Measurement system, measurement method, and measurement recipe creation method

Abstract

The purpose of the present invention is to provide a measurement system with which causes of bonding failure can be easily analyzed for two wafers to be bonded. To this end, the present invention is a measurement system for measuring a semiconductor device, the measurement system comprising a pre-bonding measurement device that measures a first wafer and a second wafer before bonding, and when a measurement target of the first wafer is taken as a first measurement point, the pre-bonding measurement device sets a second measurement point that comes into contact with the first measurement point after bonding as the measurement target of the second wafer.

Description

Measurement system, measurement method, and measurement recipe creation method

[0001] The present invention relates to a measurement system, a measurement method, and a measurement recipe creation method.

[0002] In the manufacturing process of semiconductor devices, various inspections and measurements are performed as appropriate. Patent Document 1, for example, is known as a technology for creating a program to operate equipment used for inspection and measurement. Patent Document 1 discloses that a user can automatically create a sample program for inspection and measurement by simply selecting, using an input device such as a mouse, items on a computer, such as image data to be inspected and measured, an area to be inspected and measured, the content of the inspection and measurement, and the part to be inspected and measured.

[0003] Japanese Patent Application Laid-Open No. 2006-65582

[0004] In the technology described in Patent Document 1, inspection and measurement programs are created separately for different types of wafers. Therefore, for example, when inspecting and measuring the top and bottom wafers of a bonded wafer, the layout relationship is not taken into consideration, and it may not be possible to analyze the cause of a bond failure. An object of the present invention is to provide a measurement system that makes it easy to analyze the cause of a bond failure for two bonded wafers.

[0005] In order to solve the above-mentioned problems, the present invention provides a measurement system for measuring semiconductor devices, which includes a pre-bonding measurement device that measures a first wafer and a second wafer before they are bonded, and in which, when a first measurement point is set as a measurement target of the first wafer, the pre-bonding measurement device sets a second measurement point that will come into contact with the first measurement point after bonding as a measurement target of the second wafer.

[0006] According to the present invention, it is possible to provide a measurement system that can easily analyze the cause of bonding defects between two wafers to be bonded. Problems, configurations, and effects other than those described above will become clear from the following description of the embodiments.

[0007] 1. A diagram of a process flow that extracts the portion of the semiconductor device manufacturing flow that is related to wafer bonding, and an explanatory diagram that schematically shows the structure of the wafer at that time. 2. A system configuration diagram related to the semiconductor device manufacturing process. 3. A block diagram showing the configuration of an SEM-type measurement device, which is one of the pre-bonding measurement devices. 4. An explanatory diagram of the pattern layout of the bottom wafer and top wafer that are the measurement targets. 5. An operation flow diagram of the recipe creation program. 6. An operation flow diagram of the recipe conversion program. 7. An operation flow diagram when measurement is performed by the measurement program. 8. An operation flow diagram of the analysis program. 9. A diagram showing a specific example of matching. 10. An operation flow diagram of the analysis program. 11. An example of a GUI that is displayed when the analysis program is executed. 12. A schematic diagram of the target structure and SEM image, a cross-sectional diagram of the structure, and a cross-sectional profile of the SEM image. 13. A schematic diagram of calculation of recess feature quantities. 14. An operation flow diagram when the measurement program modifies the recipe according to the measurement results and performs measurement.

[0008] An embodiment of the present invention will be described with reference to the drawings.

[0009] <Explanation of Wafer Bonding and Measurement> FIG. 1 is an explanatory diagram that shows a process flow that extracts the part related to wafer bonding from the semiconductor device manufacturing flow, and a schematic diagram of the wafer structure at that time.

[0010] Before wafer bonding, patterns are formed on the bottom wafer 103 (first wafer) and the top wafer 102 (second wafer) through multiple processes 101. To manufacture multiple semiconductor devices from a single wafer, repeated patterns are formed on the wafer. Here, the repeated unit that will ultimately be used as a semiconductor device is called a chip, which refers to a section separated by vertical and horizontal lines on the bottom wafer 103 and the top wafer 102. The repeated pattern of a portion of the chip on the bottom wafer 103 is shown in structural cross-section 133, and the repeated pattern of a portion of the chip on the top wafer 102 is shown in structural cross-section 132.

[0011] After pattern formation, copper (Cu) plating 104, which will form the interconnects, is applied. The surface is polished by CMP (Chemical Mechanical Polishing) 105 and intentionally recessed to form Cu pads 106. This recession is called a recess 107. In the subsequent bonding process 108, the top wafer 102 is inverted around a Y-axis 110, which is determined by a notch 109 that defines the wafer's rotation direction within the plane and passes through the center of the wafer, and is brought into contact with the bottom wafer 103. At this time, heating is performed while applying pressure, causing the Cu pads 106 to expand and come into contact with each other, thereby bonding and ensuring electrical continuity, resulting in the bonded wafer 112. While the following description will use an example in which the Cu pads 106 are recessed, recession is not necessary. Cu pads of other shapes, such as protruding or flat, may also be bonded. The present invention is applicable to any Cu pad shape. Note that the repeated pattern is not exposed on the bonded wafer 112 because the substrates (Si, etc.) of the bottom wafer 103 and top wafer 102 are present on both the front and back sides. As shown in a cross-sectional view 141 of the bonded wafer 112 near a bonding surface 140, a repeated pattern is formed inside. A semiconductor device is completed by subjecting the bonded wafer 112 to a plurality of processes 111.

[0012] In the manufacturing process of semiconductor devices, inspection and measurement (hereinafter simply referred to as measurement) are performed appropriately at key points in the process. In Fig. 1, an example is shown in which pre-bonding measurement 121 is performed after CMP 105, and post-bonding measurement 122 is performed after the bonding process 108. Measurements may be performed at other times before and after bonding.

