Patent Application
20110261190
The present invention relates to a technique of a defect observation device and defect observation method for observing images of defects and the like occurring in the production process of semiconductor wafers and the like. More particularly, the present invention relates to a technique for facilitating the setting of conditions for automatic observation.
Circuit patterns formed on semiconductor wafers have been increasingly miniaturized, and the influence of defects occurring in the production process on the yield becomes more obvious. Under these circumstances, process control is increasingly important to prevent defects from occurring in the production stage. Currently, a wafer inspection device and an observation device (a review tool) are used in the production filed of semiconductor wafers to control the yield. The inspection device is a device for checking the presence or absence of a defect on a wafer at high speed. At this time, images of the state of the wafer surface are acquired by optical means (a bright-field wafer inspection device or a dark-field wafer inspection device) or electron beam. Then, the acquired images are processed to determine whether there is a defect. High speed is an important requirement for the inspection device. In order to satisfy this requirement, the inspection device acquires images with as large a pixel size as possible (namely, with a low resolution) to reduce the amount of image data. In many cases, it is possible to confirm the presence of a defect from the detected low resolution images, but it is difficult to determine in particular the type of the defect.
On the other hand, the review tool is a device for acquiring and observing images with the pixel size being reduced (namely, with a high resolution) with respect to each defect detected by the inspection device. In the production process of highly miniaturized semiconductor devices, the size of the defect to be inspected and observed is sometimes on the order of several tens of nanometers. In order to observe and classify the defect with a high accuracy, a resolution of nanometer order is necessary. Thus, in recent years a review tool using a scanning electron microscope (hereinafter referred to as a review SEM) has been widely used.
The recent defect inspection device has a high detection sensitivity and can detect a large number (several hundreds to several thousands, or sometimes to several millions) of defects are detected from a single wafer. As a result, the need for efficiency in the review operation to observe the detected defects has increased more than before. In order to meet this need, many of the review SEMs (Scanning Electron Microscopes) recently introduced into the market are provided with a function for automatically acquiring images of the defect portion detected by the inspection device, or ADR (Automatic Defect Review), as well as a function for classifying the acquired images, or ADC (Automatic Defect Classification).
An example of the automatic image acquisition and automatic defect classification functions of the convention review SEM are disclosed in Patent document 1. Patent document 1 describes the configuration of the review SEM, the automatic image acquisition and automatic defect classification functions, the operation sequence of the functions, the display method of the acquired images and classification results, and the like.
Patent document 1: JP-A No. 331784/2001
First, a brief description will be given of the automatic defect image acquisition function in the review SEM described above. Next, the problems to be solved by the invention will be described.
The automatic defect image acquisition function of the review SEM is a function for acquiring images of different portions of a wafer to be observed, from the input of defect position information obtained as a result of the defect inspection of the particular wafer by the inspection device as described above. The basic sequence of the function is as follows:
However, it is necessary to take into account the coordinate error of the inspection device in the image acquisition of the defect portion. In general, the inspection device may include a coordinate error of about several micrometers to several tens of micrometers. In this case, if the inspection device directly images at the defect coordinate position with a high magnification of, for example, 50 thousand to 200 thousand times, the defect may not come into the field of view of the inspection device. The approach for this case includes the following steps. First, take an image of the area in which a defect is expected to be present by the review SEM with a wide field of view (FOV) (for example, 15 μm to 10 μm). Next, detect the position of the defect by applying image process to the image. Then, take an image with a narrow field of view (for example, 10 μm to 4 μm) so that the detected position is located at the center of the field of view. In this case, the image acquisition is performed twice by the review SEM in the basic sequence as described above.
Further, as a method for achieving the defect position detection in the above sequence, there is a method for comparing an image (reference image) in which no defect is present, with an image (defect image) taken in the defect portion. Semiconductor wafers are produced by repeatedly forming the same circuit pattern on each chip. Thus, in general, the above comparison method is applied to detect a defect position on each chip by using a reference image which is an image taken in the adjacent chip in the portion corresponding to the defect portion on the particular chip. In this case, it is necessary to perform the imaging process of the reference image and to perform the stage shift for the imaging process, in addition to the above sequence. It is to be noted that the image quality control process, such as auto focus control and auto brightness control, is also necessary for the image acquisition.
As described above, the image acquisition is performed a plurality of times for a single defect such as in the sequence of acquiring the reference image other than the defect image, and in the sequence of acquiring defect images with two different types of field of view. At this time, the image quality control process may be necessary for each image acquisition.
