Reduction of Defects in the Lapping Process of the Silicon Wafer

Engineering Management in Production and Services Volume 11 • Issue 2 • 2019

received: 5 December 2018 accepted: 5 June 2019 Reduction of defects in the lapping pages: 87-105 process of the silicon wafer manufacturing: the Six Sigma application

Mithun Sharma, Sanjeev P. Sahni, Shilpi Sharma

A B S T R A C T Aiming to reduce flatness (Total Thickness Variation, TTV) defects in the lapping process of the silicon wafer manufacturing, it is crucial to understand and eliminate the root cause(s). Financial losses resulting from TTV defects make the lapping process unsustainable. DMAIC (Define, Measure, Analyse, Improve and Control), which is a Six Sigma methodology, was implemented to improve the quality of the silicon wafer manufacturing process. The study design and the choice of procedures were contingent on customer requirements and customised to ensure maximum satisfaction; which is the underlying principle of the rigorous, statistical technique of Six Sigma. Previously unknown causes of high TTV reject rates were identified, and a massive reduction in the TTV reject rate was achieved (from 4.43% to 0.02%). Also, the lapping process capability (Ppk) increased to 3.87 (beyond the required standard of 1.67), suggesting sustainable long-term stability. Control procedures were also effectively implemented using the techniques of poka yoke and control charts. This paper explores the utility of Six Sigma, a quality management technique, to improve the quality of a process used Corresponding author: in the semiconductor industry. The application of the Six Sigma methodology in the current project provides an example of the root cause investigation methodology that Mithun Sharma can be adopted for similar processes or industries. Some of the statistical tools and techniques were used for the first time in this project, thereby providing new analysis O.P. Jindal Global University, India and quality improvement platform for the future. The article offers a deeper e-mail: [email protected] understanding of the factors that impact on the silicon wafer flatness in the lapping process. It also highlights the benefits of using a structured problem-solving Sanjeev P. Sahni methodology like Six Sigma. O.P. Jindal Global University, Indiea e-mail: [email protected] K E Y W O R D S Six Sigma, quality, silicon wafer, lapping, quality control, total thickness variation, wafer manufacturing Shilpi Sharma DOI: 10.2478/emj-2019-0013 O.P. Jindal Global University, India e-mail: [email protected]

Introduction business and $5000 billion in services, representing The semiconductor industry consists of compa- close to 10% of the global GDP, thus gaining the rec- nies engaged in the design and fabrication of semi- ognition for its critical role in the supply chain of the conductor devices, which form the foundation of electronics industry (Kazmierski, 2012). modern electronics. The industry began in the 1960s The silicon wafer manufacturing involves several and currently accounts for about 0.5% of the global stages. The semiconductor industry consists of three GDP ($299.5 billion). It also enables the generation of key sectors: silicon ingot growing, the silicon wafer approximately $1200 billion in the electronic systems manufacturing and the fabrication of integrated cir-

87 Engineering Management in Production and Services Volume 11 • Issue 2 • 2019 Tab. 1. Four key stages of Silicon wafer manufacturing proces Tab. 1. Four key stages of silicon wafer manufacturing process

PROCESS FLOW DESCRIPTION PROCESS OBJECTIVE PROCESS DEMERITS Slicing Silicon ingot is mounted on the slicing machine. A To generate wafer slice structure. Poor flatness High web of meatl wire along with cutting slurry passes To achieve correct wafer surface damage High through the ingot slowly to provide disc shaped thickness, orientation and warp contamination silicon wafers

Lapping Sliced wafers held in carriers are placed in between two metal plates along with abrasive slurry falling Reduce slicing damage. Establish Susceptible to surface on wafers. Rotation of plates and carriers causes optimum wafer flatness and flatness rejects mechnical removal of slicon therby reducing wafer Chemical Etching Lapped wafers are subjeted to abrasive chemicals which disolves silicon into chemical thereby thinning Reduce Previous Process Damage Degrades flatness wafers Reduce surface contamination wafer staining

Polishing Etched wafers are mounted on plates. With Establish optimum flatness. pressure applied on these plates, the wafers are High rejection rate Re- Eliminate surface damage. rubbed against fine abrasive polishing pads along work cost Minimise contamination level with polishing slurry to give a mirror like finish

cuitTab. 2(IC) Overview chips. of keyThe quality silicon management ingot growing concepts process is 1. Literature review contingentConcept on manyOrigin factors, suchAim as size, specifica- Focus Methodology and Criticisms tions and quality, and so the ingot growing time can Tools Japan (the Gap analysis of Skilled workers rangeTotal Productive from a week to a month. IncreaseThe next process stage capability is the PreventiveQuality and is an elusive and abstract concept (Hos- Maintenance 1950s) by reduction of unplanned predictive historical records, required to implement, wafer production, whereby a failures,fully-grown accidents silicon and maintenancesain, Tasnim of &cause Hasan,-effect 2017; analysis Wright, resource 1997). demanding In a study ingot is sliced into wafers of differentdefects thicknesses. processesby Evans and Dean (2002), managersand longfrom-term 86 firms in This sliced wafer is then subjected to the flattening the United States were asked to define “quality”. They found several definitions ranging from waste elimi- process,Total Quality and Control then Japan undergoes (the the fabrication process To reduce Methodology: Plan Do Vague and difficult to producing IC chips.1960s) The focus ofTo this coordinate research quality project reworknation, and conformityStudy Act to Tools:customer requirements,coordinate policy is on the silicon wafer manufacturing.maintenance and achievecompliance, consistencyStatistical techniques in output, getting it right the improvement from all maximumfirst time and customer satisfaction. Broadly, quality The silicon wafer manufacturinggroups toconsists achieve theof fourmost customer key value-adding processes: slicing,economical lapping, process chemical satisfactioncan be understood along four dimensions: excellence, etchingTotal andQuality polishing. Japan The(the primaryImprove aim the of quality the slicing and Customervalue, conformanceMethodology: to specifications Plan Do Vague and and the extent to Management (TQM) 1990s) consistency of processes satisfactionwhich expectationsStudy Act are Tools: met (Yonginconsistent & Wilkinson, process is to define the crystal structure of the wafer, Statistical techniques conceptualisation, finding the best possible shape. The wafer is subjected 2002). Quality as excellence is believedexcessive to resource be immea - to high-pressure cutting, to achieve correct wafer surable, and control can only be exertedconsumption, through the investment of maximum effortsunsatisfactory and best skill sets. thickness, but on the downside, it causes high surface resu lts damage and contamination. The second stage is lap- This approach has been criticised of being of little ping,Zero Defects which removesDenver the Division surface To damageenhance the caused quality byof a Defectpractical valueExtra to attentionan organisation and Expensive (Garvin, 1988). of the Martin process outcome through eliminationQuality as a carevalue devoted definition to each focuses on external slicing. It is also criticalMarietta in definingthe elimination the flatness of any of the step of the production wafer as a failure Corporationto achieve (the the defects optimum in the wafer production flat- effectiveness process(meaning ensuring costs) no and internal efficiency. ness can lead to complete1960s) wafer processrejection. Wafers are This definition,mistakes once again, has been argued to lack then exposed to abrasive chemicals during the practical and measurable parameters for organisa- chemicalReliability etching stage that helpsTo toreduce remove failure impuri modes - Longevitytional and use (YongReliability & Wilkinson,-Centred 2002).Technical A more and requires quanti - tiesengineering from the wafer surface. Finally, wafers that have dependabilityfiable andof Maintenance, objective failuredefinition skilled staff. of Expensivequality is been manufacturedShewart to the during required flatness standards, parts,conformance products and modesto specifications; and effects, andwhich demands was dominantlong term the 1920s and systemsin the 20th century.root cause It analysis,means thatcommitment an outcome must with no surface damage,the 1930s: are cited polished on one side to condition- based give a smooth, mirror-likein Kapur and finish, on which IC chips not deviate frommaintenance the specifications set by an organisa- can be fabricated.Lamberson This process flow is further illus- tion, and any deviations are considered as lowering (1977) the quality (Reeves & Bednar, 1994). Such confor- trated in Tab. 1. The lapping process helps to achieve the maxi- mance measures have been argued to increase the mum wafer flatness. Wafer flatness measured as Total internal efficiency of the process and sale prices over Thickness Variation (TTV) in microns is one of criti- time (Topalovic, 2015). Later, as the focus shifted cal-to-quality (CTQ) requirements for a silicon wafer. from manufacturers to customers, a new definition of This project aimed to reduce the number of TTV “quality” emerged, i.e. meeting or exceeding customer rejects in the lapping process. expectations. Anything that does not satisfy the customer was considered to be of low quality (Yong &

