of shots, the defect factor of a device needs to be managed in units of shots as shown in FIG. 5. When shots having a defect factor is generated because the process is performed with the process conditions not being satisfied, that is, the strength of a light source exceeds a
regulated value in the middle of the photo process, the shots having the defect factors need to be sampled. For example, if a device marked with a point at the center of a shot in FIG. 5 is a device having a defect factor, the neighboring devices are sampled as a couple.
[61] As described above, when a device out of a regulated value of the measured result in the process where the preliminary defective sample is not detected, the ID of the device is sampled and the sampled information is used to carefully monitor the device in the subsequent process.
[62] Although a semiconductor product is mainly described as an embodiment of the present invention, it mist be noted that the above description is a mere example.
[63] All manufacturers mass-producing industrial products will understand that the technical concept of the present invention is not limited to the manufacturing of a semiconductor device but can be applied to manufacturing of general industrial products and various modifications and equivalent other embodiments are available.
[64] While this invention has been particularly shown and described with reference to prefened embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Industrial Applicability
[65] As described above, in the method of managing information on defect of a product according to the present invention, a product manufacturer can not only analyze a conelation between defective factors generated in each process and the effect by the defective factors in the subsequent process, but also analyze a link between the defective factors generated in each process and a defective device that is finally determined. Furthermore, a device having a defective factor in the preceding process is sampled to be carefully monitored in the subsequent process so that the semiconductor manufacturer can easily analyze yield according to the defective factor.