Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claims 1-2, 5-17, 20-21, 24-28, and 31-35 are presented for examination. Claims 3-4, 18-19, 22-23, 29 and 26 have been canceled. This Office Action is responsive to the amendment filed on 10/09/2025, which has been entered into the above identified application.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-2, 5, 7-17, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Kong et al. (“Qualitative and Quantitative Analysis of Multi-Pattern Wafer Bin Maps”, published 09/08/2020), hereinafter Kong; in view of Milligan (US 20180330493 A1, filed 04/30/2018); in further view of Ohmart et al. (US 20140002128 A1, filed 06/25/2013), hereinafter Ohmart. Kong and Milligan were cited in a previous Office Action.
Regarding Claim 1, Kong teaches causing a system to encode a numerical data set into a heatmap, wherein the numerical data set comprises data acquired from a plurality of die on a semiconductor wafer (Kong: “Different electrical function tests are carried out on dies (i.e., chips) to verify whether these dies meet the product standardization. These test results are marked in the corresponding die’s locations and then form the wafer bin maps (WBMs). It is a kind of graphical representation of specific wafer data and can help trace back to the causes of failures for maintaining the high yield” [I. Introduction, Para. 1]),
wherein individual data points of the numerical data set correspond to respective spatial locations and individual pixels of the heatmap correspond to the individual data points (Kong: “Different electrical function tests are carried out on dies (i.e., chips) to verify whether these dies meet the product standardization. These test results are marked in the corresponding die’s locations and then form the wafer bin maps (WBMs). It is a kind of graphical representation of specific wafer data and can help trace back to the causes of failures for maintaining the high yield” [I. Introduction, Para. 1]); and
implement an artificial intelligence (Al) model configured to provide an output comprising an indication as to whether a pattern is present in a heatmap (Kong: “Multi-pattern WBMs are then transformed into multiple single pattern WBMs. A CNN classifier is used to determine the types of patterns.” [I. Introduction, Para. 10]),
wherein the output further comprises an indication of a subset of pixels of the pixels of the heatmap that are included in the pattern, and wherein the instructions, when executed, further cause the system to decode the output to provide a spatial location of the pattern (Kong: “Multiple pattern groups on one WBM need to be located and separated apart.” [I. Introduction, Para. 4]; “The pixel values of the pattern areas and the background areas are 1 and 0. The output image is the WBM with the same resolution as the input image. It keeps original patterns along with newly-added pattern group boundaries. In the output image, different pixel values represent different areas. Specifically, the pixel values of the pattern boundaries, the pattern areas, and the background areas are 2, 1, and 0.” [II.A. Boundary Detection, Para. 3]), and
wherein the instructions, when executed, further cause the system to generate a pixel mask based on the spatial location of the pattern, wherein the pixel mask comprises a first set of pixels corresponding to die not included in the pattern and a second set of pixels corresponding to die included in the pattern (Kong: “The boundary detection stage decomposes one WBM with multiple pattern groups into multiple WBMs with single or overlapped pattern and randomly distributed noise dies background, whose general idea is shown in Fig. 3. The defective dies within each boundary belong to one pattern group.” [II.A. Boundary Detection, Para. 2]; See [Figure 3], in which a wafer map is decomposed into multiple maps to isolate each pattern from other patterns and noise; In light of Paragraph [043] of the specification, which states “The pixel mask 606 may include a set of pixels 608 corresponding to die not included in the defect 604 and a set of pixels 610 corresponding to die included in the defect 604” and [Figure 6], in which the pixel mask indicates only the die associated with a single pattern and does not include defective die that are not part of the pattern (i.e., background, noise, other patterns), BRI would support that a “pixel mask” is a binary map isolating a single pattern).
However, Kong fails to expressly disclose a system comprising: at least one processor; and at least one non-transitory medium accessible to the processor encoded with instructions; wherein the numerical data set comprises an impedance, a voltage, a current, an operation speed, a temperature, a number of failed elements on a die of the plurality of die, or combinations thereof, and wherein the individual pixels are mapped to a color and a color intensity based on a value of the individual data points; and wherein the pixel mask is used to program a sorting equipment to separate die corresponding to the first set of pixels and die corresponding to the second set of pixels when the die are removed from the semiconductor wafer.