88 Engineering Management in Production and Services Volume 11 • Issue 2 • 2019

Wilkinson, 2002). However, customer requirements tions have documented unsuccessful implementation are variable and subjective, which makes them diffi- stories of the TQM concept in the manufacturing cult to satisfy. These intricacies associated with the industry (e.g., Brown et al., 1994; Eskildson, 1994; concept of quality, highlight the need to have a man- Cao et al., 2000; Nwabueze, 2001). Based on the evi- agement system that ensures the manufacturing of dence from independent publications by consulting good-quality products resulting in monetary profits firms, it can be argued that two-thirds of TQM imple- and customer satisfaction. This leads to the concept mentation efforts failed to produce any significant of quality management. improvements in the overall quality of the product, The revolution in quality management began in financial gains or the company’s competitive situation Japan during the 1950s and gradually gained momen- in the industry (Jimoh et al., 2018). This is partly due tum in the rest of the world during the 1970s and the to the ever-changing and evolving definitions of 1980s (Foley, 2004). One of the most influential qual- TQM, which can mean different things to different ity movements during this era was the total quality people, making its implementation insufficiently management (TQM), which again owes its origin and consistent and reliable (Andersson et al., 2006; conceptual developments to Japan (Cole, 1998; Esaki, Boaden, 1997; Talapatra, Uddin & Rahman, 2018). 2016; Juran, 1995). TQM eliminated the weaknesses Two other branches of quality management were of previous quality improvement techniques, which introduced during the 1960s: reliability engineering makes it an efficient quality management tool. TQM and zero defects. The reliability engineering tech- is an integrated approach that stresses the top-down nique has roots in the disciplines of pure probability approach, staff engagement in the process, evidence- and statistics. It was mostly used in the USA and based decision-making and the consideration of cus- aimed to apply principles of probability to reduce tomer requirements (Tobin, 1990). Despite its defect rates of durable products (Dimitri, 1991). The influential and successful reputation, many publica- zero defects strategy, which originated in the USA

Tab. 2 Overview of key quality management concepts METHODOLOGY AND CONCEPT ORIGIN AIM FOCUS CRITICISMS TOOLS

Increase process capability Preventive and Skilled workers Gap analysis of Total Productive Japan (the by reduction of unplanned predictive required to implement, historical records, Maintenance 1950s) failures, accidents and maintenance of resource demanding cause‐effect analysis defects processes and long‐term

To reduce To coordinate quality rework and maintenance and Methodology: Plan Do Japan (the achieve Vague and difficult to Total Quality Control improvement from all Study Act Tools: 1960s) maximum coordinate groups to achieve the most Statistical techniques customer economical process satisfaction Vague and inconsistent Methodology: Plan Do conceptualisation, Total Quality Japan (the Improve the quality and Customer Study Act Tools: excessive resource Management (TQM) 1990s) consistency of processes satisfaction Statistical techniques consumption, unsatisfactory resu lts

Denver Division To enhance the quality of a Extra attention and of the Martin process outcome through Defect care devoted to each Zero Defects Marietta the elimination of any step of the production Expensive elimination Corporation (the defects in the production process ensuring no 1960s) process mistakes

Shewart during Reliability‐Centred the 1920s and Longevity and Maintenance, failure Technical and requires Reliability the 1930s: cited dependability of modes and effects, skilled staff. Expensive To reduce failure modes engineering in Kapur and parts, products and root cause analysis, and demands long term Lamberson systems condition‐ based commitment (1977) maintenance

