JP3781601B2 - Inspection method of semiconductor samples - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a semiconductor sample inspection method for inspecting a semiconductor wafer in a semiconductor device manufacturing process.
[0002]
[Prior art]
In a manufacturing process of a semiconductor device such as an LSI, a semiconductor wafer having a predetermined pattern formed on the surface is inspected. This inspection detects defects such as pattern short-circuits and defects, adhesion of foreign substances, etc., classifies the types of defects, and feeds back the classification results to the semiconductor device manufacturing process, so as to improve the production yield of semiconductor devices I have to.
[0003]
The above-described defect inspection was initially performed using an optical microscope at an appropriate stage of the LSI manufacturing process. Defects clarified by inspection are classified into about 15 classifications according to their types, and are associated with the cause of the occurrence of defects.
[0004]
However, such inspection has become a size that cannot be observed with an optical microscope due to miniaturization of LSI. Even in the case of VLSI, there are defects of the same size as in the past, and it is more efficient in terms of throughput to rely on an optical microscope. However, with the progress of pattern miniaturization, fine defects cause poor performance of VLSI, so inspection by a charged particle beam apparatus such as a scanning electron microscope capable of inspecting fine defects has become essential. .
[0005]
The procedure of inspection by this scanning electron microscope is as follows. First, the position and size of the defect are inspected in advance by an optical defect inspection apparatus using a laser beam or the like. In this case, the position of the defect is determined by a coordinate value in a virtual coordinate system on the wafer sample, and the size of the defect is not accompanied by accuracy in the case of a fine defect, but an approximate size is found. In this optical defect inspection apparatus, a part of the foreign matter buried in the lower layer other than the image on the surface of the wafer sample can be observed.
[0006]
Data of the position and size of the defect portion inspected by the optical defect inspection apparatus is transmitted to a control processing system of the scanning electron microscope through a medium such as a floppy disk or using a communication function like a LAN. At the same time, the wafer sample loaded in the optical defect inspection apparatus is taken out and loaded on the moving stage in the sample chamber of the scanning electron microscope.
[0007]
In a scanning electron microscope (SEM), each defect portion is scanned with an electron beam, secondary electrons obtained based on this scanning are detected, and this detection signal is used as image data of the defect portion. This image data is processed by computer image processing technology, and the defects to be inspected are identified into about 15 classifications.
[0008]
Now, when acquiring an SEM image of a fine defect portion to be inspected, it is necessary that the image is enlarged to a magnification suitable for image processing. In SEM, a large number of fine defect images of interest are designed to be captured sequentially. In this case, after the image capture of one fine defect portion is completed, the moving stage is driven to move the wafer sample to the defect position obtained as data in advance in order to capture the fine defect at the next location. .
[0009]
To move the wafer sample to the defect position means to arrange the defect portion on the electron beam optical axis of the SEM. In other words, the defect portion is located at the center of the field of view of the SEM.
[0010]
As described above, even when the wafer sample is moved to the defect position, the defect portion is often not positioned at the center of the field of view of the SEM due to a stage movement error or the like. This positional error is of the order of several μm, and if SEM image data is taken in at a magnification required for defect classification, the defect portion will be out of the field of view, image processing cannot be performed, and defect classification will be hindered. It will be.
[0011]
Therefore, when the wafer is moved, first, SEM image data is acquired at a magnification of about 5,000 to 10,000 times lower than the optimum magnification for defect classification. Then, the position and size of the defective portion are detected again from the low-magnification image data, and the defective portion is moved to the center of the field of view based on the detection result. Thereafter, the image data is captured at a high magnification required for image processing for defect classification.
[0012]
In addition, when moving a defective part to the center of a visual field, a movement stage may be moved mechanically and the scanning range of an electron beam may be moved electrically (image shift).
[0013]
The series of procedures described above are automatically executed by computer control. However, as a method for detecting the position and size of a defective portion from low-magnification image data, a defect-free image acquired in advance is used. An image comparison method is used in which data of (reference image) is compared with data of an image including a defective portion (defect image).
[0014]
[Problems to be solved by the invention]
When defect inspection of a wafer is started by SEM, first, a wafer sample is moved to a defect-free chip close to a defective chip, and image data at the same location corresponding to the defect position in this chip is detected again. The image is taken in at the same magnification as that set for use as reference image data.
[0015]
Next, the wafer sample is moved to the defect position of the chip having the defect, and the image data of the defective portion is taken in as defect image data at the magnification set for redetection of the defect.
