JP2002024204A - Method and device for correlation analysis and storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To manifest correlation between an inspection item Wi and the yield of a wafer having an LSI formed hidden behind another inspection item ('on' current) having high correlation with the yield. SOLUTION: While a selection reference value Itmp of the on current is increased gradually in steps from a minimum on current Imin to a maximum on current Imax, the correlation coefficient between the inspection item Wi and yield is calculated as to the wafer whose on current below Itmp is measured and its transition is found. When a peak appears in the transition of the correlation coefficient, it can be estimated that the inspection item Wi has high correlation with the yield within a range below the 'on' current corresponding to the peak.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for analyzing a cause of a defect in a product such as a semiconductor product.

[0002]

2. Description of the Related Art In the efficient production of industrial products, in order to reduce the rate of defective products, that is, to improve the yield, it is effective to perform a defect factor analysis for extracting the defective factors and improve the extracted factors. It is a target.

[0003] A conventional failure factor analysis technique will be described below, taking the case where a product is an LSI as an example.

In this case, a factor to be extracted in the failure factor analysis is a basic characteristic of a device on a wafer measured in an inspection process called a wafer inspection, and a yield is determined by an LSI determined in an inspection process called a probe inspection. Pass rate for the product specifications.

[0005] That is, in wafer inspection, for example, a dedicated measuring device called a wafer tester is used, and for each wafer, elements formed on the wafer, for example, basic characteristics (element characteristics) such as an ON voltage and an ON current of a transistor. To inspect. However, in practice, a wafer tester measures the characteristics of a characteristic monitoring element called TEG formed in a region called scribe between product chips on a wafer. Here, FIG. 14 shows an example of a cross-sectional view of the LSI, and
FIG. 15 shows an example of measurement items of element characteristics measured in a general wafer inspection.

Item (1) in FIGS. 14 and 15 is M
This is a measurement item of the OS transistor characteristic. In this item, for example, an ON current flowing between a source and a drain when an ON voltage is applied to a gate electrode of a MOS transistor is measured. The on-voltage and on-current are items representing the most basic characteristics of a MOS transistor, and are determined when designing the transistor. Item (2) is a measurement item of a characteristic of a through hole connecting between a word line and a wiring layer, between a diffusion layer and a wiring layer, or between a wiring layer and a wiring layer. Then, the resistance of the through hole is measured. Item (3) is a measurement item of characteristics of various layer layers (sheets) such as an oxide film. In this item, the resistance of each layer layer is measured.

As described above, in the wafer inspection, several tens to several hundreds of element characteristics are measured.

On the other hand, in the probe inspection, pass / fail judgment is performed on all product chips on a wafer with respect to product specifications such as logical operation, operating frequency, and power consumption. The percentage of good chips with respect to all the product chips on the wafer obtained as a result of the probe inspection is called a probe yield, and is used as the most basic index for performing a failure analysis.

Conventionally, a correlation analysis method for obtaining a scatter diagram of a yield and a wafer inspection result and a correlation coefficient has been used for such a failure factor analysis of an LSI. The correlation coefficient is a measure indicating the strength of the tendency of a linear correlation between two variables, and takes a value from -1 to 1. The absolute value of the correlation coefficient is 1
Is closer to, there is a linear correlation between the two variables. If it is closer to 0, it is determined that there is almost no correlation. Then, a wafer inspection item having a correlation with the yield may have degraded the yield, and is considered to be highly likely to be a cause of failure.

However, in an actual product, a plurality of element characteristics are related as a cause of a defect, and the yield is sometimes deteriorated. In such a case, even if a correlation analysis is simply performed between the yield and each wafer inspection result, there is a case where the correlation between the element characteristic, which is a failure factor, and the yield does not become apparent. If the correlation analysis does not find an element characteristic correlated with the yield, it is possible to determine whether there is no defect due to the element characteristic or whether a plurality of element characteristics are deteriorating the yield. Will be gone.

On the other hand, even when there is no simple correlation between the yield and the wafer inspection result, a technology that can be expected to manifest element characteristics having a hidden correlation with the yield is disclosed in IEEE Int. Work. on Statistical M
etrology, 1997, pp. 56-61 "A SYSTEMATIC A
PPROACH TO IDENTIFY CRITICAL YIELD SENSITIVE PARAM
ETRIC PARAMETERS ".

[0012] "A SYSTEMATIC APPROACH TO IDENTIFY CR
The technique described in "ITICAL YIELD SENSITIVE PARAMETRIC PARAMETERS" includes the following three steps, as shown in FIGS.

(1) Grouping by Yield As shown in FIG. 16A, wafers to be analyzed are classified into four groups Gr. That is, the classification is performed such that the number of wafers to be analyzed included in each of Gr.

