CN109213086B

CN109213086B - Processing system and processing method

Info

Publication number: CN109213086B
Application number: CN201710512379.8A
Authority: CN
Inventors: 周友华; 庄国胜
Original assignee: Taiwan Semiconductor Manufacturing Co TSMC Ltd
Current assignee: Taiwan Semiconductor Manufacturing Co TSMC Ltd
Priority date: 2017-06-29
Filing date: 2017-06-29
Publication date: 2020-10-23
Anticipated expiration: 2037-06-29
Also published as: CN109213086A

Abstract

A manufacturing system and a manufacturing method. The processing system includes a machine, a storage device, a first database and a processor. The machine is used for performing a semiconductor process on a first wafer. The storage device is used for storing a plurality of computer program codes. The first database is used for collecting program data from the machine. The processor is used for executing the computer program codes in the storage device to perform the following steps. The program data is corrected to generate corrected program data. The calibration program data is filtered to generate filter program data. The predicted parameter is generated according to the filter program data so as to provide at least one recipe parameter for the recipe management system adjusting machine to perform the semiconductor manufacturing process on the second wafer.

Description

Processing system and processing method

Technical Field

The present disclosure relates to a manufacturing technology, and more particularly, to a manufacturing system and a manufacturing method with virtual metrology.

Background

In semiconductor manufacturing, Integrated Circuits (ICs) are manufactured from successive semiconductor layers, and the semiconductor manufacturing process is performed by various processing tools and metrology tools. To avoid the cumbersome process of actual measurements, the semiconductor process may apply virtual measurements to predict the post-process wafer properties. However, since each machine and the sensor in the machine have errors, the accuracy of the prediction is difficult to be greatly improved.

Disclosure of Invention

A processing system includes a machine, a storage device, a first database and a processor. The machine is used for performing a semiconductor process on a first wafer. The storage device is used for storing a plurality of computer program codes. The first database is used for collecting program data from the machine. The processor is used for executing the computer program codes in the storage device to perform the following steps. The program data is corrected to generate corrected program data. The calibration program data is filtered to generate filter program data. The predicted parameter is generated according to the filter program data so as to provide at least one recipe parameter for the recipe management system adjusting machine to perform the semiconductor manufacturing process on the second wafer.

A method of manufacture comprises the following steps. The first database and the recipe management system are synchronized by a clock signal. Program data is collected from a tool performing a semiconductor process on a first wafer and stored in a first database. The program data is corrected to generate corrected program data. The method further includes generating a predicted parameter for a recipe management system to adjust at least one recipe parameter of the recipe management system for performing the semiconductor process on the second wafer.

A method of manufacture comprising the following steps. Process data is collected from a tool performing a semiconductor process on a first wafer and stored in a first database. The program data is corrected to generate corrected program data. The filter data is generated by filtering the calibration program data by calculating the standard deviation of the calibration program data. The predicted parameter is generated according to the filter program data so as to provide at least one recipe parameter for the recipe management system adjusting machine to perform the semiconductor manufacturing process on the second wafer.

Drawings

The foregoing and other objects, features, advantages and embodiments of the disclosure will be more readily understood from the following description taken in conjunction with the accompanying drawings in which:

FIG. 1 is a schematic view of a processing system according to one embodiment of the present disclosure;

FIG. 2 is a flow chart illustrating a method of processing applied to the processing system of FIG. 1 in accordance with some embodiments of the present disclosure;

FIG. 3 is a schematic view of a processing system according to one embodiment of the present disclosure;

FIG. 4 is a flow chart illustrating a method of processing applied to the processing system of FIG. 3 according to some embodiments of the present disclosure; and

FIG. 5 is a flowchart illustrating a training method applied to the processing systems of FIGS. 1 and 3 according to some embodiments of the present disclosure.

Detailed Description

For a more complete and complete description of the present disclosure, reference is made to the accompanying drawings and the various embodiments described below. The examples provided are not intended to limit the scope of the disclosure; neither is the order of execution presented to limit the scope of the steps, nor is any apparatus that, when recombined, produces equivalent functionality intended to be encompassed by the present disclosure.

