JP2001345357A - Method for measuring and manufacturing semiconductor wafer, and apparatus for measuring semiconductor wafer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for more simply and rapidly measuring a semiconductor wafer to distinguish and evaluate types of crystal defects formed on the wafer. SOLUTION: The method for measuring the semiconductor wafer comprises the steps of measuring the distribution of the crystal defects on the wafer 1 as a set of crystal defect detecting points by using a crystal defect inspecting unit, and extracting the detecting points 3 belonging to an array state peculiar to the specific crystal defect from the set of the points. The points can be efficiently measured with small unevenness by the inspecting unit, and the points belonging to the array state peculiar to the specific defect is extracted and used. Hence, the specific defects can be selectively determined.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a measuring method, a method for manufacturing a semiconductor wafer, and an apparatus for measuring a semiconductor wafer.

[0002]

2. Description of the Related Art With the recent increase in the degree of integration of semiconductor elements, it has become very important to take measures against crystal defects in a semiconductor single crystal, particularly crystal defects near the surface. Since the crystal defect of the semiconductor single crystal has a very large effect on the quality of the semiconductor device formed thereby, to improve the quality of the device, the crystal defect near the surface of the semiconductor wafer serving as the substrate is evaluated. It is necessary to understand the actual situation.

Conventionally, as a method for evaluating a semiconductor wafer, for example, a distribution shape of crystal defects appearing on the surface of the semiconductor wafer under a condensing lamp is sketched, the length of the crystal defects is measured, and the total is integrated. Thus, a visual inspection for evaluating the degree of occurrence of crystal defects has been performed. In addition, as an evaluation method using a measuring device, a wafer formed by scanning a wafer surface with a laser beam and measuring light scattering intensity from particles or the like, or irradiating the wafer with X-rays to form a crystal formed on the wafer. X-ray topographic methods for detecting the diffraction intensity of X-rays due to defects are used.

[0004] When evaluating a semiconductor wafer, it is desirable to perform the evaluation while distinguishing the types of crystal defects. By doing so, it becomes easy to pursue the cause of the crystal defect and improve it. As a method of evaluating the type of crystal defects separately, foreign matter (particles) on the surface of a semiconductor wafer and COP (Crystal Origin)
A method is known in which a particle is detected by a particle detection device that classifies and detects ated particles. For example, a particle detection device disclosed in Japanese Patent Application Laid-Open No. H11-284038 receives scattered light scattered by crystal defects on the surface of a semiconductor wafer from the front, rear, above, and side with respect to the irradiation direction. Thus, the position, size and height of the defect are detected to determine the type of crystal defect.

[0005]

Among the conventional methods for evaluating a semiconductor wafer as described above, the method based on visual inspection under a condensing lamp aims to quantify the amount of a specific crystal defect, for example, the formation of a slip dislocation. In this case, the slip dislocations observed visually are sketched, and the integrated length of the slip dislocations is determined from the sketch. In addition, because of measurement by human eyes, there are many variations in inspection, and skill is required.

On the other hand, in the method using a particle detection device based on the light scattering method, since the crystal defect is mechanically detected by the device, the problem of variation in detection is reduced to some extent. However, specific crystal defects such as slip dislocations could not be selectively quantified in the crystal defect detection points measured by this method because those derived from various crystal defects were mixed.

Further, in the X-ray topography method, defects occurring on the front surface and inside or on the back surface of the crystal are observed, and the crystal defects on the surface and the crystal defects existing inside the crystal are evaluated separately. Some aspects are difficult.

An object of the present invention is to provide a simpler and faster method for measuring a semiconductor wafer for distinguishing and evaluating the types of crystal defects formed on a semiconductor wafer, a method for manufacturing a semiconductor wafer using the same, and It is an object of the present invention to provide a semiconductor wafer measuring device capable of rationally implementing the semiconductor wafer measuring method.

[0009]

In order to solve the above-mentioned problems, a method for measuring a semiconductor wafer according to the present invention uses a crystal defect inspection apparatus to remove crystal defects generated in a semiconductor wafer. A crystal defect measurement step of measuring the distribution as a set of crystal defect detection points, and a crystal defect detection point extraction step of extracting a crystal defect detection point belonging to an array form specific to the specific crystal defect from the set of crystal defect detection points; It is characterized by having.

The semiconductor wafer measuring device of the present invention further comprises a crystal defect measuring mechanism for measuring the distribution of crystal defects occurring in the semiconductor wafer as a set of crystal defect detection points,
And a crystal defect detection point extracting mechanism for extracting data of crystal defect detection points belonging to an array form specific to a specific crystal defect from the stored set of crystal defect detection points.

In the method of measuring a semiconductor wafer according to the present invention, a crystal defect inspection apparatus is used to efficiently measure a crystal defect as a crystal defect detection point with a small variation, and furthermore, an arrangement unique to a specific crystal defect. By extracting the crystal defect detection points belonging to the morphology, the specific crystal defects can be selectively quantified. It should be noted that the measurement accuracy of the crystal defect can be enhanced by setting the semiconductor wafer in which the crystal defect has been revealed by the selective etching.

Then, based on the result of the extraction of the crystal defect detection points, it is possible to reasonably and easily evaluate the crystal defects of the semiconductor wafer. For example, when extracting a specific type of crystal defect, a crystal defect detection point of a different type is determined, and is distinguished from a set of crystal defect detection points of a specific type. Then, to determine the type of the specific type of crystal defect, the type is determined from the distribution shape of the crystal defect. In the present invention,
Crystal defects formed in the semiconductor wafer can be identified according to the arrangement of the crystal defect detection points extracted in the crystal defect detection point extraction step.

FIG. 1 is a schematic diagram showing the types of typical crystal defects appearing on the surface of a semiconductor wafer and their distribution shapes. (A) is an example of a slip dislocation generated in a silicon single crystal wafer having a plane orientation of (100), and (b) is an example of a slip dislocation generated in a silicon single crystal wafer having a plane orientation of (111). The feature is that the crystal defect detection points appear linearly along the <110> direction. Therefore, slip dislocations can be easily identified by extracting crystal defect detection points linearly arranged in a specific direction on the main surface. (C) is an example of swirl, which is characterized in that the crystal defect detection points are arranged in a spiral shape. And (d)
Is an example of oxygen precipitation induced defects, which is characterized in that crystal defect detection points are arranged concentrically. If the relationship between the type of the crystal defect and the distribution shape of the crystal defect is obtained in advance, the type of the crystal defect can be easily determined from the distribution shape of the crystal defect.