In the same field of endeavor, Milligan teaches a system comprising: at least one processor; and at least one non-transitory medium accessible to the processor encoded with instructions (Milligan: “there are one or more non-transitory computer-readable media storing computer-executable instructions, the computer-executable instructions, when executed, causing one or more processors to perform the above method” [0014]);
wherein the numerical data set comprises an impedance, a voltage, a current, an operation speed, a temperature, a number of failed elements on a die of the plurality of die, or combinations thereof (Milligan: “Circuit probe testing, a type of wafer measurements, is performed during the final phase of wafer fabrication and before wafers are scribed and cut into dies (chips). Circuit probe testing can include parametric testing, functional testing and scan testing. Parametric testing measures electrical properties of pin electronics such as delay, voltages, and currents. Iddq testing, for example, measures the supply current (Idd) in the quiescent state (i.e., the circuit is not switching and inputs are held at static values).” [0044]), and
wherein the individual pixels are mapped to a color and a color intensity based on a value of the individual data points (Milligan: “A wafer map can also be represented by bin codes. FIG. 6A illustrates such an example. In FIG. 6B, different bin codes are represented by different colors. While the integer representation is useful, the binary/integer value may be transformed into a continuous value for some purposes. FIG. 7 illustrates an example of a wafer map using a continuous value representation.” [0046]; BRI of “color intensity” given its plain meaning and paragraph [019] of the specification, which states “The heatmap may be generated by assigning a pixel having an intensity and/or hue (e.g., color) for each data point based on a value of the data point”, is that intensity (or saturation) and hue are components of color, therefore controlling the color implies controlling color intensity).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have incorporated a system comprising: at least one processor; and at least one non-transitory medium accessible to the processor encoded with instructions; wherein the numerical data set comprises an impedance, a voltage, a current, an operation speed, a temperature, a number of failed elements on a die of the plurality of die, or combinations thereof, and wherein the individual pixels are mapped to a color and a color intensity based on a value of the individual data points as taught by Milligan to the system of Kong because both of these systems are directed towards utilizing AI to analyze, detect and classify defect patterns on wafer maps. By making this combination and evaluating how defective each die might be in the heatmap, it would allow the system of Kong to easily detect “many common semiconductor manufacturing defects [that] can cause the current to increase by orders of magnitude” (Milligan: [0044]), and also offers a “continuous value representation” (Milligan: [0046]) of die quality.
Kong and Milligan still fail to expressly disclose wherein the pixel mask is used to program a sorting equipment to separate die corresponding to the first set of pixels and die corresponding to the second set of pixels when the die are removed from the semiconductor wafer.
In the same field of endeavor, Ohmart teaches wherein the pixel mask is used to program a sorting equipment to separate die corresponding to the first set of pixels and die corresponding to the second set of pixels when the die are removed from the semiconductor wafer (Ohmart: “The wafer map identifies the exact location of each die using a coordinate system that corresponds to the physical structure of the wafer. The probe test results (die quality) may be expressed as a single bit value, e.g., good (accept) or bad (reject), or a multiple bit value that provides additional information such as good first grade, good second grade, etc. The wafer map includes a plurality of bin numbers to categorize various attributes and/or properties of each die. For example, bin 1 may contain identification of all good first grade dice, bin 2 may contain identification of all good second grade dice, bin 3 may contain identification of all plug dice, bin 4 may contain identification of all bad dice, and bin 5 may contain identification of all edge bad dice.” [0003]; “At an Assembly/Test (A/T) facility, a wafer undergoes sawing to singulate the dice, and pick and place processing based on the wafer map. A wafer map, which specifies the exact location of all good dice, is used to control an accept/reject function of a typical pick and place system.” [0005]; “During a pick and place operation on a wafer, once all the good die have been picked (that is, removed from the wafer and mounted on lead frames), then the remaining die are in visually unique positions and may be visually inspected to determine if remaining die match the defective die pattern provided by the die map.” [0018]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have incorporated wherein the pixel mask is used to program a sorting equipment to separate die corresponding to the first set of pixels and die corresponding to the second set of pixels when the die are removed from the semiconductor wafer, as taught by Ohmart to the system of Kong and Milligan because both of these systems are directed towards identifying defective die in semiconductor wafers based on wafer maps. In making this combination and using the information provided by the mask/map to control sorting equipment to separate the good and bad die, it would allow the system of Milligan and Kong to “attach only Good Electrical Chips (GEC) to lead frames” to be used in the final semiconductor product (Ohmart: [0005]).
Regarding Claim 2, Kong, Milligan and Ohmart teach the system of Claim 1, wherein the Al model comprises a region-based convolutional neural network (Kong: “With the help of two stages of pattern segmentation, multi-pattern WBMs are segmented into multiple single pattern WBMs. Then, the types of patterns are inferred by the CNN classifier.” [II.C. CNN Classification Model]).
Regarding Claim 5, Kong, Milligan and Ohmart teach the system of Claim 1, further comprising a display, wherein the instructions, when executed, further cause the system to generate display information for at least one of the heatmap or the output and provide the display information to the display (Milligan: “The output devices 125 may include, for example, a monitor display” [0034], “the output may indicate whether one of the known defect patterns exists on the wafer map, how similar it is to the one of the known defect patterns and where it is on the wafer map” [0055]).
Regarding Claim 7, Kong, Milligan and Ohmart teach the system of Claim 1, wherein the output further comprises a classification of the pattern (Kong: “The input WBM is unfolded into the row vector. The outputs are the predicted probabilities of all failure pattern classes. The class with the largest predicted probability is taken as belonging class of this WBM” [II. Methodology, C. CNN Classification Model, Para. 3]).
Regarding Claim 8, Kong, Milligan and Ohmart teach the system of Claim 7, wherein the output further comprises a confidence level of the classification (Kong: “It is accompanied by Softmax function to obtain the prediction probability of pattern classes. The cost function H is defined as [Equation 5] where m is the number of WBMs. labelpredict and labeltrue are the predicted label and the target label for WBM Wi. P is the probability of the WBM Wi being classified to the target label.” [II. Methodology, C. CNN Classification Model, Para. 2]).
