Die screening system and method

Abstract

The die screening system can receive die analysis performance metrology data for a population of dies on one or more samples from one or more in-line metrology tools after one or more fabrication steps, the die analysis performance metrology data including images generated using one or more measurement configurations of the in-line metrology tools. Thus, the die analysis performance metrology data provides multiple measurement channels per die, with a particular measurement channel including data from a particular pixel of a particular image. The controller can then generate screening data for the population of dies from the die analysis performance metrology data, the screening data including a subset of the multiple channels for measuring the die analysis performance metrology data, and screen the plurality of dies into at least two placement classes including at least an abnormal die according to variability in the screening data.

Description

[Technical Field]

[0001] REFERENCE TO RELATED APPLICATIONS This application claims the benefit under 35 U.S.C. §119(e) of U.S. Provisional Application No. 63 / 165,155 (filed March 24, 2021, for IMAGING REFLECTOMETRY FOR INLINE SCREENING (IRIS)), which is incorporated herein by reference in its entirety.

[0002] The present disclosure relates generally to in-line outlier screening of semiconductor devices, and more particularly to in-line screening using imaging reflectometry. [Background technology]

[0003] Fabrication of semiconductor devices typically requires numerous processing steps to produce functional devices. Risk-bar users of semiconductor devices, such as those in automotive, military, aerospace, and medical applications, require increasingly stringent reliability targets. For example, expected target failure rates in the parts-per-billion (PPB) range are desirable, but far below current levels. Advanced screening techniques for identifying defective die within samples are needed to meet these stringent reliability requirements. However, typical screening techniques for identifying die to be removed from the supply chain, such as electrical testing of all or a portion of the die, can have insufficient throughput. The term "sample" will be referred to hereafter as "specimen" where appropriate. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] U.S. Patent No. 10,761,128 [Patent Document 2] U.S. Patent No. 9,710,728 [Patent Document 3] U.S. Patent No. 10,352,876 Summary of the Invention [Problem to be solved by the invention]

[0005] It is therefore desirable to provide a system and method for efficient screening. [Means for solving the problem]

[0006] A die screening system is disclosed in accordance with one or more exemplary embodiments of the present disclosure. In one exemplary embodiment, the system includes a controller communicatively coupled to one or more in-line metrology tools. In another exemplary embodiment, the controller receives die-resolved metrology data for a population of dies on one or more samples from the one or more in-line metrology tools after one or more fabrication steps. The die-resolved metrology data can include one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, where the die-resolved metrology data provides multiple measurement channels per die, and a particular measurement channel includes data from a particular pixel of a particular image. According to one or more exemplary embodiments of the present disclosure, the controller generates screening data for the multiple dies from the die-resolved metrology data, where the screening data includes a subset of the measurement channels of the die-resolved metrology data, and the measurement channels of the screening data are sensitive to process variations associated with the fabrication of the one or more samples. In another exemplary embodiment, the controller screens the plurality of dies into two or more placement classes based on variability in the screening data, the two or more placement classes including at least a set of pass dies and a set of outlier dies.

[0007] A die screening system is disclosed in accordance with one or more exemplary embodiments of the present disclosure. In one exemplary embodiment, the system includes one or more in-line metrology tools and a controller. In another exemplary embodiment, the controller receives die analysis metrology data for a population of dies on one or more samples from one or more in-line metrology tools after one or more fabrication steps. The die analysis metrology data can include one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, where the die analysis metrology data provides multiple measurement channels per die, and a particular measurement channel includes data from a particular pixel of a particular image. According to one or more exemplary embodiments of the present disclosure, the controller generates screening data for the multiple dies from the die analysis metrology data, where the screening data includes a subset of the measurement channels of the die analysis metrology data, and the measurement channels of the screening data are sensitive to process variations associated with the fabrication of the one or more samples. In another exemplary embodiment, the controller screens the multiple dies into two or more placement classes based on variability in the screening data, where the two or more placement classes include at least a set of pass dies and a set of outlier dies.

[0008] A die screening method is disclosed in accordance with one or more exemplary embodiments of the present disclosure. In one exemplary embodiment, the method includes generating die analysis metrology data for a population of dies on one or more samples from one or more in-line metrology tools after one or more fabrication steps. The die analysis metrology data can include one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, where the die analysis metrology data provides multiple measurement channels per die, and a particular measurement channel includes data from a particular pixel of a particular image. In another exemplary embodiment, the method includes generating screening data for a plurality of dies from the die analysis metrology data, where the screening data includes a subset of the multiple measurement channels of the die analysis metrology data, and the measurement channel in the screening data is sensitive to process variations associated with the fabrication of the one or more samples. In another exemplary embodiment, the method includes screening the plurality of dies into two or more placement classes based on variability in the screening data, where the two or more placement classes include at least a set of pass dies and a set of outlier dies. [Effects of the Invention]

[0009] 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. [Brief explanation of the drawings]

[0010] The many advantages of the present disclosure may be better understood by those skilled in the art by reference to the accompanying drawings. [Figure 1A] FIG. 1A is a conceptual diagram of a die screening system according to one or more embodiments of the present disclosure. [Figure 1B] FIG. 1B is a conceptual top view of a sample including multiple dies, in accordance with one or more embodiments of the present disclosure. [Figure 1C] FIG. 1C is a conceptual diagram of an in-line metrology tool configured as an imaging reflectometer, in accordance with one or more embodiments of the present disclosure. [Figure 1D] FIG. 1D is a conceptual diagram of an in-line metrology tool configured as an imaging reflectometer providing six measurement channels, in accordance with one or more embodiments of the present disclosure. [Figure 2] FIG. 2 is a flow diagram illustrating steps performed in a method for die screening based on parametric variability, in accordance with one or more embodiments of the present disclosure. [Figure 3] FIG. 3 is an image of a sample with a single film produced using an imaging reflectometer in accordance with one or more embodiments of the present disclosure. [Figure 4] FIG. 4 is a block diagram illustrating aggregate die-level outlier screening using parametric variability screening combined with I-PAT screening, in accordance with one or more embodiments of the present disclosure. [Figure 5] FIG. 5 is a block diagram illustrating multiple interaction paths between traditional metrics and parametric variation screening, in accordance with one or more embodiments of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0011] Reference will now be made in detail to the disclosed subject matter, which is illustrated in the accompanying drawings. The present disclosure has been particularly shown and described with reference 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.

[0012] Embodiments of the present disclosure are directed to systems and methods for die placement (e.g., screening) based on parametric variability using multi-channel die analysis metrology data from one or more in-line metrology measurements that provide complete sampling of the die (e.g., from multiple measurements to multiple measurements). Die placement may include sorting fabricated die on one or more samples as pass-thru die suitable for integration into the supply chain, outlier die unsuitable for integration into the supply chain, borderline die requiring further testing, etc. For purposes of this disclosure, the term "die resolved" is used to refer to metrology measurements of a die in which the die and various features thereon are resolved. In this manner, multiple measurements are generated for each die. For example, die analysis metrology data may include images of the die in a population generated by one or more in-line metrology tools after one or more processing steps during the fabrication of the die. In this manner, each pixel in the die image may correspond to a measurement of a different location on the die and, therefore, to an imaging channel. Furthermore, any number of in-line metrology tools using any number of different measurement configurations can be used to generate the various images after any number of processing steps, so that the die-analyzed metrology data can include many (e.g., up to a million orders of magnitude or more) measurement channels, each corresponding to a particular pixel of a particular image generated by a particular in-line metrology tool after a particular processing step.

[0013] While various strategies have been developed to monitor or control the reliability of devices (e.g., fabricated die) entering a supply or distribution chain, it is recognized herein that existing strategies are insufficient to meet stringent reliability requirements (e.g., part / billion failure rates) while maintaining acceptable throughput.

[0014] For example, electrical testing of a die is typically performed to evaluate the functionality of one or more aspects of the die as data for reliability analysis. This electrical testing can be performed at any point in the manufacturing process, including, but not limited to, pre-burn-in electrical wafer sort and final test (e.g., e-test) or post-burn-in electrical testing. Devices that fail the electrical testing step can be "inked off" and isolated from the remaining devices in the population. However, electrical testing alone may not provide sufficient information to meet stringent reliability standards while maintaining cost and throughput targets. For example, post-burn-in electrical testing may provide an accurate analysis of die functionality because the die is in a near-final state, but may not be practical in high volume due to cost, time requirements, or the potential for introducing long-term reliability issues. Furthermore, electrical testing during any step of production generally only provides pass / fail information suitable for identifying devices that have already exhibited complete or partial failure, and may not be suitable for identifying devices that may fail later (e.g., devices with latent defects). Additionally, it is often impractical, or in some cases impossible, to fully characterize each die using electrical testing, resulting in gaps in electrical testing. For example, there may be theoretically possible defects in a particular circuit layout that may not be detectable using electrical testing, even with a "complete" testing strategy.