After the step of extracting crystal defect detection points,
If the number of extracted crystal defect detection points is counted, it is possible to quantitatively evaluate crystal defects occurring in the semiconductor wafer based on the counting result. Such quantitative evaluation of crystal defects can be performed by counting the number of crystal defect detection points for each crystal defect identified according to the arrangement form of the crystal defect detection points. As a result, the degree and frequency of occurrence of each type of crystal defect can be accurately grasped, so that it becomes easier to identify a cause of a defect in a process associated with the type of the generated crystal defect.

The mechanism for extracting crystal defect detection points can be configured to include a data processing device such as a computer. By taking the crystal defect detection data into a data processing device in the form of coordinate data, extraction and measurement of crystal defect detection points, and further defect classification, etc., can be performed more efficiently. The evaluation can be performed more easily and quickly.

For example, the measuring device of the above-mentioned basic invention has a counting mechanism for counting the number of crystal defect detection points extracted by the crystal defect detection point extracting mechanism, and a counting mechanism for counting the number of crystal defect detection points. And a crystal defect evaluation mechanism that quantitatively evaluates crystal defects based on the counting result by the above, and a crystal defect evaluation result output mechanism that outputs the number of counted crystal defect detection points as an evaluation result. it can. According to this configuration, counting of the extracted crystal defect detection points, quantitative evaluation of the crystal defects, and output of the results can be performed extremely efficiently. In addition, the output of the evaluation result may include a mode in which the result is displayed and output on a screen of a display device, or a mode in which the result is printed out by a printer, but is not limited thereto, for example, a semiconductor wafer manufacturing apparatus, It also includes a mode of transmitting and outputting evaluation result data to a manufacturing management device by wire or wirelessly.

In the measuring apparatus of the present invention, the type of crystal defect identified according to the arrangement of the crystal defect detection points extracted by the crystal defect detection point extraction mechanism is determined by the crystal defect evaluation mechanism. An evaluation result storage mechanism that stores the evaluation result in association with the content of the evaluation result can be provided. By classifying and storing and managing the evaluation results for each type of crystal defect,
In the process, it is possible to more easily grasp the tendency such as what kind of crystal defects are likely to occur.

Next, in the step of extracting crystal defect detection points, a set of measured crystal defect detection points is displayed on the screen of the display device, and the crystal defect detection points specific to the specific crystal defect are displayed on the screen. When an array is generated, it can be identified at a glance, so that the crystal defect detection point of interest can be easily extracted. As a specific method, an extraction region can be set on a screen of a display device that displays a set of crystal defect detection points, and crystal defect detection points within the set region can be extracted.

Further, in order to rationally implement such a method, the measuring apparatus of the present invention uses the coordinates of the crystal defect detection points for storing aggregate data of each crystal defect detection point measured by the crystal defect measurement mechanism. It can be configured to include a data storage mechanism and a display mechanism for displaying the distribution shape of the crystal defect detection points on the semiconductor wafer on a screen based on the stored set data of the crystal defect detection points. In this case, the crystal defect detection point extraction mechanism refers to the distribution shape of the crystal defect detection points displayed on the screen of the display mechanism, and converts the data of the crystal defect detection points arranged in a specific form into a specific crystal. A selection input mechanism for selecting as data corresponding to the defect detection points, and coordinates of the crystal defect detection points in a state where the data of the selected crystal defect detection points can be distinguished from the data of the non-selected crystal defect detection points. And a storage control mechanism for storing the data in the data storage mechanism. The counting mechanism of the extracted defect detection points is configured to count the number of data points of the selected crystal defect detection point.

For example, by setting an area so that only a set of crystal defect detection points arranged in a specific form is included and other crystal defect detection points are not included, crystal defect detection of the arrangement form of interest is performed. Point extraction can be performed intuitively and efficiently. Specifically, for example, the extraction region can be set so as to surround a crystal defect detection point arranged in a form specific to a specific crystal defect among a set of crystal defect detection points displayed on the screen.

Further, for example, the selection input mechanism includes an extraction area setting input mechanism for setting an extraction area on the screen of the display mechanism, and the storage control mechanism stores data of crystal defect detection points in the set extraction area. May be stored in the coordinate data storage mechanism of the crystal defect detection point as data of the selected crystal defect detection point.
With this configuration, the user of the apparatus merely sets the extraction area on the screen of the display device, and a series of processes of extracting and storing the data of the crystal defect detection points in the area are automatically performed. The operation is easy and the data management can be performed reliably and smoothly. Note that the selection input mechanism can be configured with a pointing device such as a mouse, so that a desired area on the screen can be set very easily.

Next, in the crystal defect measuring step, a light scattering type wafer surface inspection device can be used as a crystal defect detection device. The light scattering type wafer surface inspection device can be, for example, a particle detection device.
The particle detector scans the surface of the wafer with laser light and measures the intensity of light scattering from particles, etc., so that the position and size of foreign matter and COP can be recognized. Device. By using such a light scattering method, only defects near the surface of the semiconductor wafer can be accurately measured.

The method for manufacturing a semiconductor wafer according to the present invention further comprises a crystal defect measuring step of measuring a distribution of crystal defects occurring in the semiconductor wafer as a set of crystal defect detection points by using a crystal defect inspection apparatus. Extracting a crystal defect detection point belonging to an array form specific to a specific crystal defect from a set of crystal defect detection points, extracting a crystal defect detection point, and extracting a semiconductor defect based on the extraction result of the crystal defect detection point. And a sorting step of sorting semiconductor wafers based on the evaluation results of the crystal defect evaluating step.