[0016]
Therefore, the time for redetecting the defect requires both the time for moving the moving stage to the chip having no defect and the time for capturing the image data in order to capture the reference image data. Even when the position of one defect and the next defect is close to each other, such an operation is performed for all the defective portions, so that there is a problem that the inspection time is long.
[0017]
The present invention has been made in view of these points, and an object thereof is to realize a semiconductor sample inspection method capable of inspecting a defective portion of a semiconductor wafer sample in a short time.
[0018]
[Means for Solving the Problems]
According to a first aspect of the present invention, there is provided a method for inspecting a semiconductor sample on a sample by a first inspection apparatus having an optical microscope function for a semiconductor sample to be inspected having a plurality of specific regions in which patterns having the same shape are arranged. At least the position of the defective part of the sample is recognized, and the electron beam is scanned at the position of the defective part on the sample recognized by the first inspection apparatus by the second inspection apparatus having a scanning electron microscope function. A method of inspecting a semiconductor sample in which a predetermined inspection of a defective portion is performed based on a received image signal, wherein the second inspection device is identified as having no defect based on information obtained by the first inspection device. An area is found, the specific area is virtually divided, and scanning of the electron beam is performed for each of the divided areas at a predetermined magnification, and an image signal obtained from the sample by each scan is transmitted to the specific area. The image is stored as a quasi-image, and the second inspection apparatus scans the electron beam at a predetermined magnification at the defect portion recognized by the first inspection apparatus, detects a signal from the sample, acquires image data, and Cut out the image data of the region in the reference image data corresponding to the part, compare the cut out reference image data with the image data including the defective part, recognize at least the position of the defective part, and based on the recognized position The defect portion is arranged in the vicinity of the scanning center of the electron beam, the electron beam is scanned at a predetermined magnification to obtain image data, and the defect portion is inspected based on the image data. The scanning range of the electron beam when acquiring image data of a specific area that is not present is made larger than the scanning range of the electron beam when acquiring image data of the defective portion. It is set to.
[0019]
According to the first aspect of the present invention, reference image data of a predetermined area is stored in advance, and the image data of the defective portion and the image data of the region in the reference image data corresponding to the defective portion are cut out and cut out. Since the data and the image data including the defective part are compared, it is not necessary to acquire the reference image data by electron beam scanning in the defect-free area for each inspection of many defective parts, and the inspection time is remarkably shortened. Can do. Further, since the scanning range of the electron beam when acquiring the image data of the specific area having no defect is made larger than the scanning range of the electron beam when acquiring the image data of the defective portion, the reference image data of the specific area The time to acquire can be shortened.
[0023]
In the invention of claim 2 wherein, the pixel density of the scanning range of the electron beam when acquiring the image data without specific area defect, the pixel density of the scanning range of the electron beam when acquiring the image data of the defective portion On the other hand, since it is roughened, the time for acquiring the reference image data of the specific area can be shortened, and the memory of the image data can be saved.
[0026]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 shows an example of a basic system configuration for carrying out an inspection method according to the present invention. In the figure, reference numeral 1 denotes an optical inspection apparatus as a first inspection apparatus. In this optical inspection apparatus, a semiconductor wafer sample to be inspected is irradiated with laser light, light scattered from the sample is detected, and the position and size of a defect on the sample are inspected based on this detection signal. The defect of the wafer sample is an unintended shape (for example, pattern short circuit or defect) in a pattern formed on the sample surface, adhesion of foreign matters, and the like.
[0027]
Data on the position and size of the defective portion recognized by the optical inspection apparatus 1 is transmitted to the control processing unit 2 constituting a part of the second inspection apparatus via the network. For the data transfer, a medium such as a floppy disk may be used in addition to the network transfer.
[0028]
The control processing unit 2 controls the SEM included in the second inspection apparatus. The SEM includes an electron gun 3, a condenser lens 4, an objective lens 5, a deflector 6, and a moving stage 8 on which a sample 7 is placed.
[0029]
The accelerated electron beam EB generated from the electron gun 3 is finely converged on the sample 7 by the condenser lens 4 and the objective lens 5. The irradiation position of the electron beam on the sample 7 is scanned two-dimensionally by the deflector 6. Further, the electron beam irradiation region on the sample can be moved by arbitrarily moving the moving stage 8 in the X and Y directions. The drive circuit 9 for the deflector 6 and the drive circuit 10 for the moving stage 8 are controlled by the control processing unit 2.
[0030]
For example, secondary electrons generated by irradiation of the sample 7 with the electron beam EB are detected by the secondary electron detector 11. The detection signal of the detector 11 is supplied to the control processing unit 2 through the amplifier 12 and the AD converter 13 as image data representing the sample surface shape.
