JP2003028930A - Device and method of test analysis, and recording medium storing test analysis execution procedure program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device and a method for test analysis that can automatically find out an optical testing condition to identify a cause of failure, even if a plurality of causes of failure coexist and are provided with new quantitative means and method that are hard to be affected by correlation between testing conditions and failure density and skill of an analyst. SOLUTION: A test analyzing device 1 is provided at least with a control means 10, a distribution type setting means 11, a testing means 12, an initial condition setting means 13, a characteristics extracting means 14, a characteristics comparing means 15, a test condition changing means 16, an analysis means 17, an output means 18, and a storing means 19. The test condition changing means 16 sets a correction testing condition which is different from a j-th inspection condition, j=j+1, when the comparison result outputted from the characteristics comparing means 15 does not agree, and the correction test conditions is newly set as the j-th test condition this is outputted to the testing means 12, and testing is executed again.

Description

Detailed Description of the Invention

[0001]

The present invention relates to a CCD (Charge C
oupled Device) image receiving device and CMOS (Complementar)
y Metal Oxide Semiconductor) Digital information input device including image sensor, liquid crystal display, plasma display, electroluminescence display and digital information output device including light emitting diode, notification information input / output device including storage element and parallel transmission element, and semiconductor The present invention relates to an inspection analysis technique for semiconductor application devices such as wafers, and particularly to an inspection analysis device and method capable of easily obtaining optimum inspection conditions for analyzing a cause of a defect.

[0002]

2. Description of the Related Art In the inspection of a semiconductor application device represented by an LSI, a technique for determining whether an inspection target is a good product or a defective product in a short time is mainly used. In this type of technology, the test conditions are predetermined, and a pass / fail judgment is made based on whether or not the test conditions are satisfied. Since the purpose of the inspection is pass / fail judgment, the inspection ends when one or more defects appear.

In some inspections, the optimum inspection condition range is determined from the correlation between the inspection condition and the defect density, and the ability of the inspection target to withstand the severeness of the test condition is grasped. Sometimes technology is used. In this type of technology, the inspection condition is gradually changed from a mild condition to a strict condition, re-inspection is continued, and a critical condition at which the quality is switched is determined. Since the purpose of inspection is to grasp the ability, in some cases, one or more defects may be inspected.

Further, as a conventional technique for analyzing inspection result data, Japanese Patent Laid-Open Nos. 11-186354 and 2000 are known.
There is a method proposed by the present inventor disclosed in Japanese Patent Laid-Open No. 200814/2008. In Japanese Laid-Open Patent Publication No. 11-186354, by analyzing the type and frequency of divisors of defective elements,
Disclosed is an inspection analysis device and method capable of qualitatively and quantitatively discriminating the cause of a defect and discriminating between a defect caused by design and a defect not so, and Japanese Patent Laid-Open No. 2000-200814 discloses
Disclosed is a failure distribution analysis system and method capable of easily and quickly analyzing whether the distribution of defective elements in a semiconductor integrated circuit includes a regular distribution or an irregular distribution.

[0005]

In order to analyze the cause of a defect by using the above-mentioned conventional technique, the inspection condition is gradually changed from a mild condition to a severe condition, re-inspection is continued, and a critical condition at which the quality is switched. And one or more defects,
Although the method of inspecting all defects is suitable, there are some problems.

Problems 1. The cause of failure under loose inspection conditions and the cause of failure under severe inspection conditions are not necessarily the same.

Problem 2. If the causes of defects are mixed,
It is difficult to find the optimum inspection condition for identifying the cause of the defect. Particularly, in order to obtain the optimum inspection condition, it is inefficient to repeat the inspection by manually changing the inspection condition while visually confirming the inspection result and the inspection time becomes huge.

Problem 3. Even if the correlation between the inspection condition and the defect density is similar, the cause of the defect is not always the same. If classification is performed based on the subjectivity of the analyst, the analysis result is likely to be affected by the skill level of the analyst, or the analysis result is likely to be a qualitative expression.

The above problems will be described more specifically below.

FIG. 8 is a view for explaining the concept of defect condition dependency of defect density, and is a schematic graph in which the abscissa represents the strictness of the inspection condition and the defect density under each inspection condition is the ordinate. . As shown in FIG. 8, the cause of the dominant defect is changed depending on the inspection condition. As an example, when an adhered matter such as dust is the cause of a defect, it causes a disconnection or a short circuit of the adhered place, and the influence does not depend on the inspection condition, and is often defective. However, semiconductor wafers and TFTs (Thin Film T
A semiconductor application device such as a ransistor) type liquid crystal panel is usually manufactured in a clean room and therefore has a very low dust adhesion probability. Therefore, the defect density due to dust is also very low. On the other hand, wiring fineness, voids, and connection holes (Via-ho
le), poor interface formation of semiconductor PN junction, deterioration of insulation of capacitor film of capacitor and gate oxide film of transistor, etc. may cause characteristic variations such as resistance, capacitance and withstand voltage of defective parts. . The effect depends on the inspection conditions, and the defect density often changes depending on the inspection conditions.
In addition, unlike the dust, the affected area often has a planar spread, so that the defect density tends to increase. If the defect density becomes too high, all defects will be saturated and the dependency on the inspection conditions will not be known. Moreover, the influence ranges of a plurality of failure causes often overlap, which makes identification difficult.

When analyzing the inspection result including the determination of the regularity of the defect distribution, for example, the inspection result of the inspection mother body including U inspection objects is arranged along a specific coordinate (Y) axis, and Often used is a method of analyzing such defects so that they can be identified by Y coordinate values. FIG. 9 is a schematic layout diagram showing the concept of the layout of the inspection object and the inspection result in the one-dimensional coordinate system of only the Y axis, and FIG. 10 is the inspection object and the inspection result in the two-dimensional coordinate system including the Y axis and the Z axis. FIG. 6 is a schematic layout diagram in the case where is arranged. 11 (a), (b), and (c) are diagrams for explaining the concept of a mixture distribution having a planar spread. Examples of irregular distribution, examples of regularity distribution, and examples of mixture distribution are shown in FIG. 12A and 12B are schematic layout diagrams shown in FIGS. 12A, 12 </ b> B, and 12 </ b> C.
This is a diagram for explaining the concept in the case of a mixed distribution having a surface spread, where individual distributions have characteristic variations and also have different dependences on inspection conditions. Examples of irregular distributions, examples of regular distributions, It is a schematic layout drawing which shows an example of each mixture distribution.

As shown in FIG. 11, for example, when the influence ranges of a defect cause showing an irregular distribution and a defect cause showing a regular distribution overlap, the defect distribution becomes a mixed distribution, and the distribution period is larger than that in the case of only the regular distribution. Becomes difficult to detect.

Further, as shown in FIG. 12, when the individual distributions have characteristic variations and the dependences on the inspection conditions are different, it becomes more difficult to detect the distribution period of the regular distribution from the mixture distributions thereof. .

