System and method for detecting statistical anomalies induced by Z-PAT defects in semiconductor reliability failures

Abstract

The present invention relates to a system and method for detecting statistical aberrations induced by Z-PAT defects of semiconductor reliability failures, comprising receiving grouped electrical test data with semiconductor chip data for multiple wafers in a batch generated by a subsystem for detecting statistical aberrations configured to perform partial average testing in the Z direction (Z-PAT) on the test data generated by the subsystem in a post-manufacturing electrical test system for multiple wafers in the batch. The system includes receiving characterization data for multiple wafers in the batch generated by a semiconductor manufacturing characterization subsystem during the manufacturing of the multiple wafers in the batch, and determining a statistical correlation between the grouped electrical test data and the characterization data at the same location (x,y) on each wafer of the multiple wafers in the batch. Based on the statistical correlation, a defect data signature is located on the multiple wafers in the batch.

Description

[Technical field]

[0001] The present disclosure relates generally to semiconductor devices. REFERENCE TO RELATED APPLICATIONS This application claims priority to U.S. Provisional Application No. 63 / 208,014 (June 8, 2021), which is incorporated by reference herein in its entirety. [Background technology]

[0002] The present disclosure relates generally to semiconductor devices, and more particularly to systems and methods for Z-direction Part Average Testing (Z-PAT) defect-guided statistical outlier detection of semiconductor reliability failures.

[0003] The manufacture of semiconductor devices may typically require tens of thousands or more processing steps to form a functioning device. During the course of these processing steps, various characterization measurements (e.g., inspection and / or metrology measurements) may be performed to identify defects and / or monitor various parameters on the device. Electrical testing may be performed in lieu of or in addition to various characterization measurements to verify or evaluate the functionality of the device. However, while some detected defects and metrology errors may be significant enough to clearly indicate device failure, smaller variations may cause premature reliability failure of the device after exposure to the working environment. Users of semiconductor devices at risk (e.g., for automotive, military, aviation, and medical applications) are currently striving for failure rates in the parts per billion (PPB) range that exceed parts per million (PPM) levels. Evaluating the reliability of semiconductor dies is important to meet the requirements of these industries as the need for semiconductor devices in automotive, military, aviation, and medical applications continues to increase. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] US Patent Application Publication US2018 / 0328868A1 [Patent Document 2] U.S. Patent 10,761,128 Summary of the Invention [Problem to be solved by the invention]

[0005] Therefore, it may be desirable to provide a system and method for reliability defect detection. [Means for solving the problem]

[0006] A system is disclosed according to one or more embodiments of the present disclosure. In an exemplary embodiment, the system includes a controller communicatively coupled to at least a semiconductor fab (fabrication)-characterization subsystem. In another exemplary embodiment, the controller includes one or more processors configured to execute program instructions that cause the one or more processors to receive electrical test bin data via a defect-guided correlation subsystem. In another exemplary embodiment, the electrical test bin data includes semiconductor die data for a plurality of wafers in a lot. In another exemplary embodiment, the electrical test bin data is generated by a statistical outlier detection subsystem configured to perform a Z-PAT on the test data. In another exemplary embodiment, the electrical test subsystem is configured to generate the test data by testing a plurality of wafers in the lot after fabrication by a semiconductor fabrication characterization subsystem. In another exemplary embodiment, the controller includes one or more processors configured to execute program instructions that cause the one or more processors to receive characterization data via the defect-guided correlation subsystem. In another exemplary embodiment, the characterization data for a plurality of wafers in a lot is generated by the semiconductor fabrication characterization subsystem during fabrication of the plurality of wafers in the lot. In another exemplary embodiment, the controller includes one or more processors configured to execute program instructions that cause the one or more processors to determine a statistical correlation between the electrical test bin data and the characterization data via the defect induced correlation subsystem at a same x,y location on each of a plurality of wafers in the lot. In another exemplary embodiment, the controller includes one or more processors configured to execute program instructions that cause the one or more processors to locate defect data signatures on a plurality of wafers in the lot based on the statistical correlation via the defect induced correlation subsystem.

[0007] A method is disclosed according to one or more embodiments of the present disclosure. In an exemplary embodiment, the method includes, but is not limited to, receiving electrical test bin data via a defect induced correlation subsystem. In another exemplary embodiment, the electrical test bin data includes semiconductor die data for a plurality of wafers in a lot. In another exemplary embodiment, the electrical test bin data is generated by a statistical outlier detection subsystem configured to perform Z-PAT on the test data. In another exemplary embodiment, the electrical test subsystem is configured to generate the test data by testing the plurality of wafers in the lot after fabrication by a semiconductor manufacturing characterization subsystem. In another exemplary embodiment, the method includes, but is not limited to, receiving characterization data via the defect induced correlation subsystem. In another exemplary embodiment, the characterization data for the plurality of wafers in the lot is generated by the semiconductor manufacturing characterization subsystem during fabrication of the plurality of wafers in the lot. In another exemplary embodiment, the method can include, but is not limited to, determining a statistical correlation between the electrical test bin data and the characterization data via the defect induced correlation subsystem at the same x,y location on each of the plurality of wafers in the lot. In another exemplary embodiment, the method includes, but is not limited to, locating defect data signatures on multiple wafers in the lot based on statistical correlation via the defect-induced correlation subsystem.

[0008] A system is disclosed according to one or more embodiments of the present disclosure. In one exemplary embodiment, the system includes a semiconductor manufacturing characterization subsystem. In another exemplary embodiment, the semiconductor manufacturing (fab) characterization subsystem is configured to manufacture a plurality of wafers in a lot. In another exemplary embodiment, the semiconductor manufacturing characterization subsystem is configured to generate characterization data for a plurality of wafers in a lot during manufacture of the plurality of wafers in the lot. In another exemplary embodiment, the system includes an electrical testing subsystem. In another exemplary embodiment, the electrical testing subsystem is configured to generate test data for a plurality of wafers in the lot after manufacture by the semiconductor manufacturing characterization subsystem. In another exemplary embodiment, the system includes a controller communicatively coupled to at least the semiconductor manufacturing characterization subsystem. In another exemplary embodiment, the controller includes one or more processors configured to execute program instructions that cause the one or more processors to receive the electrical test bin data via the defect-induced correlation subsystem. In another exemplary embodiment, the electrical test bin data includes semiconductor die data for a plurality of wafers in the lot. In another exemplary embodiment, the electrical test bin data is generated by a statistical outlier detection subsystem configured to perform Z-PAT. In another exemplary embodiment, the controller includes one or more processors configured to execute program instructions that cause the one or more processors to receive characterization data via the defect-induced correlation subsystem. In another exemplary embodiment, the controller includes one or more processors configured to execute program instructions that cause the one or more processors to determine a statistical correlation between electrical test bin data and characterization data via the defect-induced correlation subsystem at a same x,y location on each of a plurality of wafers in the lot. In another exemplary embodiment, the controller includes one or more processors configured to execute program instructions that cause the one or more processors to locate defect data signatures on a plurality of wafers in the lot based on the statistical correlation via the defect-induced correlation subsystem. Effect of the Invention

[0009] A system and method for reliability defect detection is provided. It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not necessarily restrictive of the invention as claimed. The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the invention and, together with the general description, serve to explain the principles of the invention.

[0010] Many advantages of the present disclosure may be better understood by those skilled in the art by reference to the following drawings. [Brief description of the drawings]

[0011] [Figure 1] FIG. 1 is a block diagram of a system for detecting semiconductor reliability defects in accordance with one or more embodiments of the present disclosure. [Diagram 2] FIG. 2 is a flow diagram illustrating steps performed in a method or process for detecting semiconductor reliability defects in accordance with one or more embodiments of the present disclosure. [Figure 3A] FIG. 3A is a probe map illustrating detected semiconductor reliability defects in accordance with one or more embodiments of the present disclosure. [Figure 3B] FIG. 3B is a probe map illustrating detected and inferred semiconductor reliability failures in accordance with one or more embodiments of the present disclosure. [Figure 4] FIG. 4 is a block diagram of a system for Z-PAT defect-induced statistical outlier detection of semiconductor reliability failures in accordance with one or more embodiments of the present disclosure. [Diagram 5] FIG. 5 is a flow diagram illustrating steps performed in a method or process for Z-PAT defect induced statistical outlier detection of semiconductor reliability failures in accordance with one or more embodiments of the present disclosure. [Figure 6A] FIG. 6A is a probe map showing detected semiconductor reliability defects overlaid with characterization data in accordance with one or more embodiments of the present disclosure. [Figure 6B] FIG. 6B is a probe map showing detected semiconductor reliability failures overlaid with characterization data in accordance with one or more embodiments of the present disclosure. [Figure 7A] FIG. 7A is a block diagram of a system for manufacturing, characterizing, and / or testing semiconductor devices in accordance with one or more embodiments of the present disclosure. [Figure 7B] FIG. 7B is a block diagram of a system for manufacturing, characterizing, and / or testing a semiconductor device in accordance with one or more embodiments of the present disclosure. [Figure 8] FIG. 8 is a flow diagram illustrating steps performed in a method or process for manufacturing, characterizing, and / or testing a semiconductor device according to one or more embodiments of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0013] Reference will now be made in detail to the disclosed subject matter, which is illustrated in the accompanying drawings. The present disclosure has been specifically shown and described with respect to certain embodiments and certain features thereof. The embodiments described herein are to be construed as illustrative and not restrictive. It will be readily apparent to those skilled in the art that various changes and modifications in form and detail may be made therein without departing from the spirit and scope of the present disclosure.