(2) Calculating the average value of yield and wafer inspection results (element characteristic measurement values) As shown in FIG. 16 (2), for each of Gr. Is plotted on a graph, and the average value of the yield and the average value of the wafer inspection results are calculated for each group.

(3) Correlation Analysis As shown in FIG. 16 (3), the average of the yield and the average of the wafer inspection results obtained for each group are plotted on a graph to perform a correlation analysis.

Here, the groups Gr.1 to Gr.4 are grouped irrespective of the value of the wafer inspection result. For this reason, it is considered that the effect of the decrease in the yield due to the arbitrary element characteristics is the same between the groups. Therefore,
According to this technique, even when a plurality of device characteristics are deteriorating the yield, when a correlation analysis between a specific wafer inspection result (device characteristics) and the yield is performed, the decrease in the yield due to other device characteristics is reduced. It is believed that the effects can be removed. In other words, it is considered that even if a simple correlation does not exist between the yield and the wafer inspection result, the wafer inspection result (element characteristic) having a hidden correlation can be made obvious.

[0017]

[Problems to be Solved by the Invention] However, "A SY
STEMATIC APPROACH TO IDENTIFY CRITICAL YIELD SENSI
According to the technology described in “TIVE PARAMETRIC PARAMETERS”, what is obtained is the correlation coefficient and regression line between the average yield of each group and the average value of the wafer inspection results, and these are the correlations between the actual yield and the wafer inspection results. There is no guarantee that they correspond to relation numbers or regression lines. Therefore, when there are a plurality of failure factor characteristics, it is necessary to rank the degree of influence of each wafer inspection item on the yield based on the obtained correlation coefficient and regression line, and to predict or improve the yield improvement effect by the improvement. I can't get an estimate of the amount. In other words, according to the conventional technique, when there is no simple correlation between the product yield and the characteristic item, it is expected that a characteristic item having a hidden correlation with the yield can be realized. The details of the correlation, such as the actual correlation coefficient between the characteristic item and the yield and the regression equation, cannot be revealed.

Accordingly, an object of the present invention is to make it possible to clarify the contents of a predetermined inspection item of a product, for example, even if a simple correlation does not exist between the yield and other characteristic items. .

[0019]

In order to achieve the above object,
The present invention provides a correlation analysis method for analyzing a correlation between a plurality of test items performed on each of a plurality of analysis targets and a predetermined test item.

Specifically, one of the plurality of inspection items is defined as a first inspection item, and a plurality of selection criteria are set between the minimum value and the maximum value of the inspection value of the first inspection item. A first step of setting a value, and selecting one of the plurality of inspection items other than the first inspection item as a second inspection item, and selecting each of the plurality of selection reference values set in the first inspection item The test value of the second test item and the test value of the predetermined test item for each analysis object in which the test value of the first test item belongs to the sample range, And a third step of outputting or analyzing the transition of the correlation coefficient calculated for each of the plurality of selection reference values with respect to an increase or decrease in the selection reference value.

Alternatively, at least two of the plurality of inspection items are set as first inspection items, and a plurality of selection reference values are set between the minimum value and the maximum value of the inspection values of each first inspection item. A first step of setting, and setting one of the plurality of inspection items other than the first inspection item as a second inspection item, and for each of the first inspection items, For each of the selection reference value sets formed by extracting one of the selection reference values from the above, the inspection range of each of the first inspection items is set as a sample range that is equal to or more than each selection reference value included in the selection reference value set. A second step of calculating a correlation coefficient between an inspection value of the second inspection item and an inspection value of the predetermined inspection item for each analysis target whose value belongs to the sample range; and for each of the plurality of selection reference values. Calculated phase Coefficient, a third step of outputting or analyzing the trends for increasing or decreasing the selected reference value, is carried out.

According to such a correlation analysis method, since the correlation between the first test item and the predetermined test item is strong, the correlation between the second test item and the predetermined test item depends on the value of the first test item. If there is a range where the correlation is hidden, as detailed below,
In the transition of the correlation coefficient between the second test item and the predetermined test item with respect to the increase or decrease of the selection reference value, the existence of such a correlation and the range in which the correlation is not hidden clearly appear. Therefore, by targeting such a range, the content of the correlation between the second inspection item and the predetermined inspection item can be made obvious.

[0023]

Embodiments of the present invention will be described below.

FIG. 1 shows a hardware configuration of the characteristic failure analyzing apparatus according to the present embodiment.

As shown in the figure, a characteristic failure analyzer 700 to which the present embodiment is applied includes a CPU 701 and a main memory 71.
1, input devices such as a keyboard 702 and a mouse 703,
A computer having a general configuration including an output device such as a display 704, a storage device 705 such as a hard disk, a storage device 713 handling a portable storage medium 710, a communication control device 712 controlling communication via a network 709, and the like. Can be used.