[0031]
The image data supplied to the control processing unit 2 is supplied to and stored in the image memory 14. The data stored in the image memory 14 is read out and supplied to the image processing unit 15, and defect redetection based on the image data (hereinafter, defect detection by the second inspection apparatus is referred to as redetection). And classification processing is performed. The operation of such a configuration will be described next.
[0032]
First, when the sample 7 is a semiconductor wafer, the wafer sample is loaded into the optical inspection apparatus 1 which is a first inspection apparatus. In this optical inspection apparatus 1, the entire surface of a wafer sample on which a large number of patterns are formed is irradiated with laser light, and the scattered light is detected.
[0033]
With this scattered light, data on the position and size of the defective portion is obtained for each pattern. The obtained data is supplied to the control processing unit 2 by the data transfer means. On the other hand, the wafer sample loaded in the optical inspection apparatus 1 is taken out and placed on the moving stage 8 of the SEM that constitutes a part of the second inspection apparatus.
[0034]
Next, the image data is acquired based on the electron beam scanning of the defective portion using the SEM function, and the defect re-detection and defect classification processing are performed. That is, the operation of positioning the defective portion at the center of the electron beam scanning range (field of view) when acquiring image data is performed. This alignment process is performed by recognizing the position and size of the defective portion by comparison with a defect-free pattern corresponding to the pattern of the defective portion.
[0035]
FIG. 2 schematically shows an example of the result of inspection by the optical inspection apparatus 1, and W is a semiconductor wafer to be inspected. A large number of chips C are formed on the wafer W, and a region marked with an X in the figure is a defective region found by the optical inspection apparatus 1.
[0036]
In the present invention, one of the chips C0 having no defect is arbitrarily selected, and reference image data is acquired based on this chip. In addition, it is desirable to select a defect-free chip in the vicinity of a defective chip.
[0037]
Here, in the chip formed on the semiconductor wafer sample to be inspected, the LSI pattern arrangement in the chip is generally arranged separately for each functional block. FIG. 3 shows a chip C formed on the wafer sample, and the chip C has a plurality of functional blocks Ba to Bn.
[0038]
The functional block Ba is composed of a plurality of areas a1, a2,..., But the same pattern is formed in each area. Another functional block Bb is composed of a plurality of regions b1, b2,..., And the same pattern is formed in each region. Note that not all functional blocks are composed of a plurality of regions, and in some cases, there are functional blocks composed of a single region.
[0039]
When acquiring the reference image data, for the functional block Ba, one specific area among the areas a1, a2,... Is selected as the representative pattern area. For example, the area a1 is selected, and the image data of the area a1 is acquired using the SEM function. More specifically, the functional block Ba is observed at an appropriate SEM magnification, and the origin (Xa10, Ya10) of the block Ba and the range (Xa11, Ya11), (Xa12, Ya12), (Xa13) of the block Ba are observed. , Ya13) are read as coordinates from the origin (Xa, Ya) of the block Ba. Further, the origin of the block Ba is a coordinate from the origin of the chip.
[0040]
Thereafter, the image of the area a1 is captured at the same magnification (corresponding to the scanning range of the electron beam) as the magnification (reference magnification) used when acquiring the image data of the defective portion.
[0041]
Usually, since the range of the area a1 is wider than one field of view of the SEM based on the reference magnification, the area a1 is virtually divided and image data is acquired for each divided area. FIG. 4 shows a region a1, and the region a1 is virtually divided like a dotted line, and image data is acquired for each divided region.
[0042]
That is, first, two-dimensional scanning of the electron beam is performed in the divided region a11, and secondary electrons generated based on this scanning are detected by the secondary electron detector 11. The detection signal of the detector 11 is amplified by the amplifier 12, converted into a digital signal by the AD converter 13, and supplied to the control processing unit 2. The control processing unit 2 supplies and stores the supplied image data to the image memory 14 in association with the scanning position of the electron beam, for example, the center coordinates of the scanning region.
[0043]
After the image data is acquired in one divided area in this way, the visual field is moved by the movement of the moving stage 8 in the X and Y directions using a distance equal to the size of the visual field as a pitch. After the field of view is moved, the image data is acquired and stored in the image memory 14 in the same procedure as the first divided area. Note that the field of view can be moved by using an image shift function that electrically moves the scanning range of the electron beam regardless of the movement of the moving stage.
[0044]
In this way, the image data of all the divided areas a11 to a1n included in the area a1 are sequentially fetched, and this data is supplied to the image memory 14 via the control processing unit 2, and in this memory, each data The coordinates of the center of the field of view (divided area) are stored together with the coordinates converted into the coordinates from the origin of the area a1.