[0015] As another example, part average testing (PAT) techniques, such as, but not limited to, traditional PAT, parametric PAT (P-PAT), geographic PAT (G-PAT), or composite PAT (C-PAT), may utilize statistical methods to identify die with abnormal characteristics (e.g., outlier die) based on identifying defects within the die. For example, PAT methods may statistically analyze the defects or defect activity of die within a common sample or sample lot to identify outlier die. However, PAT methods are susceptible to "overkill," in which good die that would otherwise have a low probability of premature life failure are inappropriately inked, resulting in reduced production line throughput and increased cost per device.

[0016] As another example, existing inline screening techniques utilize various inline defect inspection or metrology measurements after one or more fabrication steps for die screening. For example, inline PAT (I-PAT) extends PAT technology to consider defects identified through inline sample analysis measurements. This I-PAT technique, referred to as inline part average test screening technology, is generally described in U.S. Patent 10,761,128 (September 1, 2020) and U.S. patent application Ser. No. 17 / 101,856 (November 23, 2020), both of which are incorporated herein by reference in their entireties. However, the inline measurements utilized by I-PAT techniques are typically provided by measurements performed for process control purposes and may not provide adequate data for robust die screening. In particular, I-PAT techniques typically utilize the results of dedicated inline measurements of specific aspects of the die or sample to identify defects for screening purposes. These defects can be broadly of two types: First, in the form of objects or patterns that should not be present (e.g., embedded particles, bridges, gaps, etc.), or second, in the form of parameter deviations of an otherwise properly fabricated sample layer or features thereon (e.g., variations in measurements of overlay (overlay measurements), critical dimension (CD), film composition, film thickness, etc.).

[0017] It is contemplated herein that reliance on dedicated measurements of specific parameters for in-die screening may be insufficient to provide robust and sensitive screening at acceptable throughput. In particular, the first type of defect is readily amenable to high-throughput optical inspection. For example, optical inspection data of a test die can be compared to reference data to identify such defects. However, the second type of defect is typically measured at dedicated locations on the specimen (e.g., dedicated targets or selected sampling locations). When higher sampling density data is desired, measurements made at the sampling locations are typically interpolated.

[0018] It is further contemplated herein that in-line die analysis metrology measurements generated after one or more process steps can provide a rich data set for die screening based on parametric variability. Die analysis metrology measurements can be generated using a variety of metrology tools, including, but not limited to, optical metrology tools, X-ray metrology tools, or particle-based metrology tools (e.g., electron beam (e-beam) metrology tools, ion beam metrology tools, etc.). In some embodiments, parametric variability screening is performed using imaging reflectometry data. For example, an imaging reflectometer may, but need not, have multiple imaging channels to provide multi-channel images (e.g., bright-field images at one or more wavelengths, dark-field images at one or more wavelengths, etc.). However, it should be understood that the systems and methods disclosed herein can be extended to any type of die analysis metrology technique.

[0019] Furthermore, die analysis metrology measurements used herein for die screening may differ from those used for alternative in-line screening techniques, such as I-PAT, in that die analysis metrology measurements may not necessarily provide as much data as traditional, dedicated in-line metrology measurements. For example, SpectraShape measurements offered by KLA Corp. offer multiple wavelengths, polarizations, or the like, and can provide highly sensitive measurements of a wide range of sample geometries through robust models that can handle significant variability and allow many parameters (e.g., 16 or more parameters) to float in the measurement. In contrast, die analysis metrology measurements used herein for die screening may provide relatively few configurations or measurement channels. However, it is contemplated herein that die screening based on parametric variability may rely on the relative variation between dies rather than the specific value of any metrology measurement. In this way, the amount of data generated by die analysis optical metrology datasets can be balanced with measurement throughput to provide highly sensitive die screening and achieve high throughput. Furthermore, die screening based on parametric variability through die analysis metrology measurements as disclosed herein can be combined with complementary die screening techniques, such as, but not limited to, I-PAT, to provide robust die screening.

[0020] In some embodiments, only a subset of the die analysis metrology data for a population of dies is used for die screening. For example, it may be the case that meaningful parametric variations between various dies can be determined using a subset of available measurement channels. Therefore, various techniques may be utilized to generate screening data that includes a subset of die analysis metrology data (e.g., a subset of available measurement channels) for die screening. For example, the screening data may be generated based on inclusion of measurement channels with known or expected variability to process variations, selection of regions of interest, dimension reduction techniques, etc.

[0021] In some embodiments, a single measurement per die is sufficiently representative of the die, and die screening can be performed through empirical analysis of this measurement for the die of interest. In some embodiments, screening is performed by generating one or more metrology measurements of process parameters and screening based on the process parameter measurements. The process parameter measurements may include, but are not limited to, measurements of overlay, CD, film thickness, film composition, film refractive index, film uniformity, feature height, sidewall angle, lithography exposure parameters (e.g., focus position, dose, energy density, etc.), or chemical mechanical planarization (CMP) polishing time. For example, the process parameter metrology measurements may be generated using any of a variety of model-based or model-less (e.g., machine learning-based) approaches. In some embodiments, die screening is performed by machine learning-based outlier analysis, in which all of the die analysis optical metrology data for the die is analyzed to determine die-level parameter metrics for the die of interest. For example, machine learning outlier analysis can include dimensionality reduction to determine which of many measurements provided by an optical metrology tool or a combination thereof are indicative of die-to-die parameter variation, and can further generate die-level parameter metrics based on this dimensionality reduction.

[0022] Furthermore, screening may be performed using only screening data associated with a single process step, or may be based on differential signals associated with different process steps, different in-line metrology tools, and / or simulated data. For example, die screening may be performed based on differential signals between screening data associated with a particular die and one or more additional dies. As another example, die-level parameter metrics may be generated based on differential signals between a particular die and simulated data. As another example, die-level parameter metrics may be generated based on differential signals associated with a particular die at different process steps. In this manner, the systems and methods disclosed herein may track signal variability between process steps. As another example, die-level parameter metrics may be generated based on differential signals for a particular die from different optical metrology tools.

[0023] Further embodiments of the present disclosure are directed to classifying die in a population for deposition. For example, die may be sorted into two or more classes, including, but not limited to, passing die, failing die, or borderline die requiring additional testing. Furthermore, die screening may be generated based solely on die-level parametric variations as disclosed herein, or using the systems and methods disclosed herein, or in combination with additional screening techniques. For example, a diaggregator can aggregate data related to the results of parameter variability screening or any step in the screening process with additional screening data from additional screening techniques to deposit die based on the aggregated data. As another example, die-level parameter metrics may be used to supplement additional screening techniques, such as, but not limited to, the I-PAT technique. For example, defects in die in a population identified using the I-PAT technique can be at least partially weighted using data related to the results of parameter variability screening or any step in the process.

[0024] 1A-3, systems and methods for in-line parametric variability screening, according to one or more embodiments of the present disclosure, are described in more detail.

[0025] 1A is a conceptual diagram of a die screening system 100 in accordance with one or more embodiments of the present disclosure. FIG. 1B is a conceptual top view of a sample 102 including multiple dies 104 in accordance with one or more embodiments of the present disclosure. For example, the various dies 104 may be separated by scribe lines 106 such that multiple devices on the various dies 104 may be fabricated simultaneously on the sample 102 and later separated (e.g., in a dicing step) to provide distinct devices.

[0026] In one embodiment, the die screening system 100 includes at least one in-line metrology tool 108 (e.g., an in-line sample analysis tool) for generating die analysis metrology measurements related to one or more layers of the specimen 102 after one or more process steps. The die screening system 100 may generally include any number or type of in-line metrology tools 108. For example, the in-line metrology tool 108 may include an optical metrology tool configured to perform metrology measurements based on interrogation of the specimen 102 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 in-line metrology tool 108 may include a particle beam metrology tool configured to perform metrology measurements based on interrogation of the specimen with one or more particle beams, such as, but not limited to, an electron beam, an ion beam, or a neutral particle beam.