In this case, the characteristic failure analysis processing described below is realized as a process embodied on an electronic computer by the central processing unit 701 executing a program loaded in the main memory 711. This program is stored in the storage device 705 in advance, and is loaded into the main storage 711 as needed. Alternatively, it is stored in a portable storage medium 710 and directly loaded into the main storage 711 via the storage device 713 as necessary. Alternatively, once the storage device 70 is
After that, the data is loaded into the main memory 711 as needed. Alternatively, after being appropriately received by the communication control device 712 or the like via a communication medium and stored in the storage device 705, it is loaded into the main storage 711 as necessary.

The characteristic failure analysis processing performed by the characteristic failure analysis device 700 will be described below.

Prior to the characteristic failure analysis processing, a network test is performed from the wafer tester 706 and the probe tester 707.
9 or information about the yield or wafer inspection result is stored in the storage device 705 of the characteristic failure analysis device 700 via the recording medium 710.

FIG. 2 shows an example of the wafer inspection result information sent from the wafer tester 706, and FIG. 3 shows an example of the probe inspection result information sent from the probe tester 707.

2, the wafer inspection result information of FIG. 2 has a format in which a lot number 801, a wafer number 802, a wafer inspection item name 803, and a measured value 804 of the wafer inspection item 803 are described in one record. Have. The probe test result information in FIG.
1, a wafer number 902 and a yield 903 of the wafer specified by the wafer number 902 are described in one record.

Thereafter, when the characteristic failure analysis processing is started, first, the probe inspection result information and the wafer inspection result information stored in the storage device 705 are shaped, and
The correlation trace data as shown in FIG. 4 is created.

As shown, the correlation trace data includes lot number 1001, wafer number 1002, yield 1003, and measured value 1004 for each wafer inspection item.
1006 have the format described in one record.

Next, in the characteristic failure analysis processing, a correlation trace condition setting screen 1101 as shown in FIG.

On this correlation trace condition setting screen 1101, the user can select from the scatter diagram vertical axis designation menu 1102.
The parameter on the vertical axis of the scatter diagram, that is, the parameter to be analyzed is designated as the vertical axis parameter. The vertical axis parameters include the yield, the occurrence rate of an arbitrary defect obtained by a probe test, and the yield
S or a wafer inspection item can be designated.

Next, from the sampling parameter menu 1103, the user designates a parameter used as a reference in the analysis as a sampling parameter. With the sampling parameters, it is possible to specify a wafer inspection item such as an on-current or a leak current.

Next, the user sets the trace direction menu 1
From 104, a trace direction is specified as to whether to set the wafer to be analyzed in ascending order of the measured value when sampling the wafer or to set the wafer to be analyzed in ascending order of the measured value.

Finally, the user sets the trace condition 1105
Is set, and the start button 1106 is selected.

When the start button 1106 is selected,
The characteristic failure analysis processing is performed on the correlation trace condition setting screen 110.
The correlation trace processing is executed according to the condition set in step 1.

Hereinafter, the correlation trace processing will be described.

Here, as an example, a case where the cause of the characteristic failure analysis is the yield related to the operating frequency of the LSI and the cause of the failure is analyzed will be described.

Now, among the wafer inspection items, FIG.
The ON current of the MOS transistor shown in (1) is LS
Dominantly affects the operable frequency of I. This is because the larger the on-state current, the faster the load capacitance such as the gate capacitance and the overlap capacitance can be charged, so that the operating frequency of the LSI is considered to be improved.

The wafer inspection items other than the on-current are also:
This is considered to affect the operable frequency of the LSI.
However, since the on-current strongly influences the operable frequency of the LSI, the correlation between the other wafer inspection items and the yield related to the operating frequency of the LSI is hidden by this, and cannot be revealed by the conventional correlation analysis. There are cases.

Therefore, in the present embodiment, the model shown in FIG. 6 is assumed.

This model is provided between an on-current and a certain wafer inspection item Wi other than the on-current, and an operating frequency.
This indicates that there is a correlation.

A horizontal line 301 in the figure represents a pass / fail judgment criterion relating to the operating frequency of the LSI, and a probe whose operable frequency does not satisfy the pass / fail judgment criterion is judged to be defective by the probe test. Also, Imin and Imax in the figure are the minimum value and the maximum value of the on-current, respectively, and Ith represents the condition of the on-current (the minimum value of the on-current) at which the product always operates above the pass / fail judgment reference value. .

The variation of the operating frequency with respect to the same on-current value indicates the influence of the wafer inspection item Wi on the operating frequency. In the probe inspection at the mass production stage, usually, a binary pass / fail judgment (GO / NO GO test) as to whether or not a product operates at an operating frequency higher than a pass / fail judgment standard in consideration of test cost and test time. It is carried out. Therefore, at the stage of mass production, the operating frequency of each chip is unknown.

