PRODUCTIVffY & YIELD
`
`Factory-wide run-to-run process control
`
`Mark Yelverton*, Advanced Micro Devices, Austin, Texas
`Koatae Taakalls, Arizona State University, Tempe, Arizona
`Kevin stoddanl*, SEMY Engineering Inc., Phoenix, Arizona
`
`Over the last several yeara. run-to-run process con(cid:173)
`trol has only been applied to select processes.
`Now, through advances in process-eugineering(cid:173)
`friendly software tools, It can be used acroesa
`wafer fab to maintain process repeatablllty auto(cid:173)
`matically and compeilRte for upstream process
`varlabllity, achieve better device yields and speeds,
`and greatly enhance factory productivity.
`
`A common methodology for monitoring batch
`
`Automatic....- control
`Most problems aswciated with manual control of sernicon•
`ductor processes can be eliminated with automatic process
`
`• Additional authors are listed in the Acknowledgme11ts.
`
`processes uses x-bar/s or x-bar/r plot~ from sta(cid:173)
`tistical process control (SPC) software. Normally
`distribured process data is monitored using a set of rules
`(i.e., "Western Electric") to determine if a process is in
`control. Manual investigation and adjustment of the
`process are necessary when a data point is out of con(cid:173)
`trol. A large percentage of these adju.mnents are made
`to compensate for run-to-run variations attributed to
`process equipment drift.
`Unfortunately, there are many problems with manu•
`ally adju.qted processes based on SPC charts. A typical
`wafer fab has ~2500 on-line SPC charts. If all Western
`Electric rules are used and if two new points are added The screen shows a faull•deteclian chart from run-to-run process control software.
`The software automatically maintains rnpeatablll!y for better device yields and factory
`to each chart/ day, there could be an average 82 false
`alanru,/day [lj. Only processes with the most signifi-
`productivity. (Comput!lr illustration courtesy of SEMY Engineering)
`cant excursions tend to be maintained due to the sheer magni-
`control. All areas in a wafer fab can show significant processcon(cid:173)
`tude of faults. In some cases, the opposite is true and too much
`trol improvement after implementing even simple automated
`attention is given to a chart and overadjustment occurs, result-
`feedback process controllers [2, 3].
`feed(cid:173)
`ing in processes "ringing." Additional process variation can be
`There are two types of run-to-run process control -
`back and feed-forwaid. A foedbackcontrulsy:;tem makes adjust(cid:173)
`introduced between shifts or individuals as they try to com-
`men ts to recipe or process-tool parameters to maintain the
`pensatc for t?ach other's process adjustments (Fig. 1).
`desired end-of-run or in situ metrology results. Compensa(cid:173)
`tion for incoming post-process variations from previous steps
`is achieved using feed-forward process control; an open-loop
`relationship or model between process steps adjusts the process
`target in the current step.
`
`Copyright © 1999. All rights reserved.
`www.solid-stale.com • ~ 19ff • Soiid State Techno;ogy ff
`
`cootfnued on page 49
`
`PDF Solutions v Ocean Semiconductor, IPR2022-01196
`PDF Exhibit 1007, Page 1 of 4
`
`

`

`Run-fo.nm ,,..,.,_CMtnJloon!lnu6dfrom pege45
`Benefits from automatic run-to-run process control are mnner(cid:173)
`ous. Greater precision and aa:urac:y are po551"ble with smaller,
`more frequent adjustments. Control algorithms lhatignon! "flier"
`data points can be tuned for maximum and repeatable perfor(cid:173)
`mance. Variables that once were considered too complex can be
`controlled. Human error can also be eliminated via a consis(cid:173)
`tent adjustment methodology.
`In contrast, manual adjustments are inherent approximations
`basedonsimplerelationships,can.bebiasedbypastexperience,
`orcanbeinfluencedbye.rrors
`in reading or entering data.
`Once automatic control is
`implemented,
`personnel
`requirements to maintain a
`process are reduced, freeing
`engineers and operators to
`work on other issues.
`Various obstacles must be
`overcome for successful imple(cid:173)
`mentation of a run-to-run con(cid:173)
`trol system. The automated
`acquisition of metrology results
`can, at times, be very difficult
`and must be extremely flexible.
`Some metrology tools do not
`provide proper connectivity to
`a factury MES system. Even in
`cases where metrology results
`are integrated into a centralized SPC database, custom interfaces
`are often required to communicate metrology data to the run(cid:173)
`to-run control system. Moreover, metrology results must be
`
`,..... s. SPC chart showing evidence of manual overadjustment of a CVO
`process deposition time.
`
`acquired in a timely manner to be useful, especially in a feed.
`forward application.
