`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 1 of 19 PageID #: 192
`
`EXHIBIT G
`
`EXHIBIT G
`
`
`
`(12) United States Patent
`Purdy
`
`(10) Patent No.:
`(45) Date of Patent:
`
`US 8,676,538 B2
`Mar. 18, 2014
`
`USOO8676538B2
`
`(54) ADJUSTING WEIGHTING OFA
`PARAMETER RELATING TO FAULT
`DETECTION BASED ON ADETECTED
`FAULT
`
`(75) Inventor: Matthew A. Purdy, Austin, TX (US)
`(73) Assignee: Advanced Micro Devices, Inc.,
`Sunnyvale, CA (US)
`
`(*) Notice:
`
`Subject to any disclaimer, the term of this
`patent is extended or adjusted under 35
`U.S.C. 154(b) by 1564 days.
`
`(21) Appl. No.: 10/979,309
`
`(22) Filed:
`
`Nov. 2, 2004
`
`(65)
`
`Prior Publication Data
`US 2006/OO95232A1
`May 4, 2006
`
`(51) Int. Cl.
`G06F II/30
`G06F 7/40
`G06F 9/00
`B23O 1700
`(52) U.S. Cl.
`USPC ................ 702/183; 73/865.9; 438/5; 438/14:
`700/96; 700/110; 700/121; 702/182: 702/185:
`702/187; 702/189
`
`(2006.01)
`(2006.01)
`(2011.01)
`(2006.01)
`
`(58) Field of Classification Search
`USPC ........... 702/185, 1,33, 34, 35, 36, 40, 57,58,
`702/59, 81, 82, 83, 84, 108, 113, 114, 115,
`702/117, 118, 127, 179, 181, 182, 183,187,
`702/189: 73/865.8, 865.9; 438/5, 6, 7, 8, 9,
`438/10, 11, 12, 13, 14, 15, 16, 17, 18;
`700/1, 11, 21, 79,90, 95, 96, 108, 109,
`700/110, 117, 118, 119, 120, 121, 159, 174,
`700/175; 708/100, 105, 200; 714/1, 25, 37,
`714/48, 100
`
`IPC ..................... B23B 49/00; B23Q 15/00,15/007,
`B23Q 15/12, 17/00, 17/904, 17/952, 17/10,
`B23Q 17/12, 17/20: G05B 13/00; G06F 11/00,
`G06F 11/30, 11/3058, 11/32, 17/00, 17/40,
`GO6F 19/OO
`See application file for complete search history.
`
`(56)
`
`References Cited
`
`U.S. PATENT DOCUMENTS
`
`2,883,255 A * 4, 1959 Anderson ....................... 346/34
`2,897,638 A * 8/1959 Maker ........
`451.5
`3,461,547 A * 8/1969 Di Curcio ......................... 438.7
`(Continued)
`
`FOREIGN PATENT DOCUMENTS
`
`T 2003
`WOO3,058699 A1
`WO
`1, 2004
`WO WO 2004/003822 A1
`WO WO 2004/105101 A2 12/2004
`
`OTHER PUBLICATIONS
`
`Cinar, A. etal. “Statistical Process and Controller Performance Moni
`toring. A Tutorial on current methods and future directions' Ameri
`can Control Conference, vol. 4, Jun. 2, 1999: pp. 2625-2639,
`XPO 10344696.
`
`(Continued)
`Primary Examiner — Edward Cosimano
`(57)
`ABSTRACT
`A method, apparatus and a system, for provided for perform
`ing a dynamic weighting technique for performing fault
`detection. The method comprises processing a workpiece and
`performing a fault detection analysis relating to the process
`ing of the workpiece. The method further comprises deter
`mining a relationship of a parameter relating to the fault
`detection analysis to a detected fault and adjusting a weight
`ing associated with the parameter based upon the relationship
`of the parameter to the detected fault.
`31 Claims, 8 Drawing Sheets
`
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 2 of 19 PageID #: 193
`
`so N.