[0014] The fabrication of semiconductor devices may typically require tens of thousands or more processing steps to form a functioning device. During these processing steps, various characterization measurements (e.g., inspection and / or metrology measurements) may be performed to identify defects and / or monitor various parameters on the device. Electrical testing may be performed instead of or in addition to various characterization measurements to verify or evaluate the functionality of the device.

[0015] However, while some detected defects and metrology errors may be significant enough to clearly indicate device failure, smaller variations may cause premature reliability failure of the device after exposure to the working environment. Defects occurring during the manufacturing process may have a wide-ranging effect on the performance of devices in the field. For example, "killer" defects occurring at known or unknown locations in the design may result in immediate device failure. For example, killer defects at unknown locations may be particularly problematic if they are susceptible to reliability escapes in the test gap and the semiconductor device may be functionally dead after processing, but the device manufacturer is unable to make this determination due to limitations in testing. As another example, minor defects may have little or no effect on device performance throughout the device's lifetime. As another example, a class of defects known as latent reliability defects (LRDs) may not lead to failure during manufacturing / testing or may not lead to immediate device failure during operation, but may lead to premature end-of-life failure of the device during operation when used in the working environment. It should be noted that, as used herein, the terms "manufacturing process" and "fabrication process" may be considered equivalent for purposes of this disclosure, along with their respective variations on the terms (e.g., "manufacturing line" and "fabrication line").

[0016] Hazardous users of semiconductor devices (e.g., automotive, military, aviation, and medical applications) are now looking for failure rates in the parts per billion (PPB) range that exceed parts per billion (PPM) levels. Evaluating semiconductor die reliability and identifying the causes of reliability failures is critical to meeting the requirements of these industries as the need for semiconductor devices in automotive, military, aviation, and medical applications continues to increase.

[0017] Semiconductor devices that play quality critical roles can be subjected to electrical testing during wafer sort and again at final test after singulation and packaging. In addition, the semiconductor devices can be subjected to a method configured to determine systematic defectivities occurring at the same x,y location on multiple wafers within a given lot. In traditional semiconductor wafer processing methods, the x and y dimensions localize the die location on the wafer, and the z dimension refers to individual wafers that are stacked on top of each other in a wafer cassette.

[0018] Component Average Testing (PAT) has been adopted by most automotive semiconductor companies primarily to help meet the stringent requirements of the automotive industry and increasingly high-end mobile devices. Reliability studies have shown that semiconductor components with abnormal electrical characteristics are more prone to contributing to long-term quality and reliability issues. For example, a device that originally passed all manufacturing tests but may be considered an "outlier" compared to others in the same population may be more likely to fail in the field. The PAT methodology proactively identifies these outliers for removal from production shipments.

[0019] PAT methods may include, but are not limited to, geographic part average testing (G-PAT) (e.g., involving testing good die in defective regions), parametric part average testing (P-PAT) (e.g., involving parametric signals outside a threshold or norm but within specification limits), synthetic part average testing (C-PAT) (e.g., involving multiple repairs on a die), in-line defective part average testing (I-PAT), and Z-direction part average testing (Z-PAT). It is noted herein that systems and methods for 1-PAT are described in U.S. Patent Application Publication US2018 / 0328868A1 (November 15, 2018). In addition, it is noted herein that systems and methods for 1-PAT are described in U.S. Patent 10,761,128 (September 1, 2020) and U.S. Application 17 / 101,856 (November 23, 2020), both of which are incorporated herein in their entirety.

[0020] Z-PAT involves partial average testing in the z-direction and traditionally relies on test data only. Semiconductor suppliers can ink out a die that tests electrically "good" if the same x,y location tests bad on multiple wafers in the same lot. This inking out, or "overkill", is based on the observation that many systematic factors that lead to quality problems at a particular location on a wafer are frequently replicated for die locations on all wafers in the lot.

[0021] In one non-limiting example, particles adhering to the chuck of a wafer processing tool may result in a consistently raised protrusion on the front surface at that location. In another non-limiting example, an etching process "dead center" problem may involve a die in the middle of a wafer being consistently under-etched. In another non-limiting example, a process tool may consistently deposit particles at a specific location around the edge of a wafer. It is noted herein that the above examples are illustrative and are not intended to be limiting with respect to systematic problems at wafer locations.

[0022] FIG. 1 illustrates a system 100 for detecting semiconductor reliability defects in accordance with one or more embodiments of the present disclosure.

[0023] In some embodiments, the system 100 includes a semiconductor fab characterization subsystem 102, hereafter referred to as the semiconductor fab characterization subsystem 102. The semiconductor fab characterization subsystem 102 may include a plurality of characterization tools configured to perform characterization measurements on semiconductor devices in a lot (e.g., semiconductor wafers 104, or wafers 104 for purposes of this disclosure). For example, the plurality of characterization tools may include, but are not limited to, one or more in-line defect inspection tools and / or metrology tools configured to characterize the semiconductor devices. As another example, the characterization measurements may include, but are not limited to, in-line defect inspection measurements and / or metrology measurements. For example, the inspection measurements may include baseline inspection (e.g., sampling-based inspection), screening inspection at key semiconductor device layers, and the like. For purposes of this disclosure, "characterization" may refer to either in-line defect inspection or in-line metrology measurements. It is noted herein that the in-line defect inspection tool and / or metrology tool may perform a standard characterization process or a non-standard (eg, proprietary) characterization process.

[0024] Characterization measurements may be performed during (e.g., before, during, and / or after) the manufacturing of one or more semiconductor devices (e.g., wafers 104) in a lot through multiple semiconductor manufacturing processes performed by multiple semiconductor manufacturing tools. For example, one or more semiconductor manufacturing characterization subsystems 102 may include, but are not limited to, one or more process tools configured to manufacture semiconductor devices including 1, 2, .... N layers are manufactured according to several steps (e.g., tens of thousands of times) performed by several semiconductor manufacturing processes.

[0025] In some embodiments, the system 100 includes an electrical test subsystem 106. The semiconductor manufacturing characterization subsystem 102 provides the wafers 104 and / or die material to the electrical test subsystem 106. The electrical test subsystem 106 can be configured to output test data 109 including data after probing a lot of the wafers 104, where the probed wafers are referred to as a lot of wafers. For example, the electrical test subsystem 106 may include, but is not limited to, one or more electrical test tools, one or more stress test tools, or the like. The electrical test subsystem 106 can be configured to test semiconductor devices manufactured by one or more semiconductor manufacturing processes executed via the semiconductor manufacturing characterization subsystem 102. For purposes of this disclosure, "tests" can be understood to refer to a process of electrically evaluating device functionality at the end of a fabrication manufacturing process (e.g., an electrical wafer sort (EWS) process, etc.), at the end of packaging, and / or at the end of final test (e.g., after burn-in and other quality check processes). It is noted herein that non-passing semiconductor dies or wafers may be segregated from passing semiconductor dies or wafers and / or flagged for further testing. The data 109 representing the probed lot of wafers may be in the form of a probe map or may be used to generate a probe map.

[0026] In some embodiments, the system 100 includes a statistical outlier detection subsystem 110. The electrical test subsystem 106 can output test data 109 to the statistical outlier detection subsystem 110. The statistical outlier detection subsystem 110 can output outlier data or electrical test bin data 112, where the electrical test bin data 112 includes semiconductor die data for the wafers 104 in the lot. For example, the statistical outlier detection subsystem 110 can include and / or be configured to perform Z-PAT methodology. As another example, the statistical outlier detection subsystem 110 can be configured to include and / or perform other PAT methods or other known statistical outlier determination techniques.

[0027] 2 illustrates a method or process 200 for detecting semiconductor reliability defects in accordance with one or more embodiments of the present disclosure. It is noted herein that the steps of the method or process 200 may be implemented in whole or in part by the system 100 illustrated in FIG 1. However, it is further recognized that the method or process 200 is not limited to the system 100 illustrated in FIG 1 in that additional or alternative system level embodiments may perform all or a portion of the steps of the method or process 200.

[0028] In step 202, a semiconductor device is received from a semiconductor manufacturing characterization subsystem. In some embodiments, the semiconductor manufacturing characterization subsystem 102 is configured to fabricate multiple wafers 104. For example, the semiconductor manufacturing characterization subsystem 102 may include, but is not limited to, one or more process tools configured to fabricate semiconductor devices including 1, 2, .... N layers are fabricated according to several steps (e.g., tens of thousands) performed by several semiconductor manufacturing processes.

[0029] In step 204, the semiconductor devices are tested in an electrical test subsystem to generate test data. In some embodiments, the electrical test subsystem 106 receives a lot of wafers 104. For example, the electrical test subsystem 106 may perform electrical tests and / or stress tests to generate the test data 109.