Further, FIGS. 13 (a), 13 (b), 13 (c),
(D) is a diagram for explaining a problem when a distribution cycle is obtained with respect to a defect cause in which the degree of influence on the characteristic value varies, and includes U inspection target examples, defect cause distribution examples, and inspection conditions. FIG. 6 is a layout diagram schematically showing each arrangement of an inspection result example when the inspection condition is strict, an inspection result example when the inspection condition is slightly loose, and an inspection result example when the inspection condition is loose. Moreover, FIG.
4 (a), (b), and (c) are graphs in which the interval between two defects is the horizontal axis and the content rate of each interval is the vertical axis. When the inspection conditions are strict, the inspection conditions are slightly loose, and the inspection is performed. 15 (a), (b), and 15 (a), 15 (b),
(C) is a graph in which the horizontal axis is the divisor included in the interval between two defects and the vertical axis is the content rate of each interval. When the inspection conditions are severe, the inspection conditions are slightly loose, and the inspection conditions are very loose. Each case is shown.

For U inspection objects, the cause of the defect is the period λ.
= 2, the distribution is regular, but it is assumed that the degree of influence on the characteristic value varies. When the inspection conditions are sufficiently strict, all defect causes are determined to be defective regardless of the degree of influence on the characteristic value, and thus the true distribution cycle of the defect causes, λ = 2, can be clearly detected.

On the other hand, when the inspection conditions are loose, the cause of the defect having a small influence on the characteristic value may be determined to be normal.
As a result, when the distribution period is obtained by the conventional mathematical means such as Fourier transform, autocorrelation function, power spectrum, or wavelet transform, d '= 2 × λ, d''which is the interval between the parts determined to be defective. = 3 × λ, d '''
= 4 × λ is used in the same manner as the true distribution period λ, and these become noise for the analysis result, and the detection sensitivity of the true distribution period λ = 2 is reduced.

In the above-mentioned Japanese Laid-Open Patent Publication No. 11-186354, the cause of the defect is qualitatively and quantitatively distinguished by analyzing the divisors of the intervals of the defective elements and the frequencies thereof, and defects caused by the design are analyzed. To distinguish between defects that are not
Further, in Japanese Patent Laid-Open No. 2000-200814, it is determined whether the distribution of defective elements is an irregular distribution or a regular distribution by analyzing the types of divisors of the intervals of defective elements, their frequencies, and expected value functions. However, even if the defect density is extremely small or large, it is possible to change the inspection conditions and easily identify the cause of the defect by simply analyzing the obtained inspection results. Was not considered.

Therefore, an object of the present invention is to improve the conventional inspection analysis apparatus and method, and to provide an inspection mother having a configuration in which a plurality of unit inspection objects are regularly arranged, in the case where a plurality of defect causes are mixed. However, it is possible to automatically find the optimum inspection condition for identifying the cause of the defect, and in addition to the correlation between the inspection condition and the defect density, a new quantitative cause of defect that is not easily affected by the skill of the analyst. It is an object of the present invention to provide an inspection analysis device and method including the identification means and method.

[0019]

Therefore, in the inspection analysis apparatus according to the present invention, when an arbitrary inspection object population including a plurality of inspection units has some arrangement rule regarding the arrangement of the inspection units in the population. , The defect density calculated from the inspection result including the quality judgment result of each of the inspection units of the population, and the inspection results are arranged on a specific coordinate system according to the rule according to the population, and all the defects are Distribution type definition that sets a distribution type group including a plurality of distribution types using the distribution type defined by using the type and number of combinations of two defect intervals and the type and number of divisors included in the interval A means and an analysis target, each of the inspection object in a predetermined inspection mother including a plurality of inspection objects is inspected under predetermined inspection conditions, and the inspection result including the quality judgment result is associated with the inspection object. The inspection means to be applied and the initial inspection condition for each of the inspection objects are set based on the setting information given from the outside, and the initial condition setting is output to the inspection means as the j-th inspection condition when j = 1. Means and
All the inspection objects of the inspection mother are inspected by the inspection means according to the jth inspection condition, and the jth inspection result is arranged on the specific coordinate system according to the rule according to the inspection mother, and the defect distribution of the inspection mother is arranged. And a characteristic extraction unit for extracting the characteristic of the inspection mother, the characteristic of the defective distribution of the inspection mother extracted by the characteristic extraction unit is compared with any distribution type characteristic included in the distribution type group, and the comparison result is compared. When the feature comparison means to output and the comparison result do not match, the j-th
When a modified inspection condition different from the inspection condition is set to j = j + 1 and the modified inspection condition is output to the inspection means as a new jth inspection condition, the comparison result is coincident with each other. Characterized in that it has at least an analysis means for analyzing a defect density and a distribution of defect distribution of the inspection target mother based on the distribution type, an output means for outputting an analysis result, and a control means for controlling the whole. To do.

Further, in the inspection analysis method of the present invention, when an arbitrary inspection object population including a plurality of inspection units has some arrangement rule regarding the arrangement of the inspection units in this inspection population, The defect density calculated from the inspection result including the quality determination result of the inspection unit, and the inspection result are arranged on the specific coordinate system according to the rule according to the population, and the interval between two defects for all defects is A distribution type defining step of setting a distribution type group including a plurality of distribution types using a distribution bias defined by using the kind and number of combinations and the kind and number of divisors included in the interval, and U ( However, U is an integer greater than or equal to 2), and the initial inspection condition for inspecting each of the inspection objects in the inspection mother including the inspection object is j
= 1, the initial condition setting step of setting based on setting information given from the outside as the j-th inspection condition and outputting to a predetermined inspection means, and all the inspections in the inspection mother according to the j-th inspection condition An inspection step of inspecting an object by the inspection means and outputting a j-th inspection result including a quality determination result in association with the inspection object, and the j-th inspection result according to a rule according to the inspection mother A feature extraction step that is arranged on the top and extracts the feature of the defect distribution of the inspection mother, and the feature of the defect distribution of the inspection mother extracted in the feature extraction step is any distribution type included in the distribution type group. Characteristic comparison step of comparing with the characteristic of No. 2 and outputting matching information or non-matching information as a comparison result; and if the comparison result does not match, the j-th inspection condition is different. Set the modified test conditions j = j + 1
When the inspection condition changing step of outputting the corrected inspection condition as a new jth inspection condition to the inspection means and returning to the inspection step and the comparison result are the same, based on the matched distribution type, The method is characterized by including an analysis step of analyzing a defect density and a distribution of defect distribution of the inspection object mother body, and an output step of outputting an analysis result.

At this time, the distribution type defining step includes a first definition process for setting a plurality of distribution types determined according to the defect density of the population to be inspected, and the defect density in which the defect density of the population is specific. When it is within the range, the type and number of combinations of intervals between two defects for all the defects included in the population are examined, and a plurality of distribution biases defined using the types and numbers of divisors included in the interval are used to determine a plurality of types. And a second definition process for setting the distribution type of the above, and the distribution type group may include the distribution types set by the first definition process and the second definition process, respectively.

Further, when the specific coordinate system for arranging the inspection result of the inspection mother body including U inspection objects is a k-dimensional coordinate system including the Y coordinate axis (where k is an integer of 1 or more), the feature extracting step is performed. Includes an arrangement process of arranging the j-th inspection result of the inspection base on a coordinate space including at least the Y coordinate axis, and a calculation process of calculating a defect density Q (j) of the inspection base. A first comparison process for confirming whether the defect density Q (j) calculated in the feature extraction step is included in a specific defect density range; and the defect density Q (j) is included in a specific defect density range. If all the defects included in the inspection result,
The number of combinations and the number N (j) of d = | Δy |, which is the Y coordinate axis component of the interval between two defects, the number uy (j) of the combinations with the interval d = 0, and the maximum value dmax (j of the interval d. ) And the number of combinations Ny (j) with the interval d> 0
= N (j) -uy (j) defect interval calculation processing,
It is checked whether or not dmax (j) ≤ Ny (j) is satisfied, and if it is satisfied, the matching information is output as a comparison result. If not satisfied, the first is if dmax (j)> Ny (j). Second comparison processing for outputting the mismatch information of No. 1 as the comparison result, and the inspection condition changing step, when the comparison result is the first mismatch information, the modified inspection condition becomes stricter than the jth inspection condition. Can be set as follows.

