JP5767836B2 - Inspection system, inspection method, and inspection program - Google Patents

Description

  Embodiments described herein relate generally to an inspection system, an inspection method, and an inspection program.

A technique is known in which a plurality of members (for example, a semiconductor element, a magnetic head, a semiconductor light emitting element, etc.) are collectively manufactured using a so-called semiconductor process (semiconductor manufacturing technique).
And in the inspection of the member manufactured collectively using the semiconductor process, sampling inspection is performed as needed. In such a sampling inspection, an inspection is performed for every certain number, or an inspection is performed based on a sampling rule predetermined for each product type.
However, since there are variations in the characteristic values of members manufactured in a batch using a semiconductor process, it is not necessary to perform sampling inspections for a certain number or sampling inspections based on a sampling rule predetermined for each product type. There is a risk that proper grading cannot be performed.

JP 2002-76087 A

  Embodiments of the present invention provide an inspection system, an inspection method, and an inspection program that can perform proper grading.

According to the embodiment, the sampling condition setting unit that inspects a part of the inspection target area and optimizes the setting of the sampling points based on the change in the characteristic value obtained by the inspection of the part of the inspection target area. A sampling inspection unit for collecting characteristic values of members by sampling inspection using the optimized sampling points, and an inspection for determining characteristic values of uninspected members that have not been subjected to the sampling inspection using an interpolation method An interpolating unit, and a classifying unit that creates information on a group of members for each class based on the collected characteristic values of the members by the sampling inspection and the characteristic values of the members obtained by the inspection interpolating unit; An information creating unit that creates desired information based on information about the grade and the group of members, and the sampling condition setting unit performs a first-order differential process on the change in the characteristic value, By increasing the density of the sampling point in the region where the calculated value by the first derivative process is equal to or higher than a predetermined value, and decreasing the concentration of the sampling point in the region where the calculated value by the first derivative process is less than the predetermined value , have rows optimization of setting of the sampling points, the test interpolation unit obtains based prior distribution of characteristic values of the untested members in the past information, and the prior distribution, the sampling inspected member The posterior distribution of the characteristic value of the uninspected member is determined from the probability density determined from the characteristic value of the characteristic value, and the probability density of the characteristic value of the uninspected member is determined based on the posterior distribution. An inspection system is provided.

It is a block diagram for illustrating the inspection system concerning a 1st embodiment. It is a schematic diagram for illustrating a mode that a sampling inspection is performed for every inspection block. FIG. 3 is a schematic graph for illustrating variations in characteristic values in an X direction of an A block illustrated in FIG. 2. It is a schematic graph for demonstrating a mode that the characteristic value of an uninspected member is calculated | required using a spline interpolation method. It is a schematic graph for demonstrating a mode that the characteristic value of an uninspected member is calculated | required using a spline interpolation method. It is a schematic diagram for illustrating the case where the characteristic value of an uninspected member is calculated | required using a spline interpolation method. It is a schematic diagram for illustrating calculation of probability density p (likelihood). It is a schematic graph for exemplifying a state in which a probability of a desired grade is estimated using a predicted distribution w (t ′ | a). It is a flowchart for demonstrating the effect | action of a classification part. (A)-(e) is a schematic diagram for illustrating the effect | action of a classification part. It is a flowchart for demonstrating the effect | action of a sampling condition setting part. (A)-(d) is a schematic process diagram for illustrating the effect | action of a sampling condition setting part. It is a flowchart for demonstrating the effect | action of a test | inspection block division part. (A)-(c) is a schematic process diagram for illustrating the effect | action of a test | inspection block division part. It is a flowchart for demonstrating about the inspection method which concerns on 2nd Embodiment. It is a block diagram for illustrating the computer system which can execute the inspection program concerning a 3rd embodiment.

Hereinafter, embodiments will be illustrated with reference to the drawings. In addition, in each drawing, the same code | symbol is attached | subjected to the same component and detailed description is abbreviate | omitted suitably.
[First embodiment]
FIG. 1 is a block diagram for illustrating an inspection system according to the first embodiment.
As shown in FIG. 1, the inspection system 1 includes a sampling inspection unit 2, an inspection interpolation unit 3, a classification unit 4, an information creation unit 5, a sampling condition setting unit 6, and an inspection block division unit 7.

(Sampling inspection unit 2)
The sampling inspection unit 2 collects the characteristic values of the members obtained by the sampling inspection.
The characteristic values of the members obtained by the sampling inspection can be collected from the sampling inspection apparatus 100.
The sampling inspection apparatus 100 performs a sampling inspection based on a predetermined sampling rule and information provided from a sampling condition setting unit 6 and an inspection block dividing unit 7 described later.

  For example, in a sampling inspection, a plurality of members manufactured in a batch using a semiconductor process can be divided into a plurality of predetermined inspection blocks, and inspection can be performed for each of the divided inspection blocks. .

FIG. 2 is a schematic diagram for illustrating a state in which a sampling inspection is performed for each inspection block.
A to H, J, and K in the figure represent divided inspection blocks, and points in the figure represent sampling points (positions for sampling). Moreover, arrows X and Y in the figure represent two directions orthogonal to each other.
Sampling points can be for a single member or can be for a region containing multiple members. In this case, if the region is a region including a plurality of members, a plurality of members are sampled, so that the inspection accuracy of the sampling inspection can be improved.
As sampling inspection, for example, a member at a sampling point is cut out, and electrical characteristics (for example, voltage characteristics, current characteristics, etc.) of the cut-out member are inspected, or electrical characteristics are measured while the cut-out member is incorporated in a product. Can be exemplified.