`It is also nece!iSllry to implement proper fault detection and
`classification logic to deal with faulty metrology measurements
`caused by drifting metrology tools, operator error, or bad warers.
`Proper classification of metrology data is essential to ensure that
`correct measurements are provided to the run-to-run controller.
`Further integration with the process tool (i.e., recipe manage(cid:173)
`ment) is also required to provide the process adjustment
`mechanism for tools that do
`not directly support adjust•
`ing process parameters. This
`should include boundaries of
`adjustment and means to han(cid:173)
`dle a control system failure.
`
`Run-to-run
`proc . . . control
`The implementation of a run(cid:173)
`to-nm processcontrollercan be
`achieved in.several ways. Direct
`implementation on a process
`tool allows for wafer-to-wafer
`process adjustments, if neces,(cid:173)
`sary, and ease of execution.
`It does not allow feed.forward
`control across different pro-
`cesses, however, and requires
`costly dedicated or on-board metrology measurement.
`Factory-wide implementation of nm-to-run process control,
`such as SEMATECH's Advanced Process Control (APC)
`
`Copyright © 1999. All rights reserved.
`
`PDF Solutions v Ocean Semiconductor, IPR2022-01196
`PDF Exhibit 1007, Page 2 of 4
`
`

`

`Framework, is gaining acceptance in a few companies. Such a
`framework, which provides a communicati0Il8 bus and stan(cid:173)
`dards for building control and visuallmtion modules, is extremely
`flexible, providing feedback and feed-forward functionality and
`the ability to exchange information between many suppliers.
`While the development and implementation of APC Framework
`modules are left entirely up to the user, the sheer magnitude and
`expense of such development may limit actual use in factories
`with a large diversity in products [4J.
`Run-to-run process control can also be implemented with off(cid:173)
`the-shelf hardware and software, such as the Equipment Super(cid:173)
`visor Workstation (ESW) developed bySEMY Engineering. This
`supervisory solution provides integration to virtually any process
`or rnetrology tool, and with SEMY's Advanced Run-to-Run
`Control (ARRC) module provides both feed-forward or feed(cid:173)
`back run-to-run process control (see sidebar "Advanced run(cid:173)
`to-run control (ARRC} tools" on p. 46). Connectivity between
`FSW andAPC Framework is designed around a Common Object
`Request Broker Architecture {CORBA). This solution is applic(cid:173)
`able to a single process tool, making it attractive for tool man(cid:173)
`ufacturers or an entire process area [5]. An example of the lat(cid:173)
`ter is the implementation of this supervisory system at White
`Oak Semiconductor [6].
`
`Pl-Geese modeHIIIC and control
`A typical methodology for implementing nm-to-run processcon-
`1rol uses a "black box" approach, where process-modeling and
`control-systems experts develop complex schemes to control the
`ptua'5Saron"atel.y [6]. This approach is very powerful and can usu(cid:173)
`allyprovide the bestresults; however, itcan also be time-consuming
`to develop and implement, especially if a suitable control infra(cid:173)
`oiructure is not in place. This infrastructure is provided by ARRC
`via "plug--in" algorithms using MATLAB- a high-performance
`numeric computation and visualization software package.
`
`The "black box" approach does possess several drawbacks.
`Once developed, it is difficult or, in some cases, impossible to be
`modified or adjusted by production~. Also, current sys(cid:173)
`tems d.o not provide graphical user interfaces that allow easy
`gauging of the perfoill'.anceand accuracy of adjustments. ARRC,
`however, has a revolutionary set of tools that allows process
`engineers with little orno experience to develop their own mod(cid:173)
`els and t.'Ontrol systems easily and analyze the performance of
`the control.
`
`Feedback control exampl-
`Simple feed.back controllers can significantly improve process
`performance and productivity in every area of the production
`environment [7-9]. This method is useful for improving process
`control and automating routineadjustmems made by operators,
`engineers,and maintenance personnel. The examples presented
`here are only a small part of what is achievable in a factory-wide
`implementation.
`For example, in CMP operations, the polish time may be
`adjusted to control !he remaining thickness of the film, and may
`be changed between wafers or between batches depending oo
`the stability of the process. When wafer-to-wafer adjustments
`are required, in situ metrology is needed to provide measure(cid:173)
`ments in time to close the loop. If the pt'OL"e&S drift is understood,
`a feedback model can be used to predict and adjust the polish
`time required for each wafer in a batch. The polish time can be
`changed from wafer to wafer based on model estiniates, and
`then verified withmetrology afrerthe batch has beenrompleted.
`Feedback controllers can also be used to compensate auto(cid:173)
`matically for lhe changes in the slurry and degradation of ihe
`polish pads.
