KR20220074089A - Process Update and Defect Cause Analysis Method of Manufacturing Process of Semiconductor - Google Patents

Abstract

The process update and failure cause analysis method of the semiconductor device manufacturing process according to the present invention compares new sensor data acquired from one or more sensors used in the semiconductor device manufacturing process with the range of reference sensor data in a preset normal process In step (a) of determining whether or not each process is abnormal, when it is determined that the new sensor data is satisfied within a preset allowable range with the reference sensor data in step (a), the process is determined as a normal process to update the new sensor data as the reference sensor data to the process reference database, and when it is determined that the new sensor data is out of the reference sensor data and a preset allowable range, the process is determined as an abnormal process (b ), in the step (c) of searching for the cause of the abnormal process determined as the abnormal process in the step (b) from the process defect database, and the abnormal cause of the process determined as the abnormal process in the step (c) is searched , follow-up actions are taken for the cause of the abnormality, and if the cause of the abnormality in the process determined to be an abnormal process is not searched, step (d) of analyzing the cause of the abnormality of the abnormal process and step (d) and (e) updating the abnormal cause of the abnormal process in the process defect database.

Description

Process Update and Defect Cause Analysis Method of Manufacturing Process of Semiconductor

The present invention relates to a process update and defect cause analysis method of a semiconductor device manufacturing process, and more particularly, continuously updates the normal process of the semiconductor device manufacturing process, and when an abnormal process occurs, the abnormal cause is analyzed for follow-up measures It relates to a process update of a semiconductor device manufacturing process and a method for analyzing the cause of defects that make it possible.

In a pure state, a semiconductor exhibits similar properties to an insulator, but is a material whose electrical conductivity is increased by the addition of impurities or has temporary electrical conductivity by light or thermal energy.

A pure semiconductor consists of group 14 elements, and all electrons form covalent bonds. When a group 15 element is added to this, surplus electrons are generated to form an n-type semiconductor, and when a group 13 element is added, electrons are insufficient to form a p-type semiconductor composed of holes.

At this time, if the n-type semiconductor and the p-type semiconductor are pasted together, a rectifying action occurs where the current flows well in the direction from the p-type semiconductor to the n-type semiconductor and hardly flows in the opposite direction. Such a device is called a diode, and this is the basis of a semiconductor device.

In the process of manufacturing such a semiconductor device, an FAB (fabrication) process of forming an electrical circuit on a substrate such as a wafer is performed, and the FAB process includes various detailed sub-processes.

Conventionally, in the process of performing the sub-processes of the FAB process, a prediction method based on simple statistics was used to predict the occurrence of a defective process, and such statistics were set by the engineer himself.

This simply has a simple algorithm to determine that the sensor data value used in each process exceeds the upper and lower limits as defective. In particular, there is a problem in that the subjective intention of the engineer is reflected in the process of determining the defective process.

Therefore, a method for solving these problems is required.

Korean Patent Publication No. 10-2009-0068600

The present invention is an invention devised to solve the problems of the prior art, objectively determining that a normal process/abnormal process occurs in a semiconductor device manufacturing process, and continuously updating the range when it is determined as a normal process, , has the purpose of increasing the production yield of the semiconductor device manufacturing process by enabling follow-up actions by analyzing the cause of the abnormality when it is determined that the process is abnormal.

The problems of the present invention are not limited to the problems mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the following description.

In the process update and failure cause analysis method of the semiconductor device manufacturing process of the present invention for achieving the above object, new sensor data obtained from one or more sensors used in the semiconductor device manufacturing process is used as a reference in a preset normal process. In step (a) of determining whether there is an abnormality in each process in comparison with the range of sensor data, when it is determined that the new sensor data is satisfied within the preset allowable range with the reference sensor data in step (a), the corresponding The process is determined as a normal process and the new sensor data is updated in the process reference database as the reference sensor data. Step (b) of determining as a process, (c) of searching for the cause of an abnormality in the process determined as an abnormal process in the step (b) from the process defect database, (c) of the process determined as an abnormal process in the step (c) If the cause of the abnormality is searched, follow-up actions are taken for the cause of the abnormality, and if the cause of the abnormality of the process determined to be an abnormal process is not searched, the step (d) of analyzing the cause of the abnormality of the abnormal process and the above (d) and (e) updating the abnormal cause of the abnormal process analyzed by step ) to the process failure database.