[0045]
As a result, the entire image of the area a1 is stored in the image memory 14 together with the center coordinates of the converted visual field. Similarly, the image of the representative pattern area b1 of the functional block Bb is fetched into the image memory 14 with coordinates. Further, similar processing is performed for other functional blocks, and finally, representative patterns of all functional blocks are stored in the image memory as reference image data.
[0046]
After such pre-processing is performed, actual defect inspection processing is performed. This defect inspection processing will be described below. Based on the position data of the defect portion obtained by the optical inspection apparatus 1, the drive circuit 11 is controlled so that the defect portion is positioned on the electron beam optical axis of the scanning electron microscope, and the moving stage 8 is moved.
[0047]
After the movement of the wafer sample 7, the defect portion is two-dimensionally scanned by the electron beam EB at the magnification of defect redetection, and the secondary electron signal detected by the secondary electron detector 11 along with this scanning is the amplifier 12, It is supplied to the control processing unit 2 via the AD converter 13.
[0048]
The control processing unit 2 determines in which region the defect is located from the position data of the defective portion. For example, it is determined that the defect is located in the area b2 of the functional block Bb. Further, the coordinate value in the region b2 is calculated for the position Db2 of the defective portion in the region b2.
[0049]
Thereafter, the position Db1 corresponding to Db2 is found from the origin of the area b1 from the reference image data stored and saved in the image memory 14 for the area b1 representing each area of the functional block Bb. Then, an image having the number of pixels for one field of view centered on the position Db1 is cut out from the image memory 14 as a reference image.
[0050]
The cut-out reference image data and the image data of the defective portion are sent to the image processing unit 15, where an image comparison process is performed, and the position and size of the defect are redetected by this process. The defect portion is automatically moved to the center of the field of view of the scanning electron microscope based on the re-detected defect position, and an image is captured at an optimum magnification for performing the defect classification process. As the visual field movement, movement of the stage and movement by image shift are arbitrarily used.
[0051]
In addition, in the case of a large defect in which the size of a re-detected defect consists of several tens of pixels or more, defect classification processing can be performed using the re-detected image, so an image for defect classification processing Importing can be omitted.
[0052]
By the above-described steps, comparison processing with reference image data, redetection of defect positions, and image capturing processing for defect classification are sequentially performed on other defective portions. After the defect image data has been captured for all the defective portions, the image processing unit 15 classifies which type each defect belongs to using computer image processing technology.
[0053]
Although the embodiment of the present invention has been described in detail above, the present invention is not limited to this embodiment. For example, a representative pattern corresponding to a functional block having a defect or a region having a defect may be captured and stored as a reference image. In this case, when the number of wafers to be inspected is one, the reference image capturing time can be shortened.
[0054]
In addition, although the field of view is moved by taking one field of view as a pitch for capturing the reference image, it may be moved by a distance that is, for example, about 10% smaller than the field of view. If the field of view is moved in this way, a part of the adjacent fields of view overlaps, but it is possible to prevent the loss of the image at the joint between the field of view and the field of view caused by stage play or the like.
[0055]
Further, instead of performing defect classification processing after capturing all defect images, defect processing at the first defect position is performed, and then parallel processing is performed while the wafer is moved to the second defect position. By performing the classification, the inspection throughput can be further improved.
[0056]
Furthermore, in the LSI memory chip, the memory functional block occupies about 80% of the entire chip area, and the functional block is equal to the area because the block is composed of a repeated pattern arrangement. Furthermore, a region is a repetition of a specific pattern arrangement. In such a chip, the reference image of the entire area may be stored only in a specific pattern which is a part of the area without storing the reference image.
[0057]
Furthermore, the reference image can be captured at a magnification lower than the magnification at the time of defect inspection, for example, 1/2 times, and this image can be compared with the defect image. That is, the scanning range of the electron beam in the defect-free region when acquiring the reference image data is increased. In order to compare images with different magnifications as described above, the lower magnification reference image is subjected to image processing such as expansion (interpolation between pixels), etc., enlarged to the same magnification, and then compared with the defect image. Is detected.
[0058]
In other words, when acquiring the reference image, the pixel density is made coarser than the density of the defect image, and when comparing with the defect image, a process for complementing the pixels of the acquired reference image is performed to compare with the defect image. It can be performed. This is because a pattern having no defect has continuity or regularity of a figure, so that a correct pattern shape can be maintained even when a process for complementing pixels is performed. On the contrary, it is conceivable to perform comparison by shrinking a defective image (thinning out pixels), but it is better not to employ this method because information on a defective portion may be lost.