In the drawing, in the wafer I1 whose on-current is smaller than Ith, the operating frequency changes across the pass / fail judgment criterion 301 depending on the wafer inspection item Wi. If the operating frequency is equal to or higher than the pass / fail judgment criterion 301, the product is a non-defective product, and if the operating frequency is lower than the pass / fail judgment criterion 301, it is a defective product. On the other hand, in the wafer of I2 whose on-current is larger than Ith, the on-current is sufficiently large, so that the operation frequency is always non-defective and the wafer inspection item W
The variation of the operating frequency due to i is not reflected in the yield.

In a portion where the on-current is larger than Ith, the variation in the operating frequency due to the wafer inspection item Wi is not reflected in the yield. Therefore, as described in the background art, the yield is targeted at a wafer selected at random. When the correlation analysis between the wafer inspection item Wi and the yield related to the operating frequency is weakened by performing the correlation analysis between the wafer inspection result and the wafer inspection result, it is considered that this correlation cannot be realized.

On the contrary, conversely, if the correlation analysis between the yield and the wafer inspection result Wi is performed in advance only for the wafers whose on-current is less than Ith, the correlation between the yield and the wafer inspection result Wi becomes apparent. It is considered possible.

However, unlike the model, actually, not only the on-current and the wafer inspection item Wi, but also other wafer inspection items affect the operating frequency. That is, in practice, the variation of the operating frequency with respect to the same on-current value is the total of the influence on the operating frequency by all the wafer inspection items that affect the operating frequency. For this reason,
Ith when the model of FIG. 6 is applied to each wafer inspection item as Wi cannot be derived only from the relationship between the on-current and the yield.

Therefore, in the present embodiment, the transition of the correlation coefficient between the yield and each wafer inspection result is traced while adding the correlation analysis target in order from the wafer with the smallest on-current, so that each wafer inspection item can be traced. Ith and I
A correlation tracing process for finding a correlation coefficient between each wafer inspection item and the yield in a range smaller than th is performed.

FIG. 7 shows the principle of the correlation trace processing.

FIG. 7A shows the relationship between the ON current and the operating frequency. With respect to the wafer inspection item Wi that causes a variation in the operating frequency, it is considered that the correlation with the yield becomes strongest when the correlation analysis is performed on a wafer whose on-current is less than Ith when the model of FIG. 6 is applied. .
Therefore, as shown in FIG. 2B, the transition of the correlation coefficient with the yield is plotted for each wafer inspection item while the analysis target is expanded in order from the wafer with the smallest on-current. Then, an item in which a peak appears in the correlation is extracted as an operating frequency variation factor, that is, a characteristic failure factor,
Obtain the correlation coefficient and the regression equation when the peak appears in the correlation.

Now, when performing such correlation trace processing for analyzing the cause of the characteristic failure as the yield related to the operating frequency of the LSI as the target of the characteristic failure analysis,
On the correlation trace condition setting screen 1101, specify the yield or the occurrence rate of defects related to the operating frequency as the vertical axis parameter, specify the on-current as the sampling parameter, and specify the analysis target as the trace direction in order from the wafer with the smallest measured value. Specifies to be a wafer.

In this case, in the correlation trace processing, the processing is performed according to the procedure shown by the PAD diagram in FIG.

That is, first, the parameter i is initialized to 1 (step 600). And the vertical axis parameter,
Here, the increment ΔI of the on-current is obtained (step 60).
1). When the sampling parameter step size is designated as the trace condition 1105 on the correlation trace condition setting screen 1101, ΔI is the designated value,
When the sampling parameter step size n is specified, the vertical axis parameter obtained from the correlation trace data in FIG. 4, that is, the maximum value Ima of the on-current, here
x and the minimum value Imin, (Imax−Imi
n) / n is ΔI.

Next, while incrementing i by 1 (step 6
06) Until i becomes the number m-1 of all wafer inspection items (step 602), the following processing is performed for the i-th wafer inspection item Wi among the wafer inspection items excluding the ON current specified as the sampling parameter. Do. That is, the following processing is performed for all wafer inspection items except for the ON current.

That is, at step 603, CorMax
Is initialized to 0, and since the designated trace direction is the increasing direction, in step 604, Imi is set as Itmp.
Calculate n + ΔI.

Since the designated trace direction is the increasing direction, the on-current is less than Itmp for each Itmp until Itmp> Imax (Step 605) while increasing Itmp by ΔI (Step 610). The correlation coefficient Cor between the yield and the wafer inspection item Wi and the significance level p are calculated and stored for the wafer (step 607). And the correlation coefficient Cor
Is plotted on a correlation coefficient number chart with the sampling parameter on the horizontal axis and the correlation coefficient on the vertical axis (step 6).
08). Here, the significance level p is a threshold value, for example, 5%.
If it is smaller and the absolute value of the correlation coefficient Cor is larger than CorMax (step 609), the absolute value of the correlation coefficient Cor is set to CorMax (step 611), and Itmp at that time is set to Ith (step 612). Note that the significance level is a probability that it can be declared that there is no correlation, and is used as a clue for determining the presence or absence of a correlation.

