JP2008177269A - Method of manufacturing semiconductor device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method of manufacturing a semiconductor device wherein wafer test is conducted to automatically judge whether or not a measured chip has a statistically unexpected value and also judge whether or not measurement is conducted again or that the chip is defective. <P>SOLUTION: The method of manufacturing a semiconductor device includes a step to form a plurality of chips on a wafer; a step to conduct testing for the chips in terms of the specified test items; a step to prepare a histogram that shows the number of chips for each of test items; a step to find out a range of the values of the test items as a non-defective area wherein the histogram becomes most approximate to a normal distribution; and a step to judge whether or not chips not corresponding to the non-defective area are measured again or that they are defective. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a method of manufacturing a semiconductor device that can automatically determine a chip whose value is not statistically predictable by a wafer test and perform remeasurement or determine that it is defective.

  In the wafer test, a predetermined test item is tested for each of a plurality of chips formed on a wafer (see, for example, Patent Document 1). From the wafer test result, a histogram as shown in FIG. 11 representing the number of chips for each value of the test item is created. Then, a chip whose test item value is larger or smaller than a predetermined standard value is judged as defective and is not picked up after chip separation.

Japanese Patent Laid-Open No. 05-322973

  However, when a wafer test is performed, there may be a case where a two-peak distribution of a non-defective product distribution and an abnormal distribution occurs as shown in FIG. That is, the characteristic distribution of the chips on the wafer is statistically expected to be concentrically continuous, but a value that cannot be statistically predicted may be measured. If remeasurement is performed, there may be a single mountain distribution. Possible causes of abnormal distribution include test pin contact failure, process defects, and dust on the pad.

  Thus, it is necessary to perform re-measurement or determine that the chip has a value that is statistically unpredictable. However, conventionally, since the operator has determined whether or not there is an abnormal distribution by looking at the histogram, there is a problem that it takes a lot of manpower and the construction period is long.

  The present invention has been made to solve the above-described problems, and its purpose is to automatically determine a chip whose value is not statistically predictable by a wafer test and perform remeasurement or A method of manufacturing a semiconductor device that can be determined to be defective is obtained.

  A method of manufacturing a semiconductor device according to the present invention includes a step of forming a plurality of chips on a wafer, a step of testing a predetermined test item for each of the plurality of chips, and the number of chips for each value of the test item. A step of creating a histogram to represent, a step of obtaining a value range of a test item whose histogram is closest to a normal distribution as a non-defective region, and a step of performing remeasurement or determining that the chip does not correspond to the non-defective region Have.

  Another method of manufacturing a semiconductor device according to the present invention includes a step of forming a plurality of chips on a wafer, a step of testing each of the plurality of chips for the first test item and the second test item, And a step of re-measuring or determining a chip whose correlation between the value of the first test item and the value of the second test item does not fall within a predetermined range.

  Still another method of manufacturing a semiconductor device according to the present invention includes a step of forming a plurality of chips on a wafer, a step of extracting several chips from the plurality of chips, and performing a sampling test for each predetermined test item, The step of obtaining the average value and standard deviation of the test items by sampling test, the step of testing each of the test items for a plurality of chips, and the value of the test item falls within a predetermined range determined from the average value and standard deviation And a step of re-measuring or judging a chip that is not defective. Other features of the present invention will become apparent below.

  According to the present invention, it is possible to automatically determine a chip in which a value that cannot be statistically predicted by the wafer test is measured, and perform remeasurement or determine that the chip is defective.

Embodiment 1 FIG.
A method for manufacturing a semiconductor device according to the first embodiment of the present invention will be described with reference to the flowchart shown in FIG. First, as shown in FIG. 2, a plurality of chips 12, for example, 1000 or more chips 12 are formed on the wafer 11 by a normal method (step S1). Next, the test pins 13 are brought into contact with each other, and a wafer test is performed on a plurality of chips 12 with respect to predetermined test items, for example, gate-source voltage Vgs (DC characteristics) or gain (RF characteristics) (step S2). . Then, as shown in FIG. 4, a histogram representing the number of chips Y for each value X of the test item is created (step S3).

Next, the value range of the test item whose histogram is closest to the normal distribution is obtained as a non-defective region (step S4). Specifically, first, the normal distribution of the histogram is assumed as in Expression 1.
Y = Aexp (− (X−μ _ij ) ² / 2σ _ij ² ) (Formula 1)
However, the average value of the number of chips is A, the average value of the test item values calculated in the range of m _i <X <n _j is μ _ij , and the standard deviation calculated by m _i <X <n _j is σ _ij . . If ΔX is the amount of change in X, then m _i = m ₁ + ΔX (i−1) and n _j = n ₁ + ΔX (j−1). Next, i is changed from 1 to L, j is changed from 1 to K, and the number of chips Y is obtained by Equation 1, respectively. Then, an error between the calculated value of the number of chips Y and the measured number of chips Ym is calculated by the least square method. Further, a combination of i and j that minimizes the error is obtained, and m _i <X <n _j at this time is defined as a non-defective region.

  Next, about the chip | tip which does not correspond to a non-defective area | region, it measures again or it judges that it is defect (step S5). When performing remeasurement, a histogram is created by merging remeasured data as shown in FIG. 5 and data determined to be a non-defective region in the initial measurement as shown in FIG.

  Next, a chip having a test item value larger or smaller than a predetermined standard value is determined to be defective. Thereafter, the wafer is separated into chips by dicing. Then, at the time of picking up, a chip determined to be defective is not picked up.

  As described above, according to the present embodiment, it is possible to automatically determine a chip in which a value that is not statistically predictable by a wafer test is measured, and perform remeasurement or determine that the chip is defective.