In this case, the step (b) is a step (b-1) of extracting new sensor data for a process selected during the manufacturing process of a semiconductor device, and a step (b) of visualizing the new sensor data extracted by the step (b-1) Step -2) and step (b-3) of determining whether the process selected from the new sensor data visualized through step (b-2) is normal or abnormal.

And the step (b-2) is a step (b-2-1) of first visualizing the new sensor data extracted by the step (b-1) through the Qusi Attractor technique, the step (b-2-1) (b-2-2) of converting the new sensor data first visualized by step (b-2-2) into Persist Homology and visualizing the new sensor data secondarily visualized by the step (b-2-2), Betti It may include a step (b-2-3) of tertiary visualization by expressing it as a sequence.

In addition, the (b-3) step is a (b-3-1) step of calculating an output value by applying a latent variable to the tertiary visualized image by the step (b-2-3) as an input value, the (b-3-1) step, In step (b-3-2) of deriving a cost function through the input value and the output value calculated in step b-3-1) and comparing the cost function with a preset threshold, the cost function is (b-3-3) of determining the selected process as a normal process when it is determined that it is lower than the threshold value, and determining the selected process as an abnormal process when it is determined that the cost function is higher than the threshold value can do.

In this case, the threshold value may be set differently depending on the type of the sensor.

In addition, in step (d), when the abnormal cause of the process determined to be abnormal in step (c) is searched for, the abnormal cause may be transmitted to the management terminal as a follow-up action, and a preset alarm may be performed.

Meanwhile, in the present invention, for the abnormal cause updated in the process failure database by the step (e), (f) analyzing the similarity to the abnormal cause previously stored in the process failure database, and the step (f) The method may further include the step (g) of extracting a plurality of improvement proposals recommended according to the analyzed similarity and individually applying them to the semiconductor device manufacturing process.

In addition, the present invention repeats the manufacturing process of the semiconductor device to which the improvement plan is individually applied by step (g) n times, and determines whether the cause of the abnormality is resolved by the steps (h) and (h). When it is determined that the solution is resolved, the method may further include (i) updating the process defect database by matching the corresponding improvement plan.

The process update and defect cause analysis method of the semiconductor device manufacturing process of the present invention for solving the above problems improves the conventional prediction method based on simple statistics to predict the occurrence of an abnormal process in the semiconductor device manufacturing process, The production yield of the semiconductor device manufacturing process by objectively judging the process progress through sensor data, continuously updating the range when it is determined to be a normal process, and analyzing an abnormal cause to enable follow-up actions has the advantage of greatly increasing the

Effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description of the claims.

1 is a view showing each process of a process update and failure cause analysis method of a semiconductor device manufacturing process according to an embodiment of the present invention;
2 is a view showing a detailed process of step (b) in the process update and failure cause analysis method of the semiconductor device manufacturing process according to an embodiment of the present invention;
3 is a view showing a detailed process of step (b-2) in the process update and failure cause analysis method of the semiconductor device manufacturing process according to an embodiment of the present invention;
4 is a view illustrating a process of visualizing new sensor data in a process update and failure cause analysis method of a semiconductor device manufacturing process according to an embodiment of the present invention;
5 is a view showing the detailed process of step (b-3) in the process update and failure cause analysis method of the semiconductor device manufacturing process according to an embodiment of the present invention; and
6 is a diagram illustrating a subsequent processing process for an updated abnormal cause in the process update and failure cause analysis method of the semiconductor device manufacturing process according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention in which the object of the present invention can be specifically realized will be described with reference to the accompanying drawings. In describing the present embodiment, the same names and the same reference numerals are used for the same components, and an additional description thereof will be omitted.

1 is a diagram illustrating each process of a process update and a defect cause analysis method of a semiconductor device manufacturing process according to an embodiment of the present invention.