`
`80
`
`
`
`
`
`
`
`Analyze Fault Data to determine if it is a
`significant fault
`
`Receive External Input
`regarding causes or non-causes of faults)
`
`S.:
`
`Determine whether to add, subtract, or not change
`factors relating to faults
`
`a
`
`
`
`ES
`
`Dynamically reduce weight to
`factor(s) that caused fault
`
`870
`
`
`
`US 8,676.538 B2
`Page 2
`
`(56)
`
`References Cited
`
`U.S. PATENT DOCUMENTS
`
`5,070,468 A * 12/1991 Ninomi et al. ............... 7O2,185
`5,287.284 A * 2/1994 Sugino et al. ..
`7OO/97
`5,658.423 A * 8/1997 Angell et al. ..................... 438.9
`5,711,849 A
`1/1998 Flamm et al. .............. 156,643.1
`5,786,023 A
`7, 1998 Maxwell et al. .................. 427.8
`5,825,482 A 10/1998 Nikoonahad et al.
`356/237
`6,119,074 A * 9/2000 Sarangapani ......
`702,185
`6,232,134 B1
`5, 2001 Farber et al.
`438.9
`6,238,937 B1 *
`5/2001 Toprac et al.
`438.9
`6,368,883 B1 * 4/2002 Bode et al. ..
`... 438/14
`6.405,096 B1* 6/2002 Toprac et al. ..
`700,121
`6,442.496 B1
`8/2002 Pasadyn et al. ................. TO2/83
`6,521,080 B2 *
`2/2003 Balasubramhanya
`et al. ........................ 156,345.24
`6,549,864 B1 * 4/2003 Potyrailo ........................ TO2/81
`6,564,114 B1* 5/2003 Toprac et al.
`700,121
`6,582,618 B1* 6/2003 Toprac et al. ................... 216.59
`6,590,179 B2* 7/2003 Tanaka et al. ..
`219,121.43
`6,616,759 B2 * 9/2003 Tanaka et al. ..
`... 118.63
`6,675,137 B1 *
`1/2004 Toprac et al. ..................... 703/2
`6,678,569 B2 *
`1/2004 Bunkofskeet al.
`700,108
`6,706,543 B2 * 3/2004 Tanaka et al. ......
`... 438/14
`6,740,534 B1
`5/2004 Adams, III et al.
`... 438/14
`6,789,052 B1* 9/2004 Toprac .............................. 703/2
`6,834,213 B1* 12/2004 Sonderman et al.
`700,121
`6.853,920 B2 *
`2/2005 Hsiung et al. ..................... 7O2/1
`6,859,739 B2 * 2/2005 Wegerich et al.
`TO2/32
`6,865,509 B1* 3/2005 Hsiung et al. ..
`702, 182
`6,871,114 B1* 3/2005 Green et al. .................. TOOf 110
`6,912,433 B1* 6/2005 Chong et al. .................. TOOf 110
`7,024,335 B1 * 4/2006 Parlos .........
`702, 182
`7,043,403 B1* 5/2006 Wang et al. .
`702,185
`7,054,786 B2* 5/2006 Sakano et al.
`702183
`78.33 38. Sin
`aw sy-
`ouilmi et al.
`7,151,976 B2 12/2006 Lin ............................... TOOf 108
`7.212,952 B2 * 52007 Watanabeet al.
`702/183
`7,328,126 B2 * 2/2008 Chamness ..................... TO2, 182
`2002/0062162 A1* 5, 2002 Bunkofskeet al. ........... TOOf 108
`2002/0072882 A1* 6/2002 Kruger et al. ..................... 703/2
`2002/0107858 A1* 8, 2002 Lundahl et al. ............... 7O7/1OO
`
`1/2003 Bell et al. ...................... 438,689
`2003/0008507 A1
`2003/0055523 A1* 3, 2003 Bunkofskeet all
`TOOf 108
`2003/0065462 A1* 4/2003 Potyrailo ...
`702/81
`2003/007.4603 A1* 4/2003 Bungert et al. ................. 714/37
`2003/0109951 A1* 6/2003 Hsiung et al. ................. TOOf 108
`2003/0136511 A1* 7/2003 Balasubramhanya
`et al. ........................ 156,345.25
`2003/0144746 A1* 7/2003 Hsiung et al. ................... TOO/28
`2004/0002776 A1
`1/2004 Bickford ......................... TOO.30
`2004/004.0001 A1* 2, 2004 Miller et al. .
`... 716.f4
`2004/0101983 A1* 5/2004 Jones et al. ..