In the description and claims, the terms "a" and "an" can be used broadly to refer to a single or to a plurality of elements, unless the context specifically states the article. It will be further understood that the terms "comprises," "comprising," "includes," "including," "has," "having," and similar language, when used herein, specify the presence of stated features, regions, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, regions, integers, steps, operations, elements, components, and/or groups thereof.

As used herein, to "couple" or "connect" may mean that two or more elements are in direct physical or electrical contact with each other, or in indirect physical or electrical contact with each other, and "couple" or "connect" may also mean that two or more elements are in operation or act with each other. In contrast, when an element is referred to as being "directly connected" or "directly coupled" to another element, there are no additional elements present.

As used herein, the term "about", "about" or "approximately about" generally means within about twenty percent, preferably within about ten percent, and more preferably within about five percent of the error or range of the numerical value. Unless otherwise indicated, all numbers recited herein are to be interpreted as approximations, as indicated by the error or range of values expressed as "about," about, "or" approximately about.

Please refer to fig. 1. FIG. 1 is a schematic diagram illustrating a processing system 100 according to one embodiment of the present disclosure. The processing system 100 includes a tool 110, a database 120, a processor 130, and a storage device. In some embodiments, the database 120 is a Fault Detection Control (FDC) database. In some embodiments, the storage device is integrated within the processor 130. In some embodiments, the storage device is separate from the processor 130. The machine 110 is used for performing a semiconductor process on a wafer, and the storage device is used for storing a plurality of computer program codes.

In some embodiments, the processor 130 is implemented by, for example, a central processing unit, a multiprocessor, distributed processing system, special purpose integrated circuit, a suitable processing unit, or other equivalent functional elements. The different circuits or units implementing the processor 130 described above are for illustrative purposes, and various circuits or units implementing the processor 130 are within the intended scope of the present disclosure.

In some embodiments, the storage device is a computer readable storage device. The computer readable storage means may be implemented, for example, by electronic, magnetic, optical, electromagnetic, infrared, and/or semiconductor devices (or apparatus). For example, a computer readable storage device includes a semiconductor memory, a magnetic tape, a removable computer diskette, a random access memory, a read-only memory, a rigid magnetic disk, an optical magnetic disk or other equivalent functional elements. In some embodiments, the computer readable storage device comprises a magneto-optical disk. In some embodiments, the computer readable storage device includes, for example, a cd-rom, a dvd, or the like.

In some embodiments, the computer-readable storage device stores instructions to interface with external machines and/or devices. Illustratively, the computer readable storage device is encoded with instructions for interfacing with a machine 110 comprising a sensor. Thus, the processor 130 may communicate with the tool 110 to perform the processing method 200 of FIG. 2, for example.

In some embodiments, a semiconductor manufacturing process includes several processing operations on a wafer to produce desired semiconductor integrated circuits. Illustratively, semiconductor processing includes a series of photolithography and chemical processing operations.

The processing operations described above include, for example, deposition, removal, patterning, and modification of electrical characteristics (e.g., doping). In some embodiments, deposition is a process of growing, coating, and/or transferring material onto a wafer. Deposition processes or techniques include, for example, Physical Vapor Deposition (PVD), Chemical Vapor Deposition (CVD), Electrochemical deposition (ECD), Molecular Beam Epitaxy (MBE), Atomic Layer Deposition (ALD), or the like. In some embodiments, the removal process removes components from the wafer and includes, for example, an etching process, for example, a Chemical Mechanical Planarization (CMP) process for planarizing the surface of the semiconductor structure. Patterning, also known as photolithography, in some embodiments shapes the deposited material. For example, the patterning process includes selectively masking portions of the semiconductor structure with a photoresist material, exposing the semiconductor structure to a particular wavelength of light, and then rinsing the unexposed areas with a developer. In some embodiments, the change in electrical characteristics includes doping the selected regions by diffusion and/or ion implantation. Illustratively, the doping process is followed by an annealing process, including, for example, a furnace anneal or a Rapid Thermal Anneal (RTA) to activate the implanted dopants.

As described above, the tool 110 performs one or more semiconductor manufacturing processes on a wafer. In some embodiments, different semiconductor manufacturing processes include, for example, etching, implantation, deposition, and annealing, each performed by a different tool. Such tools include, for example, a stepper photolithography tool, an etching tool, a deposition tool, a polishing tool, a rapid thermal annealing tool, an ion implantation tool, and the like. Each tool is configured to perform a semiconductor manufacturing process on a wafer according to a related process recipe. Illustratively, one tool is used to place a film on a wafer, while another tool is used to etch away a layer from the wafer.