Also, the j-th inspection matrix including U inspection objects
When the number of defects included in the inspection result is n (j) (where n (j) is an integer of 0 or more), the feature comparison step is
The first comparison process further includes an initial inspection confirmation process for confirming j = 1 and an initial defect number confirmation process for confirming 0 <n (1) <U in the case of j = 1. Is j ≧
2 or when 0 <n (1) <U, 0 <α <β <
Α and β that are 1 are set as the preset defect density thresholds, and α
It is examined whether or not <n (j) / U <β is satisfied, and n
(J) / U ≦ α, the first mismatch information, and β ≦ n
In the case of (j) / U, the second mismatch information is output as the comparison result, and the defect interval calculation process is performed by α <
2 for all n (j) defects when m / U <β
Type and number of combinations of intervals d = | Δy | between defects
(J) is checked, and in the inspection condition changing step, when the comparison result is the first disagreement information, the modified inspection condition is set to be stricter than the jth inspection condition, and the comparison is performed. When the result is the second mismatch information, the modified inspection condition may be set to be looser than the j-th inspection condition.

Further, the analysis step is Ny (j) / N
When (j)) is equal to or less than a predetermined distribution type determination value p, the defect distribution included in the jth inspection result is a specific Y
It is possible to include a first distribution type determination process for determining that the type is a homogeneous type that is concentrated on the coordinate values.

Further, in the analysis step, dmax (j) ≤
After satisfying Ny (j), the distance d between two Ny (j) defects
Of divisor f excluding 1 and the divisor calculation procedure for checking all the numbers Σmj (f) of intervals d containing the divisor for each divisor, and the divisors included in Ny (j) combinations A content rate calculation procedure for obtaining the content rate Pj (f) = Σmj (f) / Ny (j) of each number f for all divisors f, and the above Pj (f)
To the expected value function Tj (f) = f × Pj (f)
An expected value calculation procedure for obtaining all divisors f, and Tj
(F) An expected value determination procedure for checking whether or not f with 1 exists, and the expected value function Tj is obtained for all divisors f.
If (f) ≦ 1, it is possible to further include a second distribution type determination process for determining that the defective distribution included in the j-th inspection result is an irregular distribution.

Further, in the analysis step, the result of the second distribution type determination process is an expected value function T at a specific divisor f.
When j (f)> 1, the divisor f that maximizes Tj (f)
z, a maximum value extraction procedure for checking the maximum value T (f) max = T (fz) at this time, and a periodicity determination procedure for checking whether T (fz) = fz holds, T (fz) = Fz is true, the defect distribution included in the j-th inspection result is fz
Is determined to be a regular distribution having a single period, and T (fz) = fz is not established, the defect distribution included in the j-th inspection result is determined to be a vibration type. The distribution type determination process may be further provided.

Further, the output step may include a warning process of outputting predetermined warning information according to the analysis result of the analysis step.

Further, the inspection analysis method of the present invention includes a digital information input device including a CCD image receiving element and a CMOS image sensor, a liquid crystal display, a plasma display, an electroluminescent display and a digital information output device including a light emitting diode, a storage element, and a storage element. A digital information input / output device including a parallel transmission element, and a lot identification symbol indicating the manufacturing order of a semiconductor application device including a semiconductor wafer are set on the Y coordinate axis, and an identification symbol of a line or factory for manufacturing the semiconductor application device in parallel. Is set as the Z coordinate axis, U production lots are subject to inspection, the number of defective lots therein is n, the interval between any two defective lots is d, the type and number N of combinations of the intervals d,
2 The number of combinations in which the defective lot interval d = 0 is set to uy, and the presence or absence of the dependency of the defective lot on the manufacturing lot interval, and the manufacturing date, manufacturing week, manufacturing month, manufacturing season, and manufacturing year that are closely related to the manufacturing lot are included. It is also possible to check whether or not there is a manufacturing time dependency.

[0029]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the present invention will be described with reference to the drawings.

FIG. 1 is a schematic schematic block diagram of an embodiment of the inspection analysis apparatus of the present invention, and FIG. 2 is a schematic flowchart of an inspection analysis method for analyzing inspection results using this inspection analysis apparatus.

Referring to FIG. 1, the inspection analysis apparatus 1 of this embodiment has a control means 10, a distribution type setting means 11, an inspection means 12, an initial condition setting means 13, a feature extracting means 14, and a feature. Comparison means 15 and inspection condition changing means 16
, Analysis means 17, output means 18, and storage means 19
And at least. In addition, the inspection unit 12 is configured to include a test unit 12a and a determination unit 12b.

The distribution type setting means 11, for example, when an arbitrary inspection target population including a plurality of inspection units has some arrangement rule with respect to the arrangement of the inspection units in the population, the distribution type setting means 11 The defect density calculated from the inspection result including the quality judgment result of the inspection unit, and the inspection result are arranged on the specific coordinate system according to the rule according to the population, and the combination of the intervals between two defects for all the defects is set. A distribution type group including a plurality of distribution types is set using the distribution bias defined using the types and numbers and the types and numbers of divisors included in the interval.

The inspecting means 12 is provided in the test section 12a.
It is an analysis target, and the characteristics of each test target in a predetermined test base including a plurality of test targets are measured or read under a predetermined condition, and the determination unit 12b compares the measured value with a predetermined standard value to determine pass / fail. The inspection result including the quality determination result is output in association with the inspection target, and is recorded in, for example, the recording unit 19. The initial condition setting means 13 sets an initial inspection condition for each inspection object based on setting information given from the outside, and outputs it to the inspection means 12 as the j-th inspection condition when j = 1. The feature extracting unit 14 arranges the j-th inspection result obtained by inspecting all the inspection objects included in the inspection mother by the inspection unit 12 according to the j-th inspection condition on the specific coordinate system according to the rule according to the inspection mother, and the inspection mother The characteristics of the defect distribution of are extracted. The feature comparing unit 15 determines that the feature of the defect distribution of the inspection mother extracted by the feature extracting unit 14 is
The distribution type setting unit 11 compares the characteristics of any distribution type included in the distribution type group, and outputs the comparison result. The inspection condition changing unit 16 sets a modified inspection condition different from the jth inspection condition to j = j + 1 when the comparison result output from the feature comparison unit 15 does not match, and sets this modified inspection condition to the new jth. The inspection condition is output to the inspection means 12. When the comparison result output from the feature comparison unit 15 is a match, the analysis unit 17 analyzes the defect density and the bias of the defect distribution of the inspection target mother based on the matched distribution type, and outputs the relevant information from the output unit 18 together with the related information. Output. The control means 10 controls the operation of the inspection analysis apparatus 1 as a whole.