  Here, in order to reduce the manufacturing cost of the members and shorten the manufacturing time, a sampling inspection that omits the characteristic inspection of some members is effective. In this sampling inspection, it is possible to perform a quality inspection for each inspection block on the basis of a predetermined sampling rule to perform pass / fail judgment and grading.

However, there are variations in the characteristic values of members manufactured in a batch using a semiconductor process.
FIG. 3 is a schematic graph for illustrating the variation in the characteristic value in the X direction of the A block illustrated in FIG. 2.
Since the characteristic values of the members vary as illustrated in FIG. 3, for example, the grades of all the members included in the inspection block are uniformly determined based on the proportion of each grade determined by the sampling inspection. If it does so, there is a possibility that the reliability of the grade important in production management may be lowered. For example, if the grade of S class products is large as a result of sampling inspection, if the grade of all members included in the inspection block is S grade, other grades (for example, A grade and B grade) Depending on the mixing ratio, the reliability of the assigned grade may be reduced.
Therefore, if grades are assigned without considering variations in characteristic values, there may be a problem in production management or yield management in the process of incorporating a member into a product. In addition, since the added value of the product in which the member is incorporated differs depending on the member grade, grasping the member grade is important in terms of cost management, management of non-defective product rate, and construction period.

Therefore, in the present embodiment, a characteristic value (estimated value) of an uninspected member that has not been subjected to a sampling inspection is obtained.
(Inspection interpolation unit 3)
The inspection interpolation unit 3 obtains a characteristic value of an uninspected member that has not been subjected to a sampling inspection using an interpolation method.
Here, when manufacturing a member collectively using a semiconductor process, it is considered that the characteristic value of the member is continuously changing.
Therefore, the characteristic value of the uninspected member can be obtained based on the characteristic value of the member inspected by the sampling inspection. That is, the characteristic value of an uninspected member between the members subjected to the sampling inspection can be obtained by using the interpolation method.

In this case, the characteristic values of some members obtained by sampling inspection may be greatly different. For example, when damage due to static electricity or the like occurs in some members, the characteristic values of the members are greatly different. Therefore, if the characteristic value of an uninspected member is obtained based on the characteristic value of such a member, there is a possibility that the difference between the obtained characteristic value and the actual characteristic value becomes large.
Therefore, it is possible to remove inappropriate characteristic value information before obtaining the characteristic value of an uninspected member using an interpolation method.
Whether or not the characteristic value is inappropriate can be determined using, for example, an average value and a standard deviation. For example, a characteristic value that does not fall within the range of “average value ± n standard deviation” can be set as an inappropriate characteristic value. Note that n can be a positive integer and can be appropriately changed according to the stability of the semiconductor process, the type of member, and the like.
If the characteristic value of an uninspected member is obtained using an interpolation method after removing the inappropriate characteristic value in this way, the difference between the obtained characteristic value and the actual characteristic value is It is possible to suppress the increase.

"First interpolation method"
As the interpolation method, for example, a spline interpolation method can be used.
FIG. 4 is a schematic graph for illustrating a state in which a characteristic value of an uninspected member is obtained using a spline interpolation method. FIG. 4 shows a case where the characteristic value of an uninspected member in the X direction in an arbitrary inspection block illustrated in FIG. 2 is obtained. The vertical axis represents the characteristic value, and the horizontal axis represents the position of the member in the X direction within the inspection block. X1 to X6 are the positions of the respective members subjected to the sampling inspection.

As illustrated in FIG. 4, first, interpolation functions Sj1 (x) to Sj5 are obtained for each section by piecewise polynomial approximation or the like using the characteristic value of the member subjected to the sampling inspection and the average value of the characteristic values. (X) is obtained.
Similarly, an interpolation function in the X direction is obtained for each coordinate in the Y direction. For example, an interpolation function in the X direction is obtained for each row in the Y direction.

Next, an interpolation function in the Y direction is obtained.
FIG. 5 is a schematic graph for illustrating a state in which a characteristic value of an uninspected member is obtained using a spline interpolation method. FIG. 5 shows a case where the characteristic value of an uninspected member in the Y direction in the inspection block illustrated in FIG. 4 is obtained. The horizontal axis represents the characteristic value, and the vertical axis represents the position of the member in the Y direction within the inspection block. Y1 to Y6 are the positions of the respective members subjected to the sampling inspection.
As illustrated in FIG. 5, first, interpolation functions Sj1 (y) to Sj5 are obtained for each section by piecewise polynomial approximation or the like using the characteristic value of the member subjected to the sampling inspection and the average value of the characteristic values. (Y) is obtained.
Similarly, an interpolation function in the Y direction is obtained for each coordinate in the X direction. For example, an interpolation function in the Y direction is obtained for each column in the X direction.

Then, an interpolation function is obtained in each inspection block.
Note that it is not necessary to obtain an interpolation function in the other direction for a member for which an interpolation function in one direction is obtained. For example, it is not necessary to obtain an interpolation function in the Y direction for a member for which an interpolation function in the X direction is obtained.

By using the interpolation function obtained in this way, the characteristic value of the uninspected member between the members subjected to the sampling inspection can be obtained. Further, it is possible to assign a grade based on the obtained characteristic value.
That is, it is possible to estimate the characteristic values and grades of uninspected members.

"Second interpolation method"
When the characteristic value of an uninspected member is obtained, the estimation accuracy may be reduced if a spline interpolation method is used.
FIG. 6 is a schematic diagram for illustrating a case where the characteristic value of an uninspected member is obtained using a spline interpolation method. In the figure, “●” represents a member that has been inspected by sampling, and “◯” represents a characteristic value (actual characteristic value) of an uninspected member. A line connecting “●” represents a characteristic value obtained using the spline interpolation method.