Please refer to fig. 1 and fig. 2 simultaneously. FIG. 2 is a flow chart illustrating a method 200 of processing applied to the processing system 100 of FIG. 1 according to some embodiments of the present disclosure. The process method 200 includes a plurality of steps S201-S205. It will be appreciated by those skilled in the art that the steps in the above embodiments, except where specifically indicated by their order, may be performed in any order, even simultaneously or partially simultaneously, as may be desired.

In operation, in step S201, the database 120 collects program data from the tool 110. For example, the process data may be one or a combination of temperature, pressure, Flow rate (Flow rate), time, etc., and the disclosure is not limited thereto.

In step S202, the processor 130 is used to execute the computer program code in the storage device to correct the program data in the database 120 to generate corrected program data. In detail, the machine 110 includes sensors for measuring various parameters of the semiconductor process to generate the process data, and the measured values of the sensors are within an allowable error range. The processor 130 corrects the program data according to the known error to generate corrected program data with higher accuracy. Thus, the calibration process data accurately represents the parameters of the semiconductor manufacturing process.

Next, in step S203, the processor 130 is used to execute the computer program code in the storage device to filter the calibration program data to generate the filter program data. In some embodiments, the processor 130 filters the calibration program data by calculating a standard deviation of the calibration program data to generate filter program data. In detail, the processor 130 filters out the calibration program data exceeding 6 standard deviations to generate the filter program data. In this way, the processor 130 can filter out the calibration process data with larger difference and leave the calibration process data with smaller difference (i.e. the filter data).

In step S204, the processor 130 executes computer program code in the storage device to generate the prediction parameters according to the filter data. In some embodiments, the processor 130 generates the prediction parameters according to the filter data and through a prediction curve in the storage device. For convenience of illustration, the prediction curve represents the relationship between the thickness of the deposited film and process data (e.g., temperature, pressure, flow rate, time) for the deposition process. Therefore, the processor 130 can calculate the predicted deposited film thickness (i.e., the predicted parameter) by substituting the filter data into the prediction curve.

In step S205, the Recipe management system 140 (RMS) adjusts at least one Recipe parameter of the tool 110 for performing a semiconductor process on another wafer. In some embodiments, the recipe management system 140 and an Automated Process Control (APC) device are integrated into a single device. For example, in the case where the predicted thickness of the deposited film (i.e., the predicted parameter) is greater than the target thickness, the recipe management system 140 adjusts at least one recipe parameter (e.g., one or a combination of temperature, pressure, flow rate, time) of the semiconductor process to bring the thickness of the deposited film on another wafer closer to the target thickness. In detail, the recipe management system 140 transmits the adjusted recipe parameters of the semiconductor process to a Real Time Dispatch (RTD) system 150 to further provide the adjusted recipe parameters of the semiconductor process to the tool 110 for implementation on another wafer. In other embodiments, the real-time dispatch system 150 is coupled to the equipment management system 160, so that the equipment management system 160 can adjust the parameter settings of the tool 110 according to the predicted parameters.

Thus, through the calibration and screening steps, the processing system 100 appropriately adjusts the recipe parameters and/or the parameter settings of the tool 110 to achieve the target process result.

Please refer to fig. 3. FIG. 3 is a schematic diagram of a processing system 300 according to an embodiment of the present disclosure. The processing system 300 is substantially the same as the processing system 100, but includes a database 321 and a database 322. In some embodiments,

databases

321, 322 are Statistical Process Control (SPC) databases.

Please refer to fig. 3 and fig. 4 simultaneously. FIG. 4 is a flow chart illustrating a processing method 400 applied to the processing system 300 of FIG. 3 according to some embodiments of the present disclosure. The processing method 400 has a plurality of steps S401 to S405. It will be appreciated by those skilled in the art that the steps in the above embodiments, except where specifically indicated by their order, may be performed in any order, even simultaneously or partially simultaneously, as may be desired.

In operation, in step S401, the database 321 collects metrology data from the tool 110. For example, the measurement data is a film thickness on a wafer, but the disclosure is not limited thereto.