[0030] The test data is sent to the statistical outlier detection subsystem at step 206. In some embodiments, the electrical test subsystem 106 sends the test data 109 to the statistical outlier detection subsystem 110.

[0031] In step 208, the test data is processed in a statistical outlier detection subsystem to generate electrical test bin data. In some embodiments, the statistical outlier detection subsystem 110 may determine the location and / or systematic spread of presumed electrically failed die on the selected wafer 104 of the lot based on known electrically failed die on other wafers 104 of the lot in the test data 109 received from the electrical test subsystem 106.

[0032] At step 210, the reclassified electrical test bin data based on the characterization data is received via the statistical outlier detection subsystem. In some embodiments, at least some of the semiconductor die data in the electrical test bin data 112 are reclassified using one or more steps of the method or process 500 or 800, described in further detail herein. It is noted herein that one or more adjustments may be determined for at least one of the manufacturing, characterization, and / or testing of the semiconductor device based on the reclassification of the electrical test bin data 112 and / or defects newly discovered during the execution of the method or process 500 or 800. For example, the one or more adjustments may modify a manufacturing process or method, a characterization process or method, a testing process or method, or the like, provided in a feedback loop to components in the semiconductor manufacturing characterization subsystem 102. For example, the manufacturing process or method, the characterization process or method, the testing process or method, etc. may be adjusted (e.g., via one or more control signals) based on the reclassification of the electrical test bin data 112 and / or defects newly discovered during the execution of the method or process 500 or 800.

[0033] 3A and 3B generally illustrate a probe map 300 of a lot of wafers 108 in accordance with one or more embodiments of the present disclosure.

[0034] 3A, selected wafers 108 in a lot include good die 302 and electrically failed die 304 that include indicators of probed problems in the electrical test bin data 112. In the non-limiting example illustrated in FIG. 3A, nine of the twenty-four wafers 108 in the probe map 300 (e.g., W1, W4, W6, W8, W12, W16, W20, W22, W24) show electrically failing die 304 clustered at the same x,y location on their respective wafers.

[0035] 3B, Z-PAT is performed on a lot of wafers 108 by the statistical outlier detection subsystem 110. During Z-PAT, dies 306 may be ink out as potentially electrically failing (bad) due to their similarity in x,y location to known electrically failing dies 304 based on a rule set that includes a threshold for complete ink out (e.g., this may be defined per fab or determined for multiple fabs). For example, the threshold may represent an overkill limit that depends on the number of electrically failing dies 304 observed by the electrical testing subsystem 106. If the threshold is exceeded, the statistical outlier detection subsystem 110 may ink out the potentially electrically failing die 306 on the wafer 108 in the same x,y location as the known electrically failing die 304 on the other wafer 108 because the potentially electrically failing die 306 is deemed to contain defects due to relative positioning with the known electrically failing die 304. For example, the remaining 15 of the 24 wafers 108 (e.g., W2, W3, W5, W7, W9, W10, W11, W13, W14, W15, W17, W18, W19, W21, W23) may contain potentially electrically failing die 306 at the same x,y location, which may be ink-out. For example, the electrical test subsystem 106 may not be fully capturing potential reliability defects (LRDs) and / or escapes due to gaps in test coverage.

[0036] The Z-PAT methods listed above have several drawbacks. For example, the Z-PAT methods listed above may result in excessive yield loss, or excessive kill, due to the relative rarity of the presumed failures actually manifesting as reliability failures or customer returns, which risk-poor semiconductor suppliers to the automotive segment typically sacrifice. As another example, the Z-PAT methods listed above generally do not provide semiconductor manufacturing engineers with enough information about the underlying cause of the failure to prevent it from occurring in the future (or at least create a baseline to monitor the frequency of occurrence), resulting in adjustments to the system 100 and / or method or process 200 being reactive and not non-responsive. Thus, any methodology that could provide insight into the causal failure mechanism and / or the propagation of this failure mechanism to other wafers would enable better decision making and thus reduce overkill.

[0037] Embodiments of the present disclosure are directed to systems and methods for Z-PAT defect induced statistical outlier detection of semiconductor reliability failures. Embodiments of the present disclosure also relate to identifying Z-PAT defect signatures indicative of potential reliability and / or test gap defects at the same x,y location across multiple wafers in a lot using characterization data (e.g., in-line defect inspection data and / or metrology data). Embodiments of the present disclosure use statistical outlier algorithms to identify Z-PAT defect signatures indicative of potential reliability and / or test gap defects at the same x,y location across multiple wafers in a lot using characterization data (e.g., in-line defect inspection data and / or metrology data). FIELD OF THE DISCLOSURE The present disclosure relates to the identification of Z-PAT defect signatures. Embodiments of the present disclosure relate to the automatic notification of manufacturing engineers of the presence of Z-PAT defect signatures. Embodiments of the present disclosure relate to the characterization of Z-PAT defects using spatial signature analysis methods on the signatures.

[0038] Embodiments of the present disclosure relate to characterizing Z-PAT defect signatures using machine learning methods.

[0039] Embodiments of the present disclosure relate to identifying the presence or absence of Z-PAT defect signatures within a given lot.

[0040] Embodiments of the present disclosure relate to identifying Z-PAT defect signatures in adjacent lots.

[0041] Embodiments of the present disclosure relate to Z-PAT defect signature identification that cannot be detected by electrical test-based Z-PAT.

[0042] Embodiments of the present disclosure relate to reducing overkill using Z-PAT defect signatures to more accurately summarize the amount of impacted die / wafer.

[0043] Embodiments of the present disclosure relate to rapid identification of underlying roots triggered based on learning from previously characterized Z-PAT defect signatures.

[0044] Embodiments of the present disclosure relate to retrospective identification of Z-PAT defect signatures that are directly warranted and / or recalled using stored in-line defect data.

[0045] 4 illustrates a system 400 for Z-PAT defect-induced statistical outlier detection of semiconductor reliability failures in accordance with one or more embodiments of the present disclosure. It is noted herein that system 400 may enable semiconductor manufacturers to more accurately identify die at high risk for early-life reliability failures or test coverage gaps and / or enable suppliers to better locate die from wafers that contain systematic defects that occur at the same x,y location on multiple wafers within a given lot through the Z-PAT methodology.

[0046] In some embodiments, system 400 includes one or more components of system 100. System 400 may include a semiconductor manufacturing characterization subsystem 102 configured to manufacture and characterize semiconductor wafers 104. System 400 may include an electrical testing subsystem 106 configured to receive wafers 104 and perform electrical testing to generate test data 109. System 400 may include a statistical outlier detection subsystem 110 configured to receive test data 109 after application of the Z-PAT method and output outlier data 112.

[0047] In some embodiments, the system 400 includes a defect reduction subsystem 402. The defect reduction subsystem 402 can be configured to receive characterization data 404 (e.g., characterization measurements including, but not limited to, in-line defect inspection measurements and / or metrology measurements) extracted during (e.g., before, during, and / or after steps) manufacturing of one or more semiconductor devices (e.g., wafers 104) via multiple semiconductor manufacturing processes performed by multiple semiconductor manufacturing tools, where the characterization data 404 includes semiconductor die data for the wafers 104 within a lot.

[0048] The defect reduction subsystem 402 may be configured to generate filtered characterization data 406 (or filtered data 406 for purposes of this disclosure) that is a subset of the characterization data 404. The filtered characterization data 406 may be generated via one or more 1-PAT methods or processes. It is noted herein that systems and methods for inline part average testing (I-PAT) are described in U.S. Patent Application Publication US2018 / 0328868A1 (November 15, 2018). In addition, it is noted herein that systems and methods for I-PAT are described in U.S. Patent 10,761,128 (September 1, 2020) as well as U.S. Application 17 / 101,856 (November 23, 2020), both of which have been previously incorporated herein in their entirety. In this regard, the defect reduction subsystem 402 may be considered an I-PAT analyzer for purposes of this disclosure.

[0049] In some embodiments, the system 400 includes a defect induced correlation subsystem 408. The defect reduction subsystem 402 can be configured to output the filtered characterization data 406 to the defect induced correlation subsystem 408. The defect induced correlation subsystem 408 may be configured to output improved electrical test die bin data 410 (e.g., with improved semiconductor die data) to a stakeholder (e.g., a fab engineer) and / or may be configured to output reclassified electrical test die bin data 412 (e.g., with reclassified semiconductor die data) to the statistical outlier detection subsystem 110. For example, the improved electrical test die bin data 410 and / or the reclassified electrical test die bin data 412 may be determined through deterministic and / or statistical thresholding methods or processes, spatial signature analysis methods or processes, advanced deep learning or machine learning methods or processes, etc. Generally, the machine learning technique may be any technique known in the art, including but not limited to supervised learning, unsupervised learning, or other learning-based processes, such as but not limited to linear regression, neural or deep neural networks, heuristic-based models, etc. The output of the improved electrical test divin data 410 and / or the reclassified electrical test divin data 412 including Z-PAT defect signatures may be automatically performed by an interested party (e.g., a fab engineer) and / or the statistical outlier detection subsystem 110 following the determination of the improved electrical test divin data 410 and / or the reclassified electrical test divin data 412.