`Diffusion processes often require the simultaneous adjust(cid:173)
`ment of multiple variables. Low-pressure chemical vapor depo(cid:173)
`sition (LPCVD) batch processes typically require temperature
`and time adjustments. A feed.back con(cid:173)
`troller can be used to adjustend-'.rone tem(cid:173)
`peratures to minimize thickness differ(cid:173)
`ences between wafers proc-essed in the
`center and end zones of the furnace. The
`feedback controller also adju.'lts deposi(cid:173)
`tion time to center the process at the
`desired thickness target.
`Simple models are effective in both
`linear LPCVD deposition processes and
`nonlinear oxidation processes. In our work
`with oxidation processes, a process capa(cid:173)
`bility index (Cpk) improvement of 27"k
`was achieved usingasimplefeedbackcon(cid:173)
`troller (Fig. 2).
`Feedback adjustments are useful in
`etch processes to control CDs. Many etch
`processes use in situ end-point detection.
`Once end point has been established, the
`:recipe continues to etch the film for a pre(cid:173)
`defined over-etch lime. The impact of the
`end-point and over-etch times is mea(cid:173)
`sured in film thiclrnes& and CDs. Auto(cid:173)
`matic feedback control can be applied to
`adjust timed etch processes or over-etch
`time in end-point driven processes. The
`relationship between film thickness
`removed and CDs to etch process
`continued on PB,Je 52
`
`FJpre 2. Oxidation process unttormily a) without and ii) will! run-to-run process control.
`
`110 Solid State Technology • DeMndler 1899 • www.solld-state.com
`Copyright © 1999. All rights reserved.
`
`PDF Solutions v Ocean Semiconductor, IPR2022-01196
`PDF Exhibit 1007, Page 3 of 4
`
`

`

`Run-lo-nm praCNS ccnmot conlirn.100 from pBge 50
`parameters such as etch time,
`gas flow, and power can be
`modeled and controlled.
`
`FoecM'orwanll
`control ex-pies
`Although a process may be
`able to produce repeatable
`n.'Sults using feedback control
`methodology, process results
`may also be dependent on the
`initial state of wafers. This
`information can be automati(cid:173)
`cally provided through fucd(cid:173)
`furward modeling, However,
`prior to implementing any
`feed-forward t\.-chnique, the
`proL-css must be inherently sta(cid:173)
`ble or must use an effective
`feedback mechanism to pro-
`vide stability.
`One example where a fet.>d-forward control mechanism is
`useful is the adjustment of etch time to remove an interlayer
`dielectric for a via interconnect. This type of etch process
`may not be controlled with in situ end-point detection, because
`the small amount of film being removed does not provide the
`needed signal strength, Instead, the process should remove all
`of the film in the first attempt, but the initial film thickness is
`re-quired to select the target for the process.
`In CMP prtK.-esses, tht.>re are typical variations in initial sur(cid:173)
`face materia I that result in similar variations after the polish.
`By measuring wafers prior to polish, a feed-furward controller
`can adjw,1 the feedback controller target (amount of material
`lo be polished} after each run to minimize or eliminate these
`variations.
`Implant barrier variations can adversely affect the gain of a
`transistor. In this example, oxide and nitride layers are grown
`and deposited, respectively, on the wafer. Using lithography
`and etch proces.<;e;;, trenches are formed in the nitride layer. In
`forming the trench, the nitride is over-etched, resulting in
`removal of some of the initial oxide layer. A sacrificial oxide
`layer i'> grown in the trench over thi."> initial oxide, making a bar(cid:173)
`rier for the implant step. Thisirnplantbarriervariesrun-ter-run
`due to the over-etch of the nitride and the variations of the ini(cid:173)
`tial oxide growth. The incoming variation caused by these
`etch steps can be minimized by first measuring the initial oxide
`layer after the nitride etch and adjusting the target of the feed(cid:173)
`back controller on the sacrifida! oxide process to maintain a
`more consistent implant barrier (Eig. 3).
`fnphotolithography, feed-forward control can be used tocal(cid:173)
`t,-ulate alignment parameters from wafer to wafer. These are
`typically six to eight vararncters that can be predicted with firbt(cid:173)
`order models. Tilt can also be adjusred, but requires a much
`more complex model.
`
`Conclusion
`A,; factories look for new and innovative ways to reduce man(cid:173)
`ufacturing costs, nm-to-run process c-.mtrol solutions become
`increasingly important to squeeze the most performance out
`of processing tools. lt has been shown that run-to-run proces.q
`t,~mtro! provides significant process uniformity improvements,
`reduced process maintenance costs, and improved through(cid:173)
`put, leading to a lower cost of manufacturing,
`
`R&Ure 3. Feed-forward adjustment of sacrificial oitidli growth to raduCII implant
`barrier variation attributed to initial oxldatton and etch proo;ss variation.