When the correlation tracing process is completed as described above, the characteristic failure analysis process proceeds to the correlation coefficient transition chart 1 shown in FIG.
201 is displayed on the display 704.

Here, each polygonal line 1202 in the figure represents the transition of the correlation coefficient corresponding to each wafer inspection item. Therefore, the user can refer to the correlation coefficient transition chart 1201 and extract a wafer inspection item having a peak in the correlation coefficient as a failure factor. Further, when peaks appear in a plurality of wafer inspection items, the magnitude of the correlation coefficient is displayed, for example, by displaying CorMax obtained by the correlation trace processing on the correlation coefficient transition chart 1201, and the yield is increased. Can be ranked according to the degree of influence. Also, it can be assumed that the larger the change in the correlation coefficient before and after the peak, the more likely it is to be a failure factor.

Now, the correlation coefficient transition chart 1201
When the user moves the cursor 1203 to a polygonal line corresponding to a specific wafer inspection item using the mouse 703 or the like and clicks the mouse button or the like,
In the characteristic failure factor analysis processing, a detailed correlation coefficient transition chart 1302 shown in FIG. 10 is displayed.

A broken line 1303 in the correlation coefficient transition chart 1302 indicates a transition of the correlation coefficient, and a circle on the broken line indicates that the significance level p obtained and stored at the time of calculating the correlation coefficient in the correlation trace processing is 5%. In the above, there is no correlation, and the black circles correspond to those with correlation at a significance level of less than 5%.
This is a guide when judging the presence or absence of the correlation.

Next, if the user moves the cursor 1304 to an arbitrary position on the polygonal line 1303 and clicks the mouse button or the like on such a display, the characteristic failure factor analysis process performs processing according to the clicked position. A scatter diagram 1305 and a scatter diagram 1308 of the yield and the wafer inspection result shown in FIG.
Create and display

A scatter diagram 1305 is a scatter diagram in a case where a position corresponding to the peak of the correlation coefficient is designated. The scatter diagram shows an on-current value corresponding to this peak, that is, a value equal to or smaller than Ith (a trace designated by Ith). In the opposite direction)
It becomes a scatter diagram for wafers in the range. The user performs a regression analysis of the yield and the wafer inspection result based on the scatter diagram, and obtains regression expressions 1306 and 1307.
The effect of the wafer inspection result on the yield can be quantified. As a result, it is possible to predict a yield improvement effect when the element characteristics measured in the wafer inspection item are changed by a process measure or the like. For example, when the regression equation is represented by yield = −24 × Wi + 300,
If the on-current is less than Ith and the value of the wafer inspection item Wi is 10
By changing the value of Wi from 10 to 9, the yield can be predicted to be improved by 24% from 60% to 84% for the wafer of No. 1.

The embodiment of the present invention has been described above.

In the above, the case where the cause of the characteristic failure analysis is analyzed as the yield related to the operating frequency of the LSI and the cause of the failure is analyzed has been described as an example. However, the present embodiment is similarly applied to other analysis. be able to.

For example, when the target of characteristic failure analysis is the yield related to the power consumption of the LSI and the cause of the failure is analyzed, for example, the model shown in FIG. 11 can be applied.

That is, in this case, as a wafer inspection item in which the leak current has a large effect on power consumption, a correlation trace process for obtaining a correlation coefficient is performed while expanding the analysis target in order from a wafer having a large leak current. Can be extracted as a cause of failure.

More specifically, in this case, on the correlation trace condition setting screen 1101, the user specifies the yield or the rate of occurrence of power-related failures as the vertical axis parameter, and specifies the leak current as the sampling parameter. Then, as the tracing direction, it is specified that the wafers to be analyzed are to be analyzed in order from the wafer having the largest measured value.

In this case, the correlation tracing process is performed according to the procedure shown in FIG.

That is, first, the parameter i is initialized to 1 (step 600). And the vertical axis parameter,
Here, the increment ΔL of the leakage current is obtained (step 60).
1). ΔL is the trace condition 1105 on the correlation trace condition setting screen 1101, and when the sampling parameter step width is designated, the designated value is ΔL;
When the sampling parameter step width number n is designated, the vertical axis parameter obtained from the correlation trace data in FIG. 4, that is, the maximum value Lma of the leak current here
Using x and the minimum value Lmin, (Lmax−Lmin) /
Let n be ΔL.

Next, while incrementing i by 1 (step 6
06) Until i becomes the total number m-1 of wafer inspection items (step 602), the following processing is performed for the i-th wafer inspection item Wi among the wafer inspection items excluding the leak current specified as the sampling parameter. Do. That is, the following processing is performed for all the wafer inspection items except for the leakage current.