That is, by employing the measuring method of the present invention, the identification and quantitative evaluation of crystal defects occurring in a semiconductor wafer can be performed accurately and efficiently, and the semiconductor wafer is sorted based on the evaluation result. By doing so, for example, the defect rate of the shipped semiconductor wafer product lot is reduced, and the quality can be improved.

[0025]

Embodiments of the present invention will be described below. (Embodiment 1) FIG. 2 is a schematic view showing schematic steps of a method for manufacturing a silicon single crystal wafer according to the present invention. First, a silicon single crystal ingot is manufactured by a known method such as the FZ method or the CZ method. The single crystal ingot thus obtained is cut into blocks having a certain resistivity range, and further subjected to outer diameter grinding. An orientation flat or an orientation notch is formed in each block after the outer diameter grinding. The block thus finished is sliced by a slicer such as an inner peripheral blade cutter as shown in FIG. The outer peripheral edges of both surfaces of the silicon single crystal wafer after slicing are chamfered by bevel processing.

As shown in FIG. 2 (b), the silicon single crystal wafer after the chamfering is lapped on both sides using free abrasive grains to form a wrapped wafer. Next, FIG.
As shown in (c), by immersing the wrapped wafer in an etchant, both sides are chemically etched to become a chemically etched wafer. The chemical etching process is performed to remove a damaged layer generated on the surface of the silicon single crystal wafer in the previous machining process. After the chemical etching step, a mirror polishing step is performed as shown in FIG. When a silicon single crystal layer is epitaxially grown on the main surface of the mirror wafer based on a known vapor phase growth method, a silicon epitaxial wafer (hereinafter, also simply referred to as an epitaxial wafer) is obtained.

FIG. 2E shows a flow of an evaluation / selection process of an epitaxial wafer which is a kind of a semiconductor wafer. In A1, on the surface of an epitaxial wafer randomly extracted from a certain product lot, for example, about 1
By performing selective etching of μm, crystal defects are revealed on the surface of the wafer. An etchant for clarifying a crystal defect utilizes a fact that an etching rate largely depends on a strain field generated around the crystal defect or a difference in electrical properties. Due to such etching, defects due to dislocations are observed as pits such as pyramids or ovals, and crystal defects occurring near the surface can be made apparent. Slip dislocations are observed as a linear collection of point-like crystal defects. Note that a heat treatment may be performed before the selective etching in order to make crystal defects apparent on the wafer surface.

Next, at A2, the distribution of crystal defects on the surface of the epitaxial wafer is measured by the above-described particle detection device which is one of the light scattering type wafer surface inspection devices. In the particle detection device, the position and size of particles on the wafer surface can be measured by scanning the surface of the wafer with laser light and measuring the intensity of light scattering from irregularities such as particles. And since the above-mentioned crystal defects also have irregularities,
The scattered light intensity is detected as a particle higher than the surrounding part, and the coordinate data indicating the particle position is
Obtained as crystal defect detection point data. In A3, data processing is performed on the acquired crystal defect detection point data in order to quantify crystal defects, and in A4, an epitaxial wafer is evaluated based on the processing result. Then, based on the evaluation result, a product lot to which the epitaxial wafer belongs is sorted into a non-defective product or a defective product.

FIG. 3 shows a semiconductor wafer measuring apparatus (hereinafter, simply referred to as a measuring apparatus) 100 according to an embodiment of the present invention for performing the above-described measurement of A2 and data processing of A3.
FIG. 2 is a block diagram showing an electrical configuration of the embodiment. Measuring device 100
Is roughly divided into two components: a particle detection device 150 serving as a crystal defect measurement mechanism, and a data processing computer 200 serving as a crystal defect detection point extraction mechanism.

The particle detection device 150 has a control computer 111 and a measurement system 101 connected thereto. The control computer 111 is connected to the I / O port 10
8 and the CPU 104, ROM 105, R
An AM 106, a hard disk drive (hereinafter abbreviated as HDD) 107 as a storage device, and a keyboard 109 and a mouse 110 as input devices are connected. In the HDD 107, a control program 107a for controlling the operation of the particle detection device, a data capturing program 107b for capturing crystal defect detection point data, and a data file 107c for the captured crystal defect detection points
Is stored. In the RAM 106, work areas 106a and 106b for the control program 107a and the data acquisition program 107b and a data storage area 106c for the acquired crystal defect detection point are formed.

Next, the measuring system 101 includes a laser beam probe (not shown) and the laser beam probe.
A drive unit for scanning on a main surface of a semiconductor wafer to be measured (hereinafter, also simply referred to as a wafer). In the present embodiment, the scanning drive is performed by moving a holder (not shown) for mounting a wafer, but the laser beam probe may be driven. As the scanning method, any of the XY scanning method and the spiral scanning method may be adopted, but in the present embodiment, the XY scanning method is adopted. Therefore, the drive unit
X drive motor 113 and Y drive motor 11 for independently driving the laser beam probe in the X and Y directions
5 and a pulse generator (hereinafter, referred to as X-PG) 114 and a pulse generator (hereinafter, referred to as Y-PG) 116 for detecting the servo control and rotation angle position of the motors. The X drive motor 113 and the Y drive motor 115 are connected to the I / O port 108 of the control computer 111 via a motor driver (not shown), and are driven and controlled by executing the control program 107a.

On the other hand, the laser light scattered on the surface of the wafer is detected by a scattered light detecting section (comprising a photomultiplier, a photodiode or a CCD sensor) 117 connected to the I / O port 108. . Then, the detection level output of the scattered light is input to the data interface 103. On the other hand, the pulse generator 11
The outputs of 4 and 116 are input to the I / O port 108 via the data interface 103.

The data interface 103 is, for example,
As shown in FIG. 4, X-PG114 and Y-PG116
-Counter 103a and Y-counter 103b that count up upon receiving a pulse signal from
103c and a comparator 103d. The count output of each of the counters 103a and 103b uniquely gives a measurement position by the laser beam probe. That is, when a particle corresponding to a crystal defect is detected at the measurement position, the X coordinate and the Y coordinate of the crystal defect detection point are given, and the respective counter outputs are sent to the I / O port 108 via the gate IC 103c. Is entered. The outputs of the counters 103a and 103b are a plurality of bits, but are drawn by one line for simplification.

