JPH10107102A - Inspection method and device for manufacturing semiconductor device and semiconductor manufacturing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an inspection method, an inspection device for manufacturing semiconductor devices stably, and a semiconductor manufacturing method. SOLUTION: A method has a visual inspection device 102 for detecting defects on a semiconductor wafer, a review station 108 for observing the detected defects, and an analysis station 110 for extracting defects in a dense state from among the detected defects as a group of defects and for selecting at least one arbitrary defect out of a plurality of defects that constitute a group of defects as a representative defect. Only the representative defect is observed with respect to a group of defects at the review station 108. In this case, the type of the representative defect is judged and at the same time, all defects that constitute the group of defects are judged to be of the same type as the representative defect.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an inspection method, an inspection apparatus, and a semiconductor manufacturing method in semiconductor device manufacturing.

[0002]

2. Description of the Related Art At present, a semiconductor device manufacturing process includes a technique of forming a predetermined film on a semiconductor wafer, a technique of forming an organic photosensitive film (resist) and transferring and forming a required device pattern on the resist, and a technique of forming a resist. It is based on a variety of technologies, such as a technology for performing an etching process on a pattern and forming a required pattern on a target film.Through these steps, a target device is formed on a semiconductor wafer. You. Device patterns are becoming finer year after year as semiconductor devices become more highly integrated.
Nowadays, those with a line of 1 μm or less are commercialized.

In the course of forming a device pattern, a foreign substance may unexpectedly adhere to or be scratched on a wafer. If these affect device characteristics, the device is treated as a bad device.

[0004] Therefore, in order to stabilize the yield of a production line at a high level, it is necessary to detect defects caused by foreign matters, scratches, etc. at an early stage, to find the cause thereof, and to take effective measures for equipment and processes. is required.

That is, the found defect is usually visually checked (reviewed) by an operator to recognize what the defect is. After that, the defect is classified and the problem process / apparatus producing the defect is specified.

For example, the invention described in Japanese Patent Application Laid-Open No. Hei 6-61314 deals with a plurality of spatially dense defects. This group of defects is sometimes called a cluster defect.

[0007]

Conventionally, the above-mentioned review work has been performed for each defect found in a predetermined device (usually a device called a visual inspection device).

That is, if a large number of defects occur on a semiconductor wafer, the review work time per wafer increases,
Although the number of semiconductor wafers to be inspected per unit time decreases, the invention described in Japanese Patent Application Laid-Open No. 6-61314 does not take any measure against an increase in review time when a cluster defect including a plurality of defects exists. Has not been taken.

[0009] In view of such a problem, an object of the present invention is to provide an inspection method, an apparatus, and a semiconductor manufacturing method in semiconductor device manufacturing that can reduce unnecessary review work.

[0010]

According to one aspect of the inspection method of the present invention for achieving the above object, a defect on a semiconductor wafer is detected in a process of manufacturing a semiconductor device, and the detected defect is observed. By doing so, in the inspection method in the semiconductor device manufacturing to determine the type of the defect, to extract a defect in a dense state from the detected defects as a defect group, from among a plurality of defects constituting the defect group, By selecting one or two or more arbitrary defects, setting them as representative defects, and observing the representative defects, the type of the representative defects is determined, and the defect group other than the representative defects is configured. An inspection method in the manufacture of a semiconductor device is characterized in that all of the defects are determined to be of the same type as the representative defect.

According to one aspect of the inspection apparatus of the present invention for achieving the above object, there is provided an inspection apparatus provided with detection means for detecting a defect on a semiconductor wafer, wherein the detected defects are in a dense state among the detected defects. An inspection apparatus is provided, comprising: extraction means for extracting a plurality of defects as a defect group; and selection means for selecting one or more arbitrary defects from among a plurality of defects constituting the defect group. Is done.