In step S402, the processor 130 executes computer program code stored in the storage device to correct the metrology data in the database 321 to generate corrected metrology data. In detail, the tool 110 includes sensors for measuring various characteristic parameters of the wafer to generate the measurement data, and the measured values of the sensors are within an allowable error range. The processor 130 corrects the metrology data according to the known error to generate corrected metrology data with higher accuracy. In this way, the calibration process data accurately represents the various characteristic parameters of the wafer.

Next, in step S403, the processor 130 is configured to execute computer program code in the storage device to filter the calibration measurement data to generate filtered measurement data. In some embodiments, the processor 130 filters the calibration metrology data by calculating a standard deviation of the calibration metrology data to generate filtered metrology data. In detail, the processor 130 filters the corrected metrology data exceeding 6 standard deviations to generate filtered metrology data. In this way, the processor 130 may filter out the calibration measurement data with larger differences and leave the calibration measurement data with smaller differences (i.e., the filter data).

In step S404, the processor 130 executes computer program code in the storage device to generate prediction parameters according to the filtered metrology data. For convenience of illustration, in the case of a deposition process, the filtered metrology data includes a thickness of the deposited film, and the processor 130 may compare the measured, calibrated and filtered thickness of the deposited film with a target thickness to further generate predicted parameters for adjusting settings of the process hardware.

In step S405, the equipment management system 360 adjusts the parameter settings of the tool 110. In detail, if the thickness of the deposited film is different from the target thickness, the equipment management system 360 adjusts the parameter setting of the tool 110 according to the predicted parameter generated by the processor 130. For example, the tool management system 360 adjusts the power setting of the tool 110 based on the predicted parameters generated by the processor 130 to bring the deposited film thickness closer to the target thickness. In other embodiments, the real-time dispatch system 150 is coupled to the equipment management system 160 such that the real-time dispatch system 150 may adjust at least one recipe parameter of the tool 110 for performing a semiconductor process on another wafer based on the predicted parameter.

Thus, through the calibration and screening steps, the processing system 300 appropriately adjusts the recipe parameters and/or the parameter settings of the tool 110 to achieve the target process result.

In some embodiments, the

processing methods

200, 400 include tool parameter settings, recipe parameter settings, parameter calculations, and methods of performing semiconductor processing on wafers of the

processing systems

100, 300.

In some embodiments, the database 322 collects the predicted parameters from the processor 130 for the equipment management system 360 to adjust the parameter settings of the tool 110. The database 321 collects unprocessed metrology data, while the database 322 collects filtered metrology data that is calibrated and filtered.

In some embodiments, the tool 110, the

databases

120, 321, 322, the processor 130, the recipe management system 140, the real-time dispatch system 150, and the equipment management system 360 are synchronized via the same clock signal. As a result, the accuracy of the prediction data generated by the processor 130 is improved.

In some embodiments, the processor 130 generates the prediction curve through machine learning. Referring to fig. 5, fig. 5 is a flowchart illustrating a training method 500 applied to the

processing systems

100 and 300 of fig. 1 and 3 according to some embodiments of the present disclosure. The training method 500 includes a plurality of steps S501 to S507. It will be appreciated by those skilled in the art that the steps in the above embodiments, except where specifically indicated by their order, may be performed in any order, even simultaneously or partially simultaneously, as may be desired.

In step S501, the processor 130 may be configured to define tasks and set targets. For example, parameters related to the filter, error tolerance, format of prediction data, etc. are set.

In step S502, the processor 130 may be configured to select the sensing parameter for subsequent training. In detail, the sensing parameters correspond to sensors of the tool 110, and the sensing data generated by the sensors during the semiconductor manufacturing process is used for the subsequent training step of the processor 130. For example, for ease of illustration, the processor 130 may be configured to select a portion of the sensing parameters that significantly affect the thickness of the deposited film for subsequent training, for example, during the deposition process. Alternatively, for another example, the processor 130 may be configured to select all of the sensed parameters for subsequent training.

In step S503, the processor 130 synchronizes all devices in the processing system 300 with the same clock in order to improve the accuracy of the prediction. For example, the tool 110, the

databases

120, 321, 322, the processor 130, the recipe management system 140, the real-time dispatching system 150, and the equipment management system 360 are synchronized through the same clock signal.