[0027] Additionally, die-analyzable metrology measurements may include any type of data that provides multiple measurements of the die 104 on the sample 102. In one embodiment, the die-analyzed metrology data includes imaging data, which may be generated in one or more data capture steps. For example, the in-line metrology tool 108 may include or be configured to provide an image of the die 104 (or a portion thereof) within a field of view in a particular data capture step. Die-analyzable data related to the die 104 or the sample 102 as a whole may then be generated by stitching or otherwise combining the multiple images. As another example, the in-line metrology tool 108 may generate die-analyzable data point-wise (e.g., at periodic intervals or in a selected measurement pattern).

[0028] FIG. 1C is a conceptual diagram of an in-line metrology tool 108 configured as an imaging reflectometer, in accordance with one or more embodiments of the present disclosure.

[0029] In one embodiment, the in-line metrology tool 108 includes an illumination source 110 configured to generate at least one illumination beam 112. The illumination from the illumination source 110 may include one or more selected wavelengths of light, including, but not limited to, ultraviolet (UV) radiation, visible light, or infrared (IR) radiation.

[0030] The illumination source 110 may include any type of illumination source suitable for providing at least one illumination beam 112. In one embodiment, the illumination source 110 is a laser source. For example, the illumination source 110 may include, but is not limited to, one or more narrowband laser sources, broadband laser sources, supercontinuum laser sources, white light laser sources, or the like. In this regard, the illumination source 110 may provide an illumination beam 112 having high coherence (e.g., high spatial coherence and / or temporal coherence). In another embodiment, the illumination source 110 includes a laser-sustained plasma (LSP) source. For example, the illumination source 110 may include, but is not limited to, an LSP lamp, an LSP bulb, or an LSP chamber suitable for housing one or more elements capable of emitting broadband illumination when excited into a plasma state by a laser source. In another embodiment, the illumination source 110 includes a lamp source. For example, the illumination source 110 may include, but is not limited to, an arc lamp, a discharge lamp, an electrodeless lamp, etc. In this regard, the illumination source 110 may provide an illumination beam 112 having low coherence (eg, low spatial and / or temporal coherence).

[0031] In another embodiment, the in-line metrology tool 108 directs the illumination beam 112 to the sample 102 via an illumination path 114. The illumination path 114 may include one or more optical components suitable for modifying and / or conditioning the illumination beam 112 and for directing the illumination beam 112 to the sample 102. In one embodiment, the illumination path 114 includes one or more illumination path lenses 116 (e.g., to collimate the illumination beam 112, to relay a pupil plane and / or a field plane, etc.). In another embodiment, the illumination path 114 includes one or more illumination path optics 118 for shaping or otherwise controlling the illumination beam 112. For example, the illumination path optics 118 may include, but are not limited to, one or more field stops, one or more pupil stops, one or more polarizers, one or more filters, one or more beam splitters, one or more diffusers, one or more homogenizers, one or more apodizers, one or more beam shapers, or one or more mirrors (e.g., being static mirrors, translatable mirrors, scanning mirrors, etc.).

[0032] In another embodiment, the in-line metrology tool 108 includes an objective lens 120 that focuses the illumination beam 112 onto the sample 102 .

[0033] In another embodiment, the in-line metrology tool 108 includes one or more detectors 122 configured to capture light emanating from the sample 102 or other light (e.g., collected light 124) through a collection path 126. The collection path 126 may include one or more optical elements suitable for modifying and / or conditioning the collected light 124 from the sample 102. In one embodiment, the collection path 126 includes one or more collection path lenses 128 (e.g., to collimate the illumination beam 112, to relay a pupil plane and / or a field plane, etc.), which may, but need not, include the objective lens 120. In another embodiment, the collection path 126 includes one or more collection path optics 130 that shape or otherwise control the collected light 124. For example, the collection path optics 130 may include, but are not limited to, one or more field stops, one or more pupil stops, one or more polarizers, one or more filters, one or more beam splitters, one or more diffusers, one or more homogenizers, one or more apodizers, one or more beam shapers, or one or more mirrors (e.g., static mirrors, translatable mirrors, scanning mirrors, etc.).

[0034] The in-line metrology tool 108 can generally include any number or type of detectors 122 suitable for capturing light from the sample 102. In one embodiment, the detector 122 includes one or more sensors suitable for characterizing a static sample. In this regard, the in-line metrology tool 108 can operate in a static mode in which the sample 102 is static during measurement. For example, the detector 122 can include, but is not limited to, a two-dimensional pixel array such as a charge-coupled device (CCD) or a complementary metal-oxide semiconductor (CMOS) device. In this regard, the detector 122 can generate a two-dimensional image (e.g., a field plane image or a pupil plane image) in a single measurement. In another embodiment, the detector 122 includes one or more sensors suitable for characterizing a moving sample (e.g., a scanning sample). In this regard, the in-line metrology tool 108 can operate in a scanning mode in which the sample 102 is scanned relative to a measurement field during measurement. For example, the detector 122 may include a 2D pixel array having a capture time and / or refresh rate sufficient to capture one or more images during a scan within selected image tolerances (e.g., image blur, contrast, sharpness, etc.). As another example, the detector 122 may include a line-scan detector that sequentially generates images of one line of pixels at a time. As another example, the detector 122 may include a time-delay integration (TDI) detector in which the movement of the sample 102 is synchronized to a charge-transfer clock signal within the TDI detector. In another embodiment, the detector 122 includes a spectrometer or spectroscopic sensor suitable for providing wavelength-analysis data.

[0035] In another embodiment, the in-line metrology tool 108 includes a specimen stage 132 suitable for securing the specimen 102 and further configured to position the specimen 102 relative to the illumination beam 112. For example, the specimen stage 132 may include any combination of linear actuators, rotary actuators, or angular actuators (e.g., to control the tip and / or tilt of the specimen 102). In another embodiment, not shown, the die screening system 100 includes one or more beam scanning optics (e.g., rotatable mirrors, galvanometers, etc.) for scanning the illumination beam 112 relative to the specimen 102.

[0036] The illumination path 114 and collection path 126 of the in-line metrology tool 108 can be oriented in a wide variety of configurations suitable for illuminating the sample 102 with the illumination beam 112 and collecting light emanating from the sample 102 in response to the incident illumination beam 112. For example, as shown in FIG. 1C , the in-line metrology tool 108 can include a beam splitter 134 oriented such that a common objective lens 120 can simultaneously direct the illumination beam 112 onto the sample 102 and collect light from the sample 102. In this regard, the in-line metrology tool 108 can provide bright-field imaging of the sample 102, without limitation. For example, Figure 1C shows a first illumination source 110a providing on-axis illumination of the sample 102 (e.g., via a beam splitter 134) and on-axis collection of light from the sample 102 to provide bright-field imaging, and a second illumination source 110b providing illumination of the sample 102 outside the collection numerical aperture (NA) of the objective lens 120 to provide dark-field imaging. However, it should be understood that Figure 1C is provided for illustrative purposes only and should not be construed as limiting. For example, dark-field imaging may be implemented in a wide variety of configurations, including, but not limited to, using one or more apertures in the illumination and / or collection pupil planes.

[0037] 1C and 1D , the in-line metrology tool 108 can be configured as a multi-channel tool. In this manner, the die analysis metrology data includes multi-channel data associated with each location on the die 104. Furthermore, each channel can be associated with a different set of illumination conditions, collection conditions, wavelengths, etc.

[0038] FIG. 1D is a conceptual diagram of an in-line metrology tool 108 configured as an imaging reflectometer providing six measurement channels, in accordance with one or more embodiments of the present disclosure. In particular, FIG. 1D illustrates a configuration including bright-field and dark-field measurements at three wavelengths. For example, the in-line metrology tool 108 may include three detection channels 136a-c coupled to a first illumination beam 112a configured to illuminate the sample 102 within the collection NA of the objective lens 120 to provide bright-field imaging and a second illumination beam 112b configured to illuminate the sample 102 outside the collection NA of the objective lens 120 to provide dark-field imaging. The detection channels 136a-c may be formed using any technique known in the art, including, but not limited to, a series of dichroic beam splitters 138 as shown in FIG. 1D or wavelength-insensitive beam splitters coupled with spectral filters.