On the other hand, the comparator 103d compares the detection level output of the scattered light with the threshold level Vref, and when the detection level output becomes larger than Vref, the gate IC 103d.
A data capture permission signal is output to the inhibit input terminal of c. Accordingly, the gate IC 103c sends the count output of the counters 103a and 103b to the control computer 111, that is, the X coordinate and the Y coordinate of the crystal defect detection point.
Coordinate data is allowed to be captured via the I / O port 108. The data capturing process is performed by the data capturing program 107b, and the captured data is stored in the crystal defect detection point data file 107c. Although not shown in FIG. 4, a flip-flop that latches and holds the counter output until the data fetch is completed can be provided at the subsequent stage of the counter output.

Here, the data fetch permission signal is an I / O
It is also possible to give the signal to the CPU 104 as a signal for permitting the interrupt access to the data port of the port 108. In this case, the gate IC 103c in FIG. 4 can be omitted. Also, if the measurement point interval is uniform,
The crystal defect detection point data including the measurement points where no crystal defects were detected is converted into binary graphic bitmap data (for example, the pixel corresponding to the measurement point where the crystal defects are detected is set to “1”, Pixel corresponding to point is "0"
), And the coordinates of each crystal defect detection point can be generated from the pixel address.

Next, the data processing computer 200
Is the I / O port 208 and the CPU 20 connected thereto.
1, ROM 202, RAM 203, H as a storage device
It has a DD 204, a mouse 206 and a keyboard 207 as an input device, and a monitor 205 as a display device. The communication line (or a communication network such as a LAN) 220 via the communication interfaces 209 and 112 allows the particle detection device 150 to operate. Control computer 1
11 is connected.

In the HDD 204, a crystal defect detection point data file 204b obtained from the particle detection device 150 via the communication line 220, a data processing program 204b for performing an evaluation process on the obtained crystal defect detection point data, and An evaluation result data file 204c indicating the evaluation result obtained by the processing is stored. The RAM 203 also includes a work area 203a for the data processing program 204a, a storage area 203b for the acquired crystal defect detection point data, a display memory 203c for displaying the crystal defect detection point data in a coordinate plot,
Further, a storage area 203d for evaluation result data is formed. The transfer of the crystal defect detection point data from the particle detection device 150 to the data processing computer 200 may be performed via an external storage medium such as a floppy (registered trademark) disk.
Further, the functions of the data processing computer 200 can be integrated with the control computer 111 of the particle detection device 150.

Hereinafter, the flow of the operation of the measuring apparatus 100 will be described using a flowchart. FIG. 5 shows a flow of the inspection process in the particle detection device 150. This process is performed by the CPU 104 of the control computer 111 executing the detection device control program 107a. First, in S1, wafer identification data such as a product number, a lot number, and a production date of a wafer to be measured are input. Next, in S2 and S3, the wafer is mounted on the holder of the apparatus, and when the mounting is completed normally, the process proceeds to the measurement processing. FIG. 10 shows an example of an epitaxial wafer W on which linear slip dislocation defects DF are formed. In the example of this drawing, the plane orientation of the epitaxial wafer W is (100), and the orientation flat position indicates the [110] direction. And this orientation flat shows
The wafer W is mounted and fixed on the holder such that the [110] direction coincides with the Y movement direction, and the [1-10] direction orthogonal to the Y movement direction coincides with the X movement direction. In the present specification, the negative sign of the exponent in the Miller index is indicated in a form added before the number representing the exponent for convenience.

Returning to FIG. 5, in S4, the X drive motor 113 and the Y drive motor 115 shown in FIG. 3 are operated to move the holder position to a predetermined origin position on (X, Y) coordinates. The X-reset signal and the Y-reset signal shown in FIG.
Reset 03b. Hereinafter, while scanning the irradiation position of the laser beam probe on the XY coordinate plane with the X direction as the horizontal direction and the Y direction as the vertical direction, each position is irradiated with a laser beam, and particles or crystals are measured by scattered light measurement. Perform defect detection. That is, in S5, the Y coordinate value Yk is set to be the first scanning position Y1 in the Y direction, and in S6, the Y coordinate value Xk is set to be the first scanning position X1 in the X direction. As a result, the laser beam probe moves to the first measurement position (X1, Y1).

In step S7, the wafer is held at the measurement position for a certain period of time, and the main surface is irradiated with a laser beam to detect the scattered light. As already described with reference to FIG. 4, if the intensity signal level of the detected scattered light is higher than the reference value (threshold) Vref (that is, if a crystal defect point exists at the measurement position), X- The counter values of the counter 103a and the Y-counter 103b are
The data is taken in as X coordinate data and Y coordinate data of the crystal defect detection point (S8, S9). The captured data is
It is stored in the crystal defect detection point data storage area 106c of the RAM 106 in FIG. On the other hand, if the intensity signal level of the detected scattered light is lower than Vref, no data is captured (S9 is skipped). That is, it is determined that no defect point exists at the measurement position. Note that, in this flowchart, the step of comparing with the threshold is represented as S8 for convenience, but this comparison step is actually performed by hardware in the data interface 103 of FIG.
Needless to say, it may be performed by software in a).

In S10, it is determined whether the measurement position has reached the limit position XN in the X direction. If it has not arrived, the process proceeds to S11, in which the value of the measurement position coordinate Xk in the X direction is increased to the next coordinate value Xk + 1 while the value of the measurement position coordinate Yk in the Y direction is fixed, and the measurement position is moved. And returns to S7 to repeat the following processes up to S9. On the other hand, if the measurement position has reached the limit position XN in the X direction in S10, the process proceeds to S12, and it is similarly determined whether or not the measurement position has reached the limit position YN in the Y direction. S1 if not reached
In step 3, the value of Yk is increased to the next coordinate value Yk + 1.
6, the value of Xk is returned to the initial value X1. Thereafter, the process returns to S7 and repeats the following processes up to S9. On the other hand, if the value of Yk has reached YN in S13, it means that the measurement at all the measurement positions has been completed.
As shown in FIG. 7, a set of the coordinate data (x, y) of the taken crystal defect detection point is stored in the HDD 107 as a crystal defect detection point data file 107c in a form corresponding to the wafer specifying data.