`438/14
`2004/0259276 A1* 12, 2004 Yue et al. .....
`... 438.5
`2005, OO60103 A1* 3, 2005 Chamness
`702/30
`2005, 0071034 A1* 3, 2005 Mitrovic
`TOOf 121
`2005/0071035 A1* 3/2005 Strang.
`... 700,121
`2005, 0071036 A1* 3, 2005 Mitrovic
`... 700,121
`2005/0071037 A1* 3/2005 Strang ........................... TOOf 121
`2005/0071038 A1* 3/2005 Strang ........................... TOOf 121
`2005, 0071039 A1* 3, 2005 Mitrovic .........
`... 700,121
`2005, 0141782 A1* 6/2005 Guralnik et al. .............. 382.276
`2005/0146709 A1
`7/2005 Oh et al.
`TO2,189
`2005. O149297 A1* 7, 2005 Guralnik et al.
`... 702/182
`2005/0159922 A1* 7/2005 Hsiung et al. ...
`... 700,121
`2005, 0171627 A1* 8, 2005 Funk et al. ...
`... 700,121
`2005. O187649 A1* 8, 2005 Funk et al. ......
`... 701 114
`2005/0203696 A1* 9, 2005 Watanabe et al.
`TOOf 108
`2005/02161 14 A1* 9/2005 Hsiung et al. ...
`... 438/14
`2005/0221514 A1* 10/2005 Pasadyn et al. .
`T14f746
`2005/0268.197 A1* 12/2005 Wold ...........
`2006/0025879 A1* 2/2006 Purdy ...................... TOOf 101
`2006/0074590 A1* 4/2006 Bailey et al. .................. TO2, 182
`2006/011 1804 A1* 5/2006 Lin ............................... TOOf 110
`OTHER PUBLICATIONS
`Yue, H.H. et al.: “Weighted Principal Component Analysis and its
`Applications to Improve FDC Performance” Decision and Control.
`2004. CDC. 43 IEEE Conference on Nassau, Bahamas Dec. 14-17,
`2004; vol. 4, pp. 4262-4267, XPO 10794.793.
`& 8
`H. Yue and R. Lam; “Monitoring Etch Tool Health Using Weighted
`PCA'; AEC/APC Symposium XV, Sematech, Sep. 13-18, 2003;
`Colorado Springs, CO; XP009060799.
`PCT International Search Report; Feb. 9, 2006.
`
`
`
`* cited by examiner
`
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 3 of 19 PageID #: 194
`
`
`
`U.S. Patent
`
`Mar. 18, 2014
`
`Sheet 1 of 8
`
`US 8,676,538 B2
`
`
`
`3
`
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 4 of 19 PageID #: 195
`
`lf
`O
`y
`
`C
`l
`wa
`
`
`
`U.S. Patent
`
`Mar. 18, 2014
`
`Sheet 2 of 8
`
`US 8,676,538 B2
`
`
`
`
`
`SJØJeAA SS33OJAI
`
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 5 of 19 PageID #: 196
`
`
`
`U.S. Patent
`
`Mar. 18, 2014
`
`Sheet 3 of 8
`
`US 8,676,538 B2
`
`
`
`
`
`
`
`
`
`
`
`
`
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 6 of 19 PageID #: 197
`
`
`
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 7 of 19 PageID #: 198
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 7 of 19 PageID #: 198
`
`U.S. Patent
`
`Mar. 18, 2014
`
`Sheet 4 of 8
`
`US 8,676,538 B2
`
`Sfiw839:—
`
`Sam—
`
`:EEoUat:
`
`3Sam388m
`
`Q3de$8on
`
`5Sam£88m
`
`3END38on
`
`
`
`8a..Bouéaw
`
`«:5
`
`Sun3am5am
`
`5:530Ev:EEoUEM.5530camE
`
`«Sagan—oakmagi—E:9:535
`
`
`
`
` A:xEVN«its:«Om
`
`
`
`38an5Baaéécmuseumafl:Efifii€233;323:2
`
`
`
`
`
`
`
`
`
`658.5E88Bomémo85809:“?ANVQEESE@353on555flawm
`
`
`
`
`
`can
`
`
`
`BEBoméfiu2390983.AQEMEESI935$onsumac—2*”Him
`A9A0EaBo
`
`
`
`vEDUFA
`
`
`
`
`
`3E88BocéwO05:23thQGbEESE@353on8.5Boy—
`
`
`
`howflgSH:n—ufl
`
`
`
`
`
`
`
`
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 8 of 19 PageID #: 199
`8e
`aPO
`7.