[0039] As previously described herein, the in-line metrology tool 108 can include any type of metrology tool suitable for generating die analysis metrology data, including, but not limited to, an X-ray metrology tool or a particle-based metrology tool. Accordingly, Figures 1C and 1D are provided for illustrative purposes only and should not be construed as limiting. For example, the in-line metrology tool 108 can include, but is not limited to, a hyperspectral imaging tool or an X-ray ptychography tool.

[0040] 1A , in one embodiment, the die screening system 100 includes a controller 140. The controller 140 may include one or more processors 142 configured to execute program instructions maintained on a memory 144 (e.g., a storage medium, a storage device, etc.). Additionally, the controller 140 may be communicatively coupled to any of the components of the die screening system 100, including, but not limited to, the in-line metrology tool 108 or any component therein. In this regard, the one or more processors 142 of the controller 140 may perform any of the various process steps described throughout this disclosure, such as, but not limited to, receiving die analysis metrology data, generating die-level parameter metrics based on the die analysis metrology data, or placing a die based at least in part on the die-level parameter metrics.

[0041] The one or more processors 142 of the controller 140 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 microprocessor devices, one or more application-specific integrated circuit (ASIC) 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 142 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 142 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 the die screening system 100, as described throughout this disclosure.

[0042] The memory 144 may include any storage medium known in the art suitable for storing program instructions executable by the associated one or more processors 142. For example, the memory 144 may include a non-transitory storage medium. As another example, the memory 144 may include, but is not limited to, read-only memory (ROM), random access memory (RAM), magnetic or optical memory devices (e.g., disks), magnetic tape, solid-state drives, etc. It is further noted that the memory 144 may be housed within a common controller housing along with one or more processors 142. In one embodiment, the memory 144 may be located remotely relative to the physical location of the one or more processors 142 and the controller 140. For example, one or more processors 142 of the controller 140 may access a remote memory (e.g., a server) accessible via a network (e.g., the Internet, an intranet, etc.).

[0043] In one embodiment, the user interface 146 is communicatively coupled to the controller 140. In one embodiment, the user interface 146 may include, but is not limited to, one or more desktops, laptops, tablets, etc. In another embodiment, the user interface 146 includes a display used to display data from the die screening system 100 to a user. The display of the user interface 146 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 will recognize that any display device that can be integrated with the user interface 146 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 146.

[0044] Referring now to Figure 2, Figure 2 is a flow diagram illustrating steps performed in a method 200 for die screening based on parametric variability, in accordance with one or more embodiments of the present disclosure. Applicant notes that the embodiments and enabling techniques previously described herein in the context of die screening system 100 should be construed as extending to method 200. However, it is further noted that method 200 is not limited to the architecture of die screening system 100.

[0045] In one embodiment, method 200 includes generating 202 multi-channel die analysis metrology data for dies on one or more samples from one or more in-line metrology tools after one or more fabrication steps, where the die analysis metrology data includes multiple measurement channels per die. For example, the die analysis metrology data may include one or more images of each die generated using one or more measurement configurations of the one or more in-line metrology tools. For example, the measurement configurations may include, but are not limited to, illumination incidence angle (e.g., azimuth and / or polar angle), collection angle (e.g., azimuth and / or polar angle), illumination polarization, collection polarization, illumination wavelength, or collection wavelength. In this manner, a particular measurement channel may correspond to data from a particular pixel in a particular image. In other words, a particular measurement channel may correspond to a particular location on the die (e.g., each pixel is associated with a different location on the die) and a particular measurement configuration of a particular in-line metrology tool.

[0046] It is contemplated herein that multi-channel die analysis metrology data can be distinguished from traditional in-line metrology techniques. For example, typical in-line metrology techniques, such as, but not limited to, those used for process control, can generate highly sensitive measurements of specific process parameters (e.g., overlay, CD, film thickness, film composition, film refractive index, film uniformity, feature height, sidewall angle, lithography exposure parameters, CMP polishing time, etc.) based on measurements of specific features on a sample, such as a dedicated metrology target or selected device feature. In some instances (e.g., target-based overlay), a dedicated target or set of targets may be constructed to provide deterministic metrology measurements based on specific measurements of the targets. In some cases, models, libraries, and / or machine learning techniques are used to relate measurements generated using specific measurement conditions to physical metrology parameters. In either case, in-line metrology tools generally may have many degrees of freedom, but measurements must be performed at highly controlled measurement conditions for specific sample locations. Thus, full-die (eg, die analysis) measurements are generated by interpolation of measurements made at fixed locations on the sample.

[0047] In contrast, multi-channel die analysis metrology data, as used herein, may be generated by an in-line metrology tool that provides measurements in a fixed sampling pattern at many locations across each die rather than at dedicated sampling locations. For example, an image of a die (which may be generated in a single capture or sequential captures) may be associated with measurements at different locations across the die. Furthermore, multiple die images may be generated using multiple measurement configurations or using multiple in-line metrology tools. In this manner, die analysis metrology data may include multiple channels, each associated with a particular pixel of a particular image and therefore representing measurement data at a particular location using particular measurement conditions.

[0048] Additionally, die analysis metrology data may include any type of image generated using any type of light collection scheme, including, but not limited to, full sample images, grid level images, or die level images. Further, die analyzed metrology data may include raw images generated by the tool, or processed images (e.g., normalized, calibrated, etc.) at either full-scale resolution or down-sampled resolution provided by the in-line metrology tool. As an illustrative example, in the case of an in-line metrology tool including an imaging reflectometer, die analysis metrology data may include raw images, debug images, macro-overview images (MOIs), etc.

[0049] It is further contemplated herein that die screening may be performed based on die-to-die variability rather than the specific value of a metrology measurement. For example, die screening may identify a die that is different from the rest of the population, where the specific value of any given metrology measurement is not necessarily taken into account. In this manner, while the multi-channel die analysis metrology data used herein may not be tailored to provide dedicated, highly accurate measurements of specific process parameters (e.g., overlay, CD, film thickness, film composition, film refractive index, film uniformity, feature height, sidewall angle, lithography exposure parameters, CMP polishing time, etc.), this data may nevertheless vary in response to variations in these physical properties across the die.

[0050] As a non-limiting example, Figure 3 is an image of a sample 102 having a single film produced using an imaging reflectometer in accordance with one or more embodiments of the present disclosure. Note that the image of the sample 102 in Figure 3 does not include the die 104 for clarity. In Figure 3, various process parameters such as film thickness / uniformity 302 (e.g., visible as gradual signal variations across the sample 102), pinning 304, and chucking / thermal deformation 306 are clearly visible.

[0051] Therefore, such multi-channel die analysis metrology data may be suitable for parametric variability screening to identify die that deviate from the rest of a given population and therefore may have a higher likelihood of failure. The use of such die analysis metrology data may also beneficially provide many measurements across a die at high throughput (thus providing better spatial coverage than traditional in-die metrology techniques). For example, in-die metrology techniques that require measurements at specific locations often have relatively low throughput due to the time required to align the sample to a specific location for measurement and are therefore limited to less than one measurement per die on the sample. Referring again to the non-limiting illustration of FIG. 3, the image contained 69 million pixels and was generated in approximately 90 seconds. In contrast, an equivalent image generated using a traditional metrology tool that uses a move-acquire-move (MAM) collection scheme to acquire signals at selected locations on the sample would take approximately two years to capture.

[0052] Such die analysis metrology data may be generated using a variety of metrology tools, including, but not limited to, optical metrology tools, X-ray metrology tools, or particle beam metrology tools. In one embodiment, the die analysis metrology data is generated using an imaging reflectometer. For example, as described in the context of die screening system 100, the imaging reflectometer may provide various imaging techniques (e.g., bright-field imaging, dark-field imaging, etc.) at multiple wavelengths to provide multi-channel data.

[0053] In another embodiment, the die analysis metrology data is generated using a hyperspectral imaging tool, which can capture data from many (or all) pitches across the die and can also capture multiple diffraction orders (e.g., associated with the edge profile of isolated edge features). In another embodiment, the die analysis metrology data is generated using an X-ray ptychography tool, which can also capture data associated with a wide range of pitches on the die.

[0054] Additionally, the die analysis metrology data may include images generated by multiple in-line metrology tools and / or after different process steps. For example, incorporating measurements from multiple tools can increase the number of measurement channels suitable for monitoring parametric variability using method 200. As another example, incorporating measurements from different process steps may facilitate tracking variability between process steps.