Regarding Claim 9, Kong, Milligan and Ohmart teach the system of Claim 1, wherein the output further comprises a confidence level of the pattern, wherein the confidence level indicates a probability that the pattern is present in the heatmap (Kong: “It is accompanied by Softmax function to obtain the prediction probability of pattern classes. The cost function H is defined as [Equation 5] where m is the number of WBMs. labelpredict and labeltrue are the predicted label and the target label for WBM Wi. P is the probability of the WBM Wi being classified to the target label.” [II. Methodology, C. CNN Classification Model, Para. 2]; “The input WBM is unfolded into the row vector. The outputs are the predicted probabilities of all failure pattern classes. The class with the largest predicted probability is taken as belonging class of this WBM” [II. Methodology, C. CNN Classification Model, Para. 3]).
Regarding Claim 10, Kong, Milligan and Ohmart teach the system of Claim 1, wherein the spatial locations correspond to individual die of the plurality of die on the semiconductor wafer (Kong: “Different electrical function tests are carried out on dies (i.e., chips) to verify whether these dies meet the product standardization. These test results are marked in the corresponding dies locations and then form the wafer bin maps (WBMs)” [I. Introduction, Para. 1]).
Regarding Claim 11, Kong, Milligan and Ohmart teach the system of Claim 10, wherein the pattern comprises a defect in the semiconductor wafer (Kong: “Defective dies on WBMs usually cluster into specific spatial patterns. Diverse patterns typically contain essential information for root cause identification.” [I. Introduction, Para. 2]).
Regarding Claim 12, Kong teaches a method comprising:
encoding a numerical data set into a heatmap, wherein the numerical data comprises a plurality of data points corresponding to a plurality of spatial locations, wherein the data points are acquired from a plurality of die on a semiconductor wafer (Kong: “Different electrical function tests are carried out on dies (i.e., chips) to verify whether these dies meet the product standardization. These test results are marked in the corresponding die’s locations and then form the wafer bin maps (WBMs). It is a kind of graphical representation of specific wafer data and can help trace back to the causes of failures for maintaining the high yield” [I. Introduction, Para. 1]; “The model used for boundary detection adopts the U-net structure [26] and has a pixel-wise output. The input image is binary WBM. The pixel values of the pattern areas and the background areas are 1 and 0.” [II.A. Boundary Detection]),
wherein the heatmap comprises a plurality of pixels, and wherein the plurality of pixels correspond to the plurality of data points (Kong: “Different electrical function tests are carried out on dies (i.e., chips) to verify whether these dies meet the product standardization. These test results are marked in the corresponding die’s locations and then form the wafer bin maps (WBMs). It is a kind of graphical representation of specific wafer data and can help trace back to the causes of failures for maintaining the high yield” [I. Introduction, Para. 1]);
providing an output from an artificial intelligence (Al) model based at least in part, on the heatmap, wherein the output comprises an indication as to whether a pattern is present in the heatmap (Kong: “The model used for boundary detection adopts the U-net structure and has a pixel-wise output. The input image is binary WBM. The pixel values of the pattern areas and the background areas are 1 and 0. The output image is the WBM with the same resolution as the input image. It keeps original patterns along with newly-added pattern group boundaries. In the output image, different pixel values represent different areas. Specifically, the pixel values of the pattern boundaries, the pattern areas, and the background areas are 2, 1, and 0.” [II.A. Boundary Detection, Para. 3]; “With the help of two stages of pattern segmentation, multi-pattern WBMs are segmented into multiple single pattern WBMs. Then, the types of patterns are inferred by the CNN classifier.” [II.C. CNN Classification Model, Para. 1);
decoding the output to provide location information for the pattern, wherein the location information comprises the spatial location of the plurality of spatial locations (Kong: “Multiple pattern groups on one WBM need to be located and separated apart.” [I. Introduction, Para. 4]; “The pixel values of the pattern areas and the background areas are 1 and 0. The output image is the WBM with the same resolution as the input image. It keeps original patterns along with newly-added pattern group boundaries. In the output image, different pixel values represent different areas. Specifically, the pixel values of the pattern boundaries, the pattern areas, and the background areas are 2, 1, and 0.” [II.A. Boundary Detection, Para. 3]); and
generating a pixel mask, based, at least in part, on the location information, wherein the pixel mask comprises a first set of pixels corresponding to die not included in the pattern and a second set of pixels corresponding to die included in the pattern (Kong: “The boundary detection stage decomposes one WBM with multiple pattern groups into multiple WBMs with single or overlapped pattern and randomly distributed noise dies background, whose general idea is shown in Fig. 3. The defective dies within each boundary belong to one pattern group.” [II.A. Boundary Detection, Para. 2]; See [Figure 3], in which a wafer map is decomposed into multiple maps to isolate each pattern from other patterns and noise; In light of Paragraph [043] of the specification, which states “The pixel mask 606 may include a set of pixels 608 corresponding to die not included in the defect 604 and a set of pixels 610 corresponding to die included in the defect 604” and [Figure 6], in which the pixel mask indicates only the die associated with a single pattern and does not include defective die that are not part of the pattern (i.e., background, noise, other patterns), BRI would support that a “pixel mask” is a binary map isolating a single pattern).