89 Engineering Management in Production and Services Volume 11 • Issue 2 • 2019 during the 1960s, was considered to be the most yearly savings (Su & Chou, 2008; Yang & Hsieh, optimistic approach in the quality management field 2009). as it aimed to achieve the complete elimination of Besides, in the last decade or so, there has been defects and process failures (Crosby, 1979). Both a massive uptake and implementation of the Six concepts, however, received criticism for being Sigma technique as a process change, management impractical and expensive (Crosby, 1984). It can, and improvement strategy by global industries, which thus, be argued that the different quality management include the manufacturing process (Al-Aomar, 2006; concepts differ in their origin, aim, definitions, meth- Gangidi, 2019; Valles et al., 2009), financial organisa- odology and focus, thereby confusing rather than tions (Brewer & Eighme, 2005), engineering firms informing the readers. Furthermore, each of the (Bunce et al., 2008), hospitals and intervention clinics techniques cited above has limitations that reduce (Craven et al., 2006), banking, hospitality, pharma- their applicability and anticipated benefits (Kedar et ceutical companies (Cupryk et al., 2007), chemical al., 2008). Many organisations have reported difficul- industries (Doble, 2005), educational institutions ties in the implementation of quality management (Bandyopadhyay & Lichtman, 2007), software indus- programmes (Brown et al., 1994; Eskildson, 1994; try (Arul & Kohli, 2004), call centres (Schmidt Harari, 1997; Nwabueze, 2001); they suggested the & Aschkenase, 2004), utility service providers (Agar- lack of inherent connectivity between parts and also wal & Bajaj, 2008), the automobile sector (Gerhorst et reported some missing information about some of al., 2006), information technology (Edgeman et al., the relevant sections, as shown in Tab. 2. 2005), human resources departments (Wyper & Har- None of the quality management tools discussed rison, 2000), military administration units (Chappell so far had global success, and quality managers were & Peck, 2006) and even government departments still in search of a complete quality management (Furterer & Elshennawy, 2005). A summary is pre- programme when Six Sigma arrived. Six Sigma is sented Tab. 3. a systematic set of guidelines that aims to significantly The origins of Six Sigma can be found in statis- improve the quality of a manufacturing process and tics. The term Six Sigma owes its origin to the termi- reduce costs by minimising process variation and nology employed in the statistical modelling of reducing defects. It utilises statistical tools that can manufacturing processes. A Six Sigma process is the either be applied to facilitate a new product develop- one that produces 3.4 defects or non-conformances ment or strategic process improvement (Breyfogle et per million opportunities (DPMO). A defect is any- al., 2001). Six Sigma is the edge that helps win the thing that does not conform to the manufacturer’s market competition as it provides financial, business guidelines or customer’s specifications, and an oppor- and personal benefits; financial — by optimal and tunity is any chance for this defect to happen. The efficient use of resources; business — through ensur- sigma level, also known as a Z-value, is used as ing maximum customer satisfaction; and personal a capability index for the process, which indicates — enhancing skills of an individual and, thereby, how well that process can meet the customer’s increasing their employability. In the last decade or requirements (Bothe, 2001; Da Silva et al., 2019). so, there has been a rapid uptake of the Six Sigma Each sigma level corresponds to a certain number of technique as a process change, management and defects/non-conformances that are associated with improvement strategy by global industries. This the process, as shown in Fig. 1. helped them beat market competition and maximise

Sigma (σ) Defects / Non-Conformances per Level Million Oppurtunity (DPMO) 2 σ 308,537 3 σ 66,807 4 σ 6,210 5 σ 233 6 σ 3.4

Fig. 1. Sigma levels depending on DPMO

Fig. 2. Y = f (x) cascade

Process Capability of TTV for Tight Spec Material Process Capability of TTV for Loose Spec Material

LB USL LB USL Process Data Within Process Data Within LB 0 Ov erall LB 0 Ov erall Target * Target * USL 2 Potential (Within) C apability USL 3.5 Potential (Within) C apability Sample Mean 1.21547 Cp * Sample Mean 2.11004 Cp * Sample N 5623 CPL * Sample N 28120 CPL * StDev (Within) 1.11571 CPU 0.23 StDev (Within) 0.78286 CPU 0.59 StDev(Overall) 1.22691 C pk 0.23 StDev(Overall) 0.88975 C pk 0.59 O v erall C apability O v erall C apability Pp * Pp * PPL * PPL * PPU 0.21 PPU 0.52 Ppk 0.21 Ppk 0.52 C pm * C pm *

-0.0 1.4 2.8 4.2 5.6 7.0 8.4 9.8 0.0 1.3 2.6 3.9 5.2 6.5 7.8 9.1

O bserv ed P erformance Exp. Within Performance E xp. O v erall P erformance O bserv ed P erformance Exp. Within Performance E xp. O v erall P erformance PPM < LB 0.00 PPM < LB * PPM < LB * PPM < LB 0.00 PPM < LB * PPM < LB * PPM > USL 98168.24 PPM > USL 240975.32 PPM > USL 261269.05 PPM > USL 33499.29 PPM > USL 37907.98 PPM > USL 59120.69 PPM Total 98168.24 PPM Total 240975.32 PPM Total 261269.05 PPM Total 33499.29 PPM Total 37907.98 PPM Total 59120.69 Fig. 3. Process capability of TTV for Tight and Loose Spec Material

Engineering Management in Production and Services Volume 11 • Issue 2 • 2019

Tab. 3. Selected success stories of the Six Sigma implementation in industries

AUTHORS NAME OF AN ORGANISATION BENEFITS OF IMPLEMENTING THE SIX SIGMA TECHNOLOGY Financial sector Citibank group Numerous benefits have been reported across different organisations of this group. They successfully halved their credit processing time and reduced internal call‐back time by 80% and external by 85%. Reduced the time between a customer first placing an order till the actual service delivery and the credit Rucker decision cycle from 3 days to just 1 day. The time taken to process a statement (2000) was also decreased from 28 days to 15 days only JP Morgan Chase (Global Investment Improved customer experience in using bank's services such as account Banking) opening, balance enquiry, making transfers and payments via online or cheque mode; leading to increased customer satisfaction and a reduction in process cycle time by more than 30% Antony British Telecom wholesale Financial benefits of over $100 million, greater customer satisfaction, error (2006) reduction Bank of America 24% reduction in customer complaints and a 10.4% increase in customer Roberts satisfaction (2004) Sun Trust Banks Significant improvement in customer satisfaction American Express Improved the external vendor related processes and reduced the numbers of Bolt et al. (2000) non‐received renewal cards Manufacturing sector Motorola (1992 and 1999) 1992: Achieved dramatic reduction in the defect levels of their process by about 150 times 1999: Huge financial gains of about $15 billion over 11 years Honeywell Profit of $1.2 billion Antony Texas Instruments Achieved a financial gain of over $ 600 million (2006) Johnson and Johnson The financial gain of about $500 million Telefonica de Espana (2001) Whooping increase in revenue by about 30 million in the first 10 months and gain in savings too Dow chemical/rail delivery project Reported substantial savings in capital expenditures: of over $2.4 million AlliedSignal/ Bendix IQ brake pads The cycle time of their production‐shipment process decreased by 10 months (18 to 8 months) AlliedSignal/ Laminates plant in South Reaped a range of benefits: their capacity almost doubled, and punctuality in Carolina delivering goods reached 100% threshold level. Their cycle time and inventory had a reduction of 50% each DuPont/Yerkes plant in New York Increase in yearly savings of over $25 million McClusky (2000) (2000) Seagate Technology Gained financial profits of about £132 million in just 2 years General Electric Increase in yearly financial savings by about $2 billion Hughes Aircraft's Missiles Systems Group / Wave soldering op. The quality of their yields improved by about 1000% and productivity by 500% Raytheon/ Aircraft Integration Achieved a significant reduction (approx. 88%) in the inspection time spent on Systems the depot maintenance process GE/ Railcar leasing business Achieved a 62% reduction in the time spent at repair stops McClusky (2000) Healthcare sector Radiology film library, Anderson Service quality improved Preparation and waiting times for patients reduced from 45 min to 5 min Benedetto cancer centre Dramatic increase by 45% in daily numbers of examinations without an increase (2003) Outpatient CT exam lab at the in workforce/ equipment University of Texas Engineering and Construction sector General Electric 1997: made a profit of $320 million which was more than double their goal of Byrne(1998) $150 million, further in 1999, annual savings of 2 billion