[0013] In pre-bonding measurement 121, the surface of the Cu pad, one of the measurement targets, is exposed, allowing direct measurement. Pre-bonding measurement methods include foreign matter inspection, length measurement using a scanning electron microscope (SEM), step (recess) measurement using an SEM, atomic force microscope (AFM) measurement, and optical shape measurement. Foreign matter inspection is performed for randomly occurring defects, while length and shape measurement are suitable for identifying systematic defects that occur repeatedly in repeating patterns. In post-bonding measurement 122, the bonding surface between the Cu pads, which are the measurement targets, is not exposed, so measurements must be made from the backside (non-patterned surface) of the bottom wafer 103 or top wafer 102. Post-bonding measurement methods include ultrasonic defect inspection, X-ray defect inspection, and infrared defect inspection. After polishing the backside of the bonded wafer 112 to expose the pattern, measurements similar to those before bonding and measurements using electrical characteristics are also possible.

[0014] Pre-bonding measurement 121 involves measuring, for example, the depth of the recess 107 in the Cu pad 106 (referred to as the recess depth). As mentioned above, the Cu pad 106 overheats and expands during the bonding process 108. If the recess depth is too large, the Cu pads on the bottom wafer 103 and top wafer 102 may not contact or the contact area may be insufficient, resulting in poor electrical continuity. This results in the formation of gaps called voids between the wafers. Post-bonding measurement 122, for example, can be performed using ultrasonic defect inspection to detect voids between the bottom wafer 103 and top wafer 102. Conversely, if the recess depth is too small, the expansion of the Cu pad 106 creates a repulsive force between the wafers, resulting in the formation of voids around the Cu pad 106. Larger voids can affect the insulation performance and the electrical continuity of the surrounding Cu pads. Thus, the recess depth is an important dimension that determines device continuity. For example, the SEM used in the pre-bonding measurement 121 has a resolution (0.1 nm or less) sufficiently higher than the Cu pad size (several hundred nm or more), making it possible to measure the recess depth of each pad. However, the ultrasonic testing used in the post-bonding measurement 122 has a resolution of at least 10 μm, making it difficult to determine the presence or absence of voids on a Cu pad-by-pad basis. Therefore, it is important to link the measurement results of the post-bonding measurement 122 and the pre-bonding measurement 121 to analyze the cause of voids. Therefore, when the measurement target of the bottom wafer 103 is the first measurement point, the second measurement point that will contact the first measurement point after bonding is set as the measurement target of the top wafer 102. This makes it possible to analyze (match) whether the recess depth of the bottom wafer 103 or the top wafer 102 is abnormal. Even when checking continuity in the post-bonding measurement 122, the post-bonding measurement 122 checks each chip individually or a partial area of ​​the chip, resulting in a different resolution than the pre-bonding measurement 121. Since not only the resolution but also the accuracy of position identification differs, it is necessary to take into account the resolution and accuracy of position identification when performing analysis (comparison). For example, if the accuracy and resolution of position identification in post-bonding measurement 122 are low, it is necessary to analyze (comparison) the results of pre-bonding measurement 121 more broadly. Understanding the recess amount within the wafer surface is useful for improving the CMP process.

[0015] Below, we will discuss measurements other than recess depth. Pre-bonding measurements 121 measure not only the recess but also the surface roughness of the Cu pad, its crystal orientation, and the size and type of foreign matter. Roughness and crystal orientation affect conductivity because they affect the amount of expansion and the shape after expansion. Foreign matter generates voids near the foreign matter during bonding, resulting in poor conductivity. The size and type of foreign matter affect the size of the voids, which in turn affects conductivity. Therefore, the criticality of foreign matter can be determined based on the size and type of foreign matter. These measurements are essential for stable mass production of semiconductor devices. Wafer bonding requires understanding the relationship between the wafers to be bonded. Therefore, in both measurements, measuring the pair of wafers that will be in contact during bonding is useful for process development and process management.

[0016] 2 is a system configuration diagram related to the manufacturing process of semiconductor devices, focusing in particular on the measurement system. An MES (Manufacturing Execution System) 200 issues execution instructions to a pre-bonding measuring device 201, a post-bonding measuring device 202, and a group of other manufacturing devices 203 including CMP devices and bonding devices. The pre-bonding measuring device 201 performs pre-bonding measurement 121, and the post-bonding measuring device 202 performs post-bonding measurement 122. The number of measuring devices is not limited to two, and there may be one or three or more.

[0017] The pre-bonding measuring device 201 and the post-bonding measuring device 202 read a recipe that specifies the measurement conditions for executing a measurement program from a DB (Data Base) 204 via a network, perform measurement, and save the measurement results in the DB 204 via the network.

[0018] An analysis server 206 (analysis device) that stores an analysis program 205 acquires data from the DB 204, performs analysis (for example, collating measurement results), and displays the results on a display unit 207. The display unit 207 is equipped with an input / output device (not shown), receives analysis conditions from the user, and sends them to the analysis server 206.

[0019] 3 is a block diagram showing the configuration of an SEM-type measuring device, which is one of the pre-bonding measuring devices. As shown in Fig. 3, the SEM-type measuring device 3001 includes an electron microscope 3100 (measuring unit), a control unit 3120, power supply units 3121 and 3122, and a computer 300.

[0020] In the electron microscope 3100, an electron beam 3103 (electron beam) is irradiated onto a sample. The electron microscope 3100 outputs a detection signal obtained based on the irradiation of the electron beam 3103. The SEM measuring device 3001 includes components necessary for forming a signal waveform and an image based on the detection signal from the electron microscope 3100. First, an example of the electron microscope 3100 will be described in detail with reference to FIG. 3 .

[0021] An electron beam 3103 extracted from an electron source 3101 by an extraction electrode 3102 is accelerated by an acceleration electrode (not shown). The accelerated electron beam 3103 is focused by a condenser lens 3104, which is a type of converging lens. The focused electron beam 3103 is scanned one-dimensionally or two-dimensionally over a sample 3108 by a scanning electrode 3105. The electron beam 3103 is decelerated by a negative voltage applied to an electrode built into a sample stage 3109, and is focused by the lens action of an objective lens 3106, and is irradiated onto the sample 3108.