In step S504, the processor 130 corrects and filters the sensing data. The calibration and screening steps are as described above and are not repeated here. Therefore, the corrected and screened sensing data can accurately reflect various parameters of the semiconductor manufacturing process, and the training accuracy is improved.

In step S505, the processor 130 performs training through machine learning. In detail, wafer metrology data corresponding to the sensing data is also collected and transmitted to the processor 130, and the processor 130 calculates correlation coefficients between the sensing data and the metrology data.

In step S506, the processor 130 determines the strength of the correlation according to the calculated correlation coefficient, and selects the sensing data with strong correlation for generating the prediction curve.

In step S507, the processor 130 generates a prediction curve by using the strongly correlated sensing data. In addition, the processor 130 may further improve the prediction curve. In some embodiments, using Design of experiments (DOE), the processor 130 may change the sensing parameter range corresponding to the strongly correlated sensing data to further update the prediction curve. In other embodiments, the processor 130 updates the prediction curve through machine learning according to the calibration process data and the calibration measurement data. Therefore, the processor 130 increases the amount of data used to generate the predicted curve by using the calibration procedure data and the calibration measurement data, thereby improving the Confidence level (Confidence level) of the predicted curve.

As such, through synchronization of the calibration and equipment within the

processing system

100, 300, the

processing system

100, 300 trained by the training method 500 may adjust the recipe parameters and/or the parameter settings of the tool 110 appropriately to achieve the target process result. In addition, the

processing systems

100, 300 are updated according to the calibration process data and the calibration metrology data collected and processed in the actual semiconductor process to further accumulate data volume to improve the accuracy of the

processing systems

100, 300.

In some embodiments, the processing system further comprises a second database for collecting metrology data from the tools. The metrology data corresponds to the first wafer. The processor is further configured to execute the computer program code in the storage device to perform the following steps. The measurement data is corrected to generate corrected measurement data. The calibration measurement data is filtered to generate filtered measurement data. And generating a prediction parameter according to the filtered measurement data so as to adjust the parameter setting of the machine by the equipment management system.

In some embodiments, the processing system further comprises a third database for collecting the predicted parameters from the processor for the equipment management system to adjust the parameter settings of the tool.

In some embodiments, the processor is further configured to execute computer program code in the storage device to perform the following steps. Generating a prediction parameter from the filter data and through the prediction curve.

In some embodiments, the processor is further configured to execute computer program code in the storage device to perform the following steps. Updating the prediction curve by machine learning according to the filter process data and the filter measurement data.

In some embodiments, the processor, the first database, the second database, the recipe management system, and the equipment management device are synchronized via a clock signal.

In some embodiments, the calibration process data is filtered to generate filter process data. The predicted parameters are generated according to the filter data for the recipe management system to adjust at least one recipe parameter of the semiconductor process performed by the tool on the second wafer.

In some embodiments, the second database is synchronized with the equipment management system via a clock signal. Metrology data is collected from the tool and stored in a second database. The measurement data corresponds to a first wafer. The measurement data is corrected to generate corrected measurement data. And generating a prediction parameter according to the corrected measurement data so as to adjust the parameter setting of the machine by the equipment management system.

In some embodiments, the calibration metrology data is filtered to generate filtered metrology data. And generating a prediction parameter according to the filtered measurement data so as to adjust the parameter setting of the machine by the equipment management system.

In some embodiments, the prediction data is collected for the equipment management system to adjust the parameter settings of the tool.

In some embodiments, the prediction parameters are generated from the calibration process data and through a prediction curve.

In some embodiments, the prediction curve is updated by machine learning based on the calibration process data and the calibration metrology data.

In some embodiments, the first database and the recipe management system are synchronized via a clock signal.

In some embodiments, metrology data is collected from the machine and stored in a second database. The measurement data corresponds to a first wafer. The measurement data is corrected to generate corrected measurement data. The calibration measurement data is filtered by calculating a standard deviation of the calibration measurement data to generate filtered measurement data. And generating a prediction parameter according to the filtered measurement data so as to adjust the parameter setting of the machine by the equipment management system.

In some embodiments, the first database, the second database, the recipe management system, and the equipment management system are synchronized via a clock signal.

Although the present disclosure has been described with reference to exemplary embodiments, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the disclosure, and therefore, the scope of the disclosure should be limited only by the terms of the appended claims.