[0055] In another embodiment, the method 200 includes generating 204 screening data for the die from the die analysis metrology data, where the screening data includes a subset of the measurement channels of the die analysis metrology data. For example, the screening data can include selected measurement channels in the screening data that are sensitive to process variations associated with fabrication of one or more specimens. Because the measurement channels are associated with specific measurement conditions at different sample locations, the measurement channels in the screening data can be selected based on location on the die (e.g., a location associated with a selected feature of interest), measurement conditions, or a combination thereof.

[0056] In another embodiment, method 200 includes step 206 of screening die into two or more placement classes based on variability in the screening data, the two or more placement classes including at least a set of pass dies and a set of outlier dies. For example, the set of outlier dies may include dies that exhibit screening data that varies beyond a given tolerance from the rest of the dies (e.g., the rest of the population of dies of interest), and the set of pass dies may exhibit screening data with acceptable variation. However, it should be understood that step 206 may further classify the die into any number of placement classes. For example, dies may be classified into a set of pass dies that exhibit acceptable variation in screening data relative to the population, a set of outlier dies that exhibit unacceptable variation in screening data relative to the population (e.g., dies expected to fail immediately or after short-term use), and a set of outlier dies that exhibit screening variation that exceeds an acceptable tolerance for consideration as a pass die and is below a failure tolerance associated with the outlier die. In this manner, the outlier die may be associated with dies that have a potential risk of potential defects.

[0057] It is contemplated herein that robust die screening (e.g., step 206) may be performed based on the variability of the die analysis metrics data generated in step 202. For example, a die having a statistically meaningful difference in its die analysis metrics data (or portions thereof) may have a higher risk of failure when used in its intended environment (e.g., a higher short-term or long-term failure risk than other die on a sample or sample lot). For this reason, such a die may be classified as an outlier die (e.g., in step 206).

[0058] Unfortunately, not all measurements in the die analysis metrology data for a particular set of dies (e.g., pixels associated with measurement locations on the die, or measurement channels associated with measurement conditions at particular locations) will indicate variability useful for die screening. Furthermore, some measurements may be redundant. Thus, generating screening data in step 204 may reduce the amount of data considered to include only relevant measurement channels that provide meaningful variability. This step 204 may provide various benefits, including, but not limited to, reducing the computational load associated with screening in step 206, improving the speed and therefore overall throughput of screening in step 206, or improving the accuracy and / or sensitivity of screening in step 206 by removing potential noise sources from the data set used for screening.

[0059] It is contemplated herein that generating screening data from the die analysis metrology data 204 and screening the die 206 may be performed in a variety of ways. Furthermore, the screening data may generally include any number of measurement channels from the die analysis metrology data, ranging from a single measurement channel per die to tens of thousands, tens of thousands, or more measurement channels per die.

[0060] In some embodiments, generating 204 screening data for the die from the die analysis metrology data includes selecting a subset of measurement channels that have known or expected variability to process variations during fabrication.

[0061] For example, step 204 may include selecting measurement channels associated with particular locations (locations of interest) or sets of locations on the die that are known or expected to be sensitive to process variations. In one example, step 204 may include selecting particular regions of interest (ROIs) on the die. In another example, step 204 may include selection based on results from additional inspection or metrology measurements, such as, but not limited to, selection based on the FlexPoint target inspection technique developed by KLA Corporation.

[0062] As another example, step 204 may include performing a process Design of Experiments (DOE) to determine the effects of various process variations on each of the measurement channels of the die analysis metrology data and selecting a subset of the measurement channels to be included in the screening data based on the DOE. This DOE may be performed using any suitable technique, including, but not limited to, simulating or fabricating training samples with programmed process variations and measuring with an in-line metrology tool to generate the training die analysis metrology data.

[0063] As another example, step 204 may include downsampling the data based on the measurement location and / or measurement conditions to provide a particular resolution or number of measurement channels per die.

[0064] Additionally, step 204 may, but need not, be tailored to a particular sensitivity or variation of interest. For example, the screening data may include measurement channels that are sensitive to variations in a particular process variation of interest (e.g., those with a known correlation to reliability) and / or exclude measurement channels that are sensitive to variations in a particular process variation of no interest (e.g., those with a known low correlation to reliability). As another example, the screening data may include selecting measurement channels based on their expected impact on yield.

[0065] If the screening data is selected to include measurement channels with known or expected variability, step 206 can be performed in a variety of ways.

[0066] In one embodiment, step 206 may include an empirical analysis (e.g., an empirical trend chart) of the values ​​of one or more measurement channels in the screening data or statistical variations thereof. For example, effective screening may be performed by comparing values ​​received directly from a particular measurement channel with a combination (e.g., a weighted combination, an average, etc.) of values ​​from multiple measurement channels across dies of interest. As another example, step 206 may be performed based on tracking one or more statistical metrics of a particular measurement channel, such as, but not limited to, the mean or standard deviation of data from one or more measurement channels in the screening data of dies across samples or lots of samples. In this way, variations in a selected measurement channel or channels may indicate variations in associated process variations, although specific values ​​of process parameters (e.g., overlay, CD, film thickness, film composition, film refractive index, film uniformity, feature height, sidewall angle, lithography exposure parameters, CMP polishing time, etc.) may not necessarily be determined in this step.

[0067] In another embodiment, step 206 may include determining measurement values ​​of one or more process parameters based on data from the measurement channels in the screening data, and then screening the die based on the measurement values ​​of the process parameters. Thus, step 206 may be similar to I-PAT analysis based on in-line metrology data from traditional in-line metrology measurements, but may differ in that step 206 may beneficially utilize the high throughput and high spatial coverage provided by die analysis metrology data. Furthermore, as previously described herein, even if the sample parameter measurements in step 206 are not as accurate as traditional in-line metrology measurements, robust and sensitive die screening can be achieved based on parameter variations in the die analysis metrology data.

[0068] Sample parameter measurements can be generated using a variety of techniques.

[0069] In one embodiment, the sample parameter measurements are determined by first generating a model relating data provided by selected measurement channels in the screening data to one or more sample parameters of interest, and then applying the model to the data generated for the die of interest. For example, the model may include a model of the electromagnetic interaction of incident light (e.g., illumination beam 112) with a particular structure associated with the sample die location in the selected measurement channel, which may be achieved using any technique known in the art, including, but not limited to, rigorous coupled-wave analysis (RCWA) or finite element method (FEM) techniques. In some examples, the generated model may include one or more floating parameters so that data from the die of interest can be fitted to the model.

[0070] In another embodiment, the sample parameter measurements are determined using one or more machine learning techniques (e.g., machine learning models), such as, but not limited to, neural networks, deep learning networks, machine learning libraries, autoencoder networks, etc. For example, a supervised machine learning model may be trained to provide sample parameter measurements based on screening data as input. Such a machine learning model may be trained using any suitable training data, including, but not limited to, experimental data or simulation data. In one example, the training data includes simulation and / or experimental data associated with a design of experiments (DOE), including experimental and / or simulated screening data associated with samples having a range of programmed sample parameter variations.

[0071] In some embodiments, sample parameter measurements may be generated using signal response metric (SRM) techniques, which are generally described in U.S. Patent No. 9,710,728 (July 18, 2017) and U.S. Patent No. 10,352,876 (July 16, 2019), both of which are incorporated herein by reference in their entireties.

[0072] Referring again generally to step 204 of generating screening data and step 206 of screening die based on the variability of the screening data, in some embodiments, step 204 includes performing dimensionality reduction to select measurement channels that provide variability in response to process variations. Any dimensionality reduction technique may be used, including, but not limited to, linear techniques (e.g., principal component analysis (PCA) techniques), nonlinear techniques (e.g., multidimensional scaling), unsupervised machine learning techniques, or supervised machine learning techniques. For example, die analysis metrology data may include many measurement channels per die (e.g., on the order of one million or more). However, the number of channels that exhibit meaningful, non-redundant variability may be substantially smaller (e.g., on the order of tens of thousands). Thus, dimensionality reduction can identify a subset of appropriate measurement channels suitable for screening in step 206. Furthermore, dimensionality reduction techniques may be applied without prior knowledge or expectation of which measurement parameters may be most sensitive to process variations.