If the characteristic value of the uninspected member is obtained by using the spline interpolation method, there may be a large difference between the obtained characteristic value and the actual characteristic value as represented by X13 to X16 in FIG. Therefore, in the second interpolation method, the characteristic value of an uninspected member is obtained as follows.
In the second interpolation method, first, a prior distribution of characteristic values of an uninspected member is obtained based on past information, and a probability density p obtained from the obtained prior distribution and the characteristic value of a member subjected to a sampling inspection. From the (likelihood), the posterior distribution of the characteristic value of the uninspected member is obtained. Next, based on the obtained posterior distribution of the characteristic values of the uninspected member, information (predicted distribution) regarding values that can be taken by the characteristic value of the uninspected member is derived.
The past information can be provided from a storage unit (not shown) in which past information is stored. Details regarding past information will be described later.

そして、未検査の部材Ｔ'の周辺に位置する部材の特性値が図７に例示をしたものの場合には、確率密度ｐ（尤度）は以下の（２）式により算出することができる。

また、事前分布を過去情報に基づいて求める。
過去情報としては、例えば、過去において全数検査を行った際に得られた情報、過去において部材を組み込んだ製品を検査した際に得られた情報などを例示することができる。 この場合、品種が同一、部材の位置が同一、検査装置の機種が同一の場合における過去情報に基づいて事前分布を求めるようにすることができる。
未検査の部材Ｔ’と品種が同一、部材の位置が同一、検査装置の機種が同一の部材Ｔ_ｊ（j=1,2,…,m）の特性値ｔ_ｊが互いに独立して平均値θ、分散σ^２の正規分布に従っているとすると、未検査の部材Ｔ’の特性値ｔ’の事前分布ｗ（θ、σ^２）は以下の（３）式で表すことができる。

ここで、μ_０は特性値ｔ_ｊの平均値、ｎ_０は仮想サンプルサイズ、ν_０は逆カイ２乗分布の自由度、λ_０は逆カイ２乗分布のばらつきの尺度パラメータである。ｎ_０、ν_０、λ_０は、部材Ｔ_ｊ（j=1,2,…,m）の特性値ｔ_ｊ、サンプル数ｍ、および特性値ｔ_ｊの測定に対する信頼性の高さに応じて決定する。たとえば、特性値ｔ_ｊの測定に対する信頼性が高いときは、ｎ_０＝ｍ、ν_０＝ｍ−１、λ_０＝（ｍ＋１）×（特性値ｔ_ｊの分散）とする。 Here, the second interpolation method will be illustrated in more detail.
In the second interpolation method, the characteristic value a _i (i = 1, 2,..., N) of the member A _i subjected to the sampling inspection and the characteristic value t ′ of the uninspected member T ′ are the same independently of each other. Is assumed to follow a normal distribution (mean value μ, variance σ ² ).
The characteristic value of an uninspected member can be obtained by the following procedure.
First, the probability density p (likelihood) is calculated from the characteristic values of the members that are sampled and inspected around the uninspected member.
FIG. 7 is a schematic diagram for illustrating the calculation of the probability density p (likelihood).
Here, member A characteristic value of _i is _{a i (i = 1,2, ...} , n) which is sampling inspection probability density as a is assumed that a _i is in accordance with the average value mu, the normal distribution of variance sigma ² Therefore, it can be expressed by the following formula (1).

And when the characteristic value of the member located in the periphery of the uninspected member T ′ is illustrated in FIG. 7, the probability density p (likelihood) can be calculated by the following equation (2).

Further, the prior distribution is obtained based on past information.
Examples of past information include information obtained when 100% inspection has been performed in the past, and information obtained when a product incorporating a member in the past has been inspected. In this case, the prior distribution can be obtained based on past information in the case where the type is the same, the position of the member is the same, and the type of the inspection apparatus is the same.
The characteristic values t _j of the members T _j (j = 1, 2,..., M) having the same product type, the same position of the members, and the same model of the inspection apparatus as the uninspected members T ′ are average values independently of each other Assuming that the normal distribution of θ and variance σ ² is followed, the prior distribution w (θ, σ ² ) of the characteristic value t ′ of the uninspected member T ′ can be expressed by the following equation (3).

Here, μ ₀ is an average value of the characteristic value t _j , n ₀ is a virtual sample size, ν ₀ is the degree of freedom of the inverse chi-square distribution, and λ ₀ is a scale parameter of variation of the inverse chi-square distribution. n _0, ν _0, λ ₀ is member _{T j (j = 1,2, ...} , m) characteristic value _{t j} of the number of samples m, and depending on the reliability of the height to the measured characteristic value _{t j} decide. For example, when the reliability for the measurement of the characteristic value t _j is high, n ₀ = m, ν ₀ = m−1, λ ₀ = (m + 1) × (variance of the characteristic value t _j ).

ただし、ν_１、ｎ_１、λ_１、μ_１は、以下の各式で表されるものとする。

Next, as a = (a ₁ , a ₂ ,..., _An ), the posterior of the characteristic value of the unexamined member is calculated from the product of the prior distribution w (θ, σ ² ) and the probability density p (likelihood). Distribution w (θ, σ | a) is obtained.
In this case, the posterior distribution w (θ, σ | a) can be expressed by the following equation (4).

However, ν ₁ , n ₁ , λ ₁ , and μ ₁ are represented by the following equations.

Next, a predicted distribution w (t ′ | a) of the characteristic value t ′ of the uninspected member T ′ is obtained from the probability density p (likelihood) and the posterior distribution w (θ, σ | a).
The predicted distribution w (t ′ | a) can be expressed by the following equation (5).