However, Kong fails to expressly disclose wherein the numerical data set comprises an impedance, a voltage, a current, an operation speed, a temperature, a number of failed elements on a die of the plurality of die, or combinations thereof; assigning the individual pixels a color and a color intensity based on values of corresponding one of the plurality of data points; and wherein a pixel mask is used to program a sorting equipment to separate die corresponding to the first set of pixels and die corresponding to the second set of pixels when the die are removed from the semiconductor wafer.
In the same field of endeavor, Milligan teaches wherein the numerical data set comprises an impedance, a voltage, a current, an operation speed, a temperature, a number of failed elements on a die of the plurality of die, or combinations thereof (Milligan: “Circuit probe testing, a type of wafer measurements, is performed during the final phase of wafer fabrication and before wafers are scribed and cut into dies (chips). Circuit probe testing can include parametric testing, functional testing and scan testing. Parametric testing measures electrical properties of pin electronics such as delay, voltages, and currents. Iddq testing, for example, measures the supply current (Idd) in the quiescent state (i.e., the circuit is not switching and inputs are held at static values).” [0044]), and
assigning the individual pixels a color and a color intensity based on values of corresponding one of the plurality of data points (Milligan: “A wafer map can also be represented by bin codes. FIG. 6A illustrates such an example. In FIG. 6B, different bin codes are represented by different colors. While the integer representation is useful, the binary/integer value may be transformed into a continuous value for some purposes. FIG. 7 illustrates an example of a wafer map using a continuous value representation.” [0046]; BRI of “color intensity” given its plain meaning and paragraph [019] of the specification, which states “The heatmap may be generated by assigning a pixel having an intensity and/or hue (e.g., color) for each data point based on a value of the data point”, is that intensity (or saturation) and hue are components of color, therefore controlling the color implies controlling color intensity).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have incorporated wherein the numerical data set comprises an impedance, a voltage, a current, an operation speed, a temperature, a number of failed elements on a die of the plurality of die, or combinations thereof; assigning the individual pixels a color and a color intensity based on values of corresponding one of the plurality of data points, as taught by Kong to the system of Milligan because both of these systems are directed towards utilizing AI to analyze, detect and classify defect patterns on wafer maps. By making this combination and evaluating how defective each die might be in the heatmap, it would allow the system of Kong to easily detect “many common semiconductor manufacturing defects [that] can cause the current to increase by orders of magnitude” (Milligan: [0044]), and also offers a “continuous value representation” (Milligan: [0046]) of die quality.
Kong and Milligan still fail to expressly disclose wherein the pixel mask is used to program a sorting equipment to separate die corresponding to the first set of pixels and die corresponding to the second set of pixels when the die are removed from the semiconductor wafer.
In the same field of endeavor, Ohmart teaches wherein the pixel mask is used to program a sorting equipment to separate die corresponding to the first set of pixels and die corresponding to the second set of pixels when the die are removed from the semiconductor wafer (Ohmart: “The wafer map identifies the exact location of each die using a coordinate system that corresponds to the physical structure of the wafer. The probe test results (die quality) may be expressed as a single bit value, e.g., good (accept) or bad (reject), or a multiple bit value that provides additional information such as good first grade, good second grade, etc. The wafer map includes a plurality of bin numbers to categorize various attributes and/or properties of each die. For example, bin 1 may contain identification of all good first grade dice, bin 2 may contain identification of all good second grade dice, bin 3 may contain identification of all plug dice, bin 4 may contain identification of all bad dice, and bin 5 may contain identification of all edge bad dice.” [0003]; “At an Assembly/Test (A/T) facility, a wafer undergoes sawing to singulate the dice, and pick and place processing based on the wafer map. A wafer map, which specifies the exact location of all good dice, is used to control an accept/reject function of a typical pick and place system.” [0005]; “During a pick and place operation on a wafer, once all the good die have been picked (that is, removed from the wafer and mounted on lead frames), then the remaining die are in visually unique positions and may be visually inspected to determine if remaining die match the defective die pattern provided by the die map.” [0018]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have incorporated wherein the pixel mask is used to program a sorting equipment to separate die corresponding to the first set of pixels and die corresponding to the second set of pixels when the die are removed from the semiconductor wafer, as taught by Ohmart to the system of Kong and Milligan because both of these systems are directed towards identifying defective die in semiconductor wafers based on wafer maps. In making this combination and using the information provided by the mask/map to control sorting equipment to separate the good and bad die, it would allow the system of Milligan and Kong to “attach only Good Electrical Chips (GEC) to lead frames” to be used in the final semiconductor product (Ohmart: [0005]).
Regarding Claim 13, Kong, Milligan and Ohmart teach the method of Claim 12, wherein a property of a pixel of the plurality of pixels is based, at least in part, on a value of a corresponding data point of the plurality of data points (Kong: “Different electrical function tests are carried out on dies (i.e., chips) to verify whether these dies meet the product standardization. These test results are marked in the corresponding dies locations and then form the wafer bin maps (WBMs). It is a kind of graphical representation of specific wafer data and can help trace back to the causes of failures for maintaining the high yield” [I. Introduction, Para. 1]).