According to one aspect of the semiconductor manufacturing method of the present invention for achieving the above object, in a process of manufacturing a semiconductor device, a group of detected defects is extracted as a group of defects in a dense state. One or two or more arbitrary defects are selected from a plurality of defects constituting the defect group, are set as representative defects, and the type of the representative defect is determined by observing the representative defect. In addition, there is provided a semiconductor manufacturing method characterized by improving or stabilizing the yield by performing an inspection for judging that all the defects constituting the defect group are of the same type as the representative defect. You.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings.

First, the basic concept of the present embodiment is shown in FIG. In the semiconductor manufacturing line 101 shown in the figure, an appearance inspection is performed to stabilize the yield at a high level quickly. Based on the result of the appearance inspection, the type of the defect is determined, and a cause for improving or stabilizing the yield is devised by investigating the cause of the occurrence of the defect. Therefore, it is important to improve the efficiency of this visual inspection for stable production of semiconductor devices. The visual inspection device 102
A wafer stage 103 movable with a semiconductor wafer mounted thereon, a stage control mechanism 104 for controlling a movement amount of the wafer stage 103 and sequentially changing an inspection position on the semiconductor wafer, and recognizing a device pattern at the inspection position; And a pattern recognition mechanism 105 for detecting a defect. The visual inspection apparatus itself is a conventionally well-known apparatus, for example, Hitachi Tokyo Electronics Co., Ltd.
WI series. The position and size of the defect on the semiconductor wafer detected by the visual inspection device 102 are sent to the analysis station 110.

In the analysis station 110, various processes are performed based on the information sent from the visual inspection device 102. Among them, the defect number and the position of the defect (defect coordinates (X, Y)) are sequentially stored in the defect data table 107 provided in the storage device of the analysis station 110. The defect number may indicate, for example, the order in which the defect numbers are found by the visual inspection device 102. At this time, the group number and the category are not yet input. The group number is input when a cluster defect (a defect group on a semiconductor wafer, in which a plurality of defects are densely packed) is extracted and a clustering process is performed in which the defect is treated as one group. The category is input through a review operation for the detected defect. Review work is done at the review station 10
8 is performed.

The review station 108 includes a wafer stage 111 for moving a semiconductor wafer placed thereon, and a stage control mechanism 112 for controlling the amount of movement of the wafer stage 111 in accordance with a review instruction given from the analysis station 110. . Review instructions are:
This refers to a process of extracting the coordinate value of the target defect from the defect data table 107 and sending it to the stage control mechanism 112 of the review station 108. When receiving the coordinate values, the stage control mechanism 112 moves the wafer stage 111 so that the defect falls within the field of view of a microscope (not shown) installed for visual confirmation. The review operation refers to an operation of visually observing a defect within a visual field of a microscope, determining the type of the defect, and classifying the defect according to a defect classification called a category. When such a classification is performed and the cause of the defect can be specified, effective measures are taken for the semiconductor manufacturing line 101 to improve the yield. In the figure, the feedback represents an operation related to this countermeasure.

A feature of the present embodiment is that, when a review instruction is given, all defect coordinates registered in the defect data table 107 are not given to the review station 108, but a cluster defect 109 is placed on the semiconductor wafer 106.
If one exists, one or several of the plurality of defects constituting the cluster defect 109 are selected as representative defects, and their coordinate values are to be sent to the review station 108.

In the case of a cluster defect, usually, a plurality of defects as constituent elements do not have different causes, and the cluster defect itself has a certain cause (for example, during the transfer of a wafer). (Rubbing with the transfer passage, etc.). That is, if the existence of a cluster defect can be recognized, it is extremely useless to perform a cause finding operation for each of the defects constituting the cluster defect.

However, conventionally, an operator performs a review operation on each of the defects detected by the visual inspection device regardless of whether or not a cluster defect exists.

According to the present invention, it is not necessary to perform a review operation on all the defects found, so that useless operations are reduced and a decrease in throughput can be prevented.

Note that all defect coordinates registered in the defect data table 107 are stored in the review station 108.
The same effect can be obtained even if the review station 108 is configured to extract the coordinates of the representative defect.