Since the predicted distribution w (t ′ | a) obtained in this way is information on values that can be taken by the characteristic value of the uninspected member T ′, the characteristic value of the uninspected member T ′ is a desired grade. Can be estimated.
FIG. 8 is a schematic graph for illustrating a state in which the probability of a desired grade is estimated using the predicted distribution w (t ′ | a).
The characteristic values on the horizontal axis are classified into predetermined grades (S class, A class, B class, C class).

In this case, the area of the portion surrounded by the curve of the prediction distribution w (t ′ | a) included in the range of the desired grade with respect to the area of the portion surrounded by the curve of the prediction distribution w (t ′ | a). The probability that a desired grade is obtained can be estimated from the ratio.
In the case of the example illustrated in FIG. 8, the probability of class S can be estimated by the ratio of the area of the shaded portion to the area of the portion surrounded by the curve of the predicted distribution w (t ′ | a). For example, in the case of the example illustrated in FIG. 8, it can be estimated that the probability of class S is 60%, the probability of class A is 30%, and the probability of class C is 10%.
Then, the grade having the highest probability can be estimated as the grade of the uninspected member T.

Thus, according to the second interpolation method, it is possible to estimate the characteristic value and grade of the uninspected member. Furthermore, the probability of a desired grade can be estimated.
Therefore, since it becomes possible to collect uninspected members having a high probability of obtaining a desired grade, the reliability of the assigned grade can be improved.

(Grading part 4)
The grading unit 4 creates information on the group of members for each class based on the collected characteristic values of the members by the sampling inspection and the characteristic values of the members obtained by the inspection interpolation unit 3.
In the following, as an example, grading for a member for which the probability of a desired grade is calculated by the second interpolation method is illustrated.
FIG. 9 is a flowchart for illustrating the operation of the grading unit 4.
FIG. 10 is a schematic diagram for illustrating the operation of the grading unit 4.
As shown in FIG. 9, first, information regarding the probability of a desired grade is provided from the inspection interpolation unit 3 (step S1).
Next, the order of drawing contour lines to be described later is selected (step S2).
For example, contour lines to be described later can be drawn in order of grade, that is, in descending order of quality.
Next, a contour line having a probability of a desired grade is created, and a flag is attached to the member closest to the center of the region (step S3).
For example, as shown in FIG. 10A, a contour line with a probability of a desired grade is created, and a flag is attached to a member (for example, a member marked with *) that is closest to the center of the region.

Next, a member (first member) having the highest probability of being a desired grade is selected from the members with the flag (step S4).
For example, a member having the highest probability of being a desired grade (a member marked with a star shown in FIG. 10B) is selected.
That is, based on the collected member characteristic values obtained by the sampling inspection and the member characteristic values obtained by the inspection interpolation unit 3, the member having the highest probability of being a desired grade is selected.
Next, it is determined whether or not the probability that the selected member has a desired grade is greater than or equal to a predetermined value (step S5).
If the probability of a desired grade is less than a predetermined value, the above procedure is performed for the lower grade. If the probability of obtaining a desired grade in all grades is less than a predetermined value, all products can be inspected.

Next, a member (second member) that is in the vicinity of the selected member and has the highest probability of being the desired grade is selected to form one group (step S6).
That is, a member that is in the vicinity of the selected member and has the highest probability of being a desired grade is selected, and information on the group of members including the first selected member and the next selected member is displayed. create.
For example, as shown in FIG. 10C, when the member located on the right side has the highest probability of being a desired grade, this member is selected.
Then, as shown in FIG. 10D, information about a group of members including the member with the flag and the selected member is created.

Next, a member (third member) having the highest probability of being the desired grade is selected from the members in the vicinity of the group of members and added to the group (step S7).
That is, a member that is in the vicinity of the group of members and has a high probability of being a desired grade is selected, and information about the group of members that further includes this member is created.
For example, as shown in FIG. 10E, a member having the highest probability of being the desired grade is selected from the members in the vicinity of the group of members and added to the group.
Next, it is determined whether or not the number of members in the group of members is greater than or equal to a predetermined value (step S8).
For example, it is determined whether or not the packaging unit for shipping is reached.

Next, it is determined whether or not the average value of the probability of a desired grade in the group of members is less than a predetermined value (step S9).
Next, a group up to a member immediately before the average value of the probability of a desired grade becomes less than a predetermined value is defined as one group (step S10).
In order to add as many members within an appropriate range as possible to a group of members, a group up to a member immediately before the average value of the probability of obtaining a desired grade is less than a predetermined value is taken as one group.
Next, it is determined whether or not the above procedure has been performed for all members with flags (step S11).
Next, it is determined whether or not the above procedure has been performed for all grades (step S12). Next, information on the grade and group is provided to the information creation unit 5 (step S13).
In addition, the member which does not belong to all the grades can test | inspect individually, and the grade based on an inspection result can be provided.

It is also possible to estimate the risk of contamination with other grades.
That is, an average value and a standard deviation of the characteristic values are obtained from the characteristic values of the members obtained by the sampling inspection and the characteristic values of the members obtained by the inspection interpolation unit 3, and the average value and the standard deviation of the characteristic values are obtained. Based on the above, it is possible to determine the risk of mixing in other grades.
For example, when there is a boundary between grades in the range of “average value of characteristic values ± standard deviation”, it can be estimated that there is a high risk that another grade should be mixed.
Moreover, the information regarding the packaging classification based on the information regarding the danger can be created. For example, it is possible to create information for dividing the packaging according to the degree of danger that other grades are mixed.
In this case, information relating to risk and information relating to packaging based on risk information can also be provided to the information creating unit 5.
Even when the first interpolation method is used, it is possible to estimate the risk of mixing other grades from the probability of a desired grade.