Next, FIG. 6 shows the data processing computer 2
00 shows an example of the flow of a crystal defect evaluation process executed by the data processing program 204a. First, in S51, wafer identification data of a wafer to be evaluated is input. This input may be performed by directly inputting the wafer specifying data with the keyboard 207, or by selecting the data file menu or icon displayed on the screen with the mouse 206.
Then, the process proceeds to S52, in which a data file of the crystal defect detection point corresponding to the input wafer specifying data is read out from the particle detection device 100 by acquiring the data file via the communication line 220. Then, in S53, the read data of the crystal defect detection point is displayed on the monitor 205 on the screen.

FIG. 7 shows a screen display example in which each crystal defect detection point 2 is plotted on a coordinate plane set on the screen based on the X coordinate value and the Y coordinate value of the data. Is displayed with. That is, the function of the display mechanism is realized. In the present embodiment, the image 1 of the wafer outline is displayed together so that the position of occurrence of crystal defects on the wafer W can be easily grasped.

In the image of FIG. 11, it can be seen that most of the crystal defect detection points 2 are arranged in a linear form in a plurality of rows mainly along the [110] direction. These detection points are [11
0] direction is derived from slip dislocations generated in the [0] direction. Therefore, processing for selectively extracting the detection points arranged in a straight line is performed. That is, in S54 of FIG. 6, first, a crystal defect type name (for example, “slip dislocation”) is input (this is also convenient if the menu can be selected by the mouse 206).
Next, the process proceeds to S55, where an extraction area is set.

FIG. 12 shows an example of setting an extraction area.
For example, in the case of FIG. 11, the rows of the detection points 2 corresponding to the slip dislocations are densely formed in a plurality of regions on the outer edge of the wafer. Therefore, as shown in FIG. 12A, the extraction area 3 individually surrounding each dense area of the detection points 2
Setting, it is possible to exclude detection points (for example, those originating from crystal defects other than slip dislocations or from sources other than crystal defects such as noise) existing in portions other than the dense region.

The extraction area 3 is set by, for example, as shown in FIG. 12B, a pointer which moves on the screen by operating the mouse 206 (FIG. 3) on the display screen of the detection point plot. P is displayed, the area defining points J _S , J _E are input on the screen using the pointer P, and the extraction area 3 is determined based on the information of the area defining points J _S , J _E. Can be. For example, in FIG. 12B, in order to set a rectangular extraction region 3 whose opposite sides are parallel to the X axis and the Y axis, respectively, two vertices J _S and J _E located at both ends of the diagonal of the rectangular region. Is determined. As the method of operation, the mouse 206 move the pointer P to the position of the start point side vertex J _S, moves the pointer P while pressing the mouse button, not shown to the position of the end point vertex J _E, the use of a so-called mouse dragging Can be.

On the other hand, the extraction region 3 can be formed in an arbitrary polygonal shape or a closed curve shape as shown in FIG. For example, FIG. 13 (a), an example of setting the extraction area of the polygonal detection point group U ₀ to be extracted
Points J ₁ , J that give each vertex of the polygon surrounding
The _{2, J} 3, _{J 4} and plotted by a mouse click, shows an example of setting the extraction area _{A 1} square. The area determination processing is performed by clicking back to the start point J ₁ from the last point J _4. In this case, the extraction area A ₁ includes a plurality of vectors V ₁ , V ₁ connecting adjacent vertices of the polygon.
₂ , V ₃ , and V ₄ are provided as outline graphic objects. FIG.
(C) shows an example in which a group of detection points 5 arranged along a spiral path such as swirl is surrounded by a polygonal extraction region 3 set by the same method as described above. Note that a polygonal region can be set by connecting straight lines between the region defining points. However, if the connection is made by a free curve such as a spline curve, the extraction region can be set to an arbitrary closed curve shape. With the above method, the function of the extraction region setting input mechanism is realized.

Returning to FIG. 6, the set extraction area is S5
At 6 the image is drawn and displayed on the screen. The extraction area 3
Can be identifiably displayed on the screen by displaying an area boundary line or painting the area. After S57, the process of identifying the detection points inside and outside the extraction region 3 and the process of counting the detection points inside the extraction region 3 are performed. RAM2
03 (FIG. 3), as shown in FIG.
An extraction flag 250 is set in association with the coordinate data of each crystal defect detection point.

Then, in S57 of FIG. 6, the stored value of the extraction flag is reset, and the value of the counter nc is set to 0. In S58, the data point number ND of the crystal defect detection point is set to 1, and in S59, the coordinate data of that number is read. In S60, it is determined whether or not the read coordinate data is located in the set extraction area 3. If it is located in the area, the value of the extraction flag corresponding to the data point is set to the first storage value. (“1” in the present embodiment) (S61), and when it is located outside the area, the second storage value (“0” in the present embodiment). That is, the function of the selection input mechanism and the function of the storage control mechanism are realized in addition to the function of the extraction area setting input mechanism.
Each time a data point in the region is found, a counter n
The value of c is incremented (S62). And S
Proceeding to 63, it is determined whether there is a next data point. If there is, the flow proceeds to S64 to increment the data point number ND, and returns to S59 to repeat the processing up to S62. That is, the function of the counting mechanism of the extracted crystal defect detection points is realized.

When the determination for all the data points is completed, the process proceeds from S63 to S65, and as shown in FIG. 13B, the data point for which the extraction flag is "1", that is, the crystal in the extraction area. The display state of the defect detection point
A change is made so that it can be distinguished from a data point that is “0”, that is, a crystal defect detection point outside the extraction region (for example, by changing the color, brightness, or plot point shape). Thus, it is possible to confirm at a glance on the screen which of the extracted crystal defect detection points is.