`199m
`g
`
`A8#umD6,|7m6,P89S1U
`
`%mm5505
`
`_tmm0t2a4Pw.aScU.
`
`OaDM
`
`1Mmm
`
`aml0m51aew«MEmcow
`
`,m8&3qu3mm825m38m30H
`
`SmcomW32:853mm32m$0
`
`
`
`m8,830m820m830m1
`
`2:32wr.bEEBE83359:“;A
`
`
`
`
`U.S. Patent
`
`Mar. 18, 2014
`
`Sheet 6 of 8
`
`US 8,676,538 B2
`
`
`
`JºIIOIQUO O VOA
`
`
`
`
`
`
`
`
`
`
`
`
`
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 9 of 19 PageID #: 200
`
`9
`
`
`
`U.S. Patent
`
`US 8,676,538 B2
`
`AL CHRI[15)I H
`
`
`
`
`
`
`
`
`
`
`
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 10 of 19 PageID #: 201
`
`
`
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 11 of 19 PageID #: 202
`O11e
`aP
`1
`2
`._H
`%
`g
`
`0U
`
`e.
`
`wnmyC
`
`
`mM:sflEmoficmi
`139.:=2:me0280M
`
`
`DMAmzsflmo838-comgo338wEEdwob
`
`A“Pmm_za88:30“8San—:znm“$2234/
`%«D.can
`
`
`ur.ma
`
`mt2w
`
`2mB
`
`#.waEDGE
`
`
`
`HU<Omwfigofiom8m@183mBEEP)?BQZ02>on
`
`a69P009S
`
`D5|6,
`
`ma7onw
`
`n2e9
`
`mm8w
`
`dmbS
`
`mm
`
`.258552:»2:85on7.43:58mat—22£83mHmomegaHo:Ho“8253
`
`
`
`
`
`
`
`
`
`fl0communes:23mm@85835388mm:sfl@85865ZMAmvuouoflmowEEmmoB960A9EwBBmagmachn—@89898EH25EBbEBEdFAQ0262
`
`
`
`
`
`US 8,676,538 B2
`
`1.
`ADJUSTING WEIGHTING OFA
`PARAMETER RELATING TO FAULT
`DETECTION BASED ON ADETECTED
`FAULT
`
`BACKGROUND OF THE INVENTION
`
`10
`
`15
`
`25
`
`30
`
`1. Field of the Invention
`This invention relates generally to semiconductor manu
`facturing, and, more particularly, to a method, system, and
`apparatus for performing a process to improve fault detection
`reliability through feedback.
`2. Description of the Related Art
`The technology explosion in the manufacturing industry
`has resulted in many new and innovative manufacturing pro
`cesses. Today's manufacturing processes, particularly semi
`conductor manufacturing processes, call for a large number
`of important steps. These process steps are usually vital, and
`therefore, require a number of inputs that are generally fine
`tuned to maintain proper manufacturing control.
`The manufacture of semiconductor devices requires a
`number of discrete process steps to create a packaged semi
`conductor device from raw semiconductor material. The vari
`ous processes, from the initial growth of the semiconductor
`material, the slicing of the semiconductor crystal into indi
`vidual wafers, the fabrication stages (etching, doping, ion
`implanting, or the like), to the packaging and final testing of
`the completed device, are so different from one another and
`specialized that the processes may be performed in different
`manufacturing locations that contain different control
`schemes.
`Generally, a set of processing steps is performed across a
`group of semiconductor wafers, sometimes referred to as a
`lot. For example, a process layer that may be composed of a
`variety of different materials may be formed across a semi
`conductor wafer. Thereafter, a patterned layer of photoresist
`may be formed across the process layer using known photo
`lithography techniques. Typically, an etch process is then
`performed across the process layer using the patterned layer
`of photoresist as a mask. This etching process results in the
`formation of various features or objects in the process layer.