`
`Our development and u.,e of
`the ARRC system shows that it
`provides a seamless architec(cid:173)
`ture for factory-wide run-to-run
`process control. [ts tools allow
`users with little or no modeli.ng
`and control expertise to create
`input-output relation.">hips for
`any process and control them
`to a desired larg<.-'1:. System flex(cid:173)
`ibility allows users lo imple(cid:173)
`ment more complex modeling
`and control methodologies
`using nplug-in" modules.
`We have presented here a
`few example processes in
`CMP, diffusion, l-'tch, and pho(cid:173)
`tolithography that can bene(cid:173)
`fit from feed-forward and
`feedback prmcess control, but
`applications are limited only by the imagination of the process
`II
`engineer,
`
`Acknowl~ta
`Additional authors indude Mike Simpson and Brian Cusson
`of Advanced Micro Devices, and Pradeep Swamy, Brad
`Schulze, and Kevin Dimond of SEMY Engineering. This
`project was sponsored in part by SEMATEGI's Equipment
`Productivity Improvement Team program. The authors thank
`Abhljil Bora, Tony Colombo, Keith Laidlaw, Tru:ig Mutlag,
`Scott Say, Ashok Tripathl, and Renran Yiu of SEMY Engi(cid:173)
`neering, and Tom Timmons of Advanced Micro Devices for
`their invaluable contributions. MATLAB is a registered trade(cid:173)
`mark of The Math Works Inc. Equipment Supervisor Work(cid:173)
`station and Advanced Run-to-Run Control are registered trade(cid:173)
`marks of SLIMY Engineering.
`
`R.rer-•
`1. R. Patty, "A More Rebusi and Reliable Statistical Process Control Systom,"
`ITTtarr-,otional Symposium on Samiccnductor Manufat:turing, 1999.
`2. K. Stoddard et al., "llpplica1ion of Feed·f-afWard and -ptive Faedback Coo-
`1n:» to Samir,oruJuctOl Device Manufac!Uling," Aniertcan Gontml COflference.
`Baltimore, Maryland, ,ltX:e 1994.
`3. N, Zho ot al., "A Gompwalive Analysis of fiun-lo-Hun Control A!gorilhrrs 01
`the Semiconductor Mamlacluring lndu•try," IEEE/SEMI 1996 Advancod Semi-(cid:173)
`conductor Manufacturing Conforenco & Workshop, pp, 375'··361, New
`York, 19<J6.
`4. M. Miller, ·rromAPG Pik~ toFuilPmduc'.lon Sysrom-Scaling lJpis Hara to
`Do," /lff'JAPC Symposium Xl, Vai~ Colorado, Septamber 19W.
`5. M, Yolverton et al., ·11 Comotete Furnace Control Pl8tf0fm for High-Vol,,me
`Manufacturing," lntemational Symposium on Somiconc!uctor Manufacturing,
`Tokyo, Japan, Octob<lr 1998.
`6, T. Dow<l, ·Results of Af'.C/Af'G Deployment ct Whtto Oak Semiconductor,·
`Af'.G/APC Symposil•n Xl, Vaii, Colorado, September 1900,
`7. M. Hankinson et al., "Integrated Real 11me and Run-to-Run Control of Etch
`Depth in Reactive lor. Etchir:g, • IEEE Transactions Som/conductor Mant ilac(cid:173)
`tur.ng, Vo!. 10, No, 1, pp. 121-130, Febroory 1997.
`8, T ,H. Smirt, et al., "Run-by-Run Advanced Proooss Control of I\Aetal Sputter
`Dooosition,,. iEEE Transactions Sernk;onductor Manufacturing, Vol. : -~, No.
`2, pp, 276-284, May 1998.
`Using f¼al-llmo Tool Da~~.- lffE Tmnsactions SemicD(llR./Clor Manttfuc:l!Jr(cid:173)
`9, S.F. Leo. C.,L Sparu:-.s, "Prodlc:lor, of Waler State After Po,s:na i'rOCP.ssing
`
`ing, Vol. 8, No. 3, Al~~lh11995,
`
`For more information, contad Kevin Stoddard at SEMY £11~eer(cid:173)
`i11g Jr,c., 2340 West Shangri La Rd., Phoenix, A7. 85029; ph 602/861-
`9395,Jax 602/861-.1442, e-mail kslodaard®"emy.mm,
`
`Copyright © 1999. All rights reserved.
`112 Solid State "fcchnolcgy • December i999 • www.solki-stats.com
`
`PDF Solutions v Ocean Semiconductor, IPR2022-01196
`PDF Exhibit 1007, Page 4 of 4
`
`