[0022] When the electron beam 3103 is irradiated onto the sample 3108, electrons 3110 such as secondary electrons and backscattered electrons are emitted from the irradiated location (the irradiation location) and / or from locations inside the sample 3108 that are scattered from the irradiation location and are different from the irradiation location. The emitted electrons 3110 are accelerated toward the electron source 3101 by an acceleration action based on a negative voltage applied to the sample 3108, collide with the conversion electrode 3112, and generate secondary electrons 3111. The secondary electrons 3111 emitted from the conversion electrode 3112 are captured by detectors 3113 and 3114, and the detection signals output by the detectors 3113 and 3114 change depending on the amount of captured secondary electrons 3111. Note that a backscattered electron detector (not shown) may also be provided. The detector arrangement and configuration are merely examples, and multiple detectors may also be provided. When multiple detectors are provided, multiple images can be obtained by a single electron beam scan.

[0023] The detection signal output from the detector 3113 is supplied to the computer 300 by the control unit 3120. The computer 300 is equipped with a display unit (not shown). The brightness of the image displayed on this display unit changes according to the detection signal. That is, the amount of electrons captured by the detector 3113 (electron amount) is displayed as brightness on the display unit.

[0024] For example, when displaying a two-dimensional image on the display unit, the deflection signal supplied to the scanning electrode 3105 and the detection signal output from the detector 3113 are synchronized, so that the brightness of the image in the scanning area scanned by the deflection signal is displayed on the display unit.

[0025] The electron microscope 3100 shown in FIG. 3 is also provided with a deflector (not shown) that moves the scanning area of ​​the electron beam 3103. This deflector is used to display images of patterns of the same shape that exist at different positions on a display unit. This deflector is also called an image shift deflector, and can move the field of view position of the electron microscope 3100 without moving the sample 3108 using a sample stage (e.g., sample table 3109) that moves the sample 3108. The image shift deflector and the scanning electrode 3105 may be a common deflector, and an image shift signal and a deflection signal may be superimposed and supplied to the deflector.

[0026] Detection signals (image, brightness profile, brightness, etc.) from the electron microscope 3100 are supplied to the computer 300 via the control unit 3120. The computer 300 calculates values ​​related to changes in the shape of the target shape of the observation target based on the supplied detection signals, and outputs one or more of these calculated values. Note that the computer 300 may be integrated with the electron microscope 3100.

[0027] The control unit 3120 controls the power supply units 3121 and 3122 in accordance with instructions from the computer 300. By controlling the power supply unit 3122, the voltage applied to the extraction electrode 3102 and the acceleration electrode (not shown) changes. Similarly, by controlling the power supply unit 3121, the voltage applied to the sample 3108 changes. Furthermore, the control unit 3120 controls the deflection signal supplied to the scanning electrode 3105 and the signal supplied to the objective lens 3106 in accordance with instructions from the computer 300. Furthermore, the control unit 3120 supplies the detection signal output from the detector 3113 to the computer 300, as described above.

[0028] The computer 300 has an input / output device 301 for receiving information from a user and displaying results, a processor 302 for executing programs, and a memory area 303. The memory 303 stores a recipe creation program 311, a recipe conversion program 312, a measurement program 313, a recipe 314, and measurement results 315 accompanied by a code (ID) for identifying the measurement target.

[0029] <Explanation of Measurement Object> FIG. 4 is an explanatory diagram of the pattern layout of the bottom wafer and top wafer, which are the measurement objects.

[0030] The enlarged chip view 401 shows an enlarged view of one chip on the top wafer 102. The enlarged chip view 401 includes a pattern 402, which is a collection of Cu pads that will be conductive after wafer bonding. The enlarged layout view 403, which is a further enlargement of the enlarged chip view 401, shows the layout. The wafer includes a Cu pad 404, which is part of the pattern 402, a collection of Cu pads that will be conductive after wafer bonding and that are related to device performance, and a dummy pattern 405 that will not be conductive after wafer bonding and is not related to device performance. The chip also includes an alignment mark 407, which is used to align the wafer when it is mounted on the equipment and to determine the position of the measurement point within the chip. A coordinate system 408 is defined, with a characteristic point on the alignment mark 407 as the origin and X and Y axes perpendicular to the Y axis 110 of the chip.

[0031] The enlarged chip view 411 shows an enlarged view of one chip on the bottom wafer 103. The enlarged chip view 411 includes a pattern 412, which is a collection of Cu pads that will be conductive after wafer bonding. The enlarged layout view 403, which is a further enlargement of the enlarged chip view 411, shows the layout. The wafer includes a Cu pad 414, which is part of the pattern 412, a collection of Cu pads that will be conductive after wafer bonding and that are related to device performance, and a dummy pattern 415 that will not be conductive after wafer bonding and is not related to device performance. The chip also includes an alignment mark 417, which is used to align the wafer when it is mounted on the equipment and to determine the position of the measurement point within the chip. A coordinate system 418 is defined, with a characteristic point on the alignment mark 417 as the origin and X and Y axes perpendicular to the Y axis 110 of the chip.

[0032] The axis that is parallel to the Y axis 110 of the chip and passes through the center of the chip is called the inversion axis 409 of the top wafer and the inversion axis 419 of the bottom wafer. During wafer bonding, the top wafer 102 is inverted about the Y axis 110 of the chip and then bonded, so the pattern 402 of the top wafer 102 and the pattern 412 of the bottom wafer 103 are in an inverted relationship about the inversion axis 409 or the inversion axis 419. In other words, as shown in the enlarged layout views 403 and 413, the arrangement pitch of the Cu pads 404 and 414 is the same. However, the shapes of the Cu pads 404 and 414 are not necessarily the same.