Throughput is an important performance index for the automatic defect image acquisition function. The higher the throughput, the more defects can be observed per unit time. For this reason, high throughput is expected to increase the accuracy of understanding the defect occurrence state and the accuracy of determining countermeasures. In order to increase the throughput, it is necessary to reduce the time for the stage shift, the time for the image quality control function such as auto focus control, and the time for the image acquisition itself.
There are various approaches for reducing the image acquisition time. For example, the number of images used for image averaging is reduced. The SEM image has a lot of shot noises and has a low S/N ratio. Thus, in general, the review SEM acquires a plurality of images of the same portion, and averages the images to acquire an image with a high S/N ratio. If the number of images for the image averaging is reduced, the process time is reduced. Another effective approach for reducing the imaging operation time is to increase the current amount of electron beam current (hereinafter probe current) that is irradiated onto the sample. When the probe current is large, it is possible to acquire an image with a higher S/N, even if the averaging number is the same. Further, when the image size (the number of pixels) is reduced, it is expected to reduce not only the time for the image acquisition itself, but also the image process time as well as the time for transferring and storing images within the system. As a result, it is effective in increasing the throughput.
However, from the point of view of the defect detection process for the images acquired with a low magnification (namely, a wide field of view), the above approaches such as the reduction in the number of images to be averaged, the increase in the probe current, and the reduction in the image size may act in a direction that makes the detection more difficult. For example, the reduction in the number of images means that the defect detection must be performed from images with a lower S/N. As a result, the risk of false detection of noise as a defect increases.
Further, the increase in the probe current may lead to the risk of charging occurring on the sample surface, or the risk of contamination deposition on the sample by electron beam irradiation. In this case, a brightness difference occurs in the image even in the portion other than the defect portion. In this case, the normal portion is more likely to be falsely detected as the defect portion. Further, when the field of view of the imaging area is fixed, the reduction in the image size is equal to the increase in the pixel size, making it difficult to automatically detect defects of a size near or less than the pixel size. As a result, missing of micro defects is more likely to occur.
As described above, the change of the imaging conditions for the purpose of improving the throughput may increase the risk of reducing the capture rate. The relationship between the throughput and the capture rate is different depending on the process and type of the wafer to be inspected. Practically, it depends on the manual operation of an operator to determine the imaging conditions that satisfy both the throughput and the capture rate. As a result, a huge amount of time is required for the operation in a high-mix low-volume production line, and the like, in which frequent condition setting is necessary.
The object of the present invention is to solve the above problems by providing a defect observation method and defect review tool that can satisfy high throughput and high detection performance requirements.
In order to achieve the above object, an aspect of the present invention relates to a review SEM. The review SEM includes a means to store sets of images acquired with a plurality of imaging conditions, a means to store defect position information for each set o images, and a means to store the relationship candidate setting values for each of the parameters that constitute the imaging conditions, and process times for setting the individual candidate values. The parameters include optical parameters such as acceleration voltage and probe current, as well as imaging parameters such as image size and averaging number, and the like.
Further, the review SEM includes a means to set a plurality of imaging conditions by combining the candidate values for each of the individual parameters that constitute the imaging conditions, and a means to estimate the capture rate and the throughput performance with respect to the plurality of imaging conditions. Still further, the review SEM also includes a means to automatically select one or more imaging conditions from the plurality of imaging conditions, based on the capture rate and the throughput value. In addition, the review SEM has a function of displaying the capture rate and the throughput value that are calculated for each of the plurality of set imaging conditions, and have a function of selectively displaying the imaging condition automatically selected from the plurality of set imaging conditions.
According to the present invention, it is possible for a user to easily understand the relationship between the content of the condition setting and the two performance indexes of the capture rate and the throughput performance, which vary depending on the various conditions of the automatic image acquisition function set in the device. As a result, the user can easily set conditions for automatic review.
Hereinafter embodiments of a defect observation method according to the present invention will be described in detail.
In this device, changing the area (the field of view) to be scanned by the deflector 210 means changing the field of view in the image acquisition, which is equivalent to changing the magnification. Further, in the analog-digital (A/D) converter 209 for converting to a digital signal, the magnitude of the conversion clock interval (sampling interval) corresponds to the magnitude of the image size of the digital image to be acquired. For example, the reduction in the sampling interval is equal to the reduction in the image size even with the same field of view. In this case, smaller defects can be imaged. Such various setting conditions for image acquisition, as well as the optical conditions (the acceleration voltage of the electron beam 204 to be irradiated, the amount of current (probe current), and the like) for other image acquisitions, are set by instructions from the overall control unit 102 through the bus 116.