Regarding Claim 14, Kong, Milligan and Ohmart teach the method of Claim 12, wherein assigning colors to the plurality of pixels based on values of corresponding ones of the plurality of data points, is based, at least in part, on a colormap (Milligan: “A wafer map can also be represented by bin codes. FIG. 6A illustrates such an example. In FIG. 6B, different bin codes are represented by different colors” [0046]).
Regarding Claim 15, Kong, Milligan and Ohmart teach the method of Claim 14, wherein the colormap assigns colors to an entire range of the values of the plurality of data points (Milligan: “While the integer representation is useful, the binary/integer value may be transformed into a continuous value for some purposes. FIG. 7 illustrates an example of a wafer map using a continuous value representation” [0046]).
Regarding Claim 16, Kong, Milligan and Ohmart teach the method of Claim 14, wherein the colormap assigns a same color to values of the plurality of data points based on a comparison to a threshold value (Milligan: “A wafer map can also be represented by bin codes. FIG. 6A illustrates such an example. In FIG. 6B, different bin codes are represented by different colors” [0046]).
Regarding Claim 17, Kong, Milligan and Ohmart teach the method of Claim 16, wherein the colormap assigns the same color to the values of the plurality of data points when the values are equal to or above the threshold value or when the values are equal to or below the threshold value (Milligan: “A good die can pass the tests thoroughly and be assigned to the best grade and other bin code assignments indicate different degrees of quality inferiority. The number of bin codes can be less than ten or up to a few dozen” [0045]).
Regarding Claim 20, Kong, Milligan and Ohmart teach the method of Claim 12, wherein the plurality of spatial locations correspond to a plurality of die on a semiconductor wafer (Kong: “Different electrical function tests are carried out on dies (i.e., chips) to verify whether these dies meet the product standardization. These test results are marked in the corresponding dies locations and then form the wafer bin maps” [I. Introduction, Para. 1]), and
the method further comprises assigning a first grade to a first die of the plurality of die included in the pattern and assigning a second grade to a second die of the plurality of die outside the pattern, wherein the first grade and the second grade are different (Milligan: “Measurement results can be simply pass/fail or be assigned to a grade. Each grade can be presented using a unique bin code. A good die can pass the tests thoroughly and be assigned to the best grade and other bin code assignments indicate different degrees of quality inferiority. The number of bin codes can be less than ten or up to a few dozen. The spatial distribution of pass/fail or quality grades across a wafer can be represented by a wafer map. Wafer maps can contain characteristic patterns or signatures” [0045]),
wherein the first die included in the pattern is sorted differently than the second die outside the pattern (Milligan: “A statistical approach typically classifies patterns based on an extracted feature set, and an underlying statistical model for generating these patterns. For example, an averaging operation can first be performed on good or bad die images (white and black; 0 and 1), in which each die was averaged by its 3×3 or 5×5 neighboring dies. The obtained average value replaces the original binary value. As such, the binary wafer map is transformed into a smoothed grey-level wafer map. A threshold value is then selected. If the grey level of a die exceeds the threshold value, it is determined to be bad; otherwise, it is good.” [0047]).
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Kong in view of Milligan and Ohmart, as applied to Claim 5 above, in further view of Kurihara et al. (US 20060078188 A1, filed 07/28/2005), hereinafter Kurikara. Kurihara was cited in a previous Office Action.
Regarding Claim 6, Kong, Milligan and Saqlain teach the system of Claim 5. However, they fail to expressly disclose wherein the display is configured to provide the output as graphical overlays on the heatmap.
In the same field of endeavor, Kurihara teaches wherein the display is configured to provide the output as graphical overlays on the heatmap (Kurihara: “FIG. 12 shows an example of a display screen. The display screen is constituted of a wafer information area 501, a wafer map area 506, a defect class and defect data area 508, a view area 517, a detailed view area 519 and a defect class area 521.” [0057], “FIG. 14 is a diagram showing a detailed example of a wafer map display area in the embodiment shown in FIG. 12” [0032], see FIG. 14).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have incorporated wherein the display is configured to provide the output as graphical overlays on the heatmap into the system of Kong, Milligan and Saqlain as they relate to defect detection through image processing of wafer maps and displaying results to the user. In making this combination, it would allow the user to clearly see the position of the defect on the map on the display for review (Kurihara: [0059]).
Claims 21, 24, and 31 are rejected under 35 U.S.C. 103 as being unpatentable over Kong et al. (“Qualitative and Quantitative Analysis of Multi-Pattern Wafer Bin Maps”, published 09/08/2020), hereinafter Kong; in view of Milligan (US 20180330493 A1, filed 04/30/2018); in further view of Saqlain et al. (“A Voting Ensemble Classifier for Wafer Map Defect Patterns Identification in Semiconductor Manufacturing”, published 03/11/2019), hereinafter Saqlain. Kong and Milligan were cited in a previous Office Action.