[0033] The measurement points in this specification will be described. A measurement point is information indicating the position and range of the measurement target, such as the XY number of the chip to be measured on the wafer surface, or the center coordinates of the measurement and the measurement range relative to that center. In FIG. 4 , the measurement points on the top wafer 102 are indicated by measurement points 421 and 422 (black areas), and the measurement points on the bottom wafer 103 are indicated by measurement points 431 and 432 (black areas). Therefore, measurement points 421 and 431 are inverted with respect to the Y axis 110 of the chip, and measurement points 422 and 432 are inverted with respect to the inversion axis 409 or 419. In other words, when considered in terms of X coordinates, the relationship is such that X is converted to -X. This makes it possible to measure pairs of Cu pads that come into contact with each other during bonding process 108. As a result, when an abnormality such as a defect is detected during post-bonding measurement 122, it becomes possible to analyze the relationship between the top wafer 102 and the bottom wafer 103 and determine whether the cause is in the top wafer 102 or the bottom wafer 103. In this specification, analyzing the measurement results in consideration of the positional relationship of the measurement points is called matching.

[0034] If the positional relationship between the Y axis and the inversion axis 409 of the coordinate system 408 when the pre-bonding measuring device 201 measures the top wafer 102 differs from the positional relationship between the Y axis and the inversion axis 419 of the coordinate system 418 when the pre-bonding measuring device 201 measures the bottom wafer 103, it is necessary to perform the conversion taking into account the difference in positional relationship.

[0035] <Explanation of Recipe Creation and Recipe Conversion> Creation and conversion of a recipe including measurement point information will be explained.

[0036] FIG. 5 is an example of an operational flow diagram of the recipe creation program 311. The recipe creation program 311 first receives layout data 501, such as wafer design data and image data, and measurement condition information 502, such as measurement point coordinates and imaging conditions entered by the user. The recipe creation program 311 then registers alignment marks using the layout data 501 (step S510). Next, the recipe creation program 311 sets imaging conditions (FoV (Field of View), number of pixels, acceleration, current, number of integrations, and scan method) using the measurement condition information 502 (step S511). The recipe creation program 311 also sets imaging areas (start coordinates, number of XY steps, and XY movement distance) using the layout data 501 and the measurement condition information 502 (step S512). The recipe creation program 311 then sets evaluation conditions (e.g., recess evaluation conditions as shown in FIG. 14 , described later) using the measurement condition information 502 (step S513). Thereafter, the recipe creation program 311 registers the chip to be imaged (step S514), and stores the recipe 314, which describes the settings, in the DB 204.

[0037] Hereinafter, the recipe for measuring the bottom wafer 103 (first wafer) will be referred to as the first recipe, and the recipe for measuring the top wafer 102 will be referred to as the second recipe. It is possible to create the first and second recipes separately using the recipe creation program 311 described in Fig. 5 , but inputting measurement condition information 502 so as to achieve an inverted relationship of measurement points while taking into account differences in the layouts of the bottom wafer 103 and top wafer 102 is time-consuming and prone to errors. Therefore, by using the recipe conversion program 312, measurements can be performed efficiently and reliably while maintaining the inverted relationship of measurement points.

[0038] 6 is an example of an operation flow diagram of the recipe conversion program 312. Here, an example will be described in which a first recipe for the bottom wafer 103 is used as a base, and measurement point information of the first recipe is inverted to convert the first recipe into a second recipe for the top wafer 102. Note that the imaging condition settings in the first recipe are used as they are to create the second recipe.

[0039] The recipe conversion program 312 first reads the first recipe 504 for measuring the bottom wafer 103 from the DB 204 and acquires first measurement point information for the measurement target on the bottom wafer 103 (step S602). The first measurement point information is, for example, a chip position [x, y], which is coordinate information on the coordinate system 408, or a chip position [ChipX, ChipY], which indicates the chip number, with the center wafer on the bottom wafer 103 being [0, 0]. Note that the first measurement point information is not limited to information given as coordinate information, but may also be information given by other methods, such as vectors. Furthermore, the chip does not necessarily exist at the center of the wafer; for example, the center of the wafer may be the chip parting line. Meanwhile, the recipe conversion program 312 acquires coordinate system information 601 from the layout data 501 (step S601). Then, the recipe conversion program 312 performs coordinate conversion (inversion) of the intra-chip position 603 and the chip position 604 of the bottom wafer 103 acquired in step S602 using the coordinate system information 601 (step S605). Furthermore, the recipe conversion program 312 reflects the intra-chip position 606 and the chip position 607 acquired by the coordinate conversion in the second recipe 609 as second measurement point information to be measured on the top wafer 102 (step S608). Thereafter, the recipe conversion program 312 stores the acquired second recipe 609 in the DB 204 (step S610).

[0040] In the example of FIG. 6 , an example has been described in which the already created first recipe 504 is converted into the second recipe 609 in order to newly create the second recipe 609. However, the reverse is also possible. Furthermore, if not only the first recipe 504 but also the second recipe 609 already exists, it is also possible to reflect only the changed portion, i.e., the second measurement point information obtained by inverting the first measurement point information of the first recipe 504, in the second recipe 609. Furthermore, the recipe creation program 311 and the recipe conversion program 312 may be integrated, and the coordinate conversion 605 may be performed when creating the first recipe 504 without using the DB 204, and the second recipe 609 may be created in parallel with the first recipe 504. Furthermore, it is also possible to create a third recipe for measuring the bonded wafer 112 using the measurement point information of the first recipe.

[0041] 7 is an example of an operation flow diagram when measurement is performed using the measurement program 313. Here, the measurement program 313 for the bottom wafer 103 before bonding is a first measurement program 710, the measurement program 313 for the top wafer 102 before bonding is a second measurement program 720, and the measurement program 313 for the bonded wafer 112 is a third measurement program 730. In addition, the following description will be given taking as an example a case where measurement results obtained using the first to third measurement programs are stored in DB 204, but measurements in other processes can also be stored in DB 204, in which case analysis across multiple processes becomes possible.