Next, a description will be given of the process step of the automatic defect image acquisition function executed by the device shown in
Here, the defect image acquisition function is a function for automatically acquiring images of a defect present on the sample, or images of the area in which a defect may occur by using the imaging unit 101. The coordinates of the defect to be imaged are input from the outside. More specifically, the coordinates of the defect are given by a defect inspection device for the purpose of obtaining the position of the defect on the sample, or by a lithography simulator or other means. The lithography simulator estimates the shape of the circuit pattern formed on the sample, and identifies the area in which a pattern different from the desired pattern may be formed.
In the flow shown in
The purpose for acquiring the reference image is, as described above, that the defect position is detected by comparing the wide field-of-view defect image with the reference image. The semiconductor circuit pattern includes a portion in which the same circuit pattern is repeatedly formed, for example, in a memory cell unit of a flash memory device. In the case of the repeated pattern, it is possible to synthesize the reference image from the defect image taken in the portion including the defect by using the repeatability of the circuit pattern, instead of using a method of taking the image in the normal portion of the semiconductor circuit pattern shifted by one chip with respect to the defect coordinates. In this way, the position of the defect can be detected by comparing the wide field-of-view defect image with the reference image synthesized from the particular defect image. Thus, if it is possible to determine in advance in any way that a defect occurs in the repeated pattern portion such as the memory cell portion, the reference image acquisition and the associated stage shift are not necessary. It is to be noted that in this embodiment, the two images of different field of views (wide and narrow field of views) are acquired. This is because, as described above, acquiring the narrow field-of-view image alone may not ensure that the defect is included in the particular field of view due to the defect coordinate error, and the stage shift error or other factors.
From the process sequence of the automatic image acquisition described above as well as the aforementioned imaging principle of the review SEM, the process parameters to be set for achieving the automatic defect image acquisition function include the following five items.
(1) Acceleration voltage
(2) Probe current
(3) Number of images to be averaged (for each of the wide and narrow field-of-view images)
(4) Image size (for each of the wide and narrow field-of-view images)
(5) Field of view size (for each of the wide and narrow field-of-view images)
These condition values must be stored in the recipe storage unit 104 as recipes prior to the automatic image acquisition.
In the automatic defect image acquisition function, the capture rate and the throughput are the important performance indexes. The capture rate means the accuracy of the process of detecting the defect position from the wide field-of-view defect image. If the portion other than the defect portion is falsely detected as a result of a failure of the defect detection, it is quite natural that the narrow field-of-view image of the portion is meaningless. For this reason, in general, defect detection rate of 95% or more is required.
Another important performance index in the automatic defect image acquisition is throughput. The throughput is the number of defects in image acquisition per unit time. In order to improve the throughput, it is necessary to reduce the process time in each of the steps shown in
Next, the recipe setting method according to the present invention will be described.
One specific implementation of this step is as follows. First, list data shown in
In
Next, with respect to the wide field-of-view defect images of the acquired evaluation image data, the position of the defect portion in each image is instructed on the input/output unit 103 (S2).
In the instruction area 602, the position of a defect portion 603 of the image displayed on the screen is registered by using the mouse cursor 605. More specifically, a defect definition area 604 is defined and registered by using the mouse cursor 605 on the screen. In this case, the defect definition area 604 means the center of the defect (indicated by the cross in the figure) and the range of the defect (indicated by the circle in the figure). The image selection in the thumbnail portion 601 and the instruction process in the instruction area 602 are repeated to instruct for the acquired image data. The instructed image data is stored in the evaluation image storage unit 105.
It is assumed that the number of defects to be registered in the evaluation image data is N with respect to one set of imaging conditions that is set by selecting setting values for each of the parameter items (1) to (5), and that the number of sets of imaging conditions is M. In this case, the number of defects to be instructed is N×M. If M is large, it is not realistic to register all defects through the screen. In such a case, instead of instructing for all of the image data acquired by the imaging unit 101, it is possible to apply instruction data registered for the image data acquired with a certain imaging condition, to the image data acquired with the other imaging conditions.