[0059]
When the reference image data is acquired at such a low magnification, according to experiments, the reference image can be reduced to about 1/8 times the defect image. When this reference image is acquired at a low magnification, the capturing area in one electron beam scan becomes wide, so that the capturing time of all reference images can be shortened according to the ratio of the magnification.
[0060]
【The invention's effect】
As described above, in the invention of claim 1, the reference image data of the predetermined area is stored in advance, and the image data of the defective part and the image data of the area in the reference image data corresponding to the defective part are cut out, Since the cut-out reference image data and the image data including the defective portion are compared, it is not necessary to acquire the reference image data by electron beam scanning in the defect-free area every time a lot of defects are inspected. It can be significantly shortened.
[0061]
In addition, by finding the functional block and area range and saving the representative pattern, the capture time and memory capacity are dramatically reduced compared to saving an enormous number of images over the entire chip area. can do.
[0062]
For example, if an image obtained by capturing the entire 10 mm square chip at a magnification of 10,000 is stored, if the size of the observation region is 20 μm, it is necessary to capture 250,000 images. Even if the capture time for one image is 3 seconds, a total capture time of about 200 hours is required.
[0063]
On the other hand, in the present invention, although it differs depending on the type of LSI, in a memory LSI in which a repetitive pattern occupies most of the entire chip area, the reference image capturing time and storage memory capacity are about 1/100 to 1 / 200. Further, even in a logic LSI having many types of patterns, those are about 1/30 to 1/50. Further, since the scanning range of the electron beam when acquiring the image data of the specific area having no defect is made larger than the scanning range of the electron beam when acquiring the image data of the defective portion, the reference image data is acquired. Time can be shortened.
[0067]
According to the second aspect of the present invention, the pixel density of scanning with an electron beam when acquiring image data of a specific area having no defect is coarser than the pixel density of scanning with an electron beam when acquiring image data of a defective portion. Therefore, it is possible to shorten the time for acquiring the reference image data of the specific area and save the image data memory.
[Brief description of the drawings]
FIG. 1 is a diagram showing an example of a basic system configuration for carrying out an inspection method according to the present invention.
FIG. 2 is a diagram schematically illustrating an example of a result of inspection performed by the optical inspection apparatus 1;
FIG. 3 is a diagram showing a chip C formed on a wafer sample.
4 is a diagram showing a repetitive pattern area a1 of the functional block Ba in the chip C shown in FIG. 3;
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 Optical inspection apparatus 2 Control processing unit 3 Electron gun 4 Condenser lens 5 Objective lens 6 Deflector 7 Wafer sample 8 Moving stage 9 Deflector drive circuit 10 Stage drive circuit 11 Secondary electron detector 12 Amplifier 13 AD converter 14 Image Memory 15 Image processing unit

Claims (2)

For a semiconductor sample to be inspected having a plurality of specific regions in which patterns of the same shape are arranged, at least the position of a defective portion on the sample is recognized by a first inspection apparatus having an optical microscope function, and a scanning electron microscope function is provided. The electron beam is scanned at the position of the defective part on the sample recognized by the first inspection apparatus by the second inspection apparatus provided, and the predetermined inspection of the defective part is performed based on the image signal obtained by this scanning. An inspection method for a semiconductor sample,
Based on the information obtained by the first inspection apparatus, the second inspection apparatus finds a specific area having no defect, virtually divides the specific area, and scans the electron beam for each of the divided areas at a predetermined magnification. The image signal obtained from the sample by each scan is stored as a reference image of the specific region, and the second inspection apparatus scans the electron beam at a predetermined magnification at the defect portion recognized by the first inspection apparatus. To obtain the image data by detecting the signal from the sample, cut out the image data of the region in the reference image data corresponding to the defective part,
The cut-out reference image data and the image data including the defective portion are compared, at least the position of the defective portion is recognized, and the defective portion is arranged in the vicinity of the scanning center of the electron beam based on the recognized position. Scanning an electron beam at a magnification to obtain image data, and inspecting a defective portion based on this image data,
A method for inspecting a semiconductor sample, characterized in that the scanning range of an electron beam when acquiring image data of a specific area having no defect is made larger than the scanning range of an electron beam when acquiring image data of a defective portion . The pixel density in the scanning range of the electron beam when acquiring the image data of the specific area without the defect is roughened with respect to the pixel density of the scanning range of the electron beam when acquiring the image data of the defective portion. The method for inspecting a semiconductor sample according to claim 1.

Images