[0050] It is noted herein that the improved electrical test dive bin data 410 and / or the reclassified electrical test dive bin data 412 may result in reduced overkill by using Z-PAT defect signatures to more accurately demarcate the range of affected dies / wafers. Additionally, it is noted herein that the improved electrical test dive bin data 410 and / or the reclassified electrical test dive bin data 412 may be responsive to identifying the presence or absence of Z-PAT defect signatures within a given lot. It is further noted herein that the improved electrical test dive bin data 410 and / or the reclassified electrical test dive bin data 412 may be responsive to identifying Z-PAT defect signatures on adjacent lots. It is further noted herein that the improved electrical test dive bin data 410 and / or the reclassified electrical test dive bin data 412 may be responsive to identifying Z-PAT defect signatures not detected by electrical test-based Z-PAT. It is further noted herein that the improved electrical test divin data 410 and / or the reclassified electrical test divin data 412 can be for rapid identification of underlying root causes based on learning from previously characterized Z-PAT defect signatures. It is further noted herein that the improved electrical test divin data 410 and / or the reclassified electrical test divin data 412 can be for retrospective identification of Z-PAT defect signatures to guide warranty and / or recall efforts using stored in-line defect data.

[0051] It is noted herein that the characterization data 404 may be run through a recipe of a fab-wide defect management subsystem 414 (e.g., a fab yield management subsystem) to filter the characterization data 404 so that the defect-induced correlation subsystem 408 can process the partially analyzed characterization data 404 and / or the filtered data 406 (e.g., when the characterization data 404 first passes through the defect reduction subsystem 402).

[0052] Although embodiments of the present disclosure illustrate extracting characterization data 404 from the semiconductor manufacturing characterization subsystem 102, generating filtered characterization data 406, and processing it in the defect reduction subsystem 402 before providing it to the defect-induced correlation subsystem 408, it is noted herein that the characterization data 404 may be provided as raw data directly from the semiconductor manufacturing characterization subsystem 102 to the defect-induced correlation subsystem 408. In this regard, the defect reduction subsystem 402 may not be considered a required component of the system 400 for purposes of the present disclosure. Thus, the above description should not be construed as a limitation on the scope of the present disclosure, but merely as an example.

[0053] Although embodiments of the present disclosure illustrate the defect-inducing correlation subsystem 408 as separate from the defect reduction subsystem 402, it is noted herein that the defect-inducing correlation subsystem 408 can be integrated into the defect reduction subsystem 402 and vice versa. More generally, while embodiments of the present disclosure illustrate the subsystems 102, 106, 110, 402, 408 as separate or stand-alone subsystems in the system 400, it is noted herein that one or more of the subsystems 102, 106, 110, 402, 408 can be combined or integrated subsystems. Thus, the above description should not be construed as a limitation on the scope of the present disclosure, but merely as an example.

[0054] 5 illustrates a method or process 500 for Z-PAT defect induced statistical outlier detection of semiconductor reliability failures in accordance with one or more embodiments of the present disclosure. It is noted herein that the steps of the method or process 500 may be implemented in whole or in part by the system 400 illustrated in FIG. 4. However, it is further recognized that the method or process 500 is not limited to the system 400 illustrated in FIG. 4 in that additional or alternative system level embodiments may perform all or a portion of the steps of the method or process 500.

[0055] In step 502, characterization data for a semiconductor device is received from a semiconductor manufacturing characterization subsystem. In some embodiments, the semiconductor manufacturing characterization subsystem 102 is configured to perform characterization measurements during the manufacturing of one or more wafers 104. For example, the semiconductor manufacturing characterization subsystem 102 may include, but is not limited to, one or more in-line defect inspection and / or metrology tools configured to characterize the semiconductor device. For example, the one or more outputs may include, but are not limited to, baseline inspection (e.g., sampling-based inspection), screening inspection at key semiconductor device layers, etc. For purposes of this disclosure, "characterization" may refer to either in-line defect inspection or in-line metrology measurements.

[0056] In step 504, the characterization data is processed through a defect reduction subsystem to generate filtered data. In some embodiments, the defect reduction subsystem 402 may receive the characterization data 404 and generate filtered characterization data 406. The defect reduction subsystem 402 may include or be configured to perform a 1-PAT methodology. In this regard, the defect reduction subsystem 402 may be considered an I-PAT analyzer for purposes of this disclosure.

[0057] In step 506, the filtered data and / or the characterization data are transmitted to the defect induced correlation subsystem 408. In some embodiments, the filtered characterization data 406 is transmitted from the defect reduction subsystem 402 to the defect induced correlation subsystem 408 and / or the characterization data 404 is transmitted from the semiconductor manufacturing characterization subsystem 102 to the defect induced correlation subsystem 408. It should be noted that step 504 may be optional if the characterization data 404 is forwarded directly to the defect induced correlation subsystem 408 such that the filtered characterization data 406 is not used by the defect induced correlation subsystem 408.

[0058] In step 508, the electrical test bin data is received by the defect-induced correlation subsystem. In some embodiments, the electrical test bin data 112 is generated by one or more components of the system 100 via one or more steps of the method or process 200.

[0059] At step 510, a statistical correlation between the electrical test bin data and the filtered data or the characterization data is determined via the defect-induced correlation subsystem. In some embodiments, the filtered characterization data 406 and / or the characterization data 404 are overlaid on the electrical test bin data 112 by the defect-induced correlation subsystem 408. For example, the filtered characterization data 406 may be determined by the defect reduction subsystem 402 based on its association with the particular Z-PAT systematic failure mechanism under study (e.g., in the electrical test bin data 112). As another example, all of the characterization data 404 may be overlaid on the electrical test bin data 112.

[0060] In step 512, the defect data signatures are located on the semiconductor device based on the statistical correlation via the defect-induced correlation subsystem. In some embodiments, if a statistical correlation is found after overlaying the filtered characterization data 406 and / or the characterization data 404 onto the electrical test bin data 112, the system 400 (e.g., the defect-induced correlation subsystem 408, or other components of the system 400) looks for similar defect / metrology signatures on lots of wafers 104 processed before and after the lot of wafers 104 undergoes the study.

[0061] In step 514, at least some of the semiconductor die data in the electrical test bin data are reclassified based on the defect data signature via the defect-inducing correlation subsystem. In some embodiments, depending on the results and the placement logic applied, the system 400 (e.g., the defect-inducing correlation subsystem 408, or other components of the system 400) reclassifies the bin associated with the die 302 / 304 / 306 under consideration. For example, the bin may be identified by the statistical outlier detection subsystem 110 as containing a fault found in the electrically failing die 304. As another example, the bin may be altered to indicate that the fault found in the known electrically failing die 304 or the potentially electrically failing die 306 by the statistical outlier detection subsystem 110 is a good die 302. As another example, the bin may be altered to indicate that the good die 302 may contain a fault or does not contain a fault and should therefore be considered as the potentially electrically failing die 306 or the known electrically failing die 304.

[0062] In step 516, the reclassified electrical test bin data is sent via the defect-induced correlation subsystem. In some embodiments, the improved electrical test bin data 410 is transmitted or otherwise provided by the defect-induced correlation subsystem 408 to a party (e.g., a fab engineer). For example, the improved electrical test bin data 410 may include a recommendation as to whether to ink out a die (e.g., believed to be a good die 302 and / or a die 306 having a potential electrical failure) having the same x,y location as a known electrically failed die 304. For example, the recommendation may be made based on a rule set that includes a threshold for complete ink-out (e.g., this may be defined per fab or may be determined for multiple fabs).

[0063] In step 518, the new defect data signature is transmitted via the defect-induced correlation subsystem. In some embodiments, the new defect data signature may be transmitted or otherwise provided by the defect-induced correlation subsystem 408 to the statistical outlier detection subsystem 110 as part of the reclassified electrical test bin data 412. For example, the statistical outlier detection subsystem 110 may use the new defect data signature to adjust the resulting electrical test bin data 112 when processing a subsequent lot of wafers 104. As another example, the statistical outlier detection subsystem 110 may output the new defect data signature to the semiconductor manufacturing characterization subsystem 102 for adjustment of a component or method or process of the semiconductor manufacturing characterization subsystem 102. It is noted herein that the semiconductor manufacturing characterization subsystem 102 may receive the new defect data signature directly instead of or in addition to the statistical outlier detection subsystem 110.

[0064] At step 520, a representation of the statistical correlation is displayed. In some embodiments, an overlay of the filtered characterization data 406 and / or the characterization data 404 on the electrical test bin data 112 is presented on a graphical user interface. For example, the representation may be a quantitative (e.g., data list, table, etc.) or qualitative (e.g., graph, chart, image, video, etc.) representation of the data overlay and corresponding metrics. The representation may be accompanied by recommendations for improvements to various steps performed by the method or process 200 and / or 500.

[0065] 6A and 6B generally illustrate a probe map 300 of a lot of wafers 108 in accordance with one or more embodiments of the present disclosure. In FIGS. 6A and 6B, the filtered characterization data 406 and / or the characterization data 404 are overlaid on the probe map 300 as defect data 600 for a particular failure mechanism. A statistical correlation is performed between the characterization data 404 and the electrical test bin data 112 for both wafers 108 having electrically failed dies 304 (e.g., W1, W4, W6, W8, W12, W16, W20, W22, W24) and the remaining wafers 108 in the lot.