Regarding Claim 21, Kong teaches causing a system to:
encode a first portion of a numerical data set into a first heatmap, wherein the numerical data set comprises a plurality of data points corresponding to a plurality of spatial locations, and wherein the first heatmap comprises a first plurality of pixels, wherein the first plurality of pixels correspond to a plurality of data points of the first portion (Kong: “Different electrical function tests are carried out on dies (i.e., chips) to verify whether these dies meet the product standardization. These test results are marked in the corresponding dies locations and then form the wafer bin maps (WBMs). It is a kind of graphical representation of specific wafer data and can help trace back to the causes of failures for maintaining the high yield” [I. Introduction, Para. 1]);
encode a second portion of the numerical data set into a second heatmap, wherein the second heatmap comprises a second plurality of pixels, wherein the second plurality of pixels correspond to a plurality of data points of the second portion, wherein the numerical data set comprises data acquired from a plurality of die of a semiconductor wafer (Kong: “After detecting the boundaries, the defective dies in the closed boundary loop are considered to belong to one pattern group. One pattern with disconnected parts can be located in one boundary. Every pattern group forms a new WBM with the same resolution as the original input WBM.” [II. Methodology, A. Boundary Detection, Para. 13]); and
implement a model configured to provide an output comprising an indication as to whether a pattern is present in at least one of the first heatmap or the second heatmap (Kong: “Multi-pattern WBMs are then transformed into multiple single pattern WBMs. A CNN classifier is used to determine the types of patterns.” [I. Introduction, Para. 10]).
However, Kong fails to expressly disclose a system comprising: at least one processor; and at least one non-transitory medium accessible to the processor encoded with instructions; and wherein each pixel of the first and second pluralities of pixels are mapped to a color and a color intensity based on a value of the plurality of data points of the first portion and on a value of the plurality of data points of the second portion, respectively, and wherein the numerical data set comprises an impedance, a voltage, a current, an operation speed, a temperature, a number of failed elements on a die of the plurality of die, or combinations thereof, wherein the first portion comprises a first data type and the second portion comprises a second data type; and wherein the model comprises an artificial intelligence (AI) model configured to analyze the first heatmap corresponding to the first portion having the first data type and the second heatmap corresponding to the second portion having the second data type to provide the output.
In the same field of endeavor, Milligan teaches a system comprising: at least one processor; and at least one non-transitory medium accessible to the processor encoded with instructions (Milligan: “there are one or more non-transitory computer-readable media storing computer-executable instructions, the computer-executable instructions, when executed, causing one or more processors to perform the above method” [0014]);
wherein each pixel of the first and second pluralities of pixels are mapped to a color and a color intensity based on a value of the plurality of data points of the first portion and on a value of the plurality of data points of the second portion, respectively (Milligan: “A wafer map can also be represented by bin codes. FIG. 6A illustrates such an example. In FIG. 6B, different bin codes are represented by different colors. While the integer representation is useful, the binary/integer value may be transformed into a continuous value for some purposes. FIG. 7 illustrates an example of a wafer map using a continuous value representation.” [0046]; BRI of “color intensity” given its plain meaning and paragraph [019] of the specification, which states “The heatmap may be generated by assigning a pixel having an intensity and/or hue (e.g., color) for each data point based on a value of the data point”, is that intensity (or saturation) and hue are components of color, therefore controlling the color implies controlling color intensity), and
wherein the numerical data set comprises an impedance, a voltage, a current, an operation speed, a temperature, a number of failed elements on a die of the plurality of die, or combinations thereof (Milligan: “Circuit probe testing, a type of wafer measurements, is performed during the final phase of wafer fabrication and before wafers are scribed and cut into dies (chips). Circuit probe testing can include parametric testing, functional testing and scan testing. Parametric testing measures electrical properties of pin electronics such as delay, voltages, and currents. Iddq testing, for example, measures the supply current (Idd) in the quiescent state (i.e., the circuit is not switching and inputs are held at static values).” [0044]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have incorporated a system comprising: at least one processor; and at least one non-transitory medium accessible to the processor encoded with instructions; and wherein each pixel of the first and second pluralities of pixels are mapped to a color and a color intensity based on a value of the plurality of data points of the first portion and on a value of the plurality of data points of the second portion, respectively, and wherein the numerical data set comprises an impedance, a voltage, a current, an operation speed, a temperature, a number of failed elements on a die of the plurality of die, or combinations thereof as taught by Milligan to the system of Kong because both of these systems are directed towards utilizing AI to analyze, detect and classify defect patterns on wafer maps. By making this combination, Milligan provides hardware on which the system of Kong is to operate, allows the system of Kong to easily detect “many common semiconductor manufacturing defects [that] can cause the current to increase by orders of magnitude” (Milligan: [0044]), and also offers a “continuous value representation” (Milligan: [0046]) of die quality.
Kong and Milligan still fail to expressly disclose wherein the first portion comprises a first data type and the second portion comprises a second data type; and wherein the model comprises an artificial intelligence (AI) model configured to analyze the first heatmap corresponding to the first portion having the first data type and the second heatmap corresponding to the second portion having the second data type to provide the output.