As shown in FIG. 1 , in the process update and failure cause analysis method of a semiconductor device manufacturing process according to an embodiment of the present invention, new sensor data acquired from one or more sensors used in a semiconductor manufacturing process is set In step (a) of judging the presence or absence of abnormality in each process in comparison with the range of standard sensor data in the normal process, and in steps (a), it is determined that the new sensor data meets the standard sensor data within the preset allowable range. In this case, the process is judged as a normal process and new sensor data is updated in the process reference database as reference sensor data. Step (b) of determining as , step (c) of searching for the cause of the abnormality of the process determined as abnormal in step (b) from the process defect database, and the abnormality of the process determined as abnormal in step (c) If the cause is found, follow-up actions are taken for the cause of the abnormality. If the cause of the abnormality in the process determined to be an abnormal process is not searched, (d) analysis of the abnormal cause of the abnormal process; (d) and (e) updating the abnormal cause of the abnormal process analyzed by the step to the process defect database.

Each of these steps will be described in detail. First, in step (a), new sensor data is acquired from one or more sensors used in the semiconductor manufacturing process, and the process of comparing it with the range of reference sensor data in a preset normal process is performed. is done

Here, the sensor used in the semiconductor manufacturing process includes various factors generated in each process for process monitoring, that is, the inflow rate of various gases including temperature, pressure, HF power, LF power, argon, oxygen, silane gas, etc.; There may be various sensors for measuring various factors such as the inflow rate of the UPC, voltage, current, and thickness of the thin film, and one or a plurality of such sensors may be provided.

And in step (a), new sensor data generated from such one or more sensors is acquired in the newly performed semiconductor manufacturing process, and it is compared with the range of reference sensor data.

In this case, the reference sensor data is processed by processing sensor data obtained during a normal process from a plurality of semiconductor manufacturing processes performed in the prior art, and serves as a standard for judging whether the process is normal.

Next, in step (b), it is determined whether the new sensor data in step (a) satisfies the preset tolerance range of the reference sensor data or whether the new sensor data is out of the preset tolerance range of the reference sensor data. The process is determined as either a normal process or an abnormal process.

2 is a view showing a detailed process of step (b) in the process update and failure cause analysis method of the semiconductor device manufacturing process according to an embodiment of the present invention.

As shown in FIG. 2 , step (b) includes steps (b-1) of extracting new sensor data for a process selected during the semiconductor manufacturing process, and new sensor data extracted by step (b-1) (b-2) of visualizing , and (b-3) of determining whether the process selected from the new sensor data visualized through step (b-2) is normal or abnormal.

That is, step (b) extracts new sensor data for an arbitrarily selected process among the semiconductor manufacturing processes in detail, and then visualizes it so that a person can intuitively grasp the phase of the new sensor data.

In addition, FIG. 3 is a view showing the detailed process of step (b-2) in the process update and failure cause analysis method of the semiconductor device manufacturing process according to an embodiment of the present invention, and FIG. 4 is an embodiment of the present invention. In the process update and defect cause analysis method of the semiconductor device manufacturing process according to the example, it is a diagram illustrating a process of visualizing new sensor data.

3 and 4, the step (b-2) of visualizing the new sensor data is the first visualization of the new sensor data extracted by the step (b-1) through the Qusi Attractor technique (b- Step 2-1) and (b-2-2) of converting the new sensor data first visualized by step (b-2-1) into Persist Homology and visualizing it secondarily, (b-2-2) ) may include a step (b-2-3) of expressing the new sensor data visualized secondarily as a Betti Sequence and visualized in the third.

Here, the Qusi Attractor technique refers to a method of approximating an attractor by sampling time series data at regular intervals. Persist Homology draws a circle with an increasing radius around each point, and a homology relationship that changes depending on adjacent points. ) is a way to represent In addition, the Betti Sequence is a method of numerically representing points having a homology relationship for each dimension according to a change in radius.

That is, in step (b-2-1), the features of the extracted new sensor data (A in FIG. 4) are visualized using Qusi Attractor (B in FIG. 4), and in step (b-2-2), the phase is changed. Through Persist Homology, which can be expressed as a point, the secondary visualization (C in Fig. 4) is performed. And in step (b-2-3), the new sensor data is visualized according to each dimension (D in FIG. 4 ) so that the characteristics of the phase can be intuitively expressed.

Next, FIG. 5 is a view showing the detailed process of step (b-3) in the process update and failure cause analysis method of the semiconductor device manufacturing process according to an embodiment of the present invention.