  According to the grading unit 4, grading of uninspected members can be performed appropriately and efficiently. In addition, variations in the characteristic values of the members included in the classified group can be suppressed. Therefore, it is possible to improve the reliability of the grades assigned to the classified groups. In addition, the number of members that need to be inspected to give a grade can be reduced.

(Information creation part 5)
The information creation unit 5 creates desired information (information about packaging for shipping) based on information about the grade and the group of members provided from the classification unit 4.
Moreover, when the information regarding the packaging division based on the information regarding the danger is provided, the packaging can be divided according to this information. In addition, risk information can be attached to the package.

  In addition, when all the goods inspection is performed and the grade based on the inspection result is provided, it can also be made to package for shipment based on the said grade.

In addition, when packing based on the assigned grade is difficult, a block for packing is defined in advance, and the grade of the member included in this packing block or other grade should be used. It is also possible to determine the grade related to the packing block in consideration of the risk of contamination.
In this case, among the grades given to the members included in the packing block, the highest grade can be set as the grade related to the packing block. Further, the grades assigned to the members included in the packing block can be digitized to obtain the average value, and the grade closest to the obtained average value can be used as the grade related to the packing block.
Then, packing for shipping can be performed based on the determined grade.

  Further, the information creation unit 5 can record an ID (Identity Document) of the packing unit. Also, the shipping date and time can be recorded.

(Sampling condition setting unit 6)
When performing a sampling inspection, it is conceivable to simply inspect the characteristics of a member for every certain number.
However, when characteristic values vary due to process fluctuations, if a region with a large change in characteristic value and a region with a small change in characteristic value are inspected at the same sampling interval, a characteristic value estimation failure is found in the region with a large change in characteristic value. May occur. In addition, in the region where the change in the characteristic value is small, the inspection becomes redundant, and the inspection man-hour distribution is wasted. Further, when the members are classified into several classes based on the characteristic values, it is not necessary to know the individual characteristic values, and only the attributes relating to the classes need to be known, so that further inspection is wasted.

  Therefore, the sampling condition setting unit 6 inspects a part of the inspection target region (for example, one line) as a representative, and sets the sampling point based on the change in the characteristic value obtained by the inspection of the part of the inspection target region. To optimize.

FIG. 11 is a flowchart for illustrating the operation of the extraction condition setting unit 6.
FIG. 12 is a schematic process diagram for illustrating the operation of the extraction condition setting unit 6.
As illustrated in FIG. 2, the sampling inspection can be performed for each inspection block. In this case, there may be a variation in characteristic values due to process fluctuations in the inspection block. Then, there may be a difference in variation in characteristic values between the X direction and the Y direction in FIG.
In such a case, it is possible to inspect a part of the inspection block as a representative in a direction in which the characteristic value variation is large.
Here, as an example, a case where a part of the inspection block in the X direction is inspected as a representative is illustrated.

First, as shown in FIGS. 11 and 12A, all members in the uppermost row of the inspection block are extracted (step S21).
That is, all members are extracted in the X direction at the top of the inspection block.
Next, as shown in FIG. 11 and FIG. 12B, the characteristic values of all the extracted members are inspected to obtain a change in characteristic value (profile curve) (step S22).

Next, as shown in FIGS. 11 and 12C, noise removal and smoothing are performed (step S23).
When removing noise, that is, removing information on inappropriate characteristic values, the average value and the standard deviation can be used. For example, characteristic values that do not fall within the range of “average value ± n standard deviation” can be removed as inappropriate characteristic values. Note that n can be a positive integer and can be appropriately changed according to the stability of the semiconductor process, the type of member, and the like.
Smoothing can be performed using, for example, a moving average method.

Next, as shown in FIG. 11 and FIG. 12D, the primary differentiation process is performed (step S24). That is, the inflection point is extracted by first-order differentiation.
Next, an area (high sampling area H) in which the calculated value obtained by the primary differentiation process is equal to or greater than a predetermined value is specified (step S25).

  Next, in the specified high sampling region H, a sampling point is set at a predetermined sampling point density (the number of sampling points per unit area) N1 (step S26). If the value calculated by the primary differentiation process is large, the change in the characteristic value is large. For this reason, the occurrence of a characteristic value estimation failure is suppressed by increasing the density of the sampling points in the region where the calculated value obtained by the primary differentiation process is equal to or greater than a predetermined value.

Next, in a region that is not the high sampling region H, a sampling point is set with a density N2 of a predetermined sampling point (step S27).
If the value calculated by the primary differentiation process is small, the change in the characteristic value is small. For this reason, by reducing the density of the sampling points in the region where the calculated value by the first-order differential processing is less than a predetermined value, it is possible to suppress redundant inspection.
In this way, the sampling condition setting unit 6 performs a primary differentiation process on the change of the characteristic value, and sets the sampling point based on the calculated value by the primary differentiation process.

Note that, before setting the sampling points in the high sampling region H, the sampling points in the region other than the high sampling region H may be set. It is also possible to set sampling points sequentially from a desired direction.
In this case, the concentrations N1 and N2 of the sampling points can be determined in advance according to the stability of the semiconductor process, the type of member, and the like.

In this way, sampling points in the inspection block are set (step S28).
Next, sampling points in other inspection blocks are set in the same manner (step S29).
Information regarding the sampling points set in this way is provided to the sampling inspection apparatus 100 for sampling inspection.

  According to the sampling condition setting unit 6, an appropriate sampling point can be set. Therefore, it is possible to suppress the occurrence of inspection defects and to appropriately distribute the inspection man-hours.