The inspection / analysis apparatus 1 of the present embodiment includes at least the control means 10, the distribution type setting means 11, the initial condition setting means 13, the feature extracting means 14, the feature comparing means 15, the inspection condition changing means 16, and the analysis. The means 17 is a personal computer or an EWS (Engineering Workstation).
) Can be implemented as software on a computer including. Although the inspection means 12 is often composed of an LSI tester, it may be realized as software on a computer depending on the inspection object.
The output means 18 can be configured to include a printer device, a display device, a hard disk device, and the like, and the recording means 19 can be configured to, for example, a hard disk device.

Further, referring to FIG. 2, a method of analyzing the inspection result of the inspection mother to be analyzed by using the inspection analysis device 1 is a distribution type setting step S1, an initial condition setting step S2, and an inspection step S3. And feature extraction step S4
And feature comparison step S5 and inspection condition change step S
6, an analysis step S7, and an output step S8.

The distribution type defining step S1 is such that, when an arbitrary test object population including a plurality of test units has some arrangement rule regarding the arrangement of the test units in this population, the individual test units of the population are The defect density calculated from the inspection result including the pass / fail judgment result, and the inspection result are arranged on a specific coordinate system in accordance with the rule according to the population, and the types of combinations of the intervals between two defects for all defects and A distribution type group including a plurality of distribution types is set by using the number and the distribution bias defined using the type and the number of divisors included in the interval. Here, an example of a specific distribution type setting will be described.

When the defect distribution is a completely irregular distribution: Since the distribution probability is constant regardless of the location, the frequency at which the specific interval d appears even in the interval between two defects, in other words, the weight with which the interval d is distributed. The w (d) does not significantly increase. That is, w (d) is irrespective of the value of d,
It shows an almost constant value.

At this time, if ΣW (2i) is the sum of weights when d is an even number and ΣW (2i + 1) is the sum of weights when d is an odd number, both are almost the same number, and d (> 0). Let Ny be the total number of combinations of ΣW
(2i) ΣW (2i + 1) Ny / 2 may be considered. For example, Σm (2) = w (2) + w (4) + ... = ΣW
(2i) Consider Ny / 2.

Similarly, ΣW (fi) is a multiple fi of d
And sum ΣW (fi + 1), Σ
W (fi + 2), ΣW (fi + 3), ..., ΣW (fi +
f-1), where d is (fi + 1), (fi +
2), (fi + 3), ..., (fi + f−1), the sum of the weights is: ΣW (fi) ΣW (fi + 1) ... ΣW (fi
+ F-1) Ny / f, so Σm (f) = ΣW (f) Ny / f.

For example, when f = 3, ΣW (3i) ΣW (3i + 1) ΣW (3i + 2)
Ny / 3 Σm (3) = w (3) + w (6) + ... = ΣW (3i)
It becomes Ny / 3.

Therefore, the content P (f) of any divisor f,
And the expected value function T (f) is: P (f) = Σm (f) / Ny (Ny / f) / Ny = 1 / f T (f) = f × P (f) f × 1 / f = 1 Becomes As described above, when the defect distribution is a perfect irregular distribution, the expected value function T (f) at any divisor f converges to 1. Therefore, when there is no f that satisfies T (f)> 1, that is, when T (f) ≦ 1 for all f, that is, it is classified as an irregular distribution and a convergence type.

FIG. 6 is an example of a function graph in which the vertical axis represents the actual expected value function T (f) in the case of irregular distribution and the horizontal axis represents f.
In (a), (b), and (c), dmax / Ny is
It is a graph of "when it is larger than 1", "when it is almost 1", and "when it is sufficiently smaller than 1". As can be seen from this graph, even if the distribution is irregular, if dmax / Ny is 1 or more, the expected value function T (f) does not necessarily converge to 1, and the regularity / irregularity of the distribution is determined. It turns out that is difficult.

When the defect distribution is a regular distribution having a single period λ: Assuming an infinite population, the weight w (λi) when the interval d is a multiple λi of the period λ is constant, that is, k
Is an arbitrary integer of 1 or more, w (λ) = w (2λ) = w (3λ) = ... = w (kλ) =
… Is. Therefore, if ΣW (λi) is the sum of weights when d is a multiple λi of λ, the following is obtained.

For example, if there are 16 defective intervals of d = 2 in the number of combinations Ny, Σm (2) = 16, but the defective interval of d = 4 is two defective intervals of d = 2. Minutes, and Σm (4) = Σm (2 × 2) = {Σm (2)} /
2 = 8.

When r is an arbitrary integer satisfying 1 ≦ r ≦ (λ−1) and λir = (λi + r), ΣW (λ
ir) is the sum of weights when d is λir, that is, ΣW (λir) = w (λ + r) + w (2λ + r) + w (3
If λ + r) + ..., Σm (f = λir) = ΣW (λir) = 0.

Therefore, the content P (f) of an arbitrary divisor f,
And an expected value function T (f) is f (k) is a multiple of λ, P (f = k × λ) = Σm (f = k × λ) / Ny = (Ny /
k) / Ny = 1 / k T (f = k × λ) = k × λ × P (f = k × λ) = (k ×
λ) × (1 / k) = λ, and f is not a multiple λi of λ, that is, f = λir, P (f = λir) = Σm (f = λir) / Ny = 0 / Ny = 0 T (F = λir) = λir × P (f = λir) = λir × 0 = 0.

However, when λ itself has several divisors μ, s is an integer of 1 or more and divisor f is a multiple μ of μ.
Under the condition of i = (μ × s) and μi ≠ λi, Σm (f = μ × s) = ΣW ((μ × s) i) = Ny / s. Therefore, P (f = μ × s) = Σm (f = μ × s) / Ny = (Ny /
s) / Ny = 1 / s, and the expected value function T (f) is T (f = μ × s) = μ × s × P (f = μ × s) = (μ ×
s) × (1 / s) = μ. As described above, when the defect distribution is a regular distribution having a single period λ, the expected value function T at a specific divisor f
(F) has a feature of exceeding 1, and T (f) is 0 and λ.
Vibrate between and. Therefore, the case where the regular distribution is included is defined as the vibration type. Furthermore, if f (z) is fz when T (f) shows the maximum value, and if T (f) max = T (fz) = fz is satisfied, this regularity distribution has a single cycle of cycle λ = fz. It can be said that the distribution is regular.

FIG. 7 is an example of a function graph in which the vertical axis represents the actual expected value function T (f) including the regular distribution and the horizontal axis represents f.
In (a), (b), and (c), dmax / Ny is
It is a graph of "when it is larger than 1", "when it is almost 1", and "when it is sufficiently smaller than 1". As can be seen from this graph, when dmax / Ny is 1 or more, the characteristic that T (f) exceeds 1 and vibrates between 0 and λ at the specific divisor f described above becomes unclear. I understand.

In the case of a defect distribution with a small number of defects n and a large variation: For example, the maximum value dmax of the defect interval d.
Is larger than the number Ny of combinations of d (> 0), that is, when dmax / Ny is larger than 1,
As can be seen from (a) and FIG. 8 (a), the expected value function T (f) diverges with respect to the divisor f. Because dmax = a
× Ny (however, a> 1) and frequency Σm of f = dmax (f =
Assuming that dmax) = m (however, m> 1), the content ratio of dmax P (f = dmax) = Σm (f = dmax) / Ny = m
/ Ny, expected value function T (f = dmax) = f × P (f) =
(A × Ny) × (m / Ny) = a × m> 1.