[0073] In another embodiment, die screening to identify outlier die in step 206 may be performed using machine learning techniques. For example, machine learning techniques (e.g., unsupervised or supervised) may be used to identify a set of outlier die based on the variability of the screening data. In this manner, the variability of a particular measurement channel and / or the variability of a pattern or combination of measurement channels may be used to identify outlier die from a population of sampled die. Furthermore, such machine learning-based techniques may be applied to screening data generated using any technique, including, but not limited to, dimensionality reduction, selected ROI, downsampling, DOE, or a combination thereof.

[0074] Referring now generally to FIG. 2 , various steps of method 200 may be performed multiple times and / or using multiple techniques. It is contemplated herein that different techniques for performing various steps of method 200 may offer different tradeoffs between sensitivity, robustness, or ease of interpretation. For example, a technique for screening die (e.g., step 206) based on generating measurements of sample parameters (e.g., using a model-based or model-less approach) may provide insight into the root causes of variability that may result in a die being classified as an outlier die based on the sample parameter measurements. In this manner, such techniques can facilitate feedback or feedforward corrections to associated process tools to improve or control the fabrication process. However, such techniques may rely on generated models or training data that form the basis of the approach and, therefore, may not be robust to process variations not accounted for when generating the model or training data. Furthermore, such techniques may utilize a relatively small number of measurement channels associated with a particular model or training set and, therefore, may neglect to incorporate the vast amount of information provided in die-analyzed metrology data. As another example, techniques for screening die based on dimensionality reduction to identify relevant measurement channels and machine learning variability analysis of the relevant measurement channels may generally utilize all relevant information provided by die analysis metrology data and therefore may be relatively robust to many types of process variability. However, such techniques may not provide much insight into the root causes of variation because the analysis is not necessarily associated with any physical sample parameters.

[0075] In some embodiments, the set of outlier die identified using machine learning techniques may be further analyzed using one or more complementary techniques. For example, the set of outlier die may be screened (e.g., step 206) using complementary techniques such as model or model-less techniques to provide measurements of sample parameters, traditional metrology measurements (e.g., based on measurements of one or more dedicated targets), etc.

[0076] 1A-2 , various actions may be performed on die of various placement classes. In one embodiment, outlier die are removed from the supply chain with the goal of reducing the failure rate of die in the supply chain. In another embodiment, the outlier die are subjected to additional testing, such as, but not limited to, electrical testing, metrology measurements, or inspection measurements, to further evaluate the reliability of the die, such that a decision regarding whether to remove these outlier die from the supply chain may be based on the results of the additional testing.

[0077] In another embodiment, the variability screening disclosed herein is one of multiple screening steps on a die (e.g., a dietary population of interest), and the multiple screening steps may include application of method 200 using different techniques, alternative or complementary screening techniques, or a combination thereof. In this manner, each screening step may generate a set of outlier die, and the final decision on whether to remove a die from the supply chain may be based on the results of the multiple screening steps. In one example, a die may be removed from the supply chain if it is identified as an outlier die in a selected number of screening steps. In another example, a die may be removed from the supply chain based on a weighted analysis of the results of the multiple screening steps.

[0078] In another embodiment, parametric variability screening (e.g., as described herein with respect to die screening system 100 and / or method 200) may be provided as part of an additional die screening technique, such as, but not limited to, I-PAT. For example, defects identified on a particular die by I-PAT screening (e.g., via conventional in-line metrology) may be weighted or scored based at least in part on the parametric variability of the particular die as determined through method 200. For example, the defects may be weighted or scored in part based on the placement class provided by method 200 and / or any raw data associated with any step of method 200.

[0079] 4 is a block diagram illustrating aggregated die-level outlier screening using parametric variability screening in combination with I-PAT screening (e.g., block 402) in accordance with one or more embodiments of the present disclosure. For example, parametric variability screening data for a population of dies disclosed herein may be combined or aggregated with defect data for a population of dies identified using I-PAT techniques (e.g., in block 404), scored, and filtered (e.g., box 406) to provide an output set of outlier dies based on the aggregated approach.

[0080] In another embodiment, the parametric variability screening disclosed herein is used in combination with traditional metrics. Figure 5 is a block diagram illustrating several interaction paths between traditional metrics and parametric variability screening, in accordance with one or more embodiments of the present disclosure.

[0081] For example, traditional metrology techniques (box 502), such as sampling-based techniques that require measurements at specific locations on a sample, can generate die-level predictions (box 504) based on interpolation of data generated at the sampled locations (box 506). In contrast, as disclosed herein, parametric variability screening (e.g., associated with method 200) may provide die-level screening (box 508) by generating die analysis metrology data (box 510) for a population of dies (e.g., step 202). Screening data is generated based on a subset of the die analysis metrology data (e.g., step 204), and the population of dies is screened based on the variability of the screening data (step 206). Furthermore, method 200 may be implemented using different techniques, aggregated or otherwise combined as previously described herein, such that the die-level screening (box 508) may correspond to aggregated die-level screening.

[0082] 5, conventional metrology techniques (box 502) may interact with parametric variability screening (method 200) in various ways. In one embodiment, conventional metrology (box 502) is used to calibrate parametric variability screening (method 200) (arrow 512). For example, conventional metrology (box 502) may be used to calibrate process parameter measurements generated in method 200. In another embodiment, parametric variability screening (method 200) may be used to facilitate, assist, or calibrate interpolation of conventional metrology measurements (box 506) based on the increased resolution provided in the die analysis metrology data (arrow 514). In another embodiment, the two techniques may be utilized in parallel as a hybrid technique (arrow 516).

[0083] In another embodiment, the parametric variability screening disclosed herein is used to provide feedforward or feedsideway control of a fabrication process in a control system. For example, die analysis metrology data associated with various targets or regions of a specimen (e.g., metrology pads, uniform memory arrays, etc.) can be utilized to assist or improve other conventional metrology measurements. As a non-limiting example, die analysis metrology data associated with a blank film pad on a specimen can be used to generate thickness or composition information for that layer, which can then be used in a uniform CD array to provide better CD measurements, which can then be used in overlay sensitive regions to reduce ambiguity.

[0084] The subject matter described herein illustrates different components that are, in some cases, 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 combine to achieve a particular function 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” to each other to achieve the desired functionality, and any two components that can be so associated can also be considered to be “couplable” with each other to achieve the desired functionality. Specific examples of what 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.

[0085] 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 described forms are merely illustrative, and it is the intent of the following claims to embrace and include such modifications. It is further understood that the invention is defined by the appended claims.

Claims

1. 1. A die cleaning system comprising: a controller communicatively coupled to one or more in-line metering tools, the controller including one or more processors configured to execute program instructions; The one or more processors execute the program instructions to: receiving die analysis metrology data for a plurality of dies on one or more specimens from the one or more in-line metrology tools after one or more fabrication steps, the die analysis metrology data including one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, the die analysis metrology data providing a plurality of measurement channels for each die of the plurality of dies, a particular measurement channel of the plurality of measurement channels including data from a particular pixel in a particular image of the one or more images; generating screening data for a plurality of dies from the die analysis metrology data, the screening data including a subset of a plurality of measurement channels of the die analysis metrology data, the subset of a plurality of measurement channels in the screening data being sensitive to process variations associated with fabrication of one or more specimens; screening the plurality of dies into two or more placement classes based on variability in the screening data, the two or more placement classes including at least a set of pass dies and a set of outlier dies; generating the screening data for a plurality of dies from the die analysis metrology data includes down-sampling the die analyzed metrology data to provide a selected number of measurement channels for each die of the plurality of dies; Die cleaning system.

2. 2. The die screening system of claim 1, moreover, placing the set of outlier die for at least one of additional testing or removal from the supply chain; Die cleaning system.

3. 2. The die screening system of claim 1, At least one of the one or more in-line metering tools: including an imaging reflectometer; Die cleaning system.

4. 4. The die screening system according to claim 3, The imaging reflectometer comprises: a multi-channel imaging reflectometer including two or more measurement channels providing different measurement parameters, wherein the die analysis metrology data includes two or more reflectance measurement images associated with at least two of the two or more measurement channels; Die cleaning system.

5. 5. The die screening system according to claim 4, at least one of the two or more measurement channels includes a bright field image; Die cleaning system.

6. 5. The die screening system according to claim 4, at least one of the two or more measurement channels includes a dark field image; Die cleaning system.

7. 5. The die screening system according to claim 4, the one or more measurement configurations include at least one of an illumination wavelength, an illumination polarization, an illumination angle of incidence, a collection polarization, or a collection angle; Die cleaning system.