In the same field of endeavor, Saqlain teaches wherein the first portion comprises a first data type and the second portion comprises a second data type (Saqlain: “We extracted multi-types features from each WM to make them ready for ML classifiers implementation, subsequent analysis, classification using scaling, attribute aggregation, and attribute decomposition. We proposed three distinctive features such as density-, geometry-, and radon-based. Density-based features depend on the failure density of various parts of WM image. Geometry-based features are obtained from the geometric attributes of regions for each WM. Radon-based feature is created by the transformation of radon, which can create a 2-dimensional representation of WM based on a series of projections.” [III.C. Features Extraction, Para. 1]); and
wherein the model comprises an artificial intelligence (AI) model configured to analyze the first heatmap corresponding to the first portion having the first data type and the second heatmap corresponding to the second portion having the second data type to provide the output (Saqlain: “we propose a voting ensemble classifier with multi-types features to identify wafer map defect patterns in semiconductor manufacturing. Our research contents can be summarized as follows. First, three distinctive features such as density-, geometry-, and radon-based features were extracted from raw wafer images. Then, we applied four machine learning classifiers, namely logistic regression (LR), random forests (RFs), gradient boosting machine (GBM), and artificial neural network (ANN), and trained them using extracted features of original data set. Then their results were combined with a soft voting ensemble (SVE) technique which assigns higher weights to the classifiers with respect to their prediction accuracy.” [Abstract]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have incorporated wherein the first portion comprises a first data type and the second portion comprises a second data type; and wherein the model comprises an artificial intelligence (AI) model configured to analyze the first heatmap corresponding to the first portion having the first data type and the second heatmap corresponding to the second portion having the second data type to provide the output, as taught by Saqlain to the system of Kong and Milligan because both of these systems are directed towards predicting/classifying defect patterns in semiconductor wafer maps. In making this combination and evaluating defaults through multiple types of measurements with different data types, it would allow the system of Kong and Milligan to “overcome the limitation of individual classifiers” as “ensemble classifiers have shown more effective results regarding stability and robustness” (Saqlain: [I. Introduction, Para. 6]).
Regarding Claim 31, it is a method claim that corresponds to the system of Claim 21. Therefore, it is rejected for the same reason as Claim 21.
Regarding Claim 24, Kong, Milligan and Saqlain teaches the system of Claim 21, wherein the plurality of spatial locations for the plurality of data points of the first portion are the same as the plurality of spatial locations for the plurality of data points of the second portion (Saqlain: “A wafer map (WM) is a collection of visual data about the physical parameter that are collected from semiconductor wafers. It contains basic information about thickness, size, and location of defects on wafers.” [I. Introduction, Para. 1]; “We extracted multi-types features from each WM to make them ready for ML classifiers implementation, subsequent analysis, classification using scaling, attribute aggregation, and attribute decomposition. We proposed three distinctive features such as density-, geometry-, and radon-based. Density-based features depend on the failure density of various parts of WM image. Geometry-based features are obtained from the geometric attributes of regions for each WM. Radon-based feature is created by the transformation of radon, which can create a 2-dimensional representation of WM based on a series of projections. All these extracted features are manipulated together to show each WM with new representation.” [III.C. Feature Extraction, Para. 1]).
Claims 25-28 and 32-35 are rejected under 35 U.S.C. 103 as being unpatentable over Kong in view of Milligan and Saqlain, as applied to Claims 21 and 31 above, in further view of Backstrom et al. (US 20220277550 A1, foreign priority filed 02/26/2021), hereinafter Backstrom. Backstrom was cited in a previous Office Action.
Regarding Claim 25, Kong, Milligan and Saqlain teach the system of Claim 21. However, they fail to expressly disclose wherein the model further comprises a first artificial intelligence (AI) model and a second AI model, wherein the first AI model analyzes the first heatmap and provides a first output and the second AI model analyzes the second heatmap and provides a second output.
In the same field of endeavor, Backstrom teaches wherein the model further comprises a first artificial intelligence (AI) model and a second AI model, wherein the first AI model analyzes the first heatmap and provides a first output and the second AI model analyzes the second heatmap and provides a second output (Backstrom: “The cross-check module 64 may use a second AI model that may be trained to perform the same classification task as the first AI model (using, however, a larger number of inputs, namely at least the explainability signature and the data analysis output in addition to the input data)” [0290], see [Fig. 13 and 14]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have incorporated wherein the model comprises a first artificial intelligence (AI) model and a second AI model, wherein the first AI model analyzes the first heatmap and provides a first output and the second AI model analyzes the second heatmap and provides a second output as taught by Backstrom to the system of Kong, Milligan and Saqlain because they all relate to image processing through the handling of heatmap data. In making this combination, it allows the system of Kong, Milligan and Saqlain to limit incorrect analyses by cross-checking its classification “to determine whether the data analysis result is considered to be reliable, whether the data analysis result is to be flagged, whether the data analysis result may be used as training data for machine learning, ML, and/or whether the data analysis result indicates the suitability of a data analysis model used in the data analysis (e.g., another AI model) to perform the task for which it is being intended.” (Backstrom: [0006]).