As a specific example, dmax = 1050 and Ny = 2
When 10 (n = 21, uy = 0) and m = 2, dmax / Ny = 1050/210 = 5 P (f = dmax) = 2/210 T (f = dmax) = 1050 × 2/210 = 5 × 2 = 1
It becomes 0. Moreover, a divisor of dmax = 1050, f = 2,3
5,6,7,10,14,15,21,25,30,3
5,42,50,70,75,105,150,17
5,210,350,525 expected value function T (f) = f
From × 2/210, T (2) = 0.019047619 T (3) = 0.028571429 T (5) = 0.047619048 T (6) = 0.057142857 T (7) = 0.066666667 T (10) = 0.095238095 T (14) = 0.133333333 T ( 15) = 0.142857143 T (21) = 0.2 T (25) = 0.238095238 T (30) = 0.285714286 T (35) = 0.333333333 T (42) = 0.4 T (50) = 0.476190476 T (70) = 0.666666667 T (75) = 0.714285714 T (105) = 1 T (150) = 1.428571429 T (175) = 1.666666667 T (210) = 2 T (350) = 3.333333333 T (525) = 5 T (1050) = 10, which increases f Accordingly, it can be seen that T (f) tends to diverge.

When the specific coordinate system is two-dimensional or more and defects are concentrated on specific coordinate values, for example, Y coordinate values: In this case, even if the number of defects n is large, the number of combinations uy for which d = 0 holds. Since it is large and Ny = N−uy, dmax / Ny is 1
As described above, the threshold function F (n, uy) = Ny / N becomes significantly smaller than 1. Therefore, if a distribution type judgment value p (usually preferably p ≦ 0.1) that is sufficiently smaller than 1 is set in advance and F (n, uy) ≦ p, a defect is identified regardless of the value of dmax / Ny. Judge as homogeneous type with concentrated coordinate values.

In the inspection analysis method using the inspection analysis apparatus of this embodiment, as will be described in detail below, in the case of F (n, uy)> p and dmax / Ny> 1, the inspection condition is changed. Since the inspection mother is re-inspected so as to satisfy dmax / Ny ≦ 1, the defect distribution regularity / irregularity determination is not performed in the state of dmax / Ny> 1.

The initial condition setting step S2 is performed when j = 1 as an initial inspection condition for inspecting each inspection object in the inspection mother body including U inspection objects (where U is an integer of 2 or more). The j-th inspection condition is set by the initial condition setting means 13 based on the setting information given from the outside, and is output to the predetermined inspection means 12.

In the inspection step S3, all the inspection objects in the inspection mother are inspected by the inspection means 12 according to the j-th inspection condition, and the j-th inspection result including the quality judgment result is associated with the inspection object, for example, the recording means. Output to 19 and record.

In the feature extracting step S4, the j-th inspection result is arranged on a specific coordinate system in which a distribution type group is set according to a rule according to the inspection mother, and the features of the defect distribution of the inspection mother are extracted.

In the characteristic comparison step S5, does the characteristic of the defective distribution of the inspection mother extracted in the characteristic extraction step S4 match any distribution type characteristic included in the distribution type group set in the distribution type definition step S1? It compares and outputs matching information or mismatch information as a comparison result.

In the inspection condition changing step S6, when the comparison result in the feature comparison step S5 does not match, a modified inspection condition different from the jth inspection condition is set to j = j + 1, and the modified inspection condition is newly added to the jth It is output to the inspection means 12 as the inspection condition, and the process returns to the inspection step S3. If the comparison results are in agreement, the process proceeds to analysis step S7, in which the defect density and defect distribution bias of the inspection object mother are analyzed based on the inconsistent distribution type, and the analysis result is output in output step S8.

Next, the inspection analysis method by the inspection analysis apparatus 1 of this embodiment will be described more specifically.

First, in the distribution type defining step S1, for example, the first thresholds α1, β1 and the second thresholds α2, β of the defect density are set.
2 and the distribution type judgment value p are determined. However, 0 <α1 ≦
α2 <β2 ≦ β1 <1. Further, regarding the distribution type, as described above, a distribution type group including “convergence type”, “oscillation type”, “regular distribution consisting of a single period”, and “homogeneous type” is set. Further, when the number of defects when the first inspection mother body including U inspection objects is inspected under the initial inspection condition is n (1), the defect density Q (1) = n (1) / U is Q (1) =
0,0 <Q (1) ≦ α1, β1 ≦ Q (1) <1, Q
In accordance with (1) = 1, the defect distributions of the first inspection mother body are set as “total number non-defective type”, “irregular type”, “high-density defective type”, and “total number defective type”, respectively.

Next, in the initial condition setting step S2, the initial condition setting means 13 sets it as the j-th inspection condition when the initial inspection condition is j = 1 based on the setting information given from the outside, and the predetermined condition is set. Output to the inspection means 12.

Next, in the inspection step S3, an inspection is performed on each inspection object in the first inspection mother body including U inspection objects by using the predetermined inspection means 12 based on the j-th inspection condition, The j-th inspection result is recorded in the recording means 19. The number of defects detected at this time is n (j).

Next, in the feature extracting step S4, the feature extracting unit 14 reads the j-th inspection result from the recording unit 19, and the number of defects n under the j-th inspection condition of the first inspection mother body.
The defect density Q (j) = n (j) / U is calculated from (j), and the read j-th inspection result is arranged on the Y axis of the specific coordinate system. For example, when the result of quality judgment of the chips on one wafer is all arranged on the Y-axis from the head position according to a certain rule, and there are a plurality of wafers, for example, FIG.
As shown in (a), Z for each wafer is 1, 2, ...
The quality judgment results of all the wafers are shifted in the axial direction, and the quality determination results of all the wafers are 1, 2, ...
It may be arranged like C. In this case, naturally U =
It becomes C × S. As a result, each defect can be identified by the Y coordinate value.

Next, the characteristic comparison step S5 is carried out.
FIG. 3 is a flow chart showing the details of the feature comparing step S5 and the inspection condition changing step S6. Figures 2 and 3
First, in the first comparison process S51, the feature comparison unit 15 confirms whether or not j = 1. If j = 1, it is further determined whether α1 <Q (1) <β1 is satisfied. Check. If α1 <Q (1) <β1 is not satisfied, the process directly proceeds to analysis step S7, where j ≠ 1 or α1 <
If Q (1) <β1 is satisfied, the second comparison processing S53
Then, it is checked whether α2 <Q (j) <β2 is satisfied, including the case of j = 1. If not satisfied, the first mismatch information is output if Q (j) <α2, and the second mismatch information is output if β2 <Q (j), to the inspection condition changing unit 16 in either case. Then, the inspection condition changing step S6 is performed.

Further, when α2 <Q (j) <β2 is satisfied, in the defect interval calculation processing S55, regarding the first inspection result arranged on the Y axis of the specific coordinate system, 2 defects are found for all the defects. The type and the number N (j) of the combinations of the interval d = | Δy | in the Y-axis direction between them are checked, and the maximum value dmax of the interval d is extracted from the checked combinations. Incidentally, FIG.
When two or more dimensions are arranged as shown in (a), a combination with an interval d = 0 can occur, so the number uy
(J) is extracted, Ny (j) = N (j) −uy (j) is calculated, and F (n (j), uy (j)) = Fj (n, uy) = Ny (j).
The value of the threshold function Fj (n, uy) defined by / N (j) is calculated.
(Note that if it is a one-dimensional arrangement, uy (j) = 0 and Ny
(J) = N (j) and Fj (n, uy) = 1. Next, in the third comparison processing S56, Fj (n, uy) is compared with a predetermined distribution type judgment value p, and if Fj (n, uy) ≦ p, the process directly proceeds to analysis step S7.