As shown in FIG. 5, step (b-3) is detailed in step (b-2-3) of calculating an output value by applying a latent variable to the image visualized tertiary by step (b-2-3) as an input value (b-3- Step 1) and (b-3-2) of deriving a cost function through the input and output values calculated in step (b-3-1), and comparing the preset threshold value and the cost function, the cost When it is determined that the function is lower than the threshold, the selected process is determined as a normal process, and when it is determined that the cost function is higher than the threshold, (b-3-3) of determining the selected process as an abnormal process; can do.

In step (b-3-1), the output value (X') is obtained by applying the latent variable by using the tertiary visualization image derived through each of the above steps (b-2) as the input value (X). will yield

And in step (b-3-2), a cost function is derived through these input and output values, which can be expressed as the following equation.

<Formula 1>

And in step (b-3-3), it is determined whether the selected process is normal/abnormal by comparing the preset threshold and the cost function derived in step (b-3-2).

That is, when it is determined that the cost function is lower than the threshold, the selected process may be determined as a normal process, and when it is determined that the cost function is higher than the threshold, the selected process may be determined as an abnormal process.

In this case, of course, the threshold value may be set differently depending on the type of sensor, and each process of step (b-3) may be performed through an auto encoder.

And for the process determined as a normal process by step (b), new sensor data is updated as reference sensor data in the process reference database.

On the other hand, for the process determined as an abnormal process by step (b), after each process in step (b), step (c) of searching for the cause of the abnormality from the process defect database is performed.

Here, the process failure database is a database constructed from a plurality of semiconductor device manufacturing processes in the past, and may include historical data of abnormal causes and improvement plan data.

And in step (d), if the cause of the abnormal process determined as an abnormal process in step (c) is searched for in the process defect database, follow-up measures are taken for the abnormal cause, and the abnormal process determined as an abnormal process is performed. If the cause cannot be searched for in the process defect database, a new analysis of the cause of the abnormal process is performed.

In this case, the follow-up action in step (d) may include a task of transmitting an abnormal cause of the process to the management terminal and performing a preset alarm.

And step (e) is a process of updating the abnormal cause of the abnormal process analyzed by step (d) in the process defect database.

That is, in step (e), the abnormal cause newly analyzed by step (d) in the corresponding abnormal process is newly added to the process defect database, and the cause of the abnormality is searched for in step (c) of the semiconductor device manufacturing process to be performed later. make it possible

Meanwhile, FIG. 6 is a diagram illustrating a subsequent processing process for the updated abnormal cause in the process update and failure cause analysis method of the semiconductor device manufacturing process according to an embodiment of the present invention.

As shown in FIG. 6 , the present invention may further include a subsequent processing process performed after step (e).

Specifically, after step (e) of updating the abnormal cause of the abnormal process analyzed in step (d) to the process defect database, the abnormal cause updated in the process defect database by step (e) is stored in the process defect database. Step (f) of analyzing the degree of similarity to the previously stored abnormal causes, step (g) of extracting a plurality of improvements recommended according to the degree of similarity analyzed through step (f) and applying them individually to the manufacturing process of the semiconductor device; , It is determined that the abnormal cause has been resolved by the steps (h) and (h) of repeating the manufacturing process of the semiconductor device to which the improvement plan is individually applied by step (g) n times to determine whether the cause of the abnormality is resolved or not In this case, the step (i) of matching the improvement plan and updating the process defect database may be further performed.

In step (f), a process of identifying an existing abnormal cause having a high similarity to the abnormal cause updated in the process defect database during the execution of the new process is performed.

And in step (g), according to the degree of similarity analyzed in step (f), an improvement plan that matches the new abnormal cause and a plurality of existing abnormal causes with high similarity is recommended, and these are individually applied to the semiconductor device manufacturing process. will apply

Then, in step (h), it is determined whether the cause of the abnormality has been resolved by repeating the manufacturing process of the semiconductor device to which each improvement proposal is individually applied n times. It is judged whether it has contributed to the resolution of the abnormal cause.

Accordingly, in step (i), for the improvement proposal that contributed to the resolution of the abnormal cause, the improvement proposal is definitively matched with the abnormal cause, and the process defect database is updated.