[0042] When the MES 200 receives a measurement wafer ID 799 input by the user, it identifies the ID of the wafer to be bonded to it, and sends the user-input wafer ID and the wafer ID of the wafer to be bonded to the first measurement program 710 and the second measurement program 720. This makes it possible to reliably associate the wafer to be bonded and measure it.

[0043] The first measurement program 710 starts operation upon receiving the identification code (ID) of the target wafer and a measurement start command from the MES 200. Next, the first measurement program 710 reads the first recipe 504 from the DB 204 (step S711) and measures the bottom wafer 103 by executing the first recipe 504 (step S712). Thereafter, the first measurement program 710 saves the first measurement result 714 with the ID in the DB 204 (step S713).

[0044] The second measurement program 720 receives the target wafer identification code (ID) and a measurement start command from the MES 200 and begins operation. Next, the second measurement program 720 reads the second recipe 609 from the DB 204 (step S721) and executes the second recipe 609 (step S722) to measure the top wafer 102. At this time, the second measurement program 720 performs coordinate conversion on the measurement results by performing the inverse of the coordinate conversion performed in step S605 in the recipe conversion program 312 (step S725). This coordinate conversion also uses the coordinate system information 601 acquired from the layout data 501. This coordinate conversion of the results (step S725) can be skipped, but this facilitates data collation and makes it easier to visually compare the measurement results of the top and bottom wafers when displaying the data. Naturally, this does not need to be performed in the second measurement program 720; it may be executed after saving the data in the DB 204 or during collation. Thereafter, the second measurement program 720 stores the second measurement result 724 with the ID in the DB 204 (step S723).

[0045] The third measurement program 730 starts operation upon receiving the identification code (ID) of the target wafer and a measurement start command from the MES 200. Next, the third measurement program 730 reads a third recipe 735 from the DB 204 (step S731) and measures the top wafer 102 by executing the third recipe 735 (step S732). Thereafter, the third measurement program 730 saves the third measurement result 734 with the ID attached in the DB 204 (step S734).

[0046] <Explanation of Data Analysis (Matching) and Its Use> FIG. 8 is an example of an operational flow diagram of the analysis program 205. The analysis program 205 receives analysis conditions 801, for example, via user input, and queries the DB 204 for target data (step S802). Next, the analysis program 205 receives the first measurement result 714, second measurement result 724, and third measurement result 734 corresponding to the bottom wafer 103, top wafer 102, and bonded wafer 112, respectively, obtained by the query, and performs matching (step S803). The analysis program 205 then displays and saves the matching results (step S804). At this time, because wafer IDs are assigned to the measurement results, it is possible to query data on the top and bottom wafers and bonded wafers of an actual bonded pair from among the many wafer data. Note that multiple measurement results may be queried for a single wafer, and it is also possible to match data on multiple pairs and bonded wafers.

[0047] 9 to 11 are diagrams showing specific examples of the matching in step S803 of FIG.

[0048] In Figure 9, the recess, which is the amount of recession of the Cu pad, is extracted from the first measurement result 714 of the bottom wafer 103 and the second measurement result 724 of the top wafer 102, and these are used as the horizontal and vertical axes of the comparison graph, respectively. In the comparison graph of Figure 9, the open plot 901 indicates a Cu pad determined to be free of defects based on the third measurement result 734, and the closed plot 902 indicates a Cu pad determined to be defective. In this way, a recess control range 903 that is permissible to prevent defects can be determined from the relationship between the recess amount of the top wafer 102 before bonding, the recess amount of the bottom wafer 103 before bonding, and the presence or absence of defects in the bonded wafer 112. This can be used for process control during mass production, etc.

[0049] Note that the resolution of the devices is not necessarily the same, so when comparing, it is necessary to take into consideration the resolution and positioning accuracy. For example, when labeling the presence or absence of a defect (901 and 902) in the measurement results of the recess depth of each pad using an ultrasonic inspection device, the presence or absence of a defect may be expressed as a probability rather than as a binary value (901 and 902). Furthermore, in accordance with these labels of the presence or absence of a defect (901 and 902), the pads on the images of the first measurement result and the second measurement result may be labeled with the presence or absence of a defect and displayed (not shown).

[0050] FIG. 10 is an example of a comparison graph for a Cu pad with a different size and layout from the comparison graph in FIG. 9 . Even if the wafer IDs analyzed in the comparison graphs in FIG. 9 and FIG. 10 are the same, the distribution of defects caused by recess amounts and bonding is not necessarily the same, depending on the size and layout of the Cu pad. In the comparison graph in FIG. 10 , a white plot 1001 indicates a Cu pad determined to be free of defects from the third measurement result 734, and a black plot 1002 indicates a Cu pad determined to be defective. The recess management range 1003 for the Cu pad in FIG. 10 is in a different region from the recess management range 903 in FIG. 9 .

[0051] For example, recess control ranges 903 and 1003 are determined during process development, and if pre-bonding measurement 121 after CMP 105 during mass production confirms that the recess amount of the Cu pad is outside the recess control range 903, an alarm can be issued. Naturally, as data is accumulated during mass production, the recess control range can be reviewed each time. If the recess amount fluctuates only within a range that is sufficiently smaller than the recess control range, it is also possible to improve throughput by reducing the number of measurement points of the process monitor during mass production.

[0052] Furthermore, during process development, adjusting the CMP 105 process parameters can also be used to understand the relationship between the CMP 105 parameters and the presence or absence of defects in the bonded wafer 112. FIG. 11 plots the relationship between the CMP 105 process parameters and the presence or absence of defects in the bonded wafer 112. The horizontal and vertical axes of the graph in FIG. 11 represent CMP 105 parameters 1 and 2. The open plots 1101 and 1102 represent Cu pads determined to be defect-free based on the third measurement result 734, while the closed plots 1103 and 1104 represent Cu pads determined to be defective. The shape of the plots also indicates the type of pad. Even during comparisons such as this, the first measurement result 714 of the bottom wafer 103 and the second measurement result 724 of the top wafer 102 can be associated with each defect-presence plot, although this is not shown. In other words, the CMP 105 conditions, the shape and surface condition of the Cu pads before bonding, and the presence or absence of defects can be correlated for analysis, thereby shortening the process development time. Also, from the graph of FIG. 11, the allowable fluctuation range 1105 of the process parameter can be determined.