Magnusson et Volvo cars (Sweden) Profit of over 55 million euro in the years 2000 and 2002 al. (2003)

Business Unit of Transmission and Savings of over 200 million euro between the years 1997‐2003 Anderson et al. Transportation Networks at Ericsson (2006)

91 Engineering Management in Production and Services Volume 11 • Issue 2 • 2019

Each manufacturing process has set specification outcome. This gives FED an edge over the OFAT limits for a process and product quality. If six stan- technique, wherein only one factor can be evaluated dard deviations can be managed between the statisti- at one time. cal mean of a process and its nearest specification Many studies have compared OFAT and FED — limits, then all aspects of that process would meet the the two problem-solving techniques — and FED is specification criteria. This distance between the pro- considered to be more effective than OFAT for the cess mean and the specification limit is measured in following reasons (Czitrom, 1999): sigma units and is known as the process capability. • In FED, investment of comparatively fewer The process capability measurement index is the resources (time, money and material) results in process performance index (Ppk). Once a process has greater and more accurate information. This been brought under statistical control through the makes it extremely useful in industries, where implementation of a Six Sigma project, Ppk estimates time and financial costs of running a process are how stable these improvements would be in the long- extremely high; term and how closely they would meet customer • Interactions between factors cannot be identified expectations. The larger the Ppk value, the less is the using OFAT technique, which uses trial and error process variability and the higher the long-term sta- method; whereas the FED technique provides bility. To satisfy customers, the Ppk value should be a systematic procedure for estimating interac- greater than 1.67 (Kotz, 1993; Raman & Basavaraj, tions between several factors; 2019). • Each observation carried out during a factorial The literature indicated that even if a process experiment considers all the factors and their achieved a high sigma level over short-term, its per- interactions, which estimate the effect of factors formance could have declined over long-term; and much more precisely. In contrast, OFAT typically a common research finding is that it might fall from uses only two observations to measure the effect Six Sigma to the 4.5 sigma level (Alexander, Antony of one factor; these estimations are subject to & Rodgers, 2015; Pandey, 2007). This could happen greater variability. because the process may ‘drift’ over time. Such shift- The current study used the FED technique to ing can further lead the process mean to move away address the problem of high TTV rejects in the lap- from the target, thus reducing the number of standard ping process. At a broader level, the Six Sigma meth- deviations that can fit between the process mean and odology was applied for improving the output quality the closest specification limit. This is commonly of the lapping process. Six Sigma emphasises the need known as a 1.5 sigma shift. So, the standardised defi- to identify and clearly define customer requirements nition of the Six Sigma quality considers this shift and and internal industrial factors for setting goals. The guarantees that a six-sigma process will produce no data-driven rigour component of the Six Sigma more than 3.4 DPMO (Antony, Snee & Hoerl, 2017; approach delineates objective decision-making Harry, 1988). purely guided by the statistical analysis of data to Some previous attempts to resolve the problem of determine process strengths and weaknesses. It is high TTV failed. The problem of obtaining high TTV crucial to this approach that a solution is not offered rejects had been a recurrent issue at the current until the problem has been clearly and completely organisation, for a significantly long time. Earlier, the defined (Ishikawa, 1985; Kume, 1985, 1995; Hoerl, traditional problem-solving technique called One 1998; Sreedharan et al., 2019). Factor At a Time (OFAT) had been applied in an attempt to reduce TTV rejects. OFAT is an experimental technique that evaluates the impact of poten- 2. Research methods tial factors on the process outcome, one at a time while keeping other factors constant. However, these The Six Sigma methodology was applied to attempts failed to identify the root cause of the cur- resolve the issue of high TTV rejects in the wafer lap- rent problem. Further expert consultations and sys- ping process, of the semiconductor manufacturing tematic research review indicated a better potential of industry. The problem of the current project was for- the technique called Factorial Experimental Design mulated as follows: the mandatory replacement of (FED), as compared to the OFAT strategy. FED evalu- slurry in the lapping process results in poor wafer ates the effect of more than one factor together with flatness causing TTV rejects to increase from 0.1% to their interactions, simultaneously on the process 4.43% or a loss of £58k/month.

92 Engineering Management in Production and Services Volume 11 • Issue 2 • 2019

The overall aim of this research was to evaluate according to customer’s preferences, these are the effectiveness of Six Sigma aiming to improve the known as Critical to Quality (CTQ) metrics; quality output of the lapping process in the silicon • Measure: at this stage, in accordance with the set wafer manufacturing industry. goal and CTQ specifications, further measure- Specific research objectives were to test the utility ments of key elements for the concerned process of the Six Sigma methodology in: are carried out; • Identifying the factors responsible for high TTV • Analyse: this stage includes an in-depth and sci- defect rates; entific investigation of collected data, statistical • Implementing sustainable long-term process tools are utilised to identify and assess the statis- improvements that will reduce the defect rate to tical significance of associations between various <0.1%; aspects of a process. Any findings should be • Increasing the Lapping Process Performance objective, valid and justifiable by means of data. index (Ppk) to >1.67; Analyses results help reveal the root cause of the • Delivering at least a £25k saving to the company defect in the manufacturing process. Results by the end of the project. from the Analyse stage may also be treated as The rationale for the project selection was based pilot results which are documented to facilitate on the experience of one of the authors of this paper, replication of the same process design in the who worked as a process engineer in the world’s lead- future; ing semiconductor wafer manufacturing company • Improve: the results from the analyses were and received black belt level in-service training in the applied to eliminate the root cause of defects in Six Sigma methodology. The issue of high TTV a process, thereby improving the overall quality rejects had been causing huge financial losses (> of the outcomes; £50k/month) to the organisation and was topping the • Control: the methodology and results of each priority list of the senior management. It was, thus, stage are clearly spelt out in sufficient detail to considered necessary to apply the statistically vali- allow the replication in future processes to iden- dated, well-known, effective strategy of Six Sigma to tify and correct early errors and prevent a finan- address this problem. A team of the appropriately cial loss due to defects in the yield. At this stage, qualified technical staff was delegated to undertake Six Sigma tools of poka yoke, statistical process this task, under the leadership of a trained Six Sigma and quality control charts, and control plan were black belt specialist, the first author of this article. used. DMAIC, which is a Six Sigma process, was employed for achieving the above stated objectives. Six Sigma, inspired by Deming’s Plan-Do-Check- 3. Research results Act cycle, has two popular methodologies, namely, DMAIC and DFSS. The DMAIC methodology is uti- Stage 1 — Define: the key aim of this phase was to lised for improving an existent process whereas DFSS mutually agree on a clear and concise problem state- — for the development of a new product. The current ment, gain a fuller understanding of the process and investigation followed the DMAIC principles, follow- identify the manageable focus area. ing the five stages: To clearly define and communicate the issue at • Define is the most important and critical stage of hand, the following problem statement was developed the Six Sigma process. First and foremost, project based on IS-IS Not Analysis: the mandatory replace- scoping and mapping is carried out that helps ment of slurry in the lapping process is resulting in explain the basic problem to all the team mem- poor wafer flatness causing TTV rejects to increase bers. Then, current process defects are defined from 0.1% to 4.43% or a loss of £58k/month.