Next, a method of recognizing a cluster defect will be described with reference to FIG. Specifically, the flow of FIG.
It is executed by the CPU of the analysis station 110.

In step 201 (S201), defect data is read from the defect data table 107. Each defect data includes a defect number, a defect coordinate, a group number, and a category. As described above, the group number and the category have not been input at this time. Each of the read defect data is arranged using a mesh division method in order to reduce the processing time. Although details of the mesh division method will be described later, this mesh is set, for example, as shown in FIG.

In S202, arbitrary defect data (i-th defect) is selected. i increases by 1 each time S202 is executed. N indicates the total number of defects registered in the defect data table 107.

In S203, the same group number as the i-th defect is assigned to each defect existing in the same mesh as the i-th defect. As a result, all the defects in the target mesh are unconditionally recognized as the same group.

In step S204, the determination target area (an area having a predetermined size centered on the mesh of interest)
Select a defect (j-th defect) existing in another mesh within.

In S205, the distance d between the i-th defect and the j-th defect is calculated.

In S206, it is determined whether or not the distance d is equal to or less than a predetermined distance threshold value T. Distance d
Is less than or equal to the distance threshold T, the same group number as the i-th defect is assigned to each defect existing in the same mesh as the j-th defect.

For example, in FIG. 4, when the i-th defect is a defect 401 and the j-th defect is a defect 402,
These belong to the same group. Note that the alternate long and short dash line in the figure is added for explanation, and does not represent the mesh described above.

In S208, another mesh is selected as another mesh in the determination target area, which is focused on in S204. Normally, not all the meshes shown in FIG. 3A have a defect. Therefore, as shown in FIG. 3C, a list structure is created only with the meshes having the defect. It is preferable that only the mesh of the target be processed. The connection between the meshes may be stored using the mesh position information of the mesh data table 302.

If the above operation is performed for all the defects (i is incremented from 1 to N in S202), the defect groups distributed over the mesh can be grouped as one group. For example, when a defect group (cluster defect) 501 as shown in FIG. 5 exists, the same group number (here, group number 1) is given to each of the defects which are its constituent elements. When a plurality of defect groups (that is, a plurality of cluster defects) exist on a semiconductor wafer, a group is formed by the number thereof.

In S209 and S210, it is determined whether or not a cluster defect exists on the semiconductor wafer, and if a cluster defect exists, the processing of S211 to S214 is executed. Here, the linearity of the defect shape is determined for each of the cluster defects.

Next, details of the flow shown in FIG. 2 will be described.

In step S201, as described above, defect data is arranged using the mesh division method.

In the mesh division, a mesh 301 is set on a semiconductor wafer as shown in FIG. In the figure, each mesh is drawn considerably large for the sake of explanation, but actually, it is set more finely. Mesh 30
The size of one side is set to (1 / √2) times the distance threshold value T. The reason will be described later. Information on the position of each mesh is stored in the mesh data table 3
02 is stored. In the mesh position item of the table, the position of each mesh is stored in the form of columns and rows (y rows × x columns).

In step S201, the defect data is read from the defect data table 107, and at the same time, it is determined to which mesh the coordinate values included in the data belong. The defect number is stored in the mesh data table 302 and the defect number link table 304. For example, the item of the head defect number in the mesh data table 302 stores the youngest defect number in each mesh. That is, in the case of FIG. 3B, the defect number of the defect 303 is stored. The item of the number of defects in the mesh data table 302 stores the number of defects in the mesh, although not particularly used in the present embodiment.

The defect number link table 304 is prepared for each mesh. In this table, as shown in the drawing, when a certain defect number is selected, the next defect number to be selected is known.

Such a mesh data table 302
The defect number link table 304 is referred to when all the defects in the target mesh are sequentially selected.

In S202 to S208, dense defects are put together into one group, and it is determined that the defects are cluster defects.