Then, the flow advances to S65, where the extracted data points and the value of the counter nc are associated with the input crystal defect type name and wafer specifying data and evaluated as shown in FIG. 9A. The result is stored as the result data file 204c. Thus, a crystal defect evaluation mechanism for quantitatively evaluating crystal defects based on the counting result of the extraction crystal defect detection point counting mechanism is realized. In the example shown in FIG. 12, crystal defect detection points arranged in a linear form derived from slip dislocations are extracted and evaluated. However, when extracting and evaluating other types of defects, the processing in FIG. , S54, the corresponding one is input as the crystal defect type name, and in S55, the process is repeated in such a manner that the extraction area is set so as to include the detection points arranged in a form specific to the crystal defect of interest. . As a result, as shown in FIG. 9A, the extraction region data, the extraction point data, and the count value are stored in the evaluation result data file 204c in a form associated with each other for each crystal defect type.

The evaluation result can be output in various forms based on the storage contents of the evaluation result data file 204c. For example, FIG. This is an example in which the production date and the number of detection points for each type of crystal defect are output. Note that the output may be visually output to the monitor 205 in FIG. 3 or may be printed out by the printer 210. That is, the function of the output mechanism of the crystal defect evaluation result is realized. For example, when the count number of crystal defects of a specific type exceeds a predetermined limit value, the wafer or product lot can be determined to be defective and can be sorted out from non-defective products.

Hereinafter, an experiment performed to confirm the effects of the present invention will be described. First, as a semiconductor wafer, an epitaxial wafer was prepared by growing an epitaxial layer having a thickness of about 50 μm under various temperature conditions on a silicon single crystal substrate having a diameter of 125 mm and a plane orientation of (100).
Then, the crystal defects to be extracted were regarded as slip dislocations, and the evaluation was performed using the above-described measuring apparatus 100. On the other hand, for comparison, the same sample was evaluated by a conventional visual inspection under a condensing lamp. That is, slip dislocations on the epitaxial wafer surface were observed by visual inspection, and all of them were manually sketched (FIG. 10 is an example of the sketch). Furthermore, the length of the sketched slip dislocations was measured, and the degree of occurrence of the slip dislocations was determined by obtaining the cumulative length. FIG. 17 shows a comparison between the number of crystal defect detection points caused by slip dislocations obtained by the evaluation method of the present invention and the cumulative length of slip dislocations by visual inspection. As described above, it was confirmed that the visual inspection and the method of the present invention have a good correlation in quantitatively evaluating the state of occurrence of crystal defects.

(Embodiment 2) In the first embodiment, the extraction of the crystal defect detection points is performed by manually setting the extraction area. However, the crystal defect detection points arranged in a specific form are determined by program processing. It is also possible to automatically identify and extract. FIG. 14 is a schematic flowchart showing an example of a program routine for automatically extracting a slip dislocation. It is assumed that the inspection process of FIG. 5 has already been completed and that a data file of the crystal defect detection point has been created.
At 101, the plane orientation of the wafer is input. This is because the direction of the slip dislocation appearing on the main surface of the wafer coincides with the cleavage direction located in the main surface, and thus the direction in which the slip dislocation appears on the main surface differs depending on the plane orientation of the wafer. The <110> direction, which is the cleavage direction of the silicon single crystal wafer, is as shown in FIG.
The (100) wafer appears in two directions, and as shown in FIG. 3B, the (111) wafer appears in three directions, and crystal defect detection points derived from slip dislocations are also arranged along these directions. . Then, along the cleavage direction on these wafer main surfaces, the arrangement direction vector λ _k (= λ ₁ , λ _{2 }} ) of the crystal defect defined by the XY coordinate display.
Is stored in advance for each type of wafer plane orientation.

In S102 of FIG. 4, the data of the array direction vector λ of the crystal defects corresponding to the input plane orientation is read. In S103, an extraction flag and a labeling flag described later are reset. Then, in S104, the first one of the array direction vectors λk is read, and in S10
In FIG. 5, as shown in FIG.
Each coordinate data of the crystal defect detection point is subjected to, for example, rotational transformation around the center of the wafer main surface so as to be parallel to the axis. As shown in FIG. 16, the coordinate values after conversion for each λk are
The extraction flag and the labeling flag are stored in the extraction processing table 300 in a one-to-one correspondence. As a result of this conversion processing, if there are detection points arranged in a direction along λk, the arrangement of the detection points is changed in direction substantially parallel to the X axis. The extraction flag has a value of “1” or “0” indicating whether or not to perform extraction,
The labeling flag includes a labeling number M (for example, M =
0, 1, 2, 3,...: Where M = 0 indicates unlabeled).

Returning to FIG. 14, the labeling number M is set to "1" in S106, and the value of the continuous point counter N is reset in S107. Then, in S108, among the data points whose labeling flags are “0”, the data point with the largest Y coordinate is extracted, and that point is set as the reference point S. At the start of the process, since the labeling flags of all data points are "0", the data point having the largest Y coordinate among all the data points is selected. In order to more reliably select the leading one of the detection points continuously arranged in the X direction, the one having the largest Y coordinate is selected once, and the Y coordinate value is set as a reference value, and a certain value including the reference value is used. What is necessary is to extract all the detection points whose Y coordinate values are within the range, and select the detection point with the smallest X coordinate among the extracted detection points. In S107, the value of the continuous point counter N is incremented in accordance with the selection, and in S110, a reference line L passing through the reference point S and parallel to the X axis is generated as shown in FIG.

Then, in S111, as shown in FIG. 15B, a predetermined continuous limit distance is set to ΔX, and X
A condition that the coordinates are within a range of + ΔX from the X coordinate of the reference point S, and a predetermined linear arrangement limit distance is α, and a separation distance in the Y direction from the reference line L is ± α (hereinafter, a continuous point condition) ) Is determined. If there is a detection point that satisfies the condition, the process proceeds to S112 with the point T as a detection point continuously arranged on the same straight line with respect to the reference point, and the value of the continuous point counter N is incremented. Then, in S113, the labeling flag of the point T is set to the labeling number M set at that time, the point T is reset as the reference point S in S114, and the process returns to S111 to repeat the following processing. . Therefore, in the X-axis direction (that is, the array direction vector λ
The point group that satisfies the continuous point condition connected in the (k direction) is all given the same labeling number M by this processing. On the other hand, if there is no detected point that satisfies the continuous point condition in S111, the process proceeds to S115, and the value of the continuous point number counter N at that time is checked. If this value is equal to or greater than a predetermined reference continuous number Nb, it is determined that the point group is derived from slip dislocations generated in the λk direction, and extraction of those points whose labeling number is M is performed. All flag values are set to “1”.