Regarding Claim 26, Kong, Milligan, Saqlain and Backstrom teach the system of Claim 25, wherein the output comprises the first output and the second output (Backstrom: “The AI analysis outputs generated by a data analysis module 62 are passed on, in conjunction with the output of the CCAI, to a surrounding/subsequent system in order to take a final decision” [0292]).
Regarding Claim 27, Kong, Milligan, Saqlain and Backstrom teach the system of Claim 25, wherein the first output and the second output are combined to provide the output (Backstrom: “The control action that depends on the cross-check data may be or may comprise combining the data analysis output with the cross-check data to generate a consolidated data analysis result.” [0018]).
Regarding Claim 28, Kong, Milligan, Saqlain and Backstrom teach the system of Claim 25, herein the first AI model and the second AI model comprise at least one of a different architecture or a different parameter (Backstrom: “The first and second AI models 24, 25 may have different AI model configurations. For illustration, the first AI models 24 may be an artificial neural network (ANN), in particular a convolutional neural network (CNN). The second AI models 25 may include K-means clustering or ANNs, optionally in combination with principal component analysis (PCA) and/or histogram analysis.” [0245]).
Regarding Claim 32, Kong, Milligan and Saqlain fail to expressly disclose the method of Claim 31, wherein the model comprises a first artificial intelligence (AI) model and a second AI model, wherein the method further comprises: analyzing the first heatmap with the first AI model to generate a first output; and analyzing the second heatmap with the second AI model to generate a second output, wherein the output comprises the first output and the second output.
In the same field of endeavor, Backstrom teaches the system of Claim 31, wherein the model comprises a first artificial intelligence (AI) model and a second AI model, wherein the method further comprises: analyzing the first heatmap with the first AI model to generate a first output; and analyzing the second heatmap with the second AI model to generate a second output, wherein the output comprises the first output and the second output (Backstrom: “The cross-check module 64 may use a second AI model that may be trained to perform the same classification task as the first AI model (using, however, a larger number of inputs, namely at least the explainability signature and the data analysis output in addition to the input data)” [0290], see [Fig. 13 and 14]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have incorporated wherein the model comprises a first artificial intelligence (AI) model and a second AI model, wherein the method further comprises: analyzing the first heatmap with the first AI model to generate a first output; and analyzing the second heatmap with the second AI model to generate a second output, wherein the output comprises the first output and the second output as taught by Backstrom to the system of Kong, Milligan and Saqlain because they all relate to image processing through the handling of heatmap data. In making this combination, it allows the system of Kong, Milligan and Saqlain to limit incorrect analyses by cross-checking its classification “to determine whether the data analysis result is considered to be reliable, whether the data analysis result is to be flagged, whether the data analysis result may be used as training data for machine learning, ML, and/or whether the data analysis result indicates the suitability of a data analysis model used in the data analysis (e.g., another AI model) to perform the task for which it is being intended.” (Backstrom: [0006]).
Regarding Claim 33, Kong, Milligan, Saqlain and Backstrom teach the method of Claim 32, further comprising combining the first output and the second output to provide the output (Backstrom: “The control action that depends on the cross-check data may be or may comprise combining the data analysis output with the cross-check data to generate a consolidated data analysis result.” [0018]).
Regarding Claim 34, Kong, Milligan, Saqlain and Backstrom teach the method of Claim 32, further comprising: training the first AI model with a first training data set; and training the second AI model with a second training data set, wherein the second training data set is different from the first training data set (Backstrom: “By implementing a distributed architecture in which several distinct processing devices or systems perform AI model training using their local labeled datasets (or even unlabeled datasets), distributed AI model training may be performed while ensuring data privacy.” [0215]).
Regarding Claim 35, Kong, Milligan, Saqlain and Backstrom teach the method of Claim 32 further comprising: training an AI model with a training data set; and replicating the AI model to provide the first Al model and the second AI model (Backstrom: “The second AI model may be trained to perform the same classification task as the first AI model” [0009]).
Response to Arguments
The Examiner acknowledges the Applicant’s amendments to Claims 1, 9, 12, 21, 25 and 31.
Applicant’s arguments, filed 10/09/2025, traversing the rejection of Claims 1-29 and 31-36 under 35 U.S.C. § 101 have been fully considered and are persuasive. The rejections have been withdrawn.
Applicant’s arguments, filed 10/09/2025, regarding the rejection of Claims 1-2, 5-17, 20-21, 24-28, and 31-35 under 35 U.S.C. § 103 have been fully considered and found moot in view of the new grounds of rejection under 35 U.S.C. § 103 (see rejection above).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Chen et al. (“TestDNA-E: Wafer Defect Signature for Pattern Recognition by Ensemble Learning”) discusses an ensemble learner for wafer defect map classification based on sequences representing the results of a wafer test.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MEGAN E HWANG whose telephone number is (703)756-1377. The examiner can normally be reached Monday-Thursday 10:00-7:30 ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Jennifer Welch can be reached at (571) 272-7212. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/M.E.H./Examiner, Art Unit 2143
/JENNIFER N WELCH/Supervisory Patent Examiner, Art Unit 2143