That is, at step 603, CorMax
Is initialized to 0, and since the designated trace direction is the decreasing direction, at step 604, Lma is set as Ltmp.
Calculate x-ΔL.

Since the designated trace direction is the decreasing direction, while decreasing Ltmp by ΔL (step 610), the leak current is not less than Ltmp for each Ltmp until Ltmp <Lmin (step 605). The correlation coefficient Cor between the yield and the wafer inspection item Wi and the significance level p are calculated for the wafer of
It is stored (step 607). And the correlation coefficient Cor
Is plotted on a correlation coefficient number chart with the sampling parameter on the horizontal axis and the correlation coefficient on the vertical axis (step 6).
08). Here, the significance level p is a threshold value, for example, 5%.
If it is smaller and the absolute value of the correlation coefficient Cor is larger than CorMax (step 609), the absolute value of the correlation coefficient Cor is set to CorMax (step 611), and Ltmp at that time is set to Lth (step 612).

As described above, the wafer inspection items to be used as the wafer sampling parameters and the conditions of the correlation trace are different depending on the specification of the LSI. However, according to the present embodiment, the element characteristics that have the greatest influence on the yield are known. In this case, by setting appropriate conditions on the correlation trace condition setting screen, the cause of the failure can be analyzed in the same manner as described above.

In the above, the case where one wafer inspection item is used as a sampling parameter in accordance with the model in which one wafer inspection item greatly affects the yield has been described. Even when there are a plurality of wafer inspection items that have a great influence, it can be applied to the analysis of the cause of failure according to a model having a plurality of such wafer inspection items.

For example, there is a correlation between the on-current and the operating frequency and thus the yield. • There is a correlation between leakage current and power consumption and thus yield. -There is no correlation between on-current and leakage current. In such a case, both the ON current and the leak current are set as the sampling parameters.

Then, the trace direction is such that the on-current is from the wafer with the smaller on-current and the leak current is from the wafer with the larger leak current, and the correlation coefficient between the yield and the wafer inspection result is included in the correlation analysis. Find the transition of That is, for example,
While increasing tmp from the minimum value Imin of the on-current, whenever the value Itmp of the on-current is increased, the leakage current Ltmp is sequentially reduced from the maximum value Lmax of the leakage current. Then, a correlation coefficient is determined for each wafer inspection item other than the on-current and the wafer current, for a wafer having an on-current at or below the current Itmp and a wafer having a leakage current at or above the Ltmp.

Then, the obtained correlation coefficient is plotted and displayed on a correlation coefficient transition chart 1602 shown in FIG.

This correlation coefficient transition chart 1602 shows the result of a certain wafer inspection item Wi. In this correlation coefficient transition chart, the user can extract a wafer inspection item having a peak in the correlation as a failure factor, as in the above-described embodiment. Further, when peaks appear in a plurality of wafer inspection items, the degree of influence on the yield can be ranked by comparing the magnitude of the correlation coefficient. In addition, it can be estimated that the larger the change in the correlation coefficient before and after the peak, the more likely it is to be a failure factor.

The user moves the cursor 1603 to the correlation coefficient transition chart 1602 using, for example, a mouse.
If the mouse is moved to an arbitrary position above and a mouse button or the like is clicked, the distribution of the wafer yield and the wafer inspection item Wi that is equal to or less than the on-current corresponding to the clicked position and equal to or greater than the leak current corresponding to the clicked position. FIG. 1604 is displayed, and it is possible to estimate a target and an effect of the improvement of the wafer inspection item Wi by the regression analysis.

In the above description, the case where the wafer inspection item that greatly affects the yield is known, that is, the case where the wafer inspection item used as the sampling parameter is known has been described. The form can be applied to analysis of a cause of failure.

That is, in this case, for example, in the characteristic failure analysis process, all the wafer inspection items are automatically set as sampling parameters, and the direction in which the value increases and the direction in which the value decreases is set as the trace direction. A correlation trace process is performed, and as a result, a wafer inspection item having a peak correlation coefficient is extracted as a failure factor. Alternatively, in the characteristic failure analysis processing, a combination of a plurality of wafer inspection items is automatically made into a combination of a plurality of sampling parameters, and the above-described correlation trace is performed for each combination of increasing / decreasing directions of the value of each sampling parameter. The processing may be performed, and as a result, a wafer inspection item having a peak correlation coefficient may be extracted as a failure factor.

In the above embodiment, the case where the user determines the peak of the correlation coefficient from the correlation transition chart has been described. However, the characteristic failure analysis process detects the peak of the correlation coefficient and presents it to the user. You may make it.
Similarly, a regression equation from a scatter diagram may be calculated in the characteristic failure analysis processing.