8. 2. The die screening system of claim 1, the one or more in-line metrology tools include at least one of a hyperspectral imaging tool or an X-ray ptychography tool; Die cleaning system.

9. 2. The die screening system of claim 1, generating the screening data for a plurality of dies from the die analysis metrology data, A subset of the plurality of measurement channels of die analyzed metrology data is selected that provides at least one of known or expected variation to process variations associated with fabrication of the one or more specimens. Die cleaning system.

10. 2. The die screening system of claim 1, generating the screening data for a plurality of dies from the die analysis metrology data, selecting a subset of the plurality of measurement channels of the die analysis metrology data associated with a particular location or region of interest; Die cleaning system.

11. 2. The die screening system of claim 1, generating the screening data for a plurality of dies from the die analysis metrology data, Performing a process design of experiments (DOE) using dies with known process variations; selecting a subset of the plurality of measurement channels based on the experimental design; Die cleaning system.

12. 2. The die screening system of claim 1, down-sampling the die analysis metrology data to provide a selected number of measurement channels for each die of the plurality of dies down-sampling the die analysis metrology data to include one or more measurement channels associated with a single measurement location for each die of the plurality of dies. Die cleaning system.

13. 2. The die screening system of claim 1, screening the plurality of dies into two or more placement classes based on variability of the screening data, screening the plurality of dies into two or more placement classes based on empirical trends in the screening data; Die cleaning system.

14. 2. The die screening system of claim 1, screening the plurality of dies into two or more placement classes based on the variability of the screening data includes determining measurements of the one or more process parameters based on the screening data, and screening dies based on the measurements of the one or more process parameters. Die cleaning system.

15. 15. The die screening system of claim 14, The one or more process parameters are: at least one of a critical dimension measurement, an overlay measurement, a sidewall angle measurement, a feature height measurement, a film composition measurement, a film thickness measurement, a film uniformity measurement, a film refractive index measurement, a focus position during a lithography exposure, a dose associated with a lithography exposure, or a chemical mechanical planarization time; Die cleaning system.

16. 15. The die screening system of claim 14, determining measurements of one or more process parameters based on the screening data, determining measurements of the one or more process parameters based on the screening data based on a model relating the one or more process parameters to the screening data. Die cleaning system.

17. A die cleaning system comprising: a controller communicatively coupled to one or more in-line metering tools, the controller including one or more processors configured to execute program instructions; The one or more processors execute the program instructions to: receiving die analysis metrology data for a plurality of dies on one or more specimens from the one or more in-line metrology tools after one or more fabrication steps, the die analysis metrology data including one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, the die analysis metrology data providing a plurality of measurement channels for each die of the plurality of dies, a particular measurement channel of the plurality of measurement channels including data from a particular pixel in a particular image of the one or more images; generating screening data for a plurality of dies from the die analysis metrology data, the screening data including a subset of a plurality of measurement channels of the die analysis metrology data, the subset of a plurality of measurement channels in the screening data being sensitive to process variations associated with fabrication of one or more specimens; screening the plurality of dies into two or more placement classes based on variability in the screening data, the two or more placement classes including at least a set of pass dies and a set of outlier dies; screening the plurality of dies into two or more placement classes based on the variability of the screening data includes determining measurements of the one or more process parameters based on the screening data; and screening dies based on the measurements of the one or more process parameters; determining the measurement values ​​of one or more process parameters based on the screening data includes determining the measurement values ​​of the one or more process parameters based on the screening data based on a model relating the one or more process parameters to the screening data; the model includes at least one of a rigorous coupled wave analysis (RCWA) or a finite element method (FEM) model; Die cleaning system.

18. 15. The die screening system of claim 14, determining measurements of one or more process parameters based on the screening data, determining measurements of the one or more process parameters based on the screening data based on a machine learning model. Die cleaning system.

19. 20. The die screening system of claim 18, The machine learning model is including at least one of a neural network model, a deep learning model, or a signal response econometric model; Die cleaning system.

20. 15. The die screening system of claim 14, determining measurements of the one or more process parameters based on the screening data based on a machine learning model, creating one or more training samples using known values ​​of the one or more process parameters as an experimental design; generating training die analysis metric data for the one or more training samples using an imaging reflectometer; training the machine learning model with known values ​​of the one or more process parameters and the training die analysis metrology data; The screening data is used to generate measurements of the one or more process parameters using a trained machine learning model. Die cleaning system.

21. 15. The die screening system of claim 14, The one or more processors execute program instructions to: determining a root cause associated with the variation in the screening data of the set of outlier dies based on the measurements of the one or more process parameters; Die cleaning system.

22. A die cleaning system comprising: a controller communicatively coupled to one or more in-line metering tools, the controller including one or more processors configured to execute program instructions; The one or more processors execute the program instructions to: receiving die analysis metrology data for a plurality of dies on one or more specimens from the one or more in-line metrology tools after one or more fabrication steps, the die analysis metrology data including one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, the die analysis metrology data providing a plurality of measurement channels for each die of the plurality of dies, a particular measurement channel of the plurality of measurement channels including data from a particular pixel in a particular image of the one or more images; generating screening data for a plurality of dies from the die analysis metrology data, the screening data including a subset of a plurality of measurement channels of the die analysis metrology data, the subset of a plurality of measurement channels in the screening data being sensitive to process variations associated with fabrication of one or more specimens; screening the plurality of dies into two or more placement classes based on variability in the screening data, the two or more placement classes including at least a set of pass dies and a set of outlier dies; generating the screening data for a plurality of dies from the die analysis metrology data, performing a dimension reduction on a plurality of measurement channels to select as the screening data a subset of the plurality of measurement channels that exhibits variability with respect to process variations associated with fabrication of the one or more specimens; Die cleaning system.

23. 2. The die screening system of claim 1, screening the plurality of dies into two or more placement classes based on variability of the screening data, screening the plurality of dies into two or more placement classes based on variability of the screening data using a machine learning model; Die cleaning system.

24. A die cleaning system comprising: a controller communicatively coupled to one or more in-line metering tools, the controller including one or more processors configured to execute program instructions; The one or more processors execute the program instructions to: receiving die analysis metrology data for a plurality of dies on one or more specimens from the one or more in-line metrology tools after one or more fabrication steps, the die analysis metrology data including one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, the die analysis metrology data providing a plurality of measurement channels for each die of the plurality of dies, a particular measurement channel of the plurality of measurement channels including data from a particular pixel in a particular image of the one or more images; generating screening data for a plurality of dies from the die analysis metrology data, the screening data including a subset of a plurality of measurement channels of the die analysis metrology data, the subset of a plurality of measurement channels in the screening data being sensitive to process variations associated with fabrication of one or more specimens; screening the plurality of dies into two or more placement classes based on variability in the screening data, the two or more placement classes including at least a set of pass dies and a set of outlier dies; screening the plurality of die into two or more placement classes based on the variability of the screening data includes using a machine learning model to screen the plurality of die into two or more placement classes based on the variability of the screening data; The machine learning model includes an unsupervised machine learning model. Die cleaning system.

25. A die cleaning system comprising: a controller communicatively coupled to one or more in-line metering tools, the controller including one or more processors configured to execute program instructions; The one or more processors execute the program instructions to: receiving die analysis metrology data for a plurality of dies on one or more specimens from the one or more in-line metrology tools after one or more fabrication steps, the die analysis metrology data including one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, the die analysis metrology data providing a plurality of measurement channels for each die of the plurality of dies, a particular measurement channel of the plurality of measurement channels including data from a particular pixel in a particular image of the one or more images; generating screening data for a plurality of dies from the die analysis metrology data, the screening data including a subset of a plurality of measurement channels of the die analysis metrology data, the subset of a plurality of measurement channels in the screening data being sensitive to process variations associated with fabrication of one or more specimens; screening the plurality of dies into two or more placement classes based on variability in the screening data, the two or more placement classes including at least a set of pass dies and a set of outlier dies; The one or more processors execute program instructions to: aggregating the sets of outlier die as an aggregated set of outlier die using one or more additional sets of outlier die associated with the plurality of dies generated using the one or more additional screening techniques; placing the set of outlier die for at least one of additional testing or removal from the supply chain; Die cleaning system.

26. 26. The die screening system of claim 25, the one or more additional screening techniques include at least one of an electrical screening technique or an in-line part average testing (I-PAT) screening technique; Die cleaning system.