[0053] If the phases leading up to stable mass production of semiconductor devices are divided into three phases: process development, ramp-up, and stable mass production, we have previously described examples of using the comparison results during process development and stable mass production. However, they can also be used during ramp-up. For example, during process development, only a small number of tools for CMP 105 and wafer bonding are used, but during ramp-up, multiple tools may be installed in parallel and switched for each lot. In such cases, recess control ranges 903 and 1003 can be used to determine the recess control range for each combination of tools. By utilizing the comparison results in adjusting the tools, the ramp-up period can be shortened and stable mass production can be achieved more quickly. Furthermore, the comparison results can be used to derive the optimal combination of multiple tools.

[0054] Note that the resolution in the X and Y directions may not be consistent across all measurement results. Therefore, not all data necessarily corresponds one-to-one. By including the type and number of the measurement device actually used, as well as the imaging and evaluation conditions, in the measurement results, comparisons that take resolution and accuracy into account become possible. Furthermore, by combining multiple measurement devices, it becomes possible to analyze the correlation between the presence or absence of defects after bonding and the shape before bonding. By combining different devices, it becomes possible to combine and evaluate information that can only be obtained with each device. By comparing not only pre-bonding inspection and measurements but also post-bonding inspection and measurements, it is possible to better understand phenomena that cannot be understood by each measurement device alone, which is useful for improving the CMP process 105.

[0055] Figure 12 shows an example of data table items. The recipe includes the target device name, process name, alignment mark information, chip layout information, imaging conditions, measurement points, evaluation conditions, and information about the chip to be imaged. In addition to similar information copied from the recipe, the inspection results include the actual wafer ID measured, evaluation results, acquired images, and the type and number of the measurement equipment used. This data enables the aforementioned matching. Note that measurement results can take various forms in data tables, such as statistical quantities such as wafer-level averages, medians, and standard deviations, statistical quantities per chip, statistical quantities per pad, and statistical quantities within specific coordinates and ranges. The appropriate data table is referenced for matching.

[0056] FIG. 13 shows an example of a GUI (Graphical User Interface) displayed on the display unit 207 when the analysis program 205 is executed. The GUI 1300 includes an analysis condition input section 1310, a data display section 1320, an analysis result section 1330, and an execution button group 1340. The analysis conditions include, for example, a device name, a wafer ID, an analysis chip, analysis target data, a collation item, output data, and a data output path. When the image display button 1341 is operated, the analysis target data is displayed in the data display section 1320. The wafer map 1321 is a distribution of recess depths obtained by analyzing the second measurement result 724 of the top wafer 102. Because the bottom wafer 103 is measured at a position inverted with respect to the axis 1344, the wafer map 1322, which is a distribution of recess depths obtained by analyzing the first measurement result 714 of the bottom wafer 103, measures positions symmetrical with respect to the axis 1344. The display also displays a wafer map 1323 showing the presence or absence of defects obtained by analyzing the third measurement result 734 of the bonded wafer 112. When a chip is selected as indicated by arrow 1324, an arbitrary image 1326 of the chip on the bottom wafer 103, an arbitrary image 1325 of the chip on the top wafer 102, and an arbitrary image 1327 of the bonded wafer are displayed. Operating the next image display button 1329 displays the next image of the chip. Furthermore, if coordinate transformation of the results of the top wafer 102 (see step S725 in FIG. 7 ) is performed, pressing the inversion / non-inversion switch button 1328 displays the wafer map 1321 and image 1325 of the top wafer 102 inverted relative to axes 1344 and 1345. This makes it easier to recognize pairs of bonded chips and bonding pads. The analysis result section 1330 displays the analysis results. When the analysis button 1342 is operated, the analysis result section 1330 displays the recess amounts of data A and data B as an analysis graph 1331 by changing the color of the plot depending on the presence or absence of defects based on the results of data C. Furthermore, the analysis result section 1330 displays a recess management range 1332 (analysis result) calculated from data A to data C. Thereafter, when the save button 1343 is operated, the analysis graph 1331 and the recess management range 1332 are saved in the data output path.

[0057] <Specific Example of Measurement> As a specific example of pre-bonding measurement 121, recess measurement of a Cu pad using an SEM will be described. Here, an SEM device having four detectors (N detector, E detector, S detector, and W detector) at the same height and 90-degree azimuth intervals is used. The N detector and S detector are positioned opposite each other, and the E detector and W detector are positioned opposite each other.

[0058] FIG. 14 is a schematic diagram of the target structure, an SEM image, a cross-sectional view of the structure, and a cross-sectional profile of the SEM image.

[0059] The schematic top view 1400 shows the area around the Cu pad as viewed from above, and the schematic cross-sectional view 1410 shows the cross-sectional view taken along line 1404. The area around the Cu pad consists of an insulating portion 1401, a Cu pad 1402, and a barrier metal 1403 between the Cu pad and the insulating portion. As can be seen from the schematic cross-sectional view 1410, the insulating portion 1401 is generally flat, the Cu pad 1402 is generally flat but has irregularities on its surface, and the barrier metal 1403 has a slope. In this case, the difference in height between the insulating portion 1401 and the Cu pad 1402 is the recess depth to be measured. Also, as shown in the schematic cross-sectional view 1410, an E detector 1414 and a W detector 1415 are positioned opposite each other. Although not shown, an N detector and an S detector are also positioned opposite each other in the foreground and background directions of FIG. 14 .