Specifically, first, focusing on any defect,
Defects that exist in the same mesh as the defect of interest are
The same group. Further, the distance d between the defect existing in the adjacent mesh and the defect of interest is calculated using the respective coordinate values, and if the calculated distance d is equal to or smaller than the distance threshold T, All the defects present in the two meshes belong to the same group and are given the same group number. How to assign a group number is based on the following rules.

(1) When the group number of each defect existing in the above two meshes is all 0 (initial value), 2
An unused group number is assigned to each defect in one mesh.

(2) If there is a defect with a group number other than 0 in the two meshes described above, the smallest group number in the two meshes is assigned to each of the two meshes. Give to defects. In S204,
A defect in the determination target area is selected.

In this embodiment, eight meshes surrounding the mesh including the defect of interest are set as the determination target areas. In FIG. 8, reference numeral 801 denotes a mesh including a defect of interest, and reference numeral 802 denotes a determination target area. Since it is useless to determine the distance to a defect that is clearly far from the defect of interest, such a determination target area is provided in the present embodiment.
The determination target area moves whenever the mesh to which the defect of interest belongs moves.

As described above, S201 to S208 have been described. By performing these processes, the following effects can be obtained.

(1) Other defects in the mesh in which the defect of interest exists are determined to be in the same group without performing distance calculation, so that the calculation time is reduced. One side of the mesh is such that the maximum distance in the mesh is T
Since (1 / √2) T is set, there is no need to perform distance calculation.

(2) Since a determination target area is provided, and a distance determination is not performed for a defect that is clearly far from the defect of interest, the processing time can be reduced.

(3) In the determination of the distance between the defect of interest and each defect in the adjacent mesh, if at least one defect satisfying d ≦ D is found, the defect in the adjacent mesh has the same group number as the defect of interest. Is given, and the distance determination on the adjacent mesh ends at that point, so that the number of distance determinations can be reduced.

It should be noted that the actual inspection data is
When the processes from 1 to S208 were performed to distinguish cluster defects from other defects (random defects), when the threshold T was set to about 2000 to 3000 μm,
It was in good agreement with the result of visual confirmation by the operator.

Next, S209 to S214 will be described.

Here, the shape of the cluster defect is recognized.

At S201 to S208, each cluster defect is assigned a different group number.
In steps 210, S211, S212, and S214, the processing of S213 is executed by the number of groups.

In step S213, the correlation coefficient R is calculated by using the xy coordinate values of each defect constituting the cluster defect. The calculation of the correlation coefficient R is performed using (Equation 1).

[0053]

(Equation 1)

Then, | R | is a threshold value G (for example,
It is determined that cluster defects larger than about (0.9) and close to 1 are distributed substantially linearly.

However, when the defects constituting the cluster defect are distributed in parallel to the coordinate axis, | R | indicates an unstable value. For example, in the case of the cluster defect 601 shown in FIG. 6, the distribution is parallel to the y-axis.
y≫σx, and | R | indicates an unstable value. That is, when the defects constituting the cluster defect are distributed parallel to the coordinate axis, | R | ≧ G may not be satisfied even though the distribution is linear. Such a phenomenon can be explained mathematically, but in the present embodiment, assuming such a case, the phase relationship in the coordinate system obtained by rotating the reference coordinate system 603 by a predetermined angle (θ) is assumed. The number R ′ is calculated at the same time as the correlation coefficient R. The cluster defect 601 is expressed in the coordinate system 602 as:
Since it is no longer parallel to the coordinate axes, σy'≫σx 'does not
| R '| shows a stable value.

Then, the correlation coefficients R and R ′ are expressed as “| R | ≧
If G and / or | R ′ | ≧ G ”, it is determined that the cluster defects 601 are linearly distributed. It should be noted that a linear cluster defect may be empirically determined to be a scratch generated by rubbing an obstacle or the like.

On the other hand, in FIG.
After the processing of step 8 is executed, k representative defects are extracted from the plurality of defects constituting the cluster defect. As the extraction method, in the defect data table 107, (1) the first k pieces, (2) the k pieces at random,
(3) There are k last characters. Analysis station 110
Uses the extracted defect data (defect number, defect coordinates, etc.) of the extracted defect as a review instruction in the review station 1
08. At this time, even for a random defect that does not constitute a cluster defect, the defect data is sent to the review station 108.