(Inspection block division unit 7)
As described above, in the sampling inspection, a plurality of members manufactured in a batch using a semiconductor process is divided into a plurality of predetermined inspection blocks, and inspection is performed for each of the divided inspection blocks. Can do. Then, it is possible to uniformly divide the inspection target region so that the number of members included in the inspection block is approximately the same.
However, if the region to be inspected is uniformly divided so that the number of members included in the inspection block is approximately the same, there is a possibility that the distribution of characteristic values cannot be optimized. That is, if the inspection target area is divided based on the number of members included in the inspection block, the dispersion of the characteristic values for each inspection block may be greatly different.
Therefore, the inspection block dividing unit 7 optimizes the division of the inspection target region so that the variance of the characteristic values of each block becomes small. As an index for evaluating the dispersion of the characteristic values of each block, for example, the maximum value of the dispersion of the characteristic values may be used, or the difference in the dispersion of the characteristic values may be used.

FIG. 13 is a flowchart for illustrating the operation of the inspection block dividing unit 7.
FIG. 14 is a schematic process diagram for illustrating the operation of the inspection block dividing unit 7.
First, as shown in FIG. 13, the inspection target area is divided into m * n (nm) blocks (step S31).
Note that m is the number of columns in the X direction, and n is the number of rows in the Y direction.
In this case, it is not necessary to divide into equal parts, but here, a case where it is divided into eight equal parts (in the case of m = 4, n = 2) as shown in FIG. 14A will be described as an example.

Next, as shown in FIG. 14B, the dispersion of the characteristic values (σ nm, ijk [k = 1 to nm]) is obtained for each block (step S32).
The characteristic value can be a characteristic value of a member inspected by a sampling inspection and a characteristic value of an uninspected member. Moreover, it can also be set as the characteristic value when all products are inspected.

Next, an average value μnm, ij of the dispersion of characteristic values is obtained (step S33).
Next, by changing the number of divisions, the maximum value σ _MAX nm, ij of the dispersion of characteristic values in each division number is obtained (step S34).
For example, as shown in FIG. 14C, the maximum value σ _MAX nm, ij of the dispersion of characteristic values in each division number is obtained.
In this case, if the number of divisions increases, the maximum value σ _MAX nm, ij of the characteristic value variance gradually decreases.

Next, the minimum number of divisions in which the reduction rate of the maximum variance σ _MAX nm, ij of the characteristic value is less than a predetermined value is obtained (step S35).
For example, the minimum number of divisions in which the reduction rate of the maximum value σ _MAX nm, ij of the characteristic value is less than 0.5% is obtained.
In this case, the reduction rate can be appropriately changed according to the stability of the semiconductor process, the type of member, and the like.
That is, as shown in FIG. 14C, the reduction rate from the maximum dispersion value σ _MAX nm, ij at the division number nm to the maximum dispersion value σ _MAX nm + 1, ij at the division number nm + 1 is less than 0.5%. In this case, the division number nm is a required division number.
Next, inspection block information is created based on the obtained number of divisions (step S36).
The inspection block information created in this way is provided to the sampling inspection apparatus 100 for sampling inspection.
The optimization of the inspection block by the inspection block dividing unit 7 can be performed as necessary, or can be performed periodically.

[Second Embodiment]
Next, an inspection method according to the second embodiment is illustrated.
Note that, in each of the steps illustrated below, detailed description of matters similar to the operations of the inspection system 1 described above will be omitted as appropriate.
FIG. 15 is a flowchart for illustrating the inspection method according to the second embodiment.
First, a member characteristic value sampling inspection is performed based on a predetermined sampling rule (step S41).
Next, a characteristic value of an uninspected member that has not been subjected to the sampling inspection is obtained using an interpolation method (step S42).
In this case, the characteristic value of the uninspected member can be obtained based on the characteristic value of the member inspected by the sampling inspection. That is, the characteristic value of an uninspected member between the members subjected to the sampling inspection can be obtained by using the interpolation method. For example, the characteristic value of the uninspected member can be obtained using the first interpolation method or the second interpolation method described above.
Since the first interpolation method and the second interpolation method are the same as those described above, detailed description thereof will be omitted.

Next, a group of members is formed for each grade based on the characteristic values of the members subjected to the sampling inspection and the determined characteristic values of the uninspected members (step S43).
In this case, for example, grading can be performed according to the procedure illustrated in FIG.
Further, similarly to the case illustrated by the operation of the grading unit 4, it is possible to obtain a risk that another grade should be mixed.
Moreover, the information regarding the packaging classification based on the information regarding the danger can be created.

Next, desired information is created based on the information about the grade and the group of members (step S44).
For example, information for packaging for shipping is created.

The inspection method according to the present embodiment can optimize the setting of the sampling points in the same manner as the operation of the sampling condition setting unit 6 described above (step S51).
That is, a part of the inspection target area (for example, one line) is inspected as a representative, and the setting of the sampling point is optimized based on the change in the characteristic value obtained by the inspection of the part of the inspection target area. can do.
In addition, optimization of a sampling point can be performed according to the procedure illustrated in FIG. 11, for example.

Further, the inspection method according to the present embodiment can optimize the division of the inspection target region in the same manner as the operation of the inspection block dividing unit 7 described above (step S61). In other words, it is possible to optimize the division of the inspection target area so that the variance of the characteristic value for each inspection block is small.
In addition, optimization of division | segmentation of a test object area | region can be performed according to the procedure illustrated in FIG.