If Fj (n, uy)> p, then dmax (j) and Ny (j) are compared in the fourth comparison process S57, and d
If max (j)> Ny (j), the first mismatch information is output to the inspection condition changing unit 16, and the process proceeds to the inspection condition changing step S6. If dmax (j) ≦ Ny (j), Move to analysis step S7.

Next, in the inspection condition changing step S6, if the input mismatch information is the first mismatch information, the jth
The condition is set to be stricter than the inspection condition, and in the case of the second mismatch information, changed to the condition looser than the j-th inspection condition. Then, the process returns to the inspection step S3.

Next, details of the analysis step S7 will be described. FIG. 4 is a detailed flowchart of the analysis step S7, and FIG. 5 is a detailed flowchart of the first distribution type determination processing S71 included in the analysis step S7. Referring to FIGS. 1, 2, and 4, the analysis step S7 includes a first distribution type determination processing S71, an expected value calculation processing S72, and a second distribution type determination processing S73, all of which are performed by the analysis unit 17. It is processed.

Referring to FIGS. 2, 4 and 5, first, the first
In the distributed determination process S71, j = j in the first step S711.
Check if 1 and if j = 1, then α1 <Q
(1) Determine whether <β1 is satisfied.

If j ≠ 1, or if α1 <Q (1) <β1 is satisfied, the threshold function Fj is determined in the second step S712.
The value of (n, uy) is compared with the distribution type judgment value p, and Fj (n,
If uy) ≦ p, the failure distribution of the first inspection mother under the j-th inspection condition is classified into the homogeneous type concentrated on the specific Y coordinate value, and the process proceeds to the output step S8, and Fj (n , Uy)>
In the case of p, the process proceeds to the second distribution type determination processing process S72.

If the result of the first step S711 does not satisfy j = 1 and α1 <Q (1) <β1, the value of Q (1) is confirmed in the third step S713 and the first inspection is performed. In the case of Q (1) = 0, the defect distribution of the mother is set to “100% non-defective” and 0.
<Q (1) ≦ α1, “Irregular”, β1 ≦ Q
If (1) <1, it is a "high-density defective type", and Q (1) = 1
In the case of, each is classified into "total defective type", and the process proceeds to the output step S8.

In the second distribution type determination processing S72, d> 0 included in the j-th inspection result in the divisor calculation procedure S721.
The types Σmj (f) of the divisors f other than 1 and the number of intervals Σmj (f) that include the divisors f included in the spacing d of Ny (j) combinations are examined for all divisors f. Next, the content rate calculation procedure S
722 and the expected value calculation step S723, Pj (f) = Σmj (f), where Pj (f) and the expected value function Tj (f) for each value of the divisor f contained in the Ny (j) intervals are / Ny (j) Tj (f) = f × Pj (f), and calculate for all divisors f.

Next, in the expected value determining step S724, Tj
It is checked whether there is f such that (f)> 1. Tj (f)
Third distribution type determination process S if there is f that is> 1
73, and if there is not, that is, Tj for all f
In the case of (f) ≦ 1, the defective distribution of the first inspection mother under the j-th inspection condition is classified into a convergence type which is an irregular distribution, and the process proceeds to output step S8.

In the third distribution type determination processing S73, the value of Tj (f) is maximized under the condition that there is f such that Tj (f)> 1 in the maximum expected value extraction procedure S731, f =
fz and the maximum value Tj (f) max = Tj (f
z) is extracted. Next, in the periodicity determination procedure S732, T
It is examined whether or not j (fz) = fz is satisfied, and Tj (fz) = f
Corresponding to the case where z is satisfied or not satisfied, the j-th
The defect distribution in the j-th inspection result of the first inspection mother based on the inspection conditions is classified into a regular distribution consisting of a single period having a period λ of fz, or a vibration type that is limited to showing that the regular distribution is included. Then, in any case, the process proceeds to the output step S8.

In the output step S8, the analysis step S7
The output result of the analysis result in (1) is output together with the related information.

As described above, in the inspection analysis method using the inspection analysis device 1 of this embodiment, the defect density of the inspection result under the initial inspection condition is divided into a plurality of ranges, and the defect density falls outside the predetermined range. In some cases, the inspection condition was appropriately changed according to the relationship between the defect density value, the number Ny of combinations of defect intervals where d> 0, and the maximum value dmax of defect intervals, and reinspection was performed to obtain the result. Optimal for identifying the cause of defects by having an algorithm that repeats inspection condition changes, re-inspection, and comparison judgment until the defect density matches the defect density range where the regularity / irregularity of the defect distribution can be accurately determined. It has become possible to automatically obtain various inspection conditions in a short time.

It is needless to say that the present invention is not limited to the description of the above embodiment, and various modifications can be made within the scope of the gist thereof.

For example, although the analysis target is a chip on a wafer in the above embodiment, the analysis target may be provided with some regularity in the spatial or temporal arrangement of the inspection target contained in the inspection base. Good. In particular,
Digital information input device including CCD image receiving element and CMOS image sensor, liquid crystal display, plasma display, digital information output device including electroluminescent display and light emitting diode, digital information input / output device including storage element and parallel transmission element, and semiconductor It can be applied to the inspection result of the manufacturing process of a semiconductor application device including a wafer or the internal failure analysis of an individual device. In particular, when analyzing the inspection results of the manufacturing process of these semiconductor applied devices, for example: If the defect density is extremely low, that is, if Q (1) ≦ α1, then dust adheres to the device surface during manufacturing. If the defect density is extremely high, that is, if Q (1) ≧ β1, check whether or not an accident such as an abnormal operation has occurred in the manufacturing equipment or inspection equipment for each device, and If it is judged to be homogeneous type, it is checked whether it is more damaged than dust. ・ If it is judged to be convergent type, it is checked whether inappropriate conditions are set in the manufacturing equipment or inspection equipment. If the determination is made, a warning may be output from the output means 18 together with other related information, which includes, as a processing content, checking whether there is an insufficient margin securing portion on the design drawing.

Further, the lot identification code indicating the manufacturing order of the semiconductor application device is set on the Y coordinate axis, and the identification code of the manufacturing line or factory which manufactures the semiconductor application device in parallel is set on the Z coordinate axis. Each manufacturing lot is subject to inspection, the number of defective lots is n, the interval between any two defective lots is d, and the type and number N of combinations of the intervals d.
And uy is the number of combinations in which the defective lot interval d = 0, and whether or not there is a dependency on the occurrence of a defective production lot, and the production date, production week, production month, production season, and production year that are closely related to the production lot. It is also possible to check whether or not there is a manufacturing time dependency including. In this case, the above-mentioned lot data are recorded in advance in the external recording means 20 such as a hard disk via the communication line 30, and the inspection / analysis apparatus 1 is transferred from the external recording means 20 via the communication line 30. The lot data may be read and analyzed.