`Such features may be used as, for example, a gate electrode
`structure for transistors. Many times, trench isolation struc
`tures are also formed in various regions of the semiconductor
`wafer to create electrically isolated areas across a semicon
`ductor wafer. One example of an isolation structure that can
`be used is a shallow trench isolation (STI) structure.
`The manufacturing tools within a semiconductor manufac
`turing facility typically communicate with a manufacturing
`framework or a network of processing modules. Each manu
`50
`facturing tool is generally connected to an equipment inter
`face. The equipment interface is connected to a machine
`interface to which a manufacturing network is connected,
`thereby facilitating communications between the manufac
`turing tool and the manufacturing framework. The machine
`interface can generally be part of an advanced process control
`(APC) system. The APC system initiates a control script,
`which can be a Software program that automatically retrieves
`the data needed to execute a manufacturing process.
`FIG. 1 illustrates a typical semiconductor wafer 105. The
`semiconductor wafer 105 typically includes a plurality of
`individual semiconductor die 103 arranged in a grid 150.
`Using known photolithography processes and equipment, a
`patterned layer of photoresist may be formed across one or
`more process layers that are to be patterned. As part of the
`photolithography process, an exposure process is typically
`performed by a stepper on approximately one to four die 103
`
`35
`
`40
`
`45
`
`55
`
`60
`
`65
`
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 12 of 19 PageID #: 203
`
`2
`locations at a time, depending on the specific photomask
`employed. The patterned photoresist layer can be used as a
`mask during etching processes, wet or dry, performed on the
`underlying layer or layers of material, e.g., a layer of poly
`silicon, metal, or insulating material, to transfer the desired
`pattern to the underlying layer. The patterned layer of photo
`resist is comprised of a plurality of features, e.g., line-type
`features or opening-type features that are to be replicated in
`an underlying process layer.
`When processing semiconductor wafers, various measure
`ments relating to the process results on the semiconductor
`wafers, as well as conditions of the processing tool in which
`the wafers are processed, are acquired and analyzed. The
`analysis is then used to modify Subsequent processes. Turning
`now to FIG. 2, a flow chart depiction of a state-of-the-art
`process flow is illustrated. A processing system may process
`various semiconductor wafers 105 in a lot of wafers (block
`210). Upon processing of the semiconductor wafers 105, the
`processing system may acquire metrology data relating to the
`processing of the semiconductor wafers 105 from selected
`wafers in the lot (block 220). Additionally, the processing
`system may acquire tool state sensor data from the processing
`tool used to process the wafers (block 230). Tool state sensor
`data may include various tool state parameters such as pres
`Sure data, humidity data, temperature data, and the like.
`Based upon the metrology data and the tool state data, the
`processing system may perform fault detection to acquire
`data relating to faults associated with the processing of the
`semiconductor wafers 105 (block 240). Upon detecting vari
`ous faults associated with processing of the semiconductor
`wafers 105, the processing system may perform a principal
`component analysis (“PCA) relating to the faults (block
`250). Principal component analysis (PCA) is a multivariate
`technique that models the correlation structure in the data by
`reducing the dimensionality of the data. The correlation may
`take on various forms, such as correlation of problems with
`the processed wafers with problems in the processing tool.
`The PCA may provide an indication of the types of correc
`tions that may be useful in processing Subsequent semicon
`ductor wafers 105. Upon performing the PCA, the processing
`system may perform Subsequent processes upon the semicon
`ductor wafers 105 with various adjustments being based upon
`the PCA (block 260). The PCA performs an analysis to deter
`mine whether there are abnormal conditions that may exist
`with respect to a tool. Upon detecting any abnormal condi
`tions, various signals may be issued, indicating to the opera
`tors that various faults have been detected.
`One issue associated with state-of-the-art methods
`includes the fact that a determination of what constitutes an
`abnormal correlation may be based upon data used to build a
`fault detection model or a PCA model used to perform the
`fault detection analysis and the PCA. Generally, the abnormal
`conditions detected by performing the PCA may be statisti
`cally different from the data that may have been used to build
`the fault detection or the PCA model. The term “statistically
`different may mean a variety of statistical differences, such
`as differences based upon population mean, variance, etc.