As described above, the present invention improves the conventional prediction method based on simple statistics in order to predict the occurrence of an abnormal process in the semiconductor device manufacturing process, objectively determines the process progress status through sensor data, and also when an abnormal cause occurs By analyzing this and enabling follow-up actions, it is possible to significantly increase the production yield of the semiconductor device manufacturing process.

As described above, preferred embodiments according to the present invention have been reviewed, and the fact that the present invention can be embodied in other specific forms without departing from the spirit or scope of the present invention other than the above-described embodiments is a fact having ordinary skill in the art. It is obvious to them. Therefore, the above-described embodiments are to be regarded as illustrative rather than restrictive, and accordingly, the present invention is not limited to the above description, but may be modified within the scope of the appended claims and their equivalents.

Claims (8)

(a) comparing new sensor data acquired from one or more sensors used in a semiconductor device manufacturing process with reference sensor data in a preset normal process to determine whether there is an abnormality in each process;
In the step (a), when it is determined that the new sensor data meets the standard sensor data within a preset allowable range, the process is determined as a normal process and the new sensor data is used as the reference sensor data in a process reference database (b) determining that the process is an abnormal process when it is updated to and when it is determined that the new sensor data is out of the reference sensor data and a preset allowable range;
(c) searching for the cause of the abnormal process determined as the abnormal process in the step (b) from the process defect database;
If the cause of an abnormality in the process determined to be an abnormal process is searched for in step (c), follow-up measures are taken for the abnormal cause. (d) analyzing the cause of the abnormality; and
(e) updating the cause of the abnormality of the abnormal process analyzed by the step (d) in the process failure database;
containing,
Process update and defect cause analysis method in semiconductor device manufacturing process. According to claim 1,
Step (b) is,
(b-1) of extracting new sensor data for a selected process during the semiconductor device manufacturing process;
(b-2) of visualizing the new sensor data extracted by the step (b-1); and
(b-3) determining whether the process selected from the new sensor data visualized through the step (b-2) is normal or abnormal;
containing,
Process update and defect cause analysis method in semiconductor device manufacturing process. 3. The method of claim 2,
The step (b-2) is,
(b-2-1) of first visualizing the new sensor data extracted by the step (b-1) through the Qusi Attractor technique;
(b-2-2) converting the new sensor data first visualized by the step (b-2-1) into Persist Homology to make the second visualization; and
(b-2-3) expressing the new sensor data visualized secondarily by the step (b-2-2) as a Betti Sequence and visualizing the third;
containing,
Process update and defect cause analysis method in semiconductor device manufacturing process. 4. The method of claim 3,
The step (b-3) is,
(b-3-1) calculating an output value by applying a latent variable to the tertiary visualized image by the step (b-2-3) as an input value;
(b-3-2) deriving a cost function through the input value and the output value calculated in the step (b-3-1); and
By comparing a preset threshold value and the cost function, when it is determined that the cost function is lower than the threshold value, the selected process is determined as a normal process, and when it is determined that the cost function is higher than the threshold value, the selected process is selected determining the process as an abnormal process (b-3-3);
containing,
Process update and defect cause analysis method in semiconductor device manufacturing process. 5. The method of claim 4,
The threshold value is set differently depending on the type of sensor,
Process update and defect cause analysis method in semiconductor device manufacturing process. According to claim 1,
Step (d) is,
When the abnormal cause of the process determined as the abnormal process in step (c) is searched for, the abnormal cause is transmitted to the management terminal as a follow-up action, and a preset alarm is performed,
Process update and defect cause analysis method in semiconductor device manufacturing process. According to claim 1,
(f) analyzing the degree of similarity to the abnormal cause previously stored in the process failure database with respect to the abnormal cause updated in the process failure database by the step (e); and
(g) step of extracting a plurality of improvement proposals recommended according to the similarity analyzed in step (f) and individually applying them to the manufacturing process of the semiconductor device;
further comprising,
Process update and defect cause analysis method in semiconductor device manufacturing process. 8. The method of claim 7,
(h) repeating the manufacturing process of the semiconductor device to which the improvement plan is individually applied by the step (g) n times to determine whether the cause of the abnormality is resolved; and
(i) step of updating the process defect database by matching the improvement plan when it is determined that the abnormal cause has been resolved by the step (h);
further comprising,
Process update and defect cause analysis method in semiconductor device manufacturing process.

Images