The method of changing the inspection condition may be appropriately determined according to the inspection parameter. For example, the starting point and the end point of the inspection condition such as voltage, current, temperature or time (timing) are set in advance, The test conditions are changed stepwise by using the memory divided into equal parts as a memory. ・ If the change in the test results from the starting point to the test point of a few points to a dozen points is small, stepwise by multiple scales. If the inspection conditions are changed to, and if the inspection result changes significantly, the scale is divided into smaller pieces and the inspection conditions are changed little by little.-If all are judged to be non-defective, loose inspection conditions and all are judged to be defective. The gradual division between the strict inspection conditions expected to be the same is used as the scale, and the inspection is alternately performed under the loose inspection conditions and the severe inspection conditions, and the scales are changed one by one toward the midpoint. There are various methods such as predefining several inspection conditions and comparing the presence or absence of changes in the inspection results among the several points. Although not specified in the description of the above embodiment, even if the inspection condition is changed by these inspection condition changing methods, the defect density Q (j) is α2 <Q (j) <β2.
If the value is not satisfied or the value obtained by dividing the maximum value of the defect interval by the number of combinations of defect intervals where the defect interval is not 0 is not 1 or less, it is accurately identified whether the defect distribution includes a regular distribution. You can also judge that it is not possible.

Further, each step of the inspection analysis method of the present invention can be configured as a program executed by a computer, and the program is stored in a computer-readable recording medium such as a CD-ROM and distributed,
The program can be read from the recording medium distributed by the computer device and executed.

[0081]

As described above, the inspection analysis apparatus and the inspection analysis method according to the present invention, when the inspection target contained in the inspection mother has some regularity in the spatial or temporal arrangement in the inspection mother. , The analysis is performed by automatically changing the inspection conditions and re-inspecting until the defect density range suitable for determining whether the defect distribution included in the inspection result of the inspection mother is irregular or not is reached. Since it is possible to automatically guide the defect density range suitable for the above, it is possible to analyze the inspection result with high accuracy in a short time. Further, as a result, a large amount of inspection data is generated, for example, the inspection result of the manufacturing process of the semiconductor application device can be analyzed promptly, and the yield can be improved.

Also,

[Brief description of drawings]

FIG. 1 is a schematic block diagram of an embodiment of an inspection / analysis apparatus of the present invention.

FIG. 2 is a schematic flowchart of the inspection analysis method of the present invention.

FIG. 3 is a flowchart showing details of a feature comparison step and an inspection condition changing step in FIG.

FIG. 4 is a flowchart showing details of the inspection analysis step of FIG.

5 is a flowchart showing details of a first distribution type determination process of FIG. 4. FIG.

FIG. 6 is an actual expected value function T (f) in the case of irregular distribution
Is a vertical axis and f is a horizontal axis.
(B) and (c) are graphs in which dmax / Ny is "greater than 1", "nearly 1", and "well below 1", respectively.

FIG. 7 is a function graph example in which an actual expected value function T (f) including a regularity distribution is plotted on the vertical axis and f is plotted on the horizontal axis.
(B) and (c) are graphs in which dmax / Ny is "greater than 1", "nearly 1", and "well below 1", respectively.

FIG. 8 is a diagram for explaining a concept of defect condition dependency of defect density, and is a schematic graph in which the abscissa represents the strictness of the inspection condition and the defect density under each inspection condition is the ordinate.

FIG. 9 is a schematic layout diagram showing a layout concept of a test object and a test result in a one-dimensional coordinate system having only a Y-axis.

FIG. 10 is a schematic layout diagram when an inspection target and an inspection result are arranged in a two-dimensional coordinate system including a Y axis and a Z axis.

FIG. 11 is a diagram for explaining the concept of a mixture distribution having a planar spread, in which (a), (b), and (c) are examples of irregular distribution, regular distribution, and mixture distribution, respectively. It is a schematic layout drawing shown.

FIG. 12 is a diagram for explaining a concept of a mixed distribution having a planar spread, in which individual distributions have characteristic variations and have different dependences on inspection conditions.
(B), (c) is a schematic layout drawing which shows an example of irregular distribution, an example of regularity distribution, and an example of mixture distribution, respectively.

FIG. 13 is a diagram for explaining a problem in obtaining a distribution period with respect to a cause of a defect in which the degree of influence on a characteristic value varies, and (a), (b), (c), and (d) are , U inspection target examples, defect cause distribution examples, inspection result examples when the inspection conditions are severe, inspection result examples when the inspection conditions are slightly loose, and inspection result examples when the inspection conditions are loose, respectively. It is a layout showing typically.

FIG. 14 is a graph showing the defect distribution of FIG. 13 as an example, with the interval between two defects as the horizontal axis and the content rate of each interval as the vertical axis;
13 (a), (b), and (c) are when the inspection conditions are strict (corresponding to FIG. 13C) and when the inspection conditions are slightly loose (FIG. 13).
(Corresponding to (d)), when the inspection conditions are very loose (Fig. 13).
(Corresponding to (e)).

FIG. 15 is a graph in which the horizontal axis represents the divisor included in the interval between two defects, and the vertical axis represents the content rate of each interval, (a), (b),
In FIG. 13C, when the inspection conditions are strict (corresponding to FIG. 13C), when the inspection conditions are slightly loose (corresponding to FIG. 13D), and when the inspection conditions are very loose (FIG. 13E). Correspondence) is shown respectively.

[Explanation of symbols]

1 Inspection and analysis device 10 Control means 11 Distribution type setting means 12 Inspection means 12a Testing Department 12b Judgment unit 13 Initial condition setting means 14 Feature extraction means 15 Feature comparison means 16 Inspection condition changing means 17 Analytical means 18 Output means 19 Recording means 20 External recording means 30 communication lines

Claims (11)