The specific procedure is as follows. First, an ID is given to each defect on the sample in advance. Then, a set of N defect images is acquired from the sample wafer with one of imaging condition sets. Next, the defect positions for the individual N defect images are registered by using the method shown in
As the two images of the same defect are acquired with different imaging conditions, it is natural that there is a difference in the image quality due to the difference in the imaging conditions. In general, there is also a slight field of view displacement due to the error in the stage stop accuracy or other factors. Thus, the field of view displacement between the instructed image and the non-instructed image of the same defect is detected by pattern matching. Then, the detected displacement is added to the defect position of the instructed defect image, in order to estimate the position of the defect on the non-instructed image. This estimation of the defect position is performed for M sets of images. As a result, the defect position can be located with respect to N×M defect image data by performing the instruction process only N times.
Next, one condition is selected from the imaging condition set (S3). The defect detection process is performed with respect to the image data set (N images) acquired with the selected condition (S4). This defect detection process is performed by a defect detection execution unit 108 of the recipe evaluation unit 106. The defect detection execution unit 108 stores a program for executing the defect detection process. Further, the defect detection execution unit 108 has a function of executing off-line the same process as the defect detection process (T5) executed in the process flow of
Next, using the result data of the defect detection process for the N images, the capture rate is calculated by a capture rate calculating unit 109 of the recipe evaluation unit 106 in the defect review tool shown in
In the calculation of the capture rate, the capture rate calculating unit 109 compares the detect detection position of the process result in the detect detection execution unit 108, with the instructed detect detection position. Then, the capture rate calculating unit 109 evaluates the difference between the two defect positions. For example, when the defect detection position is located within the circle 604 which is the area defined in the instruction, it is determined that the defect detection is successful or otherwise failed. In this way, success or failure is determined for all of the N images of the defect image data set with the same imaging condition. Then, the ratio of the number of successes is calculated as the capture rate.
On the other hand, in the calculation of the throughput, function data is used for the process time stored in the process time data storage unit 107.
For example, it is assumed that the currently selected imaging conditions are as follows:
In this case, T1 is 600 msec, T2 is 800 msec, T3 is 400 msec, T4 is 800 msec, T5 is 1000 msec, and T6 is 400 msec, respectively. Thus, the total process time is 4000 msec. This value is converted to a throughput of about 900. Here, the throughput is, for example, the number of defects that can be automatically observed for one hour.
The data of the individual process times shown in
Here, the throughput is estimated by the method of accumulating the individual process time data shown in
At last, the capture rate and throughput values calculated by the recipe evaluation unit 106 for different sets of imaging conditions are output to the input/output unit 103.
It is to be noted that in the above example, the condition setting registered in the recipe storage unit 104 is manually set on the display screen of the input/output unit 103. However, it is also possible to automatically determine the condition setting based on the criteria defined in advance by the recipe evaluation unit 106. For example, one of the criteria is “the capture rate of 95% or more for the imaging condition with the highest throughput performance”. In this case, as shown in
Next, a second embodiment of the defect observation method according to the present invention will be described.
In the first embodiment, the evaluation data used for evaluating the capture rate is acquired through the image acquisition in the imaging unit 101 under a plurality of sets of imaging conditions set in advance. In the description of the second embodiment, the review SEM according to the present invention has a function of generating evaluation data by simulation with the other imaging conditions using the acquired image data, instead of actually taking images to acquire all the evaluation image data as in the first embodiment.
First, the process stores list data in the standard image storage unit 1001. The list data includes a list of setting value candidates for each of the parameters set as the imaging conditions, and information on whether image acquisition by the imaging unit is necessary to acquire images by changing the setting values for each of the parameters.
Then, based on the parameter candidate list shown in
The process flow will be described with reference to the flow chart shown in
Here, the determination between the image acquisition and the image simulation (S1105) is made by the following steps.
Step 1: Acquire the type of parameter in which the setting value is different between the currently selected imaging condition and the reference imaging condition. The number of parameter types is not necessarily one and may be two or more.
Step 2: Acquire the content of “need or no need for image acquisition” corresponding to the parameter extracted in Step 1 from the table shown in
Step 3: Perform image simulation if there is no parameter with “NEED” for “need or no need for the image acquisition” as a result of Step 2, or otherwise perform image acquisition by the imaging unit.