On the other hand, if the value of the continuous point counter N is equal to or smaller than the reference continuous number Nb, it is determined that those point groups are unrelated to the slip dislocation in the λk direction. All the set labeling flags are reset to the save flag value Mmax. Since the point to which Mmax is given may be derived from slip dislocation in another direction, the point can be reset at the time of executing a repetition routine in which the direction of λk is changed.

In the above process, in S108, after the reference line L is set so as to pass through the initially set reference point (hereinafter referred to as a continuous base point), a continuous point connected to the reference line L is The position as a reference point for determining a continuous point condition is successively taken over by an adjacent point, but the position of the reference line L is not changed until the row of continuous points is interrupted.
Therefore, as shown in FIG. 15D, when continuous points are generated in different directions from the reference line L, the distance of the first several points in the Y direction may be within the condition. However, as the distance from the continuous base point increases, the distance from the reference line L gradually increases.
Therefore, if the reference continuous number Nb is set to be larger than the number of continuous points that may occur due to the above-described factors, these points do not belong to the slip dislocation in the direction of interest. Can be reliably determined. Moreover, by giving the evacuation flag value Mmax and resetting the
When evaluating slip dislocations in the direction to which these points should originally belong, these points can be re-assigned to the unevaluated point group, so that the evaluation accuracy can be improved. Further, according to the above processing, as shown in FIG. 15E, an isolated detection point S that does not belong to any direction due to a factor such as noise.
Is present, no matter how the array direction vector λk is changed and evaluated, there will be no detection points determined to be connected to this, and this is surely excluded from the slip dislocation origin detection points. be able to.

Returning to FIG. 14, the program proceeds to S118, in which it is determined whether or not a point with a labeling flag of "0" still remains. If not, the process proceeds to S119, changes the labeling number M, for example, increments it, returns to S107, and repeats the following processing. In S108, a reference point is selected only for detection points whose labeling flag is "0" (not evaluated). In S111, it is determined whether or not there is a continuous point whose labeling flag is "0". It has become. Therefore, in the repetitive processing, the data that has already been labeled and evaluated is removed from the evaluation control, and the evaluation of the remaining unevaluated detection points is continued. For example, as shown in FIG. 15C, when a continuous point is interrupted on the same reference line, another labeling number is assigned to a detection point sequence appearing after the interrupted section. Therefore, if the points in the detection point sequence with the same labeling number are individually counted, the length of slip dislocations occurring on the wafer can be obtained individually based on the count value. Thus, for example, the distribution and average value of the slip dislocation length can be easily obtained.

On the other hand, in the above processing, every time the routine is repeated, the one with the maximum Y coordinate at that time is selected as the reference point. The evaluation of the detection point ends. In this state, the labeling flag of each point
This means that some labeling number other than “0” is assigned. Measurement points not originating from slip dislocations generated in the λk direction are given withdrawal labeling numbers Mmax, and the extraction flag values of those points remain “0”. Therefore, if it is found in S118 that there are no more detection points with the labeling flag of "0", the process exits the loop and proceeds to S120 to replace the array direction vector λk with the next one. S121 is a step of determining whether or not an unused array direction vector remains, and if so, the next array direction vector λk + 1
Select and set. Then, the process proceeds to S122, in which the labeling flags to which the evacuation flag value Mmax has been added in the previous evaluation routine are all reset to “0”, and the process returns to S105, where the newly set array direction vector λk and X-axis direction are set. Are again converted so that the coordinate values match.
The following processing is exactly the same except that only the measurement points whose labeling flags have been reset to “0” are evaluated.

When the above processing is completed for all the array direction vectors, the process proceeds from S121 to S123, and the values of the extraction flags at all the detection points are confirmed. It is considered that the detection point where the value of the extraction flag is “1” is derived from a slip dislocation appearing in any of the arrangement direction vectors.
In step 4, this can be determined as the total number of slip dislocation defect points (a parameter reflecting the total length of slip dislocations).

[Brief description of the drawings]

FIG. 1 is a view schematically showing various crystal defects generated in a silicon epitaxial wafer.

FIG. 2 is an explanatory view schematically showing a manufacturing process of a silicon epitaxial wafer.

FIG. 3 is a block diagram showing an example of an electrical configuration of the semiconductor wafer measuring device of the present invention. .

FIG. 4 is a circuit diagram showing a configuration example of a data interface of FIG. 3;

FIG. 5 is a flowchart showing a flow of an inspection process in the measuring device of FIG. 3;

FIG. 6 is a flowchart showing the flow of a crystal defect evaluation process.

FIG. 7 is a conceptual diagram showing the contents of a crystal defect detection point data file.

FIG. 8 is a conceptual diagram of an extraction flag.

FIG. 9 is a conceptual diagram showing the contents of an evaluation result data file together with an output example thereof.

FIG. 10 is a sketch diagram showing an example of a silicon epitaxial wafer on which slip dislocations are formed.

FIG. 11 is a view showing a detection output of a crystal defect with respect to the silicon epitaxial wafer of FIG. 10;

FIG. 12 is a diagram showing an example of setting an extraction region for a group of crystal defect detection points detected in FIG. 11;

FIG. 13 is an explanatory diagram showing some setting examples of a polygonal extraction region.

FIG. 14 is a flowchart illustrating an example of an algorithm for automatically extracting a slip dislocation.

FIG. 15 is a diagram schematically illustrating the flow of the process in FIG. 14;

FIG. 16 is a view showing the concept of an extraction processing table used in the processing of FIG. 14;

FIG. 17 is a graph showing a correlation between a crystal defect evaluation result obtained by the measurement method of the present invention and a crystal defect evaluation result obtained by a conventional method.