`These abnormal conditions may not be an accurate reflection
`of the true manner of operation in which the tool is perform
`ing. For example, if during the development of the fault
`detection model or the PCA model, the values for a pressure
`sensor were held within Small constraints, larger variations in
`the pressure during the actual processing would generally be
`identified as a significant fault. The problem with this meth
`odology is that if the larger variation of the pressure did not
`result in any negative impact to the material being processed,
`then the fault indication may be false. In other words, if the
`
`
`
`US 8,676,538 B2
`
`3
`larger variation was still Small enough that no significant
`impact to the process was present, a false-positive fault indi
`cation occurs. This false-positive introduces inefficiencies
`and idle times in a manufacturing setting.
`More recently, various efforts have been made to incorpo
`rate weighting schemes into PCA. The weighting schemes
`may provide a significant difference in weight attached to
`various parameters, such as the pressure. However, the prob
`lems associated with the state-of-the-art weighting schemes
`include the fact that prior knowledge is required to assign a
`predetermined weight to a particular parameter. For example,
`prior knowledge may indicate that a smalleramount of weight
`should be assigned to the pressure parameter during the PCA
`analysis relating to a particular process. This would reduce
`false indications due to variations in pressure that may have
`been harmless. However, this methodology can be an ineffi
`cient, cumbersome task and, at best, may involve guess work.
`Furthermore, it may not be readily clear if adjusting the
`weight to particular parameters would result in improved or
`worsened PCA relating to a particular process.
`The present invention is directed to overcoming, or at least
`reducing, the effects of, one or more of the problems set forth
`above.
`
`10
`
`15
`
`4
`between a parameter relating to the fault detection analysis
`and a detected fault. The controller also adjusts a weighting
`associated with the parameter based upon the relationship of
`the parameter to the detected fault.
`In another aspect of the present invention, a system is
`provided for performing a dynamic weighting technique for
`performing fault detection. The system comprises a process
`ing tool communicatively coupled to a controller, a metrology
`tool, and a tool state data sensor unit. The processing tool
`performs a process upon a workpiece. The metrology tool
`acquires metrology data relating to the process performed on
`the workpiece to provide metrology data. The tool state data
`sensor unit acquires tool state data. The controller performs a
`fault detection analysis relating to the processing of the work
`piece to determine a relationship between a parameter relat
`ing to the fault detection analysis and a detected fault. The
`controller also adjusts a weighting associated with the param
`eter based upon the relationship of the parameter to the
`detected fault.
`In yet another aspect of the present invention, a computer
`readable program storage device encoded with instructions is
`provided for performing a dynamic weighting technique for
`performing fault detection. The instructions perform a
`method comprising a processing tool communicatively
`coupled to a controller, a metrology tool, and a tool state data
`sensor unit. The processing tool performs a process upon a
`workpiece. The metrology tool acquires metrology data relat
`ing to the process performed on the workpiece to provide
`metrology data. The tool state data sensor unit acquires tool
`state data. The controller performs a fault detection analysis
`relating to the processing of the workpiece to determine a
`relationship between a parameter relating to the fault detec
`tion analysis and a detected fault. The controller also adjusts
`a weighting associated with the parameter based upon the
`relationship of the parameter to the detected fault.
`
`BRIEF DESCRIPTION OF THE DRAWINGS
`
`The invention may be understood by reference to the fol
`lowing description taken in conjunction with the accompany
`ing drawings, in which like reference numerals identify like
`elements, and in which:
`FIG. 1 is a simplified diagram of a prior art semiconductor
`wafer being processed;
`FIG. 2 illustrates a simplified flowchart depiction of a prior
`art process flow during manufacturing of semiconductor
`wafers;
`FIG.3 provides a block diagram representation of a system
`in accordance with one illustrative embodiment of the present
`invention;
`FIG. 4 illustrates a principal component analysis matrix
`table, which depicts a list of tool state variables being corre
`lated with data relating to various processed semiconductor
`wafers, inaccordance with one illustrative embodiment of the
`present invention;
`FIG. 5 illustrates a more detailed block diagram represen
`tation of a tool state data sensor unit of FIG. 3, in accordance
`with one illustrative embodiment of the present invention;
`FIG. 6 illustrates a more detailed block diagram represen
`tation of a dynamic PCA weighting unit of FIG. 3, in accor
`dance with one illustrative embodiment of the present inven
`tion;
`FIG. 7 illustrates a flowchart depiction of a method in
`accordance with one illustrative embodiment of the present
`invention; and
`FIG. 8 illustrates a more detailed flowchart depiction of a
`method of performing a dynamic PCA weighting process, as
`
`SUMMARY OF THE INVENTION
`
`25
`
`30
`
`35
`
`40
`
`In one aspect of the present invention, various methods are
`disclosed for employing a dynamic weighting technique in
`connection with fault detection analysis. In an illustrative
`embodiment, the method comprises processing a workpiece
`and performing a fault detection analysis relating to the pro
`cessing of the workpiece. The method further comprises
`determining a relationship of a parameter relating to the fault
`detection analysis to a detected fault and adjusting a weight
`ing associated with the parameter based upon the relationship
`of the parameter to the detected fault.