[0066] 6A , the defect induction correlation subsystem 408 determines that one or more defects for a particular defect mechanism are the root cause of defects on the affected wafer 108 (e.g., W1, W4, W6, W8, W12, W16, W20, W22, W24) based on the overlay of the defect data 600 on the electrical test bin data 112. In this non-limiting example, the defect induction correlation subsystem 408 determines that the electrical test subsystem has sufficiently captured the escape based on the overlay, such that there is no need to ink off the die 302 onto another wafer 108.

[0067] 6B, the defect induction correlation subsystem 408 determines that one or more defects for a particular defect mechanism are the root cause of defects on the affected wafer 108 (e.g., W1, W4, W6, W8, W12, W16, W20, W22, W24) based on the overlay of the defect data 600 on the electrical test bin data 112. In this non-limiting example, the defect induction correlation subsystem 408 determines based on the overlay that the electrical test subsystem 106 did not adequately capture the escape and needs to ink off the die 302 onto other wafers 108. For example, the electrical test subsystem 106 may not adequately capture the escape due to it being a potential reliability defect (LRD) and / or due to a gap in test coverage.

[0068] It is noted herein that the above examples are two non-limiting examples that represent the boundaries of possible decisions by the defect-induced correlation subsystem 408. For example, the defect-induced correlation subsystem 408 may determine that only a small number of wafers 108 contain escapes or defects per filtered characterization data 406 and / or characterization data 404, even though the wafers 108 do not contain corresponding electrical test bin data 112. However, it is noted herein that the system 400 and method or process 500, when applied conservatively, may tend to ink an entire set of dies 302 at the same x,y location for a particular failure mechanism across a lot of wafers 108, regardless of the number of escapes. Thus, the above description should not be construed as a limitation on the scope of the present disclosure, but merely as an example.

[0069] It is noted herein that, although embodiments of the present disclosure illustrate defect data 600 as including only filtered characterization data 406 and / or characterization data 404 for a particular failure mechanism overlaid on the probe map 300, the overlay is not limited to a particular failure mechanism, and any additional (or all additional) filtered characterization data 406 and / or characterization data 404 may be overlaid on the probe map 300. Thus, the above description should not be construed as a limitation on the scope of the present disclosure, but merely as an example.

[0070] While embodiments of the present disclosure illustrate an overlay of defect data 600 on electrical test bin data 112, it is noted herein that system 400 may be configured to proactively attempt to identify undetected repeating spatial signatures in defectivity across wafer 104 independent of the presence of electrical test bin data 112. For example, this may be done by or within defect induced correlation subsystem 408, defect reduction subsystem 402, or a yield management system included in or associated with system 400. Thus, the above description should not be construed as a limitation on the scope of the present disclosure, but merely as illustrative.

[0071] In this regard, the system 400 and method or process 500 can be configured to reduce potential yield loss or overkill since good die that are not part of the presumed underlying reliability issue are less likely to be inked off, potentially resulting in additional revenue for the semiconductor supplier.

[0072] Additionally, system 400 and method or process 500 may be configured to provide information regarding the root cause of a failure mechanism (failure at the same location on multiple wafers) and provide valuable feedback to quality engineers and / or semiconductor manufacturing facilities to drive elimination of the root cause in future wafers being processed.

[0073] Additionally, system 400 and method or process 500 may be configured to identify Z-PAT signatures on adjacent lots, including identifying other wafers potentially affected by the same root cause (e.g., other wafers outside the lot under consideration), which may result in quality improvements by reducing escapes on lots not identified by traditional Z-PAT ink-off. For example, identifying Z-PAT signatures on adjacent lots may catch a problem at an earlier stage of its incremental growth / propagation by providing granularity over defect data that may not be available through electrical test bin data 112.

[0074] Additionally, the system 400 and method or process 500 may be configured for pre-identification of other Z-PAT signatures that are not completely detected by wafer probe and / or final test.

[0075] 7A and 7B show block diagrams of a system 700 according to one or more embodiments of the present disclosure. It is noted herein that system 700 may be configured to perform processing steps for fabricating and / or analyzing semiconductor devices and / or components on semiconductor devices (e.g., semiconductor dies), as described throughout this disclosure. It is further noted herein that system 700 may include all or a portion of system 100 and / or system 400, as described throughout this disclosure.

[0076] In some embodiments, the system 700 includes a semiconductor manufacturing characterization subsystem 102 and an electrical test subsystem 106 .

[0077] In some embodiments, the semiconductor manufacturing characterization subsystem 102 includes one or more characterization tools configured to output characterization measurements in (or as) the characterization data 404. For example, the characterization measurements may include, but are not limited to, baseline inspection (e.g., sampling-based inspection), screening inspection at key semiconductor device layers, etc. For purposes of this disclosure, "characterization measurements" may refer to in-line defect inspection and / or in-line metrology measurements.

[0078] In one non-limiting example, the semiconductor manufacturing characterization subsystem 102 may include at least one inspection tool 702 (e.g., an in-line sample analysis tool) for detecting defects in one or more layers of a specimen 704 (e.g., a wafer 104). The semiconductor manufacturing characterization subsystem 102 may generally include any number or type of inspection tools 702. For example, the inspection tool 702 may include an optical inspection tool configured to detect defects based on interrogation of the specimen 704 with light from any source, such as, but not limited to, a laser source, a lamp source, an X-ray source, or a broadband plasma source. As another example, the inspection tool 702 may include a particle beam inspection tool configured to detect defects based on interrogation of the specimen 704 with one or more particle beams, such as, but not limited to, an electron beam, an ion beam, or a neutral particle beam. For example, the inspection tool 702 may include a transmission electron microscope (TEM) or a scanning electron microscope (SEM). For purposes of this disclosure, it is noted that, as used herein, at least one inspection tool 702 may refer to a single inspection tool 702 or a group of inspection tools 702 .

[0079] It is noted herein that the sample 704 may be a semiconductor wafer of a plurality of semiconductor wafers, each semiconductor wafer of the plurality of semiconductor wafers includes a plurality of (e.g., 1, 2, ... N number) layers fabricated according to several (e.g., tens of thousands) steps performed by a plurality of semiconductor manufacturing processes, and each layer of the plurality of layers includes a plurality of semiconductor dies. Here, each semiconductor die of the plurality of semiconductor dies includes a plurality of blocks. In addition, it is noted herein that the sample 704 may be a semiconductor die package formed from a plurality of semiconductor dies arranged in a 2.5D lateral combination of bare dies on a substrate inside an advanced die package or a 3D die package.

[0080] For purposes of this disclosure, the term "defect" may refer to a physical defect found by an in-line inspection tool, a metrology measurement outlier, or any other physical characteristic of a semiconductor device that is deemed to be abnormal. A defect may be considered to be any deviation of a fabricated layer or a fabricated pattern within a layer from a design characteristic, including but not limited to physical, mechanical, chemical, or optical properties. In addition, a defect may be considered to be any deviation in the alignment or joining of components within a fabricated semiconductor die package. Furthermore, a defect may have any size relative to a semiconductor die or features thereon. In this manner, a defect may be smaller than a semiconductor die (e.g., at the scale of one or more patterned features) or larger than a semiconductor die (e.g., as a wafer-scale scratch or part of a pattern). For example, a defect may include a deviation in the thickness or composition of a sample layer before or after patterning. As another example, a defect may include a deviation in the size, shape, orientation, or position of a patterned feature. As another example, the defect may include defects associated with lithography and / or etching steps, such as, but not limited to, bridges (or lack thereof) between adjacent structures, pits, or holes. As another example, the defect may include damaged portions of the sample 704, such as, but not limited to, scratches or chips. For example, the severity of the defect (e.g., the length of the scratch, the depth of the pit, the measured size or polarity of the defect, etc.) may be important and considered. As another example, the defect may include foreign material introduced into the sample 704. As another example, the defect may be a misaligned and / or mismated package component on the sample 704. It should therefore be understood that the examples of defects in this disclosure are provided merely for illustrative purposes and should not be construed as limiting.

[0081] In another non-limiting example, the semiconductor manufacturing characterization subsystem 102 can include at least one metrology tool 706 (e.g., an in-line sample analysis tool) for measuring one or more properties of the sample 704 or one or more layers thereof. For example, the metrology tool 706 can characterize properties such as, but not limited to, layer thickness, layer composition, critical dimension (CD), overlay, or lithography processing parameters (e.g., intensity or dose of illumination during a lithography step). In this regard, the metrology tool 706 can provide information regarding the manufacturing of the sample 704, one or more layers of the sample 704, or one or more semiconductor dies of the sample 704, which information can relate to the probability of manufacturing defects that can lead to reliability issues in the resulting manufactured device. For purposes of this disclosure, it is noted herein that at least one metrology tool 706 may be a single metrology tool 706 or may represent a group of metrology tools 706.