In the above embodiment, in the correlation tracing process, the value of the sampling parameter (Itmp or Ltmp) is increased / decreased in the tracing direction by a predetermined range, and the value of the sampling parameter that is equal to or larger than the value of the sampling parameter is targeted. The correlation coefficient was determined,
The value of the sampling parameter is small according to the tracing direction so that the number of wafers that include the value of the sampling parameter in the target for obtaining the correlation coefficient increases.
Alternatively, the correlation coefficient may be determined by selecting the larger one. In this case, the trace condition 1 on the correlation trace condition setting screen 1101 shown in FIG.
As 105, the designation of this number is accepted.

In the above embodiment, the target for which the correlation with the wafer inspection item is to be determined is the yield in any case. However, this may be any item that is further subdivided and the correlation is desired to be determined. For example, when it is desired to extract a wafer inspection item having a correlation with the above-described decrease in the yield due to the operating frequency and the content of the correlation, only the probe inspection result related to the operating frequency may be set as a target for obtaining the correlation.

In the above embodiment, the analysis target is described as an LSI. However, the present embodiment can be similarly applied to the analysis of any other semiconductor product and any non-semiconductor product. .

[0089]

As described above, according to the present invention, even when a simple correlation does not exist between a predetermined inspection item of a product, for example, the yield and other characteristic items, the contents of these correlations can be clarified. Can be

[0090]

[Brief description of the drawings]

FIG. 1 is a block diagram showing a hardware configuration of a characteristic failure analysis device according to an embodiment of the present invention.

FIG. 2 is a diagram showing an example of wafer inspection result information according to the embodiment of the present invention.

FIG. 3 is a diagram showing an example of probe test result information according to the embodiment of the present invention.

FIG. 4 is a diagram showing an example of correlation trace data according to the embodiment of the present invention.

FIG. 5 is a view showing a correlation trace condition setting screen according to the embodiment of the present invention.

FIG. 6 is a diagram showing a correlation model according to the embodiment of the present invention.

FIG. 7 is a diagram illustrating the principle of the correlation trace processing according to the embodiment of the present invention.

FIG. 8 is a diagram showing a procedure of a correlation trace process according to the embodiment of the present invention.

FIG. 9 is a diagram showing a correlation coefficient transition chart according to the embodiment of the present invention.

FIG. 10 is a diagram showing a detailed correlation coefficient transition chart according to the embodiment of the present invention.

FIG. 11 is a diagram showing a correlation model according to the embodiment of the present invention.

FIG. 12 is a diagram showing a procedure of a correlation trace process according to the embodiment of the present invention.

FIG. 13 is a diagram showing a correlation coefficient transition chart according to the embodiment of the present invention.

FIG. 14 is a diagram illustrating an example of an inspection item of an LSI.

FIG. 15 is a diagram for explaining the inspection items shown in FIG.

FIG. 16 is a diagram showing a procedure of a conventional characteristic failure analysis.

[Explanation of symbols]

700: characteristic failure analyzer, 701: central processing unit, 7
02 keyboard, 703 mouse, 704 display, 705 storage device, 706 wafer tester, 70
7: Probe tester, 709: Network 710: Storage medium, 711: Main memory, 712: Communication control device, 71
3. Storage device

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Naofumi Iwata 292, Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture Inside of Hitachi, Ltd.Production Technology Research Institute (72) Inventor Nakatsuka Suwauchi 5-chome, Josuihoncho, Kodaira-shi, Tokyo No. 20 No. 1 F-term in the Hitachi, Ltd. Semiconductor Group (Reference) 2G003 AA02 AA07 AA10 AB01 AB05 AF01 AF02 AF03 AF06 AH00 AH01 AH02 2G032 AB02 AB20 AC03 AE08 AE09 AE10 AL00 2G036 AA27 BB09 CA00 4M106 AA01 BP01 BB21 BB95

Claims (8)