Further, the standard deviation sigma _ij average value A varies but between lots may be referred to a constant. In this case, it is preferable to fix the standard deviation σ _ij of the histogram in the case of the normal distribution to a predetermined value in the step of obtaining the non-defective region. Thereby, a non-defective area can be easily obtained.

Further, in the step of obtaining the non-defective region, it is preferable to add as a condition of the non-defective region that the value of the standard deviation σ _ij of the histogram is within a predetermined range (σ _MIN <σ _ij <σ _MAX ). Thereby, a non-defective area can be easily obtained.

Embodiment 2. FIG.
A method for manufacturing a semiconductor device according to the second embodiment of the present invention will be described with reference to the flowchart shown in FIG. First, as in the first embodiment, a plurality of chips are formed on a wafer (step S11).

  Next, a plurality of chips are tested for the first test item and the second test item, for example, the gate-source voltage Vgs (DC characteristic) and the gain (RF characteristic) (step S12). FIG. 8 shows test results indicating the correlation between DC characteristics and RF characteristics.

Next, a chip whose correlation between the value of the DC characteristic and the value of the RF characteristic does not fall within a predetermined range is measured again or determined to be defective (step S13). Specifically, since the correlation between the DC characteristic X and the RF characteristic Y is often known from past data, a and b are calculated in advance by linear approximation of Y = aX + b. Then, taking into account the variation of the _{correlation, b min = b-Δb,} b max = b + Δb become as _{b min,} kept Decide _{b max.} Next, the measured DC characteristics are X _m , the RF characteristics are Y _m , Y _min = aX + b _min , Y _max = aX + b _max , and remeasurement is performed for a chip that does not fall within the range of Y _min <Y _m <Y _max. Do it or judge it to be bad.

  Thus, by taking the correlation between the DC characteristics and the RF characteristics, it is possible to eliminate defects such as a contact of the DC probe but poor contact of the RF probe. Therefore, according to the present embodiment, it is possible to automatically determine a chip in which a value that cannot be statistically predicted by the wafer test is measured, and perform remeasurement or determine that the chip is defective.

Embodiment 3 FIG.
A method for manufacturing a semiconductor device according to the third embodiment of the present invention will be described with reference to the flowchart shown in FIG. First, similarly to the first embodiment, a plurality of chips are formed on a wafer (step S21).

  Next, as shown in FIG. 10, several (for example, five) chips 14 are extracted from a plurality of chips on the wafer 11 at locations where the in-plane distribution of the wafer can be reflected. Then, a sampling test is performed on each of these chips for predetermined test items, for example, gate-source voltage Vgs (DC characteristics) or gain (RF characteristics) (step S22). Here, it is possible to use a chip 14 having a TEG (Test Element Group) pattern formed as the chip 14 for performing the sampling test.

  Next, the average value A and the standard deviation σ of the test items are obtained by a sampling test (step S23). Then, a wafer test is performed on each of the test items for a plurality of chips (step S24).

  Next, a measurement is performed again for a chip whose test item value by the wafer test does not fall within a predetermined range (for example, a range of ± 3σ from the average value A) determined from the average value A and the standard deviation σ. Judgment is made (step S25).

  As described above, according to the present embodiment, it is possible to automatically determine a chip in which a value that is not statistically predictable by a wafer test is measured, and perform remeasurement or determine that the chip is defective.

  Instead of the sampling test, the average value and standard deviation of the test items obtained by the past wafer test may be used as criteria for the quality determination.

3 is a flowchart showing a method for manufacturing the semiconductor device according to the first embodiment of the present invention. It is a perspective view which shows the wafer in which the some chip | tip was formed. It is a perspective view which shows the state which performs a wafer test. It is the histogram calculated | required from the result of the wafer test. It is a histogram which shows the data measured again. It is a histogram which shows the data judged to be a non-defective area by the first measurement. It is a flowchart which shows the manufacturing method of the semiconductor device which concerns on Embodiment 2 of this invention. It is a test result which shows the correlation with DC characteristic and RF characteristic. It is a flowchart which shows the manufacturing method of the semiconductor device which concerns on Embodiment 3 of this invention. It is a top view which shows the state which performs a sampling test. It is the histogram calculated | required from the result of the wafer test.

Explanation of symbols

11 Wafer 12 Chip 13 Test pin 14 Chip extracted by sampling test

Claims (6)

Forming a plurality of chips on the wafer;
Testing each of the plurality of chips for a predetermined test item;
Creating a histogram representing the number of chips for each value of the test item;
Determining the range of the value of the test item as the non-defective region where the histogram is closest to the normal distribution;
A method of manufacturing a semiconductor device, comprising: re-measuring a chip that does not correspond to the non-defective region or determining that the chip is defective.   2. The method of manufacturing a semiconductor device according to claim 1, wherein, in the step of obtaining the non-defective region, a standard deviation of the histogram when the distribution is normal is fixed to a predetermined value.   2. The method of manufacturing a semiconductor device according to claim 1, wherein in the step of obtaining the non-defective region, a condition that the standard deviation value of the histogram is within a predetermined range is added as a condition for the non-defective region. Forming a plurality of chips on the wafer;
Testing each of the plurality of chips for a first test item and a second test item;
And a step of re-measuring or judging a chip whose correlation between the value of the first test item and the value of the second test item does not fall within a predetermined range. A method for manufacturing a semiconductor device. Forming a plurality of chips on the wafer;
Extracting a number of chips from the plurality of chips and performing a sampling test for each predetermined test item;
Obtaining an average value and a standard deviation of the test items by the sampling test;
Testing each of the plurality of chips for the test item;
And a step of re-measuring or determining that the test item value does not fall within a predetermined range determined from the average value and the standard deviation.   6. The method of manufacturing a semiconductor device according to claim 5, wherein a chip for performing the sampling test has a TEG (Test Element Group) pattern formed thereon.

Images