[Third embodiment]
Next, an inspection program according to the third embodiment is illustrated.
The inspection program according to the third embodiment can cause a computer to execute the above-described inspection method.
FIG. 16 is a block diagram for illustrating a computer system capable of executing an inspection program according to the third embodiment.
As shown in FIG. 16, the computer system 200 includes a computer 201, an input unit 202, a display unit 203, and a sampling inspection result storage unit 204.
The computer 201 includes an arithmetic unit 201a that executes various types of information processing, a temporary storage unit 201b such as a RAM (Random Access Memory) that temporarily stores information, an input / output unit 201c that controls transmission and reception of information, and a third unit A storage unit 201d in which an inspection program according to the embodiment is stored is provided.
A known technique can be applied to each element provided in the computer 201, and thus detailed description thereof will be omitted.

  In order to execute a series of inspection methods, an inspection program according to the third embodiment is stored in a storage unit 201 d provided in the computer 201. The inspection program can be stored in a storage unit 201d provided in the computer 201 by being supplied to the computer 201 in a state of being stored in a recording medium and being read out. The inspection program may be stored in the storage unit 201d provided in the computer 201 via a LAN (Local Area Network) or the like.

  The inspection program stored in the storage unit 201d is read out to the temporary storage unit 201b, and various calculations are performed in the calculation unit 201a. At this time, necessary information is input from the input unit 202, and calculation results and the like can be displayed on the display unit 203 as necessary. In addition, a sampling inspection result storage unit 204 is connected to the input unit 202. The sampling inspection result storage unit 204 stores the result of the sampling inspection performed based on a predetermined sampling rule.

In this case, an inspection program for executing the following procedure can be stored in the storage unit 201d.
(1) A procedure for collecting characteristic values of members by sampling inspection.
In this case, the sampling inspection results can be collected from the sampling inspection result storage unit 204 via the input / output unit 201c.
(2) A procedure for calculating a characteristic value of an uninspected member that has not been subjected to a sampling inspection using an interpolation method.
(3) A procedure for calculating information on the group of members for each grade based on the collected characteristic values of the members by sampling inspection and the calculated characteristic values of uninspected members.
(4) A procedure for executing output of information on the grade and the group of members.
(5) A procedure for optimizing the setting of sampling points.
(6) A procedure for optimizing the division of the inspection target area.

Since the contents of each procedure can be the same as those exemplified in the inspection system 1 and the inspection method, detailed description thereof is omitted.
In addition, the inspection program according to the third embodiment may be executed in time series according to the above-described order, or may be executed in parallel or selectively even if not necessarily executed in time series. It may be.
Further, the inspection program according to the third embodiment may be processed by a single calculation unit, or may be distributedly processed by a plurality of calculation units.

  According to the embodiments described above, it is possible to realize an inspection system, an inspection method, and an inspection program that can perform appropriate grading.

As mentioned above, although several embodiment of this invention was illustrated, these embodiment is shown as an example and is not intending limiting the range of invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, changes, and the like can be made without departing from the spirit of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the scope of the invention described in the claims and equivalents thereof. Further, the above-described embodiments can be implemented in combination with each other.
For example, the arrangement, the number, and the like of each element included in the inspection system 1 are not limited to those illustrated, but can be changed as appropriate.

  DESCRIPTION OF SYMBOLS 1 Inspection system 2 Sampling inspection part 3 Inspection interpolation part 4 Grading part 5 Information preparation part 6 Sampling condition setting part 7 Inspection block division part 100 Sampling inspection apparatus 200 Computer system 201 Computer 202 Input Section, 203 display section, 204 sampling inspection result storage section

Claims (9)