[Claims] 1. When an arbitrary inspection target population including a plurality of inspection units has some arrangement rule regarding the arrangement of the inspection units in the population, the quality judgment result of each inspection unit of the population The defect density calculated from the inspection result including, the inspection result is arranged on a specific coordinate system according to the rule according to the population, and the type and number of combinations of intervals between two defects for all defects, and Distribution type defining means for setting a distribution type group including a plurality of distribution types by using a distribution bias defined by the type and number of divisors included in an interval, and an analysis target, including a plurality of inspection targets An inspection means for inspecting each of the inspection objects in a predetermined inspection mother under predetermined inspection conditions and outputting an inspection result including a quality determination result in association with the inspection object, and an initial inspection for each of the inspection objects The condition is set based on setting information given from the outside, and initial condition setting means for outputting to the inspecting means as the jth inspection condition when j = 1, and all of the inspection mothers according to the jth inspection condition. A feature extracting means for arranging the j-th inspection result obtained by inspecting the inspection object by the inspecting means on the specific coordinate system according to a rule according to the inspecting mother to extract a characteristic of a defect distribution of the inspecting mother; The characteristic of the defective distribution of the inspection mother extracted by the characteristic extraction means, the characteristic comparison means for comparing and comparing with any distribution type characteristic included in the distribution type group, and outputting the comparison result, the comparison result is In the case of disagreement, a correction inspection condition different from the jth inspection condition is set to j = j + 1, and the correction inspection condition is output to the inspection means as a new jth inspection condition. When the comparison result is a match, an analyzing unit that analyzes a defect density and a defect distribution bias of the inspection target mother based on the matched distribution type, an output unit that outputs an analysis result, and a control that controls the whole An inspection analysis device comprising at least means. 2. When an arbitrary inspection target population including a plurality of inspection units has some arrangement rule with respect to the arrangement of the inspection units in the population, the quality determination result of each inspection unit of the population The defect density calculated from the inspection result including, the inspection result is arranged on a specific coordinate system according to the rule according to the population, and the type and number of combinations of intervals between two defects for all defects, and Distribution type definition step of setting a distribution type group including a plurality of distribution types by using distribution bias defined using the type of divisor and number included in the interval, and U (where U is an integer of 2 or more) ), The initial inspection condition for inspecting each of the inspection objects in the inspection mother including the inspection object is set based on the setting information given from the outside as the j-th inspection condition when j = 1.
An initial condition setting step of outputting to a predetermined inspection means, all the inspection objects in the inspection mother body are inspected by the inspection means according to the jth inspection condition, and the jth inspection result including the quality judgment result is regarded as the inspection object. The inspection step of correspondingly outputting and the j-th according to the rule according to the inspection mother
The inspection result is arranged on the specific coordinate system, and the feature extraction step of extracting the characteristic of the defective distribution of the inspection mother, and the characteristic of the defective distribution of the inspection mother extracted in the characteristic extraction step are in the distribution type group. A feature comparison step of comparing any of the distribution-type features included and comparing and outputting matching information or non-matching information as a comparison result; and a correction inspection different from the j-th inspection condition when the comparison result does not match. When the condition is set to j = j + 1, the modified inspection condition is output to the inspection means as a new jth inspection condition, and the inspection condition changing step of returning to the inspection step and the comparison result are in agreement, And an analysis step of analyzing a defect density and a distribution of defect distribution of the inspection object mother based on the distribution type, and an output step of outputting an analysis result. Inspection analysis how. 3. The distribution type defining step comprises a first defining process of setting a plurality of distribution types determined according to a defect density of a population to be inspected, and a defect density range in which the defect density of the population is a specific defect density range. , The type and number of combinations of intervals between two defects for all the defects included in the population are examined,
A second definition process for setting a plurality of distribution types by using a distribution bias defined by using the type of divisor and the number included in the interval;
The inspection analysis method according to claim 2, wherein the distribution type group includes distribution types set by the first definition process and the second definition process, respectively. 4. The feature extracting step when the specific coordinate system for arranging the inspection result of the inspection mother body including U inspection objects is a k-dimensional coordinate system including the Y coordinate axis (where k is an integer of 1 or more). An arrangement process for arranging the j-th inspection result of the inspection base on a coordinate space including at least the Y coordinate axis;
The feature comparison step includes a calculation process of calculating a defect density Q (j) of the inspection mother, and whether the defect density Q (j) calculated in the feature extraction step is included in a specific defect density range. When the defect density Q (j) is included in a specific defect density range, the first comparison process to be confirmed is the Y coordinate axis component of the interval between two defects for all defects included in the inspection result. d = | Δy |, the number of combinations N (j), the number uy (j) of combinations with an interval d = 0, and the maximum value dmax (j) of the interval d, and the combination with an interval d> 0. Number Ny (j) = N
(J) -uy (j) defect interval calculation processing, and dma
It is checked whether or not x (j) ≤ Ny (j) is satisfied, and if it is satisfied, the matching information is output as a comparison result, and if not satisfied, the first if dmax (j)> Ny (j). Second comparison processing for outputting the mismatch information of No. 1 as the comparison result, and the inspection condition changing step, when the comparison result is the first mismatch information, the modified inspection condition becomes stricter than the jth inspection condition. The inspection analysis method according to claim 2 or 3, which is set as follows. 5. The number of defects included in the j-th inspection result of the inspection base including U inspection objects is n (j) (where n (j) is 0).
The integer above), the feature comparison step is j
Further includes an initial inspection confirmation process for confirming whether = 1 and an initial defect number confirmation process for confirming 0 <n (1) <U in the case of j = 1. , J ≧ 2,
Alternatively, when 0 <n (1) <U, α <n, where 0 <α <β <1, is set as a preset defect density threshold, and α <n
By checking whether or not (j) / U <β is satisfied, n (j) /
If U ≦ α, the first mismatch information is used, and β ≦ n (j) /
In the case of U, the second mismatch information is output as a comparison result, and the defective interval calculation process is performed by α <m / U <
In the case of β, the type and the number N (j) of combinations of intervals d = | Δy | between two defects are checked for all n (j) defects, and the inspection condition changing step is performed by the comparison result. Is the first mismatch information, the modified inspection condition is set to be stricter than the jth inspection condition,
The inspection analysis method according to claim 4, wherein when the comparison result is the second disagreement information, the modified inspection condition is set to be less strict than the jth inspection condition. 6. The analysis step comprises Ny (j) / N
When (j)) is equal to or less than a predetermined distribution type determination value p, the defect distribution included in the jth inspection result is a specific Y
The inspection analysis method according to claim 4 or 5, further comprising a first distribution type determination process for determining the homogeneity type that is concentrated on the coordinate values. 7. The analyzing step comprises dmax (j) ≦ Ny
After satisfying (j), the type of divisor f excluding 1 included in the space d between two Ny (j) defects and the number Σmj (f) of space d including the divisor for each divisor are calculated. A divisor calculation procedure to check all and a content Pj (f) = Σmj (f) / Ny (j) for each divisor f contained in Ny (j) combinations is calculated for all divisors f Rate calculation procedure and the Pj (f)
To the expected value function Tj (f) = f × Pj (f)
An expected value calculation procedure for obtaining all divisors f, and Tj
(F) An expected value determination procedure for checking whether or not f with 1 exists, and the expected value function Tj is obtained for all divisors f.
7. The inspection analysis method according to claim 6, further comprising a second distribution type determination process of determining that the defect distribution included in the jth inspection result is an irregular distribution when (f) ≦ 1. 8. In the analysis step, the result of the second distribution type determination process is an expected value function Tj at a specific divisor f.
When (f)> 1, divisor fz that maximizes Tj (f)
And a maximum value extraction procedure for checking the maximum value T (f) max = T (fz) at this time, and a periodicity determination procedure for checking whether T (fz) = fz holds, T (fz) = When fz is satisfied, it is determined that the defect distribution included in the j-th inspection result is a regular distribution having a single period having a distribution period λ of fz, and when T (fz) = fz is not satisfied, the above-mentioned defect distribution is determined. j
Defect distribution included in inspection result is judged to be vibration type
The inspection analysis method according to claim 7, further comprising a distributed determination process. 9. The inspection analysis method according to claim 2, wherein the output step includes a warning process of outputting predetermined warning information according to an analysis result of the analysis step. 10. A digital information input device including a CCD image receiving element and a CMOS image sensor, a liquid crystal display, a plasma display, a digital information output device including an electroluminescent display and a light emitting diode, a digital information input including a storage element and a parallel transmission element. A lot identification symbol indicating the manufacturing order of the semiconductor application device including the output device and the semiconductor wafer is set on the Y coordinate axis, and an identification symbol of a line or a factory that concurrently manufactures this semiconductor application device is set on the Z coordinate axis. The number of defective lots is n, the interval between any two defective lots is d, the type and number N of combinations of the intervals d, and the interval between two defective lots d = 0. The number of combinations is uy. Deep manufacturing date and the manufacturing lot of the relationship,
The inspection analysis method according to any one of claims 2 to 9, wherein presence or absence of manufacturing time dependency including a manufacturing week, a manufacturing month, a manufacturing season, and a manufacturing year is checked. 11. A computer-readable recording medium in which a software program for executing the inspection analysis method according to any one of claims 2 to 10 is recorded.