Claims

1. A process system, comprising:

a machine for performing a semiconductor process on a first wafer and generating sensing data;

a storage device for storing a plurality of computer program codes;

a first database for collecting a program data from the machine; and

a processor for executing the computer program codes in the storage device to perform the following steps:

correcting the program data to generate corrected program data;

screening the calibration program data to generate a filter program data;

calculating a correlation coefficient between the sensing data and a measurement data to select strongly correlated sensing data;

generating a prediction curve according to the strongly correlated sensing data; and

generating a prediction parameter according to the filter data and the prediction curve for a recipe management system to adjust at least one recipe parameter of the tool for performing a semiconductor process on a second wafer.

2. The processing system of claim 1, further comprising:

a second database for collecting the metrology data from the tool, wherein the metrology data corresponds to the first wafer; and

a third database for collecting the predicted parameters from the processor for an equipment management system to adjust a parameter setting of the equipment;

wherein the processor is further configured to execute the computer program codes in the storage device to perform the following steps:

calibrating the measurement data to generate a calibrated measurement data;

screening the calibration measurement data to generate a filtered measurement data; and

and generating the prediction parameter according to the filtering measurement data so that the equipment management system can adjust the parameter setting of the machine.

3. The processing system of claim 2, wherein the processor is further configured to execute the computer program codes in the storage device to perform the following steps:

generating the prediction parameter according to the filter program data and through the prediction curve; and

updating the prediction curve by machine learning according to the filtering process data and the filtering measurement data,

wherein the processor, the first database, the second database, the recipe management system, and the equipment management system are synchronized via a clock signal.

4. A method of manufacturing, comprising:

synchronizing a first database and a recipe management system via a clock signal;

collecting a process data from a machine performing a semiconductor process on a first wafer and storing the process data in the first database;

correcting the program data to generate corrected program data;

calculating a correlation coefficient between the sensing data collected from the tool and a measurement data to select strongly correlated sensing data;

generating a prediction parameter according to the calibration procedure data and the prediction curve for the recipe management system to adjust at least one recipe parameter of the machine for performing the semiconductor process on a second wafer.

5. The process method as claimed in claim 4, further comprising:

screening the calibration program data to generate a filter program data; and

the predicted parameter is generated according to the filter data for the recipe management system to adjust the at least one recipe parameter of the semiconductor process performed by the tool on the second wafer.

6. The process method as claimed in claim 4, further comprising:

synchronizing a second database with an equipment management system via the clock signal;

collecting the metrology data from the tool and storing the metrology data in the second database, wherein the metrology data corresponds to the first wafer;

calibrating the measurement data to generate a calibrated measurement data;

generating the prediction parameter according to the calibration measurement data to adjust a parameter setting of the equipment by the equipment management system;

7. The process method as claimed in claim 4, further comprising:

collecting the predicted parameter for an equipment management system to adjust a parameter setting of the machine;

generating the prediction parameter according to the calibration program data and through the prediction curve; and

updating the prediction curve by machine learning according to the calibration procedure data and a calibration measurement data.

8. A method of manufacturing, comprising:

collecting a process data from a machine performing a semiconductor process on a first wafer and storing the process data in a first database;

correcting the program data to generate corrected program data;

filtering the calibration program data by calculating a standard deviation of the calibration program data to generate a filter program data;

generating a prediction parameter according to the filter data and the prediction curve for a recipe management system to adjust at least one recipe parameter of the tool for performing the semiconductor process on the second wafer.

9. The process method of claim 8, further comprising:

collecting the metrology data from the tool and storing the metrology data in a second database, wherein the metrology data corresponds to the first wafer;

calibrating the measurement data to generate a calibrated measurement data;

screening the calibration measurement data by calculating a standard deviation of the calibration measurement data to generate a filtered measurement data;

generating the prediction parameter according to the filtering measurement data to adjust a parameter setting of the machine by an equipment management system; and

the first database, the second database, the recipe management system and the equipment management system are synchronized through a clock signal.

10. The process method of claim 9, further comprising:

collecting the predicted parameters for the equipment management system to adjust the parameter setting of the machine;

updating the prediction curve by machine learning according to the calibration procedure data and the calibration measurement data.