[0082] In some embodiments, the semiconductor manufacturing characterization subsystem 102 includes at least one semiconductor manufacturing tool or process tool 708. It is noted herein that the specimen 704 can be moved between one or more inspection tools 702, one or more metrology tools 706, and one or more process tools 708 during the fabrication of the specimen 704. For example, the process tool 708 can include any tool known in the art, including, but not limited to, an etcher, a scanner, a stepper, a cleaner, and the like. For example, the fabrication process can include fabricating a plurality of dies distributed across a surface of the specimen (e.g., being a semiconductor wafer, etc.), each die including a plurality of patterned layers of material forming device components. Each patterned layer may be formed by the process tool 708 through a series of steps including material deposition, lithography, etching to generate a pattern of interest, and / or one or more exposure steps (e.g., performed by a scanner, stepper, etc.). As another example, the process tool 708 can include any tool known in the art configured to package and / or bond semiconductor dies into 2.5D and / or 3D semiconductor die packages. For example, the manufacturing process may include, but is not limited to, mating the semiconductor die and / or electrical components on the semiconductor die. Additionally, the manufacturing process may include, but is not limited to, joining the semiconductor die and / or electrical components on the semiconductor die via hybrid bonding (e.g., die-to-die, die-to-wafer, wafer-to-wafer, etc.) solder, adhesives, fasteners, or the like. It should be noted that for purposes of this disclosure, at least one process tool 708 may be a single process tool 708 or may represent a group of process tools 708.It should be noted that, as used herein, the terms "fabrication process" and "manufacturing process" may be considered equivalent for purposes of this disclosure, along with respective variations of the terms (e.g., "fabrication line" and "manufacturing line", "fabrication supplier" and "manufacturing supplier", etc.).

[0083] In some embodiments, system 700 includes an electrical test subsystem 106 for testing the functionality of one or more portions of the manufactured devices. For example, electrical test subsystem 106 may be configured to generate test data 109. It is noted herein that specimen 704 may be moved from semiconductor fabrication characterization subsystem 102 to electrical test subsystem 106 after manufacturing of specimen 704 is completed.

[0084] In one non-limiting example, the electrical test subsystem 106 may include any number or type of electrical test tools 710 to complete pre-probing at the wafer level, for example, where pre-probing is not designed to attempt to force failures at the wafer level.

[0085] In another non-limiting example, the electrical test subsystem 106 can include any number or type of stress test tools 712 for testing, inspecting, or otherwise characterizing the properties of one or more portions of a manufactured device at any point in the manufacturing cycle. For example, the stress test tools 712 can include, but are not limited to, tests such as pre-bum-in electrical wafer sort and final test (e.g., e-test) or post-burn-in electrical test configured to heat the specimen 704 (e.g., in an oven or other heat source), cool the specimen 704 (e.g., in a freezer or other cold source), operate the specimen 704 at an incorrect voltage (e.g., a power supply), etc.

[0086] In some embodiments, defects are identified using any combination of semiconductor manufacturing characterization subsystems 102 (e.g., inspection tools 702, metrology tools 706, etc.), electrical testing subsystems 106 (e.g., including electrical testing tools 710 and / or stress testing tools 712, etc.) utilized before or after one or more processing steps (e.g., lithography, etching, alignment, bonding, etc.) performed by one or more process tools 708 on the semiconductor die and / or layers of interest within the semiconductor die package. In this regard, defect detection at various stages of the manufacturing process may be referred to as in-line defect detection.

[0087] In some embodiments, the system 700 includes a controller 714. The controller 714 may be communicatively coupled to any of the components of the system 700, including, but not limited to, the semiconductor manufacturing characterization subsystem 102 (e.g., including the inspection tool 702 or the metrology tool 706), the electrical testing subsystem 106 (e.g., including the electrical testing tool 710 or the stress testing tool 712), and the like. It is noted herein that the embodiment shown in FIG. 7A and the embodiment shown in FIG. 7B may be considered as parts of the same system 700 or parts of different systems 700 for purposes of this disclosure. It is further noted herein that the components in the system 700 shown in FIG. 7A and the components in the system 700 shown in FIG. 7B may communicate directly or through the controller 714.

[0088] The controller 714 may include one or more processors 716 configured to execute program instructions maintained on a memory 718 (e.g., a storage medium, a storage device, etc.). The controller 714 may be configured to execute one or all of the steps of the method or process 200, the method or process 500, and / or the method or process 800 (e.g., as described throughout this disclosure). In this regard, the subsystems 110, 402, and / or 408 may be stored in and / or configured to be executed by the controller 714. However, it is noted herein that the subsystems 110, 402, and / or 408 may be separate from the controller 714 and configured to communicate with the controller 714 (e.g., either directly or through a server or controller communicatively coupled to the controller 714, which may include a processor and memory and other communicatively coupled components as described throughout this disclosure).

[0089] The one or more processors 716 may include any processor or processing element known in the art. For purposes of this disclosure, the term "processor" or "processing element" may be broadly defined to encompass any device having one or more processing or logic elements (e.g., one or more graphics processing units (GPUs), microprocessing units (MPUs), systems on chips (SoCs), one or more application specific integrated circuits (ASICs) devices, one or more field programmable gate arrays (FPGAs), or one or more digital signal processors (DSPs)). In this sense, the one or more processors 716 may include any device configured to execute algorithms and / or instructions (e.g., program instructions stored in memory). In one embodiment, the one or more processors 716 may be embodied as a desktop computer, a mainframe computer system, a workstation, an image computer, a parallel processor, a network computer, or any other computer system configured to execute programs that operate or are configured to operate with components of systems 100, 400, and / or 700 as described throughout this disclosure.

[0090] The memory 718 may include any storage medium known in the art suitable for storing program instructions executable by each of the associated one or more processors 716. For example, the memory 718 may include a non-transitory memory medium. As another example, the memory 718 may include, but is not limited to, a read-only memory (ROM), a random access memory (RAM), a magnetic or optical memory device (e.g., disk), a magnetic tape, a solid-state drive, and the like. Furthermore, it should be noted that the memory 718 may be housed in a common controller housing with the one or more processors 716. In one embodiment, the memory 718 may be located remotely with respect to the physical location of the respective one or more processors 716. For example, the respective one or more processors 716 may access a remote memory (e.g., a server) accessible via a network (e.g., the Internet, an intranet, etc.).

[0091] In another embodiment, the system 700 includes a user interface 720 coupled (e.g., physically coupled, electrically coupled, communicatively coupled, etc.) to the controller 714. For example, the user interface 720 may be a separate device coupled to the controller 714. As another example, the user interface 720 and the controller 714 may be located within a common or shared housing. However, it is noted herein that the controller 714 may not include, require, or be coupled to the user interface 720.

[0092] The user interface 720 may include, but is not limited to, one or more desktops, laptops, tablets, etc. The user interface 720 may include a display used to display data of the systems 100, 400, and / or 700 to a user. The display of the user interface 720 may include any display known in the art. For example, the display may include, but is not limited to, a liquid crystal display (LCD), an organic light emitting diode (OLED)-based display, or a CRT display. Those skilled in the art should recognize that any display device capable of being integrated with the user interface 720 is suitable for implementation in the present disclosure. In another embodiment, a user may input selections and / or commands in response to data displayed to the user via a user input device of the user interface 720.

[0093] It is noted herein that one or more of the systems 100, 400, 700 may be configured to operate with electronic chip identification (ID) tags, markers, designators, etc. For example, electronic chip IDs may be assigned to facilitate correlation of wafer-based bin data, characterization data (e.g., in-line defect inspection data and / or metrology data), packaged test data, etc.

[0094] 8 illustrates a method or process 800 illustrating steps for manufacturing, characterizing, and / or testing a semiconductor device according to one or more embodiments of the present disclosure. It is noted herein that the steps of the method or process 800 may be implemented in whole or in part by the system 700 illustrated in Figures 7A and 7B. However, it is further recognized that the method or process 800 is not limited to the system 700 illustrated in Figures 7A and 7B in that additional or alternative system level embodiments may perform all or a portion of the steps of the method or process 800.

[0095] In step 802, a semiconductor device is fabricated. In some embodiments, the semiconductor device (e.g., wafer 104) is fabricated through multiple semiconductor fabrication processes. For example, the semiconductor fabrication characterization subsystem 102 can include, but is not limited to, one or more process tools 708 configured to fabricate semiconductor devices including 1, 2, .... N layers are fabricated according to several steps (e.g., tens of thousands of times) performed by several semiconductor fabrication processes.

[0096] In step 804, characterization measurements are obtained during the manufacturing of the semiconductor devices. In some embodiments, the characterization measurements are obtained by the semiconductor manufacturing characterization subsystem 102. For example, characterization measurements may be performed by multiple characterization tools (e.g., inspection tool 702 and / or metrology tool 706) during (e.g., before, during, and / or after) the manufacturing of one or more semiconductor devices (e.g., wafers 104) through multiple semiconductor manufacturing processes performed by multiple process tools 708.

[0097] In step 806, the semiconductor devices are provided to an electrical test subsystem. In some embodiments, the electrical test subsystem 106 receives a lot of wafers 104. For example, the electrical test subsystem 106 may perform electrical tests and / or stress tests to generate test data 109.