A sampling condition setting unit that inspects a part of the inspection target area and performs optimization of the setting of the sampling point based on the change in the characteristic value obtained by the inspection of the part of the inspection target area;
A sampling inspection unit that collects characteristic values of members by sampling inspection using the optimized sampling points;
An inspection interpolation unit for obtaining a characteristic value of an uninspected member that has not been subjected to the sampling inspection using an interpolation method;
Based on the collected characteristic values of the members by sampling inspection and the characteristic values of the members obtained by the inspection interpolation unit, a grading unit that creates information on a group of members for each grade,
An information creation unit for creating desired information based on information on the grade and the group of members;
With
The sampling condition setting unit performs a first derivative process on the change in the characteristic value, increases the concentration of the sampling point in a region where the calculated value by the first derivative process is equal to or greater than a predetermined value, and the calculated value by the first derivative process. There by lowering the concentration of the sampling points of the region is less than the predetermined value, it has rows optimization of setting of the sampling points,
The inspection interpolation unit obtains a prior distribution of characteristic values of the uninspected member based on past information, and from the prior distribution and the probability density obtained from the characteristic value of the member subjected to the sampling inspection, An inspection system , wherein a posterior distribution of characteristic values of uninspected members is obtained, and a probability density of characteristic values of the uninspected members is obtained based on the posterior distribution . A sampling inspection unit that collects characteristic values of members by sampling inspection performed for each inspection block that divides the inspection target area;
An inspection interpolation unit for obtaining a characteristic value of an uninspected member that has not been subjected to the sampling inspection using an interpolation method;
Based on the collected characteristic values of the members by sampling inspection and the characteristic values of the members obtained by the inspection interpolation unit, a grading unit that creates information on a group of members for each grade,
An information creation unit for creating desired information based on information on the grade and the group of members;
An inspection block dividing unit for optimizing the division of the inspection target area so that the variance of the characteristic value for each inspection block is reduced;
Equipped with a,
The inspection interpolation unit obtains a prior distribution of characteristic values of the uninspected member based on past information, and from the prior distribution and the probability density obtained from the characteristic value of the member subjected to the sampling inspection, An inspection system , wherein a posterior distribution of characteristic values of uninspected members is obtained, and a probability density of characteristic values of the uninspected members is obtained based on the posterior distribution . The sampling inspection is performed for each inspection block obtained by dividing the inspection target area.
The inspection system according to claim 1, further comprising: an inspection block dividing unit that optimizes the division of the inspection target region so that the variance of the characteristic value for each inspection block is small. The grading unit is a first member having the highest probability of being a desired grade based on the collected characteristic values of the members obtained by the sampling inspection and the characteristic values of the members obtained by the inspection interpolation unit. Select
Select a second member that is in the vicinity of the first member and has a high probability of being a desired grade, and information about the group of members including the first member and the second member. make,
Selecting a third member that is in the vicinity of the group of members and has a high probability of being of the desired grade;
The inspection system according to any one of claims 1 to 3 , wherein information relating to the group of members further including the third member is created. The grading unit obtains the average value and standard deviation of the characteristic values from the characteristic values of the members by the collected sampling inspection and the characteristic values of the members obtained by the inspection interpolation unit,
The inspection system according to any one of claims 1 to 4 , further comprising determining a risk of mixing with another grade based on an average value of the characteristic values and the standard deviation. Inspecting a part of the inspection target area, and performing the optimization of the setting of the sampling point based on the change of the characteristic value obtained by the inspection of the part of the inspection target area;
Performing sampling inspection of the characteristic value of the member based on the optimized sampling point and the sampling rule;
Determining a characteristic value of an uninspected member that has not been subjected to the sampling inspection using an interpolation method;
Forming a group of members for each grade based on the characteristic values of the sampling-examined members and the determined characteristic values of the unexamined members;
Creating desired information based on information about the grade and the group of members;
With
In the step of optimizing the setting of the sampling point, a change in the characteristic value is subjected to a primary differentiation process, and the concentration of the sampling point in a region where a calculated value by the primary differentiation process is equal to or greater than a predetermined value, by lowering the concentration of the sampling points of the area value calculated by the first-order differential processing is less than a predetermined value, it has rows optimization of setting of the sampling points,
In the step of obtaining the characteristic value of the uninspected member that has not been subjected to the sampling inspection using the interpolation method, the prior distribution of the characteristic value of the uninspected member is obtained based on past information, and the prior distribution and The posterior distribution of the characteristic value of the uninspected member is obtained from the probability density obtained from the characteristic value of the sampling-examined member, and the characteristic value of the uninspected member is determined based on the posterior distribution. An inspection method characterized by obtaining probability density . A step of optimizing the division of the inspection target region so that the variance of the characteristic value for each inspection block into which the inspection target region is divided is reduced;
For each of the optimized inspection blocks, performing a sampling inspection of the characteristic value of the member based on a sampling rule;
Determining a characteristic value of an uninspected member that has not been subjected to the sampling inspection using an interpolation method;
Forming a group of members for each grade based on the characteristic values of the sampling-examined members and the determined characteristic values of the unexamined members;
Creating desired information based on information about the grade and the group of members;
Equipped with a,
In the step of obtaining the characteristic value of the uninspected member that has not been subjected to the sampling inspection using the interpolation method, the prior distribution of the characteristic value of the uninspected member is obtained based on past information, and the prior distribution and The posterior distribution of the characteristic value of the uninspected member is obtained from the probability density obtained from the characteristic value of the sampling-examined member, and the characteristic value of the uninspected member is determined based on the posterior distribution. An inspection method characterized by obtaining probability density . Inspecting a part of the inspection target area, causing the setting of the sampling point to be optimized based on the change in the characteristic value obtained by the inspection of the part of the inspection target area,
Collecting characteristic values of members by sampling inspection using the optimized sampling points;
Calculating a characteristic value of an uninspected member that has not been subjected to the sampling inspection using an interpolation method;
Based on the collected characteristic value of the member by sampling inspection and the calculated characteristic value of the uninspected member, calculating information on the group of members for each grade;
Outputting information on the grade and the group of members;
When performing the optimization of the setting of the sampling point, the change of the characteristic value is subjected to a primary differentiation process, the concentration of the sampling point in a region where the calculated value by the primary differentiation process is equal to or greater than a predetermined value, By lowering the density of the sampling points in the region where the calculated value by the first-order differential processing is less than a predetermined value, the sampling point is optimized.
When calculating the characteristic value of an uninspected member that has not been subjected to the sampling inspection using the interpolation method, the prior distribution of the characteristic value of the uninspected member is obtained based on past information, and the prior distribution And, from the probability density obtained from the characteristic value of the member subjected to the sampling inspection, a posterior distribution of the characteristic value of the uninspected member is obtained, and based on the posterior distribution, the characteristic value of the uninspected member Find the probability density of
An inspection program characterized by causing a computer to perform these operations. To optimize the division of the inspection target area so that the variance of the characteristic value for each inspection block into which the inspection target area is divided is reduced;
Collecting the characteristic values of the members by sampling inspection for each of the optimized inspection blocks;
Calculating a characteristic value of an uninspected member that has not been subjected to the sampling inspection using an interpolation method;
Based on the collected characteristic value of the member by sampling inspection and the calculated characteristic value of the uninspected member, calculating information on the group of members for each grade;
Causing the output of information about the grade and the group of members to be executed;
When calculating the characteristic value of an uninspected member that has not been subjected to the sampling inspection using the interpolation method, the prior distribution of the characteristic value of the uninspected member is obtained based on past information, and the prior distribution And, from the probability density obtained from the characteristic value of the member subjected to the sampling inspection, a posterior distribution of the characteristic value of the uninspected member is obtained, and based on the posterior distribution, the characteristic value of the uninspected member Find the probability density of
An inspection program characterized by causing a computer to perform these operations.

Images