Tab. 4. Critical to Quality metrics of the lapping process TYPE NAME VOC CTQ External Polishing No sharp roll‐off at the edge of wafer For Loose Spec TTV < 3.5 µm For Tight Spec TTV < 2.0 µm Internal Lapping Comparable yield to old active agent TTV reject < 0.10 % Good Flatness For Tight Spec TTV < 2.0 µm

Tab. 5. Data collection plan ITEM NO. WHAT WHY WHEN HOW WHERE WHO

1 Wafer Traceability To correlate poor TTV Every Lot Data capture Central Database ‐ Lapping Op 93 with different application Browser Lapping ADE, Data 2 TTV KPOV, CTQ Every wafer Central Database CW Insp. Op upload Every time fresh 3 Slurry Mixing Time KPIV Machine Setting (SOP) Slurry Sheet Lapping Op slurry prepared 4 Slurry Density KPIV Every time fresh Manual Measurement Control Charts Lapping Op slurry prepared 5 Machine flowrate KPIV Every time loop Manual Measurement Control Charts Lapping Op was changed KPIV (can affect wafer Every 40 hr of 6 Plate Shape Using dial gauge Plate Shape Sheet Lapping Op shape) Opertion Time 7 Recycle Slurry To evaluate impact of Every 5 mins Data capture BMS Database PSE Op Status different slurry application Active Agent KPIV (can affect slurry Every time fresh 8 Machine Setting (SOP) Slurry Sheet Lapping Op Volume viscosity) slurry prepared KPIV (can affect wafer Any time changed 9 Sun Gear Ratio Machine Setting (SOP) QA Records Engineer rotation) by engineer KPIV (can affect wafer Any time changed 10 Bottom Plate Speed Machine Setting (SOP) QA Records Engineer rotation) by engineer KPIV (can affect wafer Any time changed 11 Exhaust Timer Machine Setting (SOP) QA Records Engineer rotation) by engineer KPIV (can affect wafer Any time changed 12 Acceleration Timer Machine Setting (SOP) QA Records Engineer rotation) by engineer Slurry KPIV (can affect slurry Every time fresh 13 Machine Setting (SOP) Slurry Sheet Lapping Op Temperature viscosity) slurry prepared KPIV (can affect slurry Lot Processing 14 Plate Temperature At the start of shift Digital thermometer Lapping Op viscosity) Sheet

Tab. 6. Key results from process capability charts

TIGHT FLATNESS LOOSE FLATNESS PARAMETERS COMBINED SPEC SPEC

Wafer Qty 5,623 28,120 33,743 TTV Reject % 9.82 3.35 4.43 DPMO 98,168 33,499 44,282 Process Capability 0.21 0.52 ‐NA‐ (Ppk)

Tab. 7. Response Table for Means

SUN GEAR BOTTOM PLATE EXHAUST ACCELERATION PLATE RECYCLE SLURRY ACTIVE AGENT SLURRY SLURRY LEVEL RATIO SPEED TIMER TIMER TEMP. STATUS CONCENTRATION TEMP. MIXING TIME 1 1.2893 1.0338 1.0053 1.0092 1.1997 0.9940 1.0120 0.9957 1.0062 2 0.7123 0.9678 0.9963 0.9925 0.8020 1.0077 0.9897 1.0060 0.9955 Delta 0.5770 0.0660 0.0090 0.0167 0.3977 0.0137 0.0223 0.0103 0.0107 Rank 1 3 9 5 2 6 4 8 7

Tab. 8. Displaying the impact of improvements on TTV rejects BEFORE IMPROVEMENT AFTER IMPROVEMENT TIGHT FLATNESS LOOSE TIGHT FLATNESS LOOSE PARAMETERS COMBINED COMBINED SPEC FLATNESS SPEC SPEC FLATNESS SPEC Wafer Qty 5,623 28,120 33,743 19,677 50,666 70,343 TTV Reject % 9.82 3.35 4.43 0.02 0.01 0.01 DPMO 98,168 33,499 44,282 152 118 127 Process Capability (Ppk) 0.21 0.52 ‐NA‐ 3.87 1.30 ‐NA‐

Engineering Management in Production and Services Volume 11 • Issue 2 • 2019

Process mapping: all the essential components All causes identified during the making of the and steps of lapping were outlined in detail, such as C&E diagram were combined in broad categories to material flow, operational activities, resources and form clusters. These clusters (five in the current sce- material required; and the desired standards of the nario) formed the highest level in the Y=f(x) cascade, outcome were established. The aim is to provide all which was drilled into lower levels till it reached the team members with the understanding of the a manageable scope, as shown in Fig 2. Key input internal requirements of the process to allow for variables to be investigated are highlighted in blue effective and quicker reviews in the case of any errors. and green. Critical to quality metrics: in Six Sigma, customer It should be noted that Six Sigma is an iterative requirements are expressed through the Voice of process, and it must continue running until the Customer (VOC), which is converted into quantifi- desired results are achieved. So, in this case, if the able Critical to Quality (CTQ) metrics. Tab. 4 displays chosen experimental variables did not result in any the results on VOC and CTQ characteristics for both improvement, remaining inputs were selected for the internal and external customers in the context of the next iterative cycle. current project. As lapping supplies two different Stage 2 — Measure: from the define stage, four- products to polishing, there were two different CTQs teen input variables were identified, for which the for the same customer requirement (VOC). These data collection plan using the Kipling’s checklist was CTQs were used throughout the project to assess developed, as shown in Tab. 5. improvements made to the lapping process. MSA — the measurement systems analysis — Cause and Effect (C&E) Diagram is also known was conducted on all the measurement devices used as the fishbone, 6M or Ishikawa diagram (Ishikawa, for data collection, which aimed to identify sources of 1968). It is a tool to facilitate brainstorming, identify- variations, induced due to the measurement process. ing the causes for an effect under six broad categories It included checks on measurement devices, person- of Measurements, Material, Man, Environment, nel engaged in the data collection process, their skill MethodsSigma and Machines.(σ) In theDefects current / Non-Conformancesproject, poor sets, peradequacy and accuracy of specifications, raw flatness or TTV reject (CTQs) was the effect and material and the measurement procedure. In the cur- Level Million Oppurtunity (DPMO) a potential factor responsible for poor flatness. The rent study, two MSA tools of Gage R & R and Gage initial project2 σ scoping identified over one 308,537hundred of Bias & Linearity Study were used. Results from both potential causes3 σ for TTV rejects. It would have66,807 been studies suggested that all measurement tools were fit excessively time consuming to analyse all these fac- for purpose. tors. So, it was4 σ important to prioritise potential6,210 causes Establish the Baseline DPMO: first and foremost, and reduce the project scope to a manageable extent, the starting line for the process was established with 5 σ 233 which was achieved by using the Y=f(x) cascade tool the help of DPMO and process capability (Ppk). Key as shown below6 σ in Fig. 2. 3.4 CTQ was TTV based on loose and tight specifica-