[0098] In step 808, the characterization measurements are sent to the defect reduction subsystem or the defect induced correlation subsystem. In some embodiments, the system 700 is configured to overlay the characterization data 404 and / or the filtered characterization data 406 on the electrical test bin data 112 via one or more steps of the method or process 500. For example, the defect reduction subsystem 402 may be configured to receive the characterization data 404 and generate the filtered characterization data 406 via one or more steps of the method or process 500 (e.g., performed by one or more components of the system 400). As another example, the defect induced correlation subsystem 408 may be configured to receive the characterization data 404 and / or the filtered characterization data 406. The defect induced correlation subsystem 408 may be configured to overlay the characterization data 404 and / or the filtered characterization data 406 on the electrical test bin data 112.

[0099] At step 810, a control signal for adjustment is generated based on the reclassified electrical test bin data. In some embodiments, following overlay characterization data 404 and / or filtered characterization data 406 on electrical test bin data 112, at least some of the electrical test bin data 112 are reclassified by defect-induced correlation subsystem 408. In addition, defects may be newly discovered by defect-induced correlation subsystem 408 based on the overlay. It is noted herein that one or more adjustments may be determined for at least one of the manufacturing, characterization, and / or testing of semiconductor devices based on the reclassification of electrical test bin data 112 and / or defects newly discovered during execution of method or process 500 or 800. For example, one or more adjustments may modify a manufacturing process or method, a characterization process or method, a testing process or method, or the like, provided in a feedback loop to components within semiconductor manufacturing characterization subsystem 102. For example, a manufacturing process or method, a characterization process or method, a testing process or method, etc. may be adjusted (e.g., via one or more control signals) based on newly discovered electrical test bin data 112 and / or reclassification of defects during performance of the method or process 500 or 800.

[0100] The adjustments are sent via a feedback loop (e.g., to adjust future semiconductor devices). Control signals may adjust components of system 100 or 400 and corresponding methods or processes based on the reclassified electrical test bin data 112. For example, improvements may be directed to adjusting one or more components of system 100 and / or steps of method or process 200. For example, improvements may be directed to adjusting one or more components of semiconductor manufacturing characterization subsystem 102. As another example, improvements may be directed to adjusting one or more components of system 400 and / or steps of method or process 500. In this regard, the manufacturing and / or characterization process may be improved, leading to reduced costs (e.g., in time, money, etc.) for the manufacturer while maintaining a desired quality level (e.g., PPB failure rate).

[0101] It should be noted herein that the methods or processes 200, 500, and 800 are not limited to the steps and / or sub-steps provided. The methods or processes 200, 500, and 800 may include more or fewer steps and / or sub-steps. The methods or processes 200, 500, and 800 may perform steps and / or sub-steps simultaneously. The methods or processes 200, 500, and 800 may perform steps and / or sub-steps sequentially, including in the order provided or in an order other than that provided. Thus, the above description should not be construed as a limitation on the scope of the present disclosure, but merely as an example.

[0102] In one non-limiting example of the systems and methods described throughout this disclosure, for reliability sensitive devices, the semiconductor manufacturing characterization subsystem 102 may begin screening tests with 4-8 critical inspection steps, which are performed on every die of every wafer 104 of every lot for a given semiconductor device, to obtain characterization data 404. The characterization data 404 may be automatically forwarded to a fab-wide defect management subsystem 414 and / or a defect reduction subsystem 402 (e.g., an I-PAT Analyzer, etc.), which may weigh and aggregate the defectivity to arrive at a die-based defectivity score, which is forwarded to an appropriate fab database as filtered characterization data 406.

[0103] After completion of the manufacturing process, the wafer 104 may undergo wafer sort electrical testing and singulation via the electrical test subsystem 106. After singulation, the die are packaged and undergo a number of electrical and stress tests to generate test data 109. Following all tests, statistical outlier algorithms are applied to the test data 109 (e.g., but not limited to, Z-PAT) by the statistical outlier detection subsystem 110. When an instance of a Z-PAT outlier is identified by the statistical outlier detection subsystem 110, the electrical test bin data 112 for the corresponding die is sent to the defect induction correlation subsystem 408 for analysis.

[0104] The defect induction correlation subsystem 408 may overlay the electrical test bin data 112 with the characterization data 404 and / or the filtered characterization data 406. Based on the overlay, the defect induction correlation subsystem 408 may determine whether defects were correctly found in the electrically failing die 304 on the wafer 104 by the electrical test subsystem 106, whether the electrical test subsystem 106 missed defects on the selected wafer 104 by declaring the corresponding die to be a good die 302, or whether the electrical test subsystem 106 mischaracterized the wafer 104 as having a presumed electrically failing die 306. The defect induction correlation subsystem 408 may determine whether the ink-out die is valid or not and provide that information to the relevant parties (e.g., as improved electrical test die bin data 410) or at least to the statistical outlier detection subsystem 110 (e.g., as reclassified electrical test die bin data 412).

[0105] In this regard, the systems and methods of the present disclosure may provide increased sampling (e.g., 100% inspection of all wafers in all lots as opposed to 10% inspection of lots at 3 wafers / lot, or other subsets of wafers and lots) while improving electrical testing and / or defect testing through identification of potential reliability and / or test gap defects. The systems and methods of the present disclosure may provide improved insights that help enable automotive semiconductor device manufacturers to reduce reliability failures from the PPM to PPB range. Semiconductor failures are a one-failure item number for automotive manufacturing, and this issue becomes more intense as semiconductor content for automobiles grows (e.g., with the implementation of autonomous driving and electric vehicles). Similarly, reliability concerns are also becoming increasingly important in industrial, biomedical, defense, aerospace, hyperscale data centers, and the like. Identifying test coverage gaps brings awareness of limitations of electrical test methods, thus driving the adoption of in-line defect screening inspection to mitigate these issues.

[0106] Advantages of the present disclosure are directed to systems and methods for Z-PAT defect induced statistical outlier detection of semiconductor reliability failures. Advantages of the present disclosure also relate to identifying Z-PAT defect signatures indicative of potential reliability and / or test gap defects at the same x,y location across multiple wafers in a lot using characterization data (e.g., in-line defect inspection data and / or metrology data). Embodiments of the present disclosure use statistical outlier algorithms. Advantages of the present disclosure are directed to systems and methods for Z-PAT defect induced statistical outlier detection of semiconductor reliability failures. Advantages of the present disclosure also relate to automatically notifying manufacturing engineers of the presence of Z-PAT defect signatures. Embodiments of the present disclosure relate to characterizing Z-PAT defects using spatial signature analysis methods for the signatures. Advantages of the present disclosure relate to characterizing Z-PAT defect signatures using machine learning methods. Advantages of the present disclosure relate to identifying the presence or absence of Z-PAT defect signatures within a given lot. Advantages of the present disclosure relate to identifying Z-PAT defect signatures in adjacent lots. Advantages of the present disclosure relate to Z-PAT defect signature identification that was not detectable by electrical test-based Z-PAT. Advantages of the present disclosure relate to reduced overkill to more accurately summarize the amount of impacted die / wafers using Z-PAT defect signatures. Advantages of the present disclosure relate to rapid identification of underlying roots triggered based on learning from previously characterized Z-PAT defect signatures. Advantages of the present disclosure relate to retroactive identification of Z-PAT defect signatures that are directly warranted and / or recalled using stored in-line defect data.

[0107] The subject matter described herein illustrates different components that are sometimes included within or connected to other components. It should be understood that such depicted architectures are merely exemplary, and that in fact many other architectures that achieve the same functionality may be implemented. In a conceptual sense, any arrangement of components to achieve the same functionality is effectively "associated" such that the desired functionality is achieved. Thus, any two components herein that are combined to achieve a particular functionality can be considered to be "associated" with each other such that the desired functionality is achieved, regardless of the architecture or intermediate components. Similarly, any two components so associated can also be considered to be "connected" or "coupled" with each other to achieve the desired functionality, and any two components capable of being so associated can also be considered to be "couplable" with each other to achieve the desired functionality. Specific examples of components that can be coupled include, but are not limited to, physically interactable and / or physically interacting components and / or wirelessly interactable and / or wirelessly interacting components and / or logically interactable and / or logically interacting components.

[0108] It is believed that the present disclosure and many of its attendant advantages will be understood from the foregoing description, and it will be apparent that various changes can be made in the form, construction and arrangement of the elements without departing from the disclosed subject matter or sacrificing all of its material advantages. The forms described are merely illustrative, and it is the intent of the following claims to embrace and include such modifications. It is to be understood, further, that the invention is defined by the appended claims.

Claims

1. 1. A system comprising: A controller communicatively coupled to at least the semiconductor manufacturing characterization subsystem, the controller including one or more processors configured to execute program instructions, the one or more processors comprising: receiving electrical test bin data via a defect induction correlation subsystem, the electrical test bin data including semiconductor die data for a plurality of wafers in a lot, the electrical test bin data generated by a statistical outlier detection subsystem configured to perform Z-PAT on the test data, the electrical test subsystem configured to generate the test data by testing the plurality of wafers in the lot after fabrication by a semiconductor manufacturing characterization subsystem; receiving characterization data via the defect-induced correlation subsystem, the characterization data for a plurality of wafers in a lot being generated by a semiconductor manufacturing characterization subsystem during manufacturing of the plurality of wafers in the lot; determining a statistical correlation between the electrical test bin data and the characterization data via a defect-induced correlation subsystem at the same x,y location on each of a plurality of wafers in the lot; identifying locations of the defect data signatures on a plurality of wafers within the lot based on the statistical correlation via the defect-induced correlation subsystem; system.