[Claims] 1. A correlation analysis method for analyzing a correlation of a plurality of test items performed on each of a plurality of analysis objects with a predetermined test item, wherein one of the plurality of test items is determined. A first step of setting a plurality of selection reference values between a minimum value and a maximum value of inspection values of the first inspection item as a first inspection item; 1 other than 1 inspection item
One as a second inspection item, and as for each of the plurality of selection criterion values set in the first inspection item, the inspection value of the first inspection item is a sample range with a sample range above or below the selection criterion value. A second step of calculating a correlation coefficient between a test value of the second test item and a test value of the predetermined test item for each analysis target belonging to the group; and a phase relationship calculated for each of the plurality of selection reference values. A third step of outputting or analyzing a transition of the number with respect to an increase or decrease of the selection reference value. 2. A correlation analysis method for analyzing a correlation between a plurality of test items performed on each of a plurality of analysis targets and a predetermined test item, wherein at least two of the plurality of test items are analyzed. As a first inspection item, a first step of setting a plurality of selection reference values between a minimum value and a maximum value of the inspection value of each first inspection item; 1 other than the first inspection item
One as a second inspection item, and for each of the first inspection items, a selection criterion value set formed by extracting one from the plurality of selection criterion values set for the inspection item. The test value of each of the first test items and the test value of the second test item with respect to each analysis object belonging to the sample range, and the test value and the predetermined test value, each of which is equal to or more than each of the selection reference values included in the value set. A second step of calculating a correlation coefficient with the inspection value of the item; and a third step of outputting or analyzing a transition of the correlation coefficient calculated for each of the plurality of selection reference values with respect to an increase or decrease in the selection reference value. And a correlation analysis method. 3. The correlation analysis method according to claim 1, wherein the object to be analyzed is a semiconductor device, and the inspection value of the predetermined inspection item is a defect rate or a yield of the predetermined item of the semiconductor device. The correlation analysis method, wherein the inspection values of the plurality of inspection items are measurement values representing characteristics of elements constituting a semiconductor device. 4. The correlation analysis method according to claim 1, wherein the first inspection item is an inspection item having a strong correlation with the predetermined inspection item, and wherein the third step includes: A correlation analysis method, wherein, when a peak is present in a transition with respect to an increase / decrease of a selection reference value of a correlation coefficient, the second inspection item is a candidate for an inspection item having a correlation with the predetermined inspection item. 5. A correlation analysis apparatus for analyzing a correlation between a plurality of test items performed on each of a plurality of analysis targets and a predetermined test item, wherein one of the plurality of test items is determined. First means for setting a plurality of selection reference values between a minimum value and a maximum value of inspection values of the first inspection item as a first inspection item; 1 other than the inspection items
One as a second inspection item, and as for each of the plurality of selection criterion values set in the first inspection item, the inspection value of the first inspection item is a sample range with a sample range above or below the selection criterion value. Second means for calculating a correlation coefficient between the test value of the second test item and the test value of the predetermined test item for each analysis target belonging to the following, and a phase relationship calculated for each of the plurality of selection reference values A third means for outputting or analyzing the transition of the number with respect to the increase / decrease of the selection reference value. 6. A correlation analysis device for analyzing a correlation between a plurality of test items performed on each of a plurality of analysis objects and a predetermined test item, wherein at least two of the plurality of test items are analyzed. A first inspection item, a first means for setting a plurality of selection reference values between the minimum value and the maximum value of the inspection value of each first inspection item, 1 other than the first inspection item
One as a second inspection item, and for each of the first inspection items, a selection criterion value set formed by extracting one from the plurality of selection criterion values set for the inspection item. The test value of each of the first test items and the test value of the second test item with respect to each analysis object belonging to the sample range, and the test value and the predetermined test value, each of which is equal to or more than each of the selection reference values included in the value set. A second means for calculating a correlation coefficient with the inspection value of the item; and a third means for outputting or analyzing a transition of the correlation coefficient calculated for each of the plurality of selection reference values with respect to an increase or decrease in the selection reference value. And a correlation analyzer. 7. An analysis system for analyzing a cause of a failure of a semiconductor device, comprising: a measurement device for measuring a plurality of characteristics of elements constituting the semiconductor device; a determination device for determining good / defective of the semiconductor device; A plurality of characteristic values measured by the measurement device for each of the semiconductor devices, a failure factor correlation analysis device for analyzing a correlation with a determination result of the determination device, the failure factor correlation analysis device, One of a plurality of characteristic values measured by the measuring device is defined as a first characteristic value, and a plurality of selection reference values are set between a minimum value and a maximum value of the first characteristic value. Means, and one of the plurality of characteristic values measured by the measuring device other than the first characteristic value is set as a second characteristic value, and each of the plurality of selection reference values set as the first characteristic value is selected. , Sample range above or below the selection criteria A second means for calculating a correlation coefficient between the second characteristic value for each semiconductor device whose first characteristic value belongs to a sample range and a determination result of the determination device; and the plurality of selection reference values. Third means for outputting or analyzing the transition of the correlation coefficient calculated for each of the selected reference values with respect to increase and decrease. 8. A storage medium storing a correlation analysis program for analyzing a correlation between a plurality of inspection items performed on each of a plurality of analysis objects and a predetermined inspection item, wherein the correlation analysis program is , Which is read and executed by the computer, on the computer, one of the plurality of inspection items is set as a first inspection item, and the minimum value of the inspection value of the first inspection item is First means for setting a plurality of selection reference values between the maximum value and one of the plurality of inspection items other than the first inspection item
One as a second inspection item, and as for each of the plurality of selection criterion values set in the first inspection item, the inspection value of the first inspection item is a sample range with a sample range above or below the selection criterion value. Second means for calculating a correlation coefficient between the test value of the second test item and the test value of the predetermined test item for each analysis target belonging to the following, and a phase relationship calculated for each of the plurality of selection reference values And a third means for outputting or analyzing the transition of the number with respect to the increase / decrease of the selection reference value.