[0060] In the E image 1420 detected by the E detector 1414, when the normal to the barrier metal 1403 faces the E detector, the signal level is higher than in the surrounding area and the area appears white (bright area 1421), and conversely, when the normal faces the opposite direction from the E detector, the signal level is lower and the area appears dark (dark area 1422). In this case, when the cross-sectional profile 1430 along the line 1424 is viewed, a bright peak 1431 and a dark peak 1432 are observed.

[0061] In the W image 1440 detected by the W detector 1415, when the normal to the barrier metal 1403 faces the W detector, the signal level is higher than in the surrounding area and the area appears white (bright area 1442), and conversely, when the normal faces the opposite direction from the W detector, the signal level is lower and the area appears dark (dark area 1441). In this case, when the cross-sectional profile 1450 along the line 1444 is examined, a bright peak 1452 and a dark peak 1451 are observed.

[0062] The EW image 1460 is obtained by taking the difference between the E and W images and normalizing the difference so that the minimum and maximum range is between 0 and 255. The EW image 1460 changes brightness depending on the direction of the slope of the barrier metal 1403. Bright regions 1461 are regions where the normal to the barrier metal 1403 faces the E detector 1414, and dark regions 1462 are regions where the normal to the barrier metal 1403 faces the W detector 1415. The intermediate color region 1463 represents a region where there is no difference between the detection signals of the E detector 1414 and the W detector 1415, i.e., a region facing the vertical information of the cross-sectional schematic diagram 1410. Typically, the image does not have three levels of brightness, darkness, and neutral, but rather has a continuous value including values ​​between them, as shown in the cross-sectional profile 1470 of the line 1464.

[0063] FIG. 15 shows a schematic diagram of recess feature calculation. An EW feature extraction schematic diagram 1500 is obtained by superimposing a bright area extraction region 1501 and a dark area extraction region 1502 on the EW image 1460 in FIG. 14 . A calculated value (e.g., difference) between a statistic (e.g., average) of the bright area extraction region 1501 and a statistic (e.g., average) of the dark area extraction region 1502 is used as an EW recess index for determining the size relationship of the recesses. An NS feature extraction schematic diagram 1510 is obtained by superimposing a bright area extraction region 1514 and a dark area extraction region 1513 on a dark area 1511 and a bright area 1512 included in an NS image, which is a differential image between the N detector and the S detector. A calculated value (e.g., difference) between a statistic (e.g., average) of the bright area extraction region 1514 and a statistic (e.g., average) of the dark area extraction region 1513 is used as an NS recess index for determining the size relationship of the recesses. By using the calculated values ​​(e.g., average) of the WE recess index and the NS recess index as the recess index, one recess depth can be calculated for each Cu pad. In this way, SEM can calculate the recess depth for each Cu pad. It is also possible to obtain data with different resolutions, such as the average recess depth of Cu pads in the peripheral area, the average for each pad type, the average for pads within a chip, the average for pads across the entire wafer, and the distribution (trend) of the average recess depth within a chip. This allows for matching to the resolution of the post-bonding measurement 122.

[0064] 14 and 15 have been described using an example in which four detectors are arranged perpendicularly and parallel to the boundary line of the Cu pad. However, if the effective arrangement of the detectors in the four directions is not perpendicularly and parallel, for example, bright regions 1421 and dark regions 1422 will also occur at the upper and lower boundaries of the Cu pad. The same applies when the Cu pad 1602 is circular rather than square. In such cases, this can be addressed by modifying the bright area extraction region 1501, the dark area extraction region 1502, the bright area extraction region 1514, and the dark area extraction region 1513. Furthermore, these extraction regions do not need to encompass the entire bright and dark areas; they may simply include a portion of them.

[0065] In this embodiment, an example has been described in which the average brightness is calculated as the feature amount. However, a value obtained by integrating the difference images (EW and NS images) of the detectors facing each other from a certain reference point toward the measurement point where the height is to be measured may also be used.

[0066] <Explanation of Modifications> In the above-described first embodiment, the bottom wafer 103, the top wafer 102, and the bonded wafer 112 were each measured once. However, the number of measurements for each wafer is arbitrary, and the measurement results can be analyzed (collated) multiple times. Multiple analyses (collations) facilitate, for example, the investigation of the cause of void defects and the improvement of the CMP process. Naturally, the same wafer can be measured using different instruments, and instruments with different measurement accuracies, measurement speeds, and resolutions can be combined. For example, a wide range of recess depth distribution can be obtained using optical measurement, which allows low-resolution, high-speed measurement, and then the local pad recess depth can be evaluated using a high-resolution SEM or AFM. This can be used to analyze the cause of void defects and the characteristics of CMP.

[0067] By evaluating the shape of each pad and comparing the average and standard deviation with the electrical characteristics, it is possible to derive the relationship between the electrical characteristics and shape. It is significant to compare statistical quantities such as the average and standard deviation when combining devices with different resolutions.

[0068] Although the example of wafer bonding has been described, this embodiment can also be used in a process called D2W (Dai to wafer), in which Dai separated from a wafer is bonded to another wafer. Although not shown, it is also possible to dynamically change the pair of wafers to be bonded by comparing the first measurement result 714 of the bottom wafer 103 before bonding with the second measurement result 724 of the top wafer 102 before bonding to determine a pair of wafers with fewer bonding defects. In the case of D2W, replacement on a Dai-by-Dai basis is also possible.

[0069] In the second embodiment, the recipe is modified in accordance with the measurement results, thereby enabling a reduction in measurement time and detailed analysis.

[0070] 16 shows an example of an operation flow when a measurement program modifies a recipe in accordance with measurement results and executes measurement. For example, the second measurement program 720 modifies the second recipe 609 read from the DB 204 based on the first measurement result 714 (step S1601), and executes measurement based on the modified second recipe. The third measurement program 730 modifies the third recipe 735 read from the DB 204 based on the first measurement result 714 and the second measurement result 724 (step S1602), and executes measurement based on the modified third recipe.