[Explanation of symbols]

W Silicon epitaxial wafer (semiconductor wafer) 2 Crystal defect detection point 3 Extraction area 100 Semiconductor wafer measurement device 150 Particle detection device (crystal defect measurement mechanism, light scattering wafer surface inspection device) 200 Data processing computer (crystal defect detection point Extraction mechanism, extraction crystal counting point counting mechanism, crystal defect evaluation mechanism, storage control mechanism) 204 Hard disk drive (evaluation result storage mechanism) 205 monitor (display mechanism, crystal defect evaluation result output mechanism) 206 mouse (selection) Input mechanism, extraction area setting input mechanism) 210 Printer (crystal defect evaluation result output mechanism)

Continued on the front page F term (reference) 2G051 AA51 AB06 AC11 BA10 BC05 CB05 DA07 EA12 EA14 EA21 EB01 ED01 ED09 4G077 AA02 BA04 CF10 GA06 4M106 AA01 BA05 CA27 CA42 CA43 CA50 DA15 DH03 DH12 DH32 DJ14 DJ20 DJ21

Claims (17)

[Claims] 1. The use of a crystal defect inspection apparatus,
A crystal defect measuring step of measuring the distribution of crystal defects occurring in the semiconductor wafer as a set of crystal defect detection points, and from the set of crystal defect detection points the crystal defect detection points belonging to an array form unique to the specific crystal defect. Extracting a crystal defect detection point to be extracted. 2. The method according to claim 1, wherein a crystal defect formed in the semiconductor wafer is identified according to an arrangement form of the crystal defect detection points extracted in the step of extracting the crystal defect detection points. A method for measuring semiconductor wafers. 3. After the step of extracting crystal defect detection points,
3. The semiconductor wafer measurement according to claim 1, wherein the number of the extracted crystal defect detection points is counted, and a crystal defect generated in the semiconductor wafer is quantitatively evaluated based on the counting result. 4. Method. 4. The quantitative evaluation of the crystal defects is performed for each crystal defect identified according to an arrangement form of the crystal defect detection points.
4. The method according to claim 3, wherein the method is performed by counting the number of crystal defect detection points. 5. In the step of extracting crystal defect detection points, a set of the measured crystal defect detection points is displayed on a screen of a display device, and the crystal defect detection points are extracted on the screen. The method for measuring a semiconductor wafer according to claim 1, wherein: 6. The method according to claim 5, wherein an extraction area is set on a screen of the display device for displaying the set of the crystal defect detection points, and the crystal defect detection points within the set area are extracted. The method for measuring a semiconductor wafer according to the above. 7. The extraction region is set so as to surround a crystal defect detection point arranged in a form specific to a specific crystal defect among a set of the crystal defect points displayed on the screen. The method for measuring a semiconductor wafer according to claim 6. 8. The method according to claim 1, wherein in the crystal defect measuring step, a light scattering type wafer surface inspection device is used as the crystal defect detection device. 9. The method for measuring a semiconductor wafer according to claim 1, wherein the semiconductor wafer has crystal defects that have been revealed by selective etching. 10. The method according to claim 1, wherein the crystal defects are slip dislocations. 11. A crystal defect measuring step of measuring a distribution of crystal defects occurring in a semiconductor wafer as a set of crystal defect detection points by using a crystal defect inspection apparatus, and an arrangement form specific to a specific crystal defect. A crystal defect detection point extracting step of extracting a crystal defect detection point belonging to a set of the crystal defect detection points, and a crystal defect for evaluating a crystal defect of the semiconductor wafer based on an extraction result by the crystal defect detection point extraction step. A method of manufacturing a semiconductor wafer, comprising: an evaluation step; and a selection step of selecting the semiconductor wafer based on an evaluation result in the crystal defect evaluation step. 12. A crystal defect measurement mechanism for measuring the distribution of crystal defects generated in a semiconductor wafer as a set of crystal defect detection points, and a specific crystal defect among a stored set of crystal defect detection points. And a mechanism for extracting crystal defect detection points for extracting data of crystal defect detection points belonging to various arrangement forms. 13. A counting mechanism of extracted crystal defect detection points for counting the number of crystal defect detection points extracted by the crystal defect detection point extraction mechanism, and a counting result by the counting mechanism of the extracted crystal defect detection points. 13. A crystal defect evaluation mechanism for quantitatively evaluating a crystal defect by performing the calculation, and a crystal defect evaluation result output mechanism for outputting the number of the counted crystal defect detection points as an evaluation result. Semiconductor wafer measuring device. 14. The type of a crystal defect identified according to the arrangement form of the crystal defect detection points extracted by the crystal defect detection point extraction mechanism corresponds to the content of the evaluation result performed by the crystal defect evaluation mechanism. 14. The apparatus for measuring a semiconductor wafer according to claim 12, further comprising a storage mechanism for storing an evaluation result that is attached and stored. 15. A crystal defect detection point coordinate data storage mechanism for storing a set of crystal defect detection points measured by the crystal defect measurement mechanism, and the stored crystal defect detection point set data. A display mechanism for displaying a distribution shape of the crystal defect detection points on the semiconductor wafer on a screen, wherein the extraction mechanism for the crystal defect detection points is a distribution of crystal defect detection points displayed on a screen of the display mechanism. While referring to the shape, the data of crystal defect detection points arranged in a specific form,
A selection input mechanism for selecting the data as the data corresponding to the specific crystal defect detection point, and the data of the selected crystal defect detection point, the crystal in a state that can be distinguished from the data of the non-selected crystal defect detection point A memory control mechanism for storing the detected crystal defect in a coordinate data storage mechanism, wherein the extracted crystal defect detection point counting mechanism counts the number of data points of the selected crystal defect detection point. 15. The measuring device for a semiconductor wafer according to any one of 12 and 14. 16. The selection input mechanism includes an extraction area setting input mechanism for setting an extraction area on a screen of the display mechanism, and the storage control mechanism determines a crystal defect detection point in the set extraction area. 16. The semiconductor wafer measuring apparatus according to claim 15, wherein data is stored in the coordinate data storage mechanism of the crystal defect detection point as data of the selected crystal defect detection point. 17. The apparatus according to claim 1, wherein the crystal defect measuring mechanism is a light scattering type wafer surface inspection apparatus.
16. The apparatus for measuring a semiconductor wafer according to any one of 2 to 15.