tions. The process capability for both types of materi- Fig. 1. Sigma levels depending on DPMO

Fig. 2. Y = f (x) cascade

94 Process Capability of TTV for Tight Spec Material Process Capability of TTV for Loose Spec Material

-0.0 1.4 2.8 4.2 5.6 7.0 8.4 9.8 0.0 1.3 2.6 3.9 5.2 6.5 7.8 9.1

Tab. 4. Critical to Quality metrics of the lapping process TYPESigma (σ)NAME Defects / Non-ConformancesVOC per CTQ External LevelPolishing No sharp roll‐off at the edge of wafer Million Oppurtunity (DPMO)For Loose Spec TTV < 3.5 µm 2 σ 308,537 For Tight Spec TTV < 2.0 µm Internal Lapping Comparable yield to old active agent EngineeringTTV reject < 0.10 % Management in Production and Services Volume 11 • Issue 2 • 2019 3 σ Good Flatness 66,807 For Tight Spec TTV < 2.0 µm

Tab. 5. Data collection plan 4 σ 6,210 ITEM NO. WHAT WHY WHEN HOW WHERE WHO 5 σ 233 1 Wafer Traceability To correlate poor TTV Every Lot Data capture Central Database ‐ Lapping Op with different application Browser 6 σ 3.4 Lapping ADE, Data 2 TTV KPOV, CTQ Every wafer Central Database CW Insp. Op upload Fig. 1. Sigma levels depending on DPMO Every time fresh 3 Slurry Mixing Time KPIV Machine Setting (SOP) Slurry Sheet Lapping Op slurry prepared 4 Slurry Density KPIV Every time fresh Manual Measurement Control Charts Lapping Op slurry prepared 5 Machine flowrate KPIV Every time loop Tab. 4. Critical to Quality meManual Measurement trics of the lapping process Control Charts Lapping Op was changed E KPIV (can affect wafer Every 40 hr of TYP NAME VOC CTQ 6 Plate Shape Using dial gauge Plate Shape Sheet Lapping Op shape) Opertion Time External Polishing No sharp roll‐off at the edge of wafer For Loose Spec TTV < 3.5 µm 7 Recycle Slurry To evaluate impact of Every 5 mins Data capture BMS Database PSE Op For Tight Spec TTV < 2.0 µm Status different slurry Internal application Lapping Comparable yield to old active agent TTV reject < 0.10 % Active Agent KPIV (can affect slurry Every time fresh 8 Machine Setting (SOP) Slurry Sheet Lapping Op Volume viscosity) slurry prepared Good Flatness For Tight Spec TTV < 2.0 µm KPIV (can affect wafer Any time changed 9 Sun Gear Ratio Machine Setting (SOP) QA Records Engineer rotation) by engineer Tab. 5. Data collection plan ITEM NO. WHAT WHY WHEN HOW WHERE WHO KPIV (can affect wafer Any time changed 10 Bottom Plate Speed Machine Setting (SOP) QA Records Engineer rotation) by engineer 1 Wafer Traceability To correlate poor TTV Every Lot Data capture Central Database ‐ Lapping Op KPIV (can affect wafer Any time changed 11 Exhaust Timer Machine Setting (SOP) QA Records with different Engineer application Browser rotation) by engineer Lapping ADE, Data 2 TTV KPOV, CTQ Every wafer Central Database CW Insp. Op KPIV (can affect wafer Any time changed upload 12 Acceleration Timer Machine Setting (SOP) QA Records Engineer rotation) by engineer Every time fresh 3 Slurry Mixing Time KPIV Machine Setting (SOP) Slurry Sheet Lapping Op Slurry KPIV (can affect slurry Every time fresh 13 Machine Setting (SOP) Slurry Sheet Lapping Op slurry prepared Temperature viscosity) slurry prepared Every time fresh Fig. 2. Y = f (x) cascade 4 Slurry Density KPIV Manual Measurement Control Charts Lapping Op KPIV (can affect slurry Lot Processing 14 Plate Temperature At the start of shift Digital thermometer Lapping Op slurry prepared viscosity) Sheet 5 Machine flowrate KPIV Every time loop Manual Measurement Control Charts Lapping Op was changed KPIV (can affect wafer Every 40 hr of Process Capability of TTV for Tight Spec Material 6 ProcessPlate Shape Capability of TTV for Loose Spec Material Using dial gauge Plate Shape Sheet Lapping Op Tab. 6. Key results from process capability charts shape) Opertion Time LB USL Recycle Slurry LB USL To evaluate impact of Data capture Process Data TIGHT FLATNESS LOOSE FLATNESS Within Process7 Data WithinEvery 5 mins BMS Database PSE Op ARAMETERS OMBINED LB P 0 COv erall LB 0 Status different slurry Ov erall application Target * SPEC SPEC Target * USL 2 Potential (Within) C apability USL 3.5 Active Agent KPIV (can affect slurry Potential (Within)Every time fresh C apability Sample Mean 1.21547 Cp * Sample Mean8 2.11004 Cp * Machine Setting (SOP) Slurry Sheet Lapping Op Sample N 5623 5,623 28,120 CPL 33,743* Sample N 28120 Volume viscosity) CPLslurr* y prepared Wafer QtyStDev (Within) 1.11571 CPU 0.23 StDev (Within) 0.78286 CPU 0.59 StDev(Overall) 1.22691 C pk 0.23 StDev(Overall) 0.88975 KPIV (can affect wafer C pk Any time changed 0.59 TTV Reject % 9.82 3.35 O v erall C apability4.43 9 Sun Gear Ratio O v erall C apability Machine Setting (SOP) QA Records Engineer Pp * rotation) Pp by en* gineer 98,168 33,499 PPL 44,282* PPL * DPMO PPU 0.21 KPIV (can affect wafer PPUAny time changed 0.52 Ppk 0.21 10 Bottom Plate Speed Ppk 0.52 Machine Setting (SOP) QA Records Engineer Process Capability C pm * rotation) C pmby en* gineer 0.21 0.52 ‐NA‐ (Ppk) KPIV (can affect wafer Any time changed 11 Exhaust Timer Machine Setting (SOP) QA Records Engineer rotation) by engineer Tab. 7. Response Table for Means -0.0 1.4 2.8 4.2 5.6 7.0 8.4 9.8 0.0 1.3 2.6 3.9 5.2KPIV (can affect wafer 6.5 7.8 9.1 Any time changed O bserv ed P erformance Exp. Within Performance E xp. O v erall P erformance O bserv ed 12 P erformance Acceleration Timer Exp. Within Performance E xp. O v erall P erformance Machine Setting (SOP) QA Records Engineer PPM < LB 0.00 PPM < LB * PPM < LB * PPM < LB 0.00 PPM < LB * PPM < rotationLB * ) by engineer PPM > USL 98168.24SUN PPMG >EAR USL 240975.32BOTTOM PPM > PUSLLATE 261269.05EXHAUST ACCELERATION PPMP LATE> USL 33499.29RECYCLE PPM >S USLLURRY 37907.98 APPMCTIVE AGENT > USL 59120.69 SLURRY SLURRY PPM Total 98168.24 PPM Total 240975.32 PPM Total 261269.05 PPM Total 33499.29 Slurry PPM Total 37907.98 PPM TotalKPIV (can affect slurry 59120.69 Every time fresh LEVEL 13 Machine Setting (SOP) Slurry Sheet Lapping Op RATIO SPEED TIMER TIMER TEMP. Temperature STATUS CONCENTRATIONviscosity) TEMP. Mslurry prepared IXING TIME KPIV (can affect slurry Lot Processing Fig1 . 