[0071] For example, if a foreign particle is found in the first measurement result 714 of the bottom wafer 103 during defect inspection during mass production, the recipe may be modified in step S1601 or step S1602 to skip measurement of the corresponding measurement points on the top wafer 102 or the bonded wafer 112. In this case, the measurement time can be shortened by not measuring locations where random defects due to foreign particles occur.

[0072] For example, during process development, if the bottom wafer 103 is measured at many measurement points and a location where a shape error has occurred is identified, the second recipe or the third recipe may be modified so that the corresponding location is measured intensively on the top wafer 102 or the bonded wafer 112. In this case, the shape defect can be correlated with the presence or absence of a defect, enabling process development in a short time.

[0073] 102...top wafer, 103...bottom wafer, 106, 404, 414, 1402...Cu pad, 110...Y axis, 112...bonded wafer, 107...recess, 207...display section, 401, 411...enlarged chip view, 402, 412...pattern, 403, 413...enlarged layout view, 405, 415...dummy pattern, 407, 417...alignment mark, 408, 418...coordinate system, 409, 419...inversion axis, 421, 422, 431, 432...measurement point, 1300...GUI, 1400...top view, 1410...cross-sectional view, 1401...insulating section, 1403...barrier metal, 1404, 1424, 1444...lines, 1414...E detector, 1415...W detector, 1420...E image, 1421, 1442, 1461, 1512...bright regions, 1422, 1441, 1462, 1511...dark regions, 1430, 1450, 1470...cross-sectional profiles, 1431, 1452...bright peaks, 1432, 1451...dark peaks, 1440...W image, 1460...EW image, 1500...EW feature extraction schematic diagram, 1501, 1514...bright portion extracted regions, 1502, 1513...dark portion extracted regions, 1510...NS feature extraction schematic diagram, 3001...SEM-type measuring device, 3100...electron microscope, 3120...control unit

Claims

1. A measurement system for measuring semiconductor devices, comprising a pre-bonding measurement device that measures a first wafer and a second wafer before bonding, wherein when a first measurement point is set as a measurement target of the first wafer, the pre-bonding measurement device measures a second measurement point that will come into contact with the first measurement point after bonding as a measurement target of the second wafer.

2. The measurement system according to claim 1, wherein the pre-bonding measurement device measures the first measurement point on the first wafer and then measures the second measurement point on the second wafer.

3. The measurement system according to claim 1, further comprising: a post-bonding measurement device that measures the first wafer and the second wafer after bonding; and an analysis device that collates the post-bonding measurement results with the measurement results at the first measurement point and the measurement results at the second measurement point.

4. The measurement system according to claim 3, wherein the measurement results collated by the analysis device are either the pad measurement values ​​themselves, the average of the measurement values ​​of some or all of the pads within a chip, or the trend of the pad measurement values ​​within a wafer.

5. The measurement system according to claim 3, wherein the analytical device determines, through comparison, a control range that is acceptable for the measurement result of the first measurement point and a control range that is acceptable for the measurement result of the second measurement point.

6. The measurement system of claim 3, wherein the analysis device displays the measurement results of the first measurement point or an image corresponding to the measurement results, or both, or the measurement results of the second measurement point or an image corresponding to the measurement results, in an inverted manner.

7. The measurement system according to claim 3, wherein the resolution of the post-bonding measurement device is lower than the resolution of the pre-bonding measurement device.

8. The measurement system according to claim 7, wherein the post-bonding measurement device is a device using SEM or AFM, and the pre-bonding measurement device is a device using any of ultrasonic waves, X-rays, and infrared rays.

9. The measurement system according to claim 1, wherein the pre-bonding measurement device comprises a computer that creates a recipe that defines conditions for measurement, and a measurement unit that performs measurement based on the recipe, wherein the computer: reads a first recipe for the first wafer and acquires first measurement point information that is to be measured on the first wafer; reads layout data for the first wafer and / or the second wafer and acquires coordinate system information from the layout data; inverts the first measurement point information using the coordinate system information; and creates a second recipe for the second wafer based on the inverted first measurement point information.

10. The measurement system according to claim 1, wherein the pre-bonding measurement device comprises a computer that creates a recipe that defines the conditions for measurement, and a measurement unit that performs measurement based on the recipe, and the computer creates a first recipe that includes first measurement point information that is the measurement target of the first wafer, based on layout data of the first wafer and measurement conditions for the first wafer input by a user, inverts the first measurement point information using coordinate system information acquired from the layout data, and creates a second recipe that includes second measurement point information that is the measurement target of the second wafer, based on the inverted first measurement point information.

11. The measurement system according to claim 1, wherein the pre-bonding measurement device comprises a computer that creates a recipe that defines the conditions for measurement, and a measurement unit that performs measurement based on the recipe, wherein the computer creates a first recipe that includes information on first measurement points that are to be measured on the first wafer, and creates a second recipe that includes information on second measurement points that are to be measured on the second wafer, and wherein the measurement unit performs measurement of the first wafer based on the first recipe, and then performs measurement of the second wafer based on the second recipe that has been modified using the measurement results of the first measurement points.

12. A measurement method for measuring a first wafer and a second wafer before bonding, comprising: a step of measuring the first wafer using a first measurement point as the measurement target; and a step of measuring the second wafer using a second measurement point that will come into contact with the first measurement point after bonding as the measurement target.

13. A measurement recipe creation method comprising the steps of: reading a first recipe for a first wafer and acquiring first measurement point information that is a measurement target for the first wafer; reading layout data for the first wafer and / or a second wafer and acquiring coordinate system information from the layout data; and inverting the first measurement point information using the coordinate system information, and creating a second recipe for the second wafer based on the inverted first measurement point information.

14. A measurement recipe creation method comprising the steps of: creating a first recipe including first measurement point information that is a measurement target on a first wafer, based on layout data of the first wafer and measurement conditions for the first wafer input by a user; and inverting the first measurement point information using coordinate system information acquired from the layout data, and creating a second recipe including second measurement point information that is a measurement target on a second wafer, based on the inverted first measurement point information.

Images