The stage control mechanism 112 of the review station 108 moves the wafer stage 111 based on the coordinate values included in the defect data, and puts a target defect in the field of view of the microscope. The operator visually confirms the defect and determines the category.

When the category of the representative defect is determined, the fact is notified to the analysis station 110 via the review station 108. In FIG. 7, each category of the representative defects 702 and 703 of the cluster defect 701 is determined to be A, and the fact is notified. In the defect data table 107 of the analysis station 110, all the defect categories constituting the cluster defect 701 are set to A.

Further, as described above, the category of each defect which is a component of the cluster defect determined to have a linear shape in S213 of FIG. 2 can be specified without performing review, as described above. The content of the category (specifically, the damage) is automatically reflected in the defect data table 107.

Although the embodiment of the present invention has been described above, the amount of communication data between the analysis station and the review station slightly increases, but as another embodiment of the present invention, for example, FIG. It may be as shown.

Here, the defect data table 9 is stored between the analysis station 901 and the review station 903.
02 is transmitted and received.

Specifically, first, the analysis station 90
1, after extracting a representative defect, the defect data table 90
At 2, review the review flag of the representative defect.
Set to. Other defects that do not constitute a cluster defect are also set to “review”. All other defects are set to "No Review".
When the setting of the review flag is completed, the defect data table 902 is sent to the review station 903.

In the review station 903, defects not to be reviewed (defects set to “not reviewed”) are skipped, and only defects to be reviewed (defects set to “review”) are sequentially read in the field of view of the microscope. Go inside.

In this embodiment, the data of the defect not to be reviewed (ie, the data of the defect set to “not reviewed”) is also held at the review station 903 side. Can be manually moved to review all defects.

Note that the method of transmitting and receiving data stored in the defect data table 902 is not limited to the network as in this example, and may be performed using a storage medium such as a floppy disk.

Further, even when the functions of the analysis station 901 (for example, a function of extracting a cluster defect, a function of selecting a representative defect, and the like) are mounted on the above-described visual inspection apparatus and review station 903, the present invention can be applied. Same effect.

[0068]

According to the present invention, since it is not necessary to perform a review operation on all the defects constituting the defect group, the inspection time is greatly reduced. If you can shorten the inspection time,
Since the number of wafers that can be inspected within a limited time increases, the inspection can be performed in more steps. Of course, the number of inspections per one process can be increased.

That is, according to the present invention, it is possible to more accurately and promptly grasp the status of a defect in a production line, and it is possible to stably produce semiconductor devices.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing an entire system according to an embodiment of the present invention.

FIG. 2 is a diagram showing a processing flow of an embodiment of the present invention.

FIG. 3 is an explanatory diagram relating to grouping of cluster defects performed in one embodiment of the present invention.

FIG. 4 is an explanatory diagram relating to distance determination performed in one embodiment of the present invention.

FIG. 5 is an explanatory diagram showing the correspondence between the contents of a defect data table used in an embodiment of the present invention and cluster defects.

FIG. 6 is an explanatory diagram related to coordinate conversion performed in an embodiment of the present invention.

FIG. 7 is an explanatory diagram showing a state of a representative defect review performed in an embodiment of the present invention.

FIG. 8 is an explanatory diagram regarding a determination target area used in an embodiment of the present invention.

FIG. 9 is a configuration diagram showing another embodiment of the present invention.