3. Process capability1.2893 of TTV for1.0338 Tight and1.0053 Loose Spec Material1.0092 1.1997 14 Plate Temperature 0.9940 1.0120 0.9957 At the start of shift 1.0062 Digital thermometer Lapping Op viscosity) Sheet 2 0.7123 0.9678 0.9963 0.9925 0.8020 1.0077 0.9897 1.0060 0.9955 Delta 0.5770 0.0660 0.0090 0.0167 0.3977 0.0137 0.0223 0.0103 0.0107 als was calculated using Minitab Software, which is Tab. 6. Key results from process capability charts Rank 1 3 9 5 2 6 4 8 7 shown in Fig. 3 and Tab. 6. TIGHT FLATNESS LOOSE FLATNESS PARAMETERS COMBINED Tab. 8. Displaying the impactAs shown in Tab. 6, the of improvements on TTV rejec combined TTV rejectts SPEC SPEC BEFORE IMPROVEMENT AFTER IMPROVEMENT was very high at 4.43%, with TTV rejects for tight Wafer Qty 5,623 28,120 33,743 TIGHT FLATNESS LOOSE TIGHT FLATNESS LOOSE flatness specificationPARAMETERS being higher than that of loose COMBINED 9.82 3.35 COMBINED4.43 SPEC FLATNESS SPEC TTV Reject % SPEC FLATNESS SPEC flatnessWafer Qty specification. For a Six Sigma5,623 project to be28,120 DPMO 33,743 98,16819,677 33,49950,666 44,28270,343 successful, it should be able to reduce DPMO to Process Capability TTV Reject % 9.82 3.35 4.43 0.210.02 0.520.01 ‐NA‐0.01 1/10th,DPMO which means the combined 98,168 TTV DPMO 33,499 (Ppk) 44,282 152 118 127 Process Capability (Ppk) 0.52 ‐NA‐ 3.87 1.30 ‐NA‐ should be less than 4428. 0.21 Tab. 7. Response Table for Means Stage 3 — Analyse: during this phase, the col- Although thereSUN G EAR are differentBOTTOM PLATE kindsEXHAUST of control ACCELERATION charts, PLATE RECYCLE SLURRY ACTIVE AGENT SLURRY SLURRY LEVEL lected data was analysed through a systematic appli- in the currentRATIO study, Xbar-SSPEED and I-MRTIMER chartsTIMER were TEMP. STATUS CONCENTRATION TEMP. MIXING TIME cation of statistical and graphical tools. used1 together 1.2893 with boxplots.1.0338 The further1.0053 investiga1.0092 - 1.1997 0.9940 1.0120 0.9957 1.0062 Control charts, also known as Shewhart charts, tion2 into out-of-control0.7123 lots0.9678 gave more0.9963 clues, from0.9925 0.8020 1.0077 0.9897 1.0060 0.9955 were used to identify key trends and generate clues. whichDelta a number0.5770 of Multi-Vari0.0660 charts0.0090 were generated0.0167 0.3977 0.0137 0.0223 0.0103 0.0107 Rank 1 3 9 5 2 6 4 8 7

Tab. 8. Displaying the impact of improvements on TTV rejects 95 BEFORE IMPROVEMENT AFTER IMPROVEMENT TIGHT FLATNESS LOOSE TIGHT FLATNESS LOOSE PARAMETERS COMBINED COMBINED SPEC FLATNESS SPEC SPEC FLATNESS SPEC Wafer Qty 5,623 28,120 33,743 19,677 50,666 70,343 TTV Reject % 9.82 3.35 4.43 0.02 0.01 0.01 DPMO 98,168 33,499 44,282 152 118 127 Process Capability (Ppk) 0.21 0.52 ‐NA‐ 3.87 1.30 ‐NA‐

Engineering Management in Production and Services Volume 11 • Issue 2 • 2019 but only two charts displayed a significant trend (Fig. • Generally, the wafers with Line Saw Mark (LSM) 4). reject have a higher Avg and Std Dev TTV than Avg TTV refers to the mean of TTV for a lot; and the wafers without LSM reject. This is an impor- it should be noted that a lot can have a wafer quantity tant finding as it indicates that due to the poor from 100 to 320. Then, this mean of Avg TTV is fur- quality of an incoming wafer, the wafer was una- ther split by three factors: Lapping Machines, lots ble to rotate freely at lapping. Wafer rotation with Line Saw Mark (LSM) Rejects and Run Order of could be affected by factors like sun gear ratio, lots during the shift. The right-hand side graph is plate speed, acceleration time and exhaust time; exactly the same, except for Std Dev of TTV on the • Generally, lapping machines have no significant Y-Axis. In summary, the following inferences could impact on Avg and Std Dev TTV. It means that be extracted from the Multi-Vari charts (Fig. 4): the problem is global and related to something • The green trend line in Fig. 4 shows that gener- that was common to all lapping machines, like ally, Avg TTV for the first lot of the shift was slurry composition, slurry type etc. comparatively higher than the rest of the lots Stage 4 — the Improve phase: based on enhanced processed during the same shift. It means that learning gained from the Analyse phase, a list of fac- something at the start of the shift was not correct, tors that can influence TTV, was generated by the which was resulting in high Avg TTV. Factors team using the brainstorming technique. To charac- that were different at the start of the shift and got terise the impact of these input variables on TTV, stabilised during the shift were the slurry tem- Design of Experiments were used. As part of DoE perature, slurry mixing time and plate tempera- planning, a CNX diagram was generated, as shown in ture; Fig. 5. Multi-Vari Chart for Avg TTV by LSM Reject - Run Order Multi-Vari Chart for Std Dev TTV by LSM Reject - Run Order