27. A die cleaning system comprising: a controller communicatively coupled to one or more in-line metering tools, the controller including one or more processors configured to execute program instructions; The one or more processors execute the program instructions to: receiving die analysis metrology data for a plurality of dies on one or more specimens from the one or more in-line metrology tools after one or more fabrication steps, the die analysis metrology data including one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, the die analysis metrology data providing a plurality of measurement channels for each die of the plurality of dies, a particular measurement channel of the plurality of measurement channels including data from a particular pixel in a particular image of the one or more images; generating screening data for a plurality of dies from the die analysis metrology data, the screening data including a subset of a plurality of measurement channels of the die analysis metrology data, the subset of a plurality of measurement channels in the screening data being sensitive to process variations associated with fabrication of one or more specimens; screening the plurality of dies into two or more placement classes based on variability in the screening data, the two or more placement classes including at least a set of pass dies and a set of outlier dies; The one or more processors execute program instructions to: performing one or more additional metrology measurements on at least some of the set of outlier dies; and determining root causes of defects in at least some of the set of outlier dies based on the one or more additional metrology measurements. Die cleaning system.

28. 2. The die screening system of claim 1, The one or more processors execute program instructions to: adjusting one or more fabrication tools for processing one or more layers of one or more additional dies based on variability of the plurality of dies using at least one of a feedback, a feedforward, or a feedsideway control process; Die cleaning system.

29. 1. A die cleaning system comprising: one or more in-line metering tools; a controller communicatively coupled to the one or more in-line metering tools and including one or more processors configured to execute program instructions; Including, The one or more processors execute the program instructions to: receiving die analysis metrology data for a plurality of dies on one or more specimens from one or more in-line metrology tools after one or more fabrication steps, the die analysis metrology data including one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, the die analysis metrology data providing a plurality of measurement channels for each die of the plurality of dies, a particular measurement channel of the plurality of measurement channels including data from a particular pixel of a particular image of the one or more images; generating screening data for a plurality of dies from the die analysis metrology data, the screening data including a subset of a plurality of measurement channels of the die analysis metrology data, the subset of a plurality of measurement channels in the screening data being sensitive to process variations associated with fabrication of the one or more samples; screening the plurality of dies into two or more placement classes based on variability in the screening data, the two or more placement classes including at least a set of pass dies and a set of outlier dies; generating the screening data for a plurality of dies from the die analysis metrology data includes down-sampling the die analyzed metrology data to provide a selected number of measurement channels for each die of the plurality of dies; Die cleaning system.

30. 30. The die screening system of claim 29, the one or more in-line metrology tools include at least one of an imaging reflectometer, a hyperspectral imaging tool, or an X-ray ptychography tool; Die cleaning system.

31. A die screening method comprising: generating die analysis metrology data for a plurality of dies on one or more samples from one or more in-line metrology tools after one or more fabrication steps, the die analysis metrology data including one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, the die analysis metrology data providing a plurality of measurement channels for each die of the plurality of dies, a particular measurement channel of the plurality of measurement channels including data from a particular pixel of a particular image of the one or more images; generating screening data for a plurality of dies from the die analysis metrology data, the screening data including a subset of a plurality of measurement channels of the die analysis metrology data, the subset of the plurality of measurement channels in the screening data being sensitive to process variations associated with fabrication of one or more samples; screening the plurality of dies into two or more placement classes based on variability in the screening data, the two or more placement classes including at least a set of pass dies and a set of outlier dies; Including, generating the screening data for a plurality of dies from the die analysis metrology data includes down-sampling the die analysis metrology data to provide a selected number of measurement channels for each die of the plurality of dies; Die screening method.

32. 32. The die screening method of claim 31, comprising: further comprising placing the set of outlier die for at least one of additional testing or removal from the supply chain. Die screening method.

33. 32. The die screening method of claim 31, comprising: generating the screening data for the plurality of dies from the die analysis metrology data, selecting a subset of a plurality of measurement channels of the die analysis metrology data to provide at least one of known or expected variations to process variations associated with fabrication of the one or more specimens; Die screening method.

34. 32. The die screening method of claim 31, comprising: generating the screening data for a plurality of dies from the die analysis metrology data, selecting a subset of the plurality of measurement channels of the die analysis metrology data associated with a particular location or region of interest; include, Die screening method.

35. 32. The die screening method of claim 31, comprising: generating the screening data for a plurality of dies from the die analysis metrology data, performing a design of experiments (DOE) for the process using dies having known process variations and selecting a subset of the plurality of measurement channels based on the DOE; Die screening method.

36. 32. The die screening method of claim 31, comprising: Screening a plurality of dies into two or more placement classes based on variability of the screening data includes: determining measurements of one or more process parameters based on the screening data; screening the die based on measurements of one or more process parameters; Including, Die screening method.

37. 37. The die screening method of claim 36, comprising: determining measurements of one or more process parameters based on the screening data, determining measurements of the one or more process parameters based on the screening data based on a model relating one or more process parameters to the screening data; Die screening method.

38. 37. The die screening method of claim 36, comprising: determining measurements of one or more process parameters based on the screening data, determining measurements of one or more process parameters based on the screening data based on a machine learning model; Die screening method.

39. 39. The die screening method of claim 38, comprising: determining measurements of one or more process parameters based on the screening data based on a machine learning model, fabricating one or more training samples having known values ​​of one or more process parameters as an experimental design; generating training die analysis metrics data for one or more training samples using an imaging reflectometer; training the machine learning model with known values ​​of the one or more process parameters and the training die analysis metrology data; using the screening data to generate measurements of one or more process parameters using a trained machine learning model; Including, Die screening method.

40. 37. The die screening method of claim 36, comprising: The one or more processors execute program instructions to: determining a root cause associated with the variation in the screening data of the set of outlier dies based on the measurements of the one or more process parameters; Die screening method.

41. A die screening method comprising: a controller communicatively coupled to one or more in-line metering tools, the controller including one or more processors configured to execute program instructions; The one or more processors execute the program instructions to: receiving die analysis metrology data for a plurality of dies on one or more specimens from the one or more in-line metrology tools after one or more fabrication steps, the die analysis metrology data including one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, the die analysis metrology data providing a plurality of measurement channels for each die of the plurality of dies, a particular measurement channel of the plurality of measurement channels including data from a particular pixel in a particular image of the one or more images; generating screening data for a plurality of dies from the die analysis metrology data, the screening data including a subset of a plurality of measurement channels of the die analysis metrology data, the subset of a plurality of measurement channels in the screening data being sensitive to process variations associated with fabrication of one or more specimens; screening the plurality of dies into two or more placement classes based on variability in the screening data, the two or more placement classes including at least a set of pass dies and a set of outlier dies; generating the screening data for a plurality of dies from the die analysis metrology data, performing a dimension reduction on the plurality of measurement channels to select a subset of the plurality of measurement channels that exhibits variability to process variations associated with fabrication of the one or more specimens as the screening data. Die screening method.

42. 32. The die screening method of claim 31, comprising: Screening a plurality of dies into two or more placement classes based on variability of the screening data includes: using a machine learning model to screen a plurality of die into two or more placement classes based on variability in the screening data. Die screening method.

43. A die screening method comprising: a controller communicatively coupled to one or more in-line metering tools, the controller including one or more processors configured to execute program instructions; The one or more processors execute the program instructions to: receiving die analysis metrology data for a plurality of dies on one or more specimens from the one or more in-line metrology tools after one or more fabrication steps, the die analysis metrology data including one or more images generated using one or more measurement configurations of the one or more in-line metrology tools, the die analysis metrology data providing a plurality of measurement channels for each die of the plurality of dies, a particular measurement channel of the plurality of measurement channels including data from a particular pixel in a particular image of the one or more images; generating screening data for a plurality of dies from the die analysis metrology data, the screening data including a subset of a plurality of measurement channels of the die analysis metrology data, the subset of a plurality of measurement channels in the screening data being sensitive to process variations associated with fabrication of one or more specimens; screening the plurality of dies into two or more placement classes based on variability in the screening data, the two or more placement classes including at least a set of pass dies and a set of outlier dies; moreover, aggregating the set of outlier die with one or more additional sets of outlier die associated with the plurality of dies generated using one or more additional screening techniques as an aggregated set of outlier die; placing the set of outlier die for at least one of additional testing or removal from the supply chain; Including, Die screening method.

Images