[Explanation of symbols]

101: semiconductor manufacturing line, 102: appearance inspection device,
103, 111: wafer stage, 104, 11
2: stage control mechanism, 105: pattern recognition mechanism,
106: semiconductor wafer, 107, 902: defect data table, 108, 903: review station, 109, 501, 601, 701: cluster defect, 110, 901: analysis station, 301:
Mesh, 302: mesh data table, 30
3: Top defect 304: Defect number link table
401: defect of interest, 402: defect to be determined, 60
2, 603: coordinate system, 702, 703: defect, 80
1: mesh of interest, 802: area to be judged

 ──────────────────────────────────────────────────続 き Continued from the front page (72) Inventor Hideki Nagai 3-2-2 Fujibashi, Ome-shi, Tokyo Hitachi Tokyo Electronics Co., Ltd.

Claims (10)

[Claims] In a process of manufacturing a semiconductor device,
In an inspection method in semiconductor device manufacturing, which detects defects on a semiconductor wafer and observes the detected defects to determine the type of the defects, a group of detected defects in a dense state is extracted as a group of defects. Out of a plurality of defects constituting the defect group, 1 or 2
By selecting any of the above defects and setting them as representative defects, by observing the representative defects, the type of the representative defects is determined, and the defects other than the representative defects, which constitute the defect group, are determined. An inspection method in the manufacture of semiconductor devices, characterized in that all are determined to be of the same type as the representative defect. 2. The method according to claim 1, wherein when a plurality of defects in a dense state are present on the semiconductor wafer, a plurality of defects are extracted as different defect groups, and a representative defect is selected for each defect group. An inspection method in manufacturing a semiconductor device, comprising: 3. The method according to claim 1, wherein a distribution state of the plurality of defects constituting the defect group is recognized, and if the recognized distribution state matches a predetermined distribution state, the plurality of defects are identified. An inspection method in semiconductor device manufacturing, wherein the type is determined to be a predetermined type. 4. The semiconductor device manufacturing method according to claim 3, wherein whether or not the distribution state of the plurality of defects matches a predetermined distribution state is determined by a correlation coefficient of coordinate values of each defect. Inspection method. 5. A coordinate system according to claim 4, wherein a coordinate system (second coordinate system) different from the coordinate system for determining the coordinate values (hereinafter referred to as a first coordinate system) is provided. The coordinate values of each of the defects are obtained, and the correlation coefficient of each of the obtained coordinate values is calculated. Whether or not the distribution state of the plurality of defects matches a predetermined distribution state is determined in the first coordinate system. An inspection method in manufacturing a semiconductor device, wherein the determination is performed using the correlation coefficient and the correlation coefficient in the second coordinate system. 6. An inspection apparatus provided with a detecting means for detecting a defect on a semiconductor wafer, comprising: an extracting means for extracting a group of detected defects in a dense state as a defect group; 1 or 2 out of the defects
An inspection apparatus comprising: a selection unit for selecting any of the above defects. 7. The method according to claim 6, wherein, when a plurality of groups in a dense state are present on the semiconductor wafer, the extracting unit extracts each of the plurality of groups as a different group of defects. An inspection apparatus for selecting a representative defect for each defect group. 8. The defect observation optical system according to claim 6, further comprising: a defect observation optical system; and control means for controlling a relative position of the defect observation optical system with respect to the semiconductor wafer. Controlling the relative position of the defect observation optical system so that only the selected one or two or more defects among the plurality of defects constituting the above sequentially enter the field of view of the observation optical system. Inspection equipment. 9. A means according to claim 6, 7 or 8, wherein means for recognizing a distribution state of a plurality of defects constituting said defect group, and judging whether or not the recognized distribution state matches a predetermined distribution state. An inspection apparatus, comprising: means; and means for determining the type of the plurality of defects as a predetermined type when it is determined that the recognized distribution state matches a predetermined distribution state. 10. In a process of manufacturing a semiconductor device, a group of detected defects in a dense state is extracted as a group of defects, and one or two of the plurality of defects constituting the group of defects are extracted.
By selecting any of the above defects as a representative defect and observing the representative defect, the type of the representative defect is determined, and all of the defects constituting the defect group other than the representative defect are determined. A semiconductor manufacturing method characterized by improving or stabilizing a yield by performing an inspection for judging that a defect is of the same type as a representative defect.