`In another aspect of the present invention, a method is
`provided for performing a dynamic weighting technique for
`performing fault detection. The method comprises processing
`a workpiece and performing a fault detection analysis relating
`to the processing of the workpiece based upon a tool state
`parameter being input into a fault detection model associated
`with the fault detection analysis. The method further com
`prises determining whether said parameter is associated with
`a detected fault as a result of performing the fault detection
`analysis and modifying a weighting of the parameter in the
`fault detection model based upon a determination that the
`parameter is associated with the detected fault.
`In yet another aspect of the present invention, a method is
`provided for performing a dynamic weighting technique for
`performing fault detection. The method comprises processing
`a workpiece and performing a fault detection analysis relating
`to the processing of the workpiece based upon a tool state
`parameter being input into a fault detection model associated
`with the fault detection analysis. The method further com
`55
`prises performing a principal component analysis (PCA) in
`conjunction with the fault detection analysis and determining
`whether the parameter is associated with a detected fault as a
`result of performing the fault detection analysis and the PCA.
`The method further comprises modifying a weighting of the
`parameter in the fault detection model based upon a determi
`nation that the parameter is associated with the detected fault.
`In another aspect of the present invention, an apparatus is
`provided for performing a dynamic weighting technique for
`performing fault detection. The apparatus comprises a con
`troller that performs a fault detection analysis relating to a
`processing of a workpiece to determine a relationship
`
`45
`
`50
`
`60
`
`65
`
`Case 4:20-cv-00991 Document 1-7 Filed 12/31/20 Page 13 of 19 PageID #: 204
`
`
`
`US 8,676,538 B2
`
`5
`indicated in FIG. 7, in accordance with one illustrative
`embodiment of the present invention.
`While the invention is susceptible to various modifications
`and alternative forms, specific embodiments thereof have
`been shown by way of example in the drawings and are herein
`described in detail. It should be understood, however, that the
`description herein of specific embodiments is not intended to
`limit the invention to the particular forms disclosed, but on the
`contrary, the intention is to cover all modifications, equiva
`lents, and alternatives falling within the spirit and scope of the
`invention as defined by the appended claims.
`
`DETAILED DESCRIPTION OF SPECIFIC
`EMBODIMENTS
`
`6
`corrections to those parameters. This may have the effect of
`reducing the number and/or the magnitude of faults caused by
`those parameters. Similarly, over time, the weighting of the
`model parameters may be modified to reduce the frequency of
`false positive fault indications, thereby reducing unnecessary
`downtime and inefficiencies during the manufacturing of
`semiconductor wafers 105.
`Turning now to FIG. 3, a block diagram depiction of a
`system 300 in accordance with illustrative embodiments of
`the present invention is illustrated. A process controller 305 in
`the system 300 is capable of controlling various operations
`relating to a processing tool 310. The process controller 305
`may comprise a computer system that includes a processor,
`memory, and various computer-related peripherals. More
`over, although a single process controller 305 is schemati
`cally depicted in FIG. 3, in practice, the function performed
`by the process controller 305 may be performed by one or
`more computers or workstations spread throughout the manu
`facturing system.