In the case shown in
The procedure of the image simulation process (S1106) will be described below. First, the image simulation procedure for the averaging number is as follows. In this case, reducing the averaging number means reducing the S/N ratio. When the signal (S) is set constant, the noise (N) increases. In other words, it is possible to generate images by simulation with a smaller value for the averaging number, by performing a process of adding random noise to the acquired reference image.
There is a statistical relationship between the averaging number and S/N. That is, when the averaging number is doubled, S/N is improved √{square root over ( )}2 times. Thus, using this relationship, the S/N of desired simulation images is obtained based on the S/N value calculated from the reference image and on the number of the specified averaging number. Then, the amount of noise to be added by the simulator 1005 is controlled to actually acquire the desired images. Further, in the case of the simulation with respect to the image size, for example, images of 1024 and 724 pixels are generated by thinning the standard image (1448 pixels in image size). It is a matter of importance for the thinning process to keep the image S/N unchanged before and after thinning of the reference image.
On the other hand, when it is determined that it is necessary to take images by the imaging unit 101 for the image acquisition under the selected imaging condition, the image acquisition is actually performed with the particular imaging condition (S1107). Probably, the field of view of such reacquired images is different from the field of view of the defect images to which the instruction data have been set. Thus, similarly to the method described in the first embodiment, the displacement of the field of view is detected by pattern matching. Then, the defect position is identified on the acquired images by adding the value of the displacement to generate the instruction data.
In the case of the image simulation, displacement of the field of view does not occur. Thus, the instruction result for the reference image directly represents the instruction result of the image simulation results. In the process flow of
As described above, instead of acquiring all the image data sets with different imaging conditions by the imaging unit 101, the image simulation can also be performed depending on the types of parameters to be changed in order to improve the efficiency of the image set generation. In a low volume high mix production line, there are many types of recipes to be generated. Thus, the reduction in the image acquisition time for generating one recipe has great influence on such a production. As the most significant case, when only the conditions allowing image simulation, namely, only the parameters of addition number and image size are changed, there is no need to perform image acquisition. As a result, the evaluation image data can be generated off-line in a location separated from the imaging unit 101, for example, in another terminal connected to a network.
Next, a third embodiment of the defect observation method according to the present invention will be described. In the first and second embodiments described above, the instruction process only instructs the location of the defect position for the defect image data set for evaluation. The third embodiment is different from the first and second embodiments in that the other defect information is also set in addition to the defect position.
In the instruction process screen described above with reference to
By providing the defect information other than the defect position with respect to each defect, the operator can set imaging conditions suitable for the automatic defect image acquisition according to the attributes of the defect. For example, it is possible to classify the evaluation sample by the defect type and the defect size, and evaluate the relationship between the imaging conditions and the performance indexes (the capture rate and the throughput) for each classification result. For example, in the case of the defect size, it is possible that defects are divided into two classes based on a certain size (for example, 1 micrometer) to be able to set imaging conditions suitable for each defect class. For example, when the defect size is large, in general, the defect detection by imaging process is easy. Thus, with respect to the class of defects with a defect size larger than the standard value, there is a low risk that the defect detection rate will decrease, even with a high throughput condition such as small averaging number or small image size. On the other hand, the smaller the size of the defect the more difficult the defect detection becomes. Thus, with respect to the class of defects with a defect size less than the standard value, the imaging conditions should be selected taking into account that the capture rate is not likely to be reduced, namely, in which the throughput is sacrificed, so as to be suitable for the image acquisition of this class.
As described above, the capture rate and the throughput for each defect class are evaluated, and the results are displayed on the screen as shown in
The easiness of the defect detection is different depending on whether the defect size is large or small. When the defect size is large, 95% of the capture rate is achieved with three conditions (conditions 2, 3, 4) of the four conditions. In this case, the maximum throughput is 2250. On the other hand, when the defect size is small, only the condition 4 can be selected to achieve 95% of the capture rate. In this case, the throughput is as low as 780.
The defect inspection device has the function of identifying the position of a defect as well as outputting the approximate size of the defect. Thus, from this result, it is possible to set imaging conditions suitable for the automatic image acquisition according to the defect size of each defect. Some defect inspection devices have the function of outputting the defect size as well as the automatic defect classification result. In this case, the imaging conditions can be changed by not only the defect size but also the defect type information according to the classification result.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2008-255902 | Oct 2008 | JP | national |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/JP2009/067293 | 9/28/2009 | WO | 00 | 7/15/2011 |