2. 2. The system of claim 1, The one or more processors are further configured to execute program instructions that cause the one or more processors to: processing the characterization data through a defect reduction subsystem to generate a subset of the characterization data as filtered characterization data, the defect reduction subsystem configured to perform an in-line defect partial average test (I-PAT), the defect reduction subsystem configured to generate the filtered characterization data by performing the I-PAT on the characterization data before a defect-induced correlation subsystem receives the filtered characterization data; system.

3. 2. The system of claim 1, The one or more processors are further configured to execute program instructions that cause the one or more processors to: overlaying the characterization data on the electrical test bin data to determine a statistical correlation between the electrical test bin data and the characterization data via the defect-induced correlation subsystem, the overlaying of the characterization data on the electrical test bin data being performed at the same x, y location on each of a plurality of wafers in the lot; system.

4. 2. The system of claim 1, The one or more processors are further configured to execute program instructions that cause the one or more processors to: reclassifying at least some of the semiconductor die data within the electrical test bin data based on the defect data signature via the defect-induced correlation subsystem; system.

5. 5. The system of claim 4, at least some of the semiconductor die data are reclassified as being good die, as being known electrically failed die, or as being potentially electrically failed die. system.

6. 6. The system of claim 5, The one or more processors are further configured to execute program instructions that cause the one or more processors to: transmit the reclassified semiconductor die data via the defect induction correlation subsystem as improved electrical test bin data, the improved electrical test bin data including one or more recommendations for inking out the reclassified semiconductor die data as either good dies or potentially electrically failing dies on selected wafers of a plurality of wafers in the lot having identical x,y locations as known electrically failing dies on other wafers of a plurality of wafers in the lot; system.

7. 7. The system of claim 6, The one or more processors are further configured to execute program instructions that cause the one or more processors to: sending the reclassified semiconductor die data via the defect induced correlation subsystem to a statistical outlier detection subsystem; system.

8. 5. The system of claim 4, The one or more processors are further configured to execute program instructions that cause the one or more processors to: determining one or more adjustments to at least one of manufacturing, characterizing, and / or testing a subsequent plurality of wafers in a subsequent lot based on the re-sorted semiconductor die data in the electrical test bin data; system.

9. 9. The system of claim 8, The one or more processors are further configured to execute program instructions that cause the one or more processors to: generating one or more control signals based on one or more adjustments to at least one of the fabrication, characterization, and / or testing of a subsequent plurality of wafers in a subsequent lot; system.

10. 2. The system of claim 1, the controller is communicatively coupled to the electrical test subsystem; system.

11. 1. A method comprising: receiving electrical test bin data via a defect induction correlation subsystem, the electrical test bin data including semiconductor die data for a plurality of wafers in a lot, the electrical test bin data being generated by a statistical outlier detection subsystem configured to perform Z-PAT on the test data, the electrical test subsystem being configured to generate the test data by testing the plurality of wafers in the lot after fabrication by the semiconductor manufacturing characterization subsystem; receiving characterization data via the defect-induced correlation subsystem, the characterization data for a plurality of wafers in a lot being generated by a semiconductor manufacturing characterization subsystem during manufacturing of the plurality of wafers in the lot; determining a statistical correlation between the electrical test bin data and the characterization data via the defect-induced correlation subsystem at the same x,y location on each of a plurality of wafers in the lot; identifying locations of the defect data signatures on a plurality of wafers within the lot based on the statistical correlation via the defect-induced correlation subsystem; The method includes:

12. 12. The method of claim 11, processing the characterization data through a defect reduction subsystem to generate a subset of the characterization data as filtered characterization data, the defect reduction subsystem configured to perform an in-line defect-partial-average test (I-PAT), the defect reduction subsystem configured to generate the filtered characterization data by performing the I-PAT on the characterization data before the defect-induced correlation subsystem receives the filtered characterization data. method.

13. 12. The method of claim 11, overlaying the characterization data with the electrical test bin data to determine a statistical correlation between the electrical test bin data and the characterization data via a defect-guided correlation subsystem, wherein overlaying the characterization data with the electrical test bin data is performed at the same x, y location on each of a plurality of wafers in the lot. method.

14. 12. The method of claim 11, reclassifying at least some of the semiconductor die data within the electrical test bin data based on the defect data signature via the defect-induced correlation subsystem. method.

15. 15. The method of claim 14, at least some of the semiconductor die data are reclassified as being good die, as being known electrically failed die, or as being potentially electrically failed die. method.

16. 16. The method of claim 15, transmitting the reclassified semiconductor die data as improved electrical test bin data via the defect induction correlation subsystem, the improved electrical test bin data including one or more recommendations for inking out the reclassified semiconductor die data as either good dies or potentially electrically failing dies on selected wafers of a plurality of wafers in the lot having identical x,y locations as known electrically failing dies on other wafers of a plurality of wafers in the lot. method.

17. 17. The method of claim 16, transmitting the reclassified semiconductor die data to a statistical outlier detection subsystem via a defect-induced correlation subsystem; method.

18. 15. The method of claim 14, determining one or more adjustments to at least one of manufacturing, characterizing, and / or testing a subsequent plurality of wafers in a subsequent lot based on the re-sorted semiconductor die data in the electrical test bin data; method.

19. 20. The method of claim 18, generating one or more control signals based on one or more adjustments to at least one of the manufacturing, characterization, and / or testing of a subsequent plurality of wafers in a subsequent lot; method.

20. 1. A system comprising: a semiconductor manufacturing characterization subsystem configured to fabricate a plurality of wafers in a lot, the semiconductor manufacturing characterization subsystem configured to generate characterization data for the plurality of wafers in the lot during fabrication of the plurality of wafers in the lot; an electrical testing subsystem configured to generate test data for the plurality of wafers in the lot after production by the semiconductor manufacturing characterization subsystem; A controller communicatively coupled to at least the semiconductor manufacturing characterization subsystem, the controller including one or more processors configured to execute program instructions, the one or more processors comprising: receiving electrical test bin data via a defect induced correlation subsystem, the electrical test bin data including semiconductor die data for a plurality of wafers in the lot, the electrical test bin data being generated by a statistical outlier detection subsystem configured to perform Z-PAT; receiving characterization data via the defect induced correlation subsystem; determining a statistical correlation between the electrical test bin data and the characterization data via the defect induced correlation subsystem at a same x,y location on each of a plurality of wafers in the lot; and locating defect data signatures on the plurality of wafers in the lot based on the statistical correlation via the defect induced correlation subsystem. system.

21. 21. The system of claim 20, The one or more processors are further configured to execute program instructions that cause the one or more processors to: processing the characterization data through a defect reduction subsystem to generate a subset of the characterization data as filtered characterization data, the defect reduction subsystem configured to perform an in-line defect partial average test (I-PAT), the defect reduction subsystem configured to generate the filtered characterization data by performing the I-PAT on the characterization data before a defect-induced correlation subsystem receives the filtered characterization data; system.

22. 21. The system of claim 20, The one or more processors are further configured to execute program instructions that cause the one or more processors to: overlaying the characterization data on the electrical test bin data to determine a statistical correlation between the electrical test bin data and the characterization data via the defect-induced correlation subsystem, the overlaying of the characterization data on the electrical test bin data being performed at the same x, y location on each of a plurality of wafers in the lot; system.

23. 21. The system of claim 20, The one or more processors are further configured to execute program instructions that cause the one or more processors to: reclassifying at least some of the semiconductor die data within the electrical test bin data based on the defect data signature via the defect-induced correlation subsystem; system.

24. 24. The system of claim 23, at least some of the semiconductor die data are reclassified as being good die, as being known electrically failed die, or as being potentially electrically failed die. system.

25. 25. The system of claim 24, The one or more processors are further configured to execute program instructions that cause the one or more processors to: transmit the reclassified semiconductor die data via the defect induction correlation subsystem as improved electrical test bin data, the improved electrical test bin data including one or more recommendations for inking out the reclassified semiconductor die data as either good dies or potentially electrically failing dies on selected wafers of a plurality of wafers in the lot having identical x,y locations as known electrically failing dies on other wafers of a plurality of wafers in the lot; system.

26. 26. The system of claim 25, The one or more processors are further configured to execute program instructions that cause the one or more processors to: sending the reclassified semiconductor die data via the defect induced correlation subsystem to a statistical outlier detection subsystem; system.

27. 24. The system of claim 23, The one or more processors are further configured to execute program instructions that cause the one or more processors to: determining one or more adjustments to at least one of manufacturing, characterizing, and / or testing a subsequent plurality of wafers in a subsequent lot based on the re-sorted semiconductor die data in the electrical test bin data; system.

28. 28. The system of claim 27, The one or more processors are further configured to execute program instructions that cause the one or more processors to: generating one or more control signals based on one or more adjustments to at least one of the fabrication, characterization, and / or testing of a subsequent plurality of wafers in a subsequent lot; system.

29. 21. The system of claim 20, the controller is communicatively coupled to the electrical test subsystem; system.

Images