`Semiconductor wafers 105 are processed by the processing
`tool 310 using a plurality of control input signals, or manu
`facturing parameters, provided via a line or network315. The
`control input signals, or manufacturing parameters, on the
`line 315 are sent to the processing tool 310 from a process
`controller 305 via machine interfaces that may be located
`inside or outside the processing tool 310. In one embodiment,
`semiconductor wafers 105 may be provided to the processing
`tool 310 manually. In an alternative embodiment, semicon
`ductor wafers 105 may be provided to the processing tool 310
`in an automatic fashion (e.g., robotic movement of semicon
`ductor wafers 105). In one embodiment, a plurality of semi
`conductor wafers 105 is transported in lots (e.g., stacked in
`cassettes) to the processing tools 310. Examples of the pro
`cessing tool used in semiconductor manufacturing processes
`may be photolithography tools, ion implant tools, steppers,
`etch process tools, deposition tools, chemical-mechanical
`polishing (CMP) tools, and the like.
`The system 300 is capable of acquiring manufacturing
`related data, Such as metrology data, related to processed
`semiconductor wafers 105, tool state data, and the like. The
`system 300 may also comprise a metrology tool 350 to
`acquire metrology data related to the processed semiconduc
`tor wafers 105. The system 300 may also comprise a tool state
`data sensor unit 320 for acquiring tool state data. The tool
`state data may include pressure data, temperature data,
`humidity data, gas flow data, various electrical data, a level of
`out-gas data, and other types of data, related to operations of
`the processing tool 310. Exemplary tool state data for an etch
`tool may include gas flow, chamber pressure, chamber tem
`perature, Voltage, reflected power, backside helium pressure,
`RF tuning parameters, etc. The tool state data may also
`include data external to the processing tool 310. Such as
`ambient temperature, humidity, pressure, etc. A more detailed
`illustration and description of the tool state data sensor unit
`320 is provided in FIG. 5 and accompanying description
`below.
`The system 300 may also comprise a database unit 340.
`The database unit 340 is provided for storing a plurality of
`types of data, such as manufacturing-related data, data related
`to the operation of the system 300 (e.g., the status of the
`processing tool 310, the status of semiconductor wafers 105,
`etc.). The database unit 340 may store parameter data relating
`to parameters used in fault detection and PCA models, as well
`as tool state data relating to a plurality of process runs per
`formed by the processing tool 310. The database unit 340 may
`comprise a database server 342 for storing tool state data
`
`10
`
`15
`
`30
`
`35
`
`Illustrative embodiments of the invention are described
`below. In the interest of clarity, not all features of an actual
`implementation are described in this specification. It will of
`course be appreciated that in the development of any Such
`actual embodiment, numerous implementation-specific deci
`sions must be made to achieve the developers specific goals,
`Such as compliance with system-related and business-related
`constraints, which will vary from one implementation to
`another. Moreover, it will be appreciated that such a develop
`ment effort might be complex and time-consuming, but
`25
`would nevertheless be a routine undertaking for those of
`ordinary skill in the art having the benefit of this disclosure.
`There are many discrete processes that are involved in
`semiconductor manufacturing. Many times, workpieces
`(e.g., semiconductor wafers 105, semiconductor devices,
`etc.) are stepped through multiple manufacturing process
`tools. Embodiments of the present invention provide for per
`forming a dynamic adjustment of the weighting of one or
`more parameters associated with fault detection and/or per
`forming a principal component analysis (PCA). The weight
`ing of various parameters that may be used in a fault detection
`model and/or a PCA model may be automatically determined
`and the weighting of the parameters may be adjusted dynami
`cally. For example, after a fault condition is identified by a
`processing system, an automatic input or a manual input may
`40
`be provided to the processing system to indicate whether the
`detected fault was a significant fault or an insignificant fault.
`Based upon this indication, a weighting fault matrix, which
`contains data correlating various tool state parameters to par
`ticular wafers, may be modified to make the detection of
`45
`similar faults more likely, or alternatively, less likely. There
`fore, in multi-variate fault detection and/or PCA models, one
`or more parameters that contributed to the fault condition and
`their relative importance to the fault may be detected and a
`dynamic adjustment of the weighting of those parameters that
`contributed the fault may be increased proportionally. Like
`wise, one or more parameters that did not significantly con
`tribute to the fault condition and their relative non-importance
`to the fault may be characterized and a dynamic adjustment of
`the weighting of those parameters may be decreased propor
`tionally. In other words, the weighting of the parameters that
`were found not to have contributed to a fault may be
`decreased. Therefore, a stronger signal would be required
`relating to those para