LA P-001 LA P-002 LA P-003 LA P-004 LA P-001 LA P-002 LA P-003 LA P-004 1st lot 2nd lot 3rd lot LSM 1st lot 2nd lot 3rd lot LSM 3.25 Reject Multi-Vari Chart for Avg TTV by LSM Reject - Run Order Multi-Vari Chart for 1.4Std Dev TTV by LSM Reject - Run Order Reject No LA P-001 LA P-002 LA P-003 LA P-004 No Yes LA P-001 LA P-002 LA P-003 LA P-004 3.00 Yes 1st lot 2nd lot 3rd lot LSM 1.2 1st lot 2nd lot 3rd lot LSM 3.25 Reject 1.4 Reject 2.75 V

No T 1.0 No T

3.00 V Yes v T 1.2 Yes T e

2.50 D

g 0.8

v V 2.75 d t T A 1.0 T 2.25 S

V 0.6 v T T e

2.50 D

g 0.8 2.00 v d 0.4 t A

2.25 S 1.75 0.6 0.2 2.00 LA P-001 LA P-002 LA P-003 LA P-004 LA P-001 LA P-002 LA P-003 LA P-004 1.50 0.4 LA P-001 LA P-002 LA P-003 LA P-004 LA P-001 LA P-002 LA P-003 LA P-004 Machine 1.75 Machine 0.2 1.50 Panel variable: Run Order LA P-001 LA P-002 LA P-003 LA P-004Panel variable: Run Order LA P-001 LA P-002 LA P-003 LA P-004 LA P-001 LA P-002 LA P-003 LA P-004 LA P-001 LA P-002 LA P-003 LA P-004 Machine Machine Fig. 4. Multi-Vari charts for Avg and Std Dev TTV by LSM reject — Run order Panel variable: Run Order Panel variable: Run Order

Fig. 4. Multi-Vari charts for Avg and Std Dev TTV by LSM reject — Run order

Fig. 5. CNX Diagram for DOE

Main Effects Plot for Means Data Means

MainSun Effects Gear Ratio Plot for MeansBottom Plate Speed Exhaust Timer Data Means 1.2 Sun Gear Ratio Bottom Plate Speed Exhaust Timer 1.0 1.2 0.8 1.0 0.43 0.70 25 35 1 5 0.8 s n A cceleration Timer Plate Temperature Recy cle S lurry S tatus a 0.43 e 0.70 25 35 1 5 s 1.2 M

A ccelerationf Timer Plate Temperature Recy cle S lurry S tatus a o 1.0

1.2 n M

f 0.8 e

o 1.0

M n 1 3 18 30 On O ff a 0.8 e A ctiv e A gent C oncentration S lurry Temperature Slurry Mixing Time M 1 1.2 3 18 30 On O ff A ctiv e A gent C oncentration S lurry Temperature Slurry Mixing Time 1.0 1.2 0.8 1.0 0.36 1.00 25 32 1 10 0.8 0.36 1.00 25 32 1 10 Fig. 6. Main effect plots for means Fig. 6. Main effect plots for means Multi-VariMulti-Vari Chart Chart for for Avg Avg TTV TTV by by LSM LSM Reject Reject - -Run Run Order Order Multi-VariMulti-Vari Chart Chart for for Std Std Dev Dev TTV TTV by by LSM LSM Reject Reject - -Run Run Order Order

LALA P-001 P-001LALA P-002 P-002LALA P-003 P-003LALA P-004 P-004 LALA P-001 P-001LALA P-002 P-002LALA P-003 P-003LALA P-004 P-004 1st1st lot lot 2nd2nd lot lot 3rd3rd lot lot LSMLSM 3.25 1st1st lot lot 2nd2nd lot lot 3rd3rd lot lot LSMLSM 3.25 RejectReject 1.4 1.4 RejectReject NoNo NoNo Yes 3.003.00 Yes Yes 1.21.2 Yes

2.75 V 2.75 V T

T 1.0

T 1.0 T

V v T v T T e T

2.50 e

2.50 D

g 0.8 D

g 0.8

v d v d t A t A

2.25 S

2.25 S 0.60.6 2.002.00 0.40.4 1.751.75 0.20.2 1.50 LALA P-001 P-001LALA P-002 P-002LALA P-003 P-003LALA P-004 P-004 LALA P-001 P-001LALA P-002 P-002LALA P-003 P-003LALA P-004 P-004 1.50 Machine LALA P-001 P-001LALA P-002 P-002LALA P-003 P-003LALA P-004 P-004 LALA P-001 P-001LALA P-002 P-002LALA P-003 P-003LALA P-004 P-004 Machine MachineMachine PanelPanel variable: variable: Run Run Order Order PanelPanel variable: variable: Run Run Order Order FigFig. .4 4. .Multi Multi-Vari-Vari charts charts for for Avg Avg and and Std Std Dev Dev TTV TTV by by LSM LSM reject reject — — Run Run order order

Tab. 5. Data collection plan ITEM NO. WHAT WHY WHEN HOW WHERE WHO

1 Wafer Traceability To correlate poor TTV Every Lot Data capture Central Database ‐ Lapping Op with different application Browser Lapping ADE, Data 2 TTV KPOV, CTQ Every wafer Central Database CW Insp. Op upload Every time fresh 3 Slurry Mixing Time KPIV Machine Setting (SOP) Slurry Sheet Lapping Op slurry prepared

4 Slurry Density KPIV Every time fresh Manual Measurement Control Charts Lapping Op FigFig. .5 5. .CNX CNX Diagram Diagram for for DOE DOE slurry prepared 5 Machine flowrate KPIV Every time loop Manual Measurement Control Charts Lapping Op was changed KPIV (can affect wafer Every 40 hr of 6 Plate Shape Using dial gauge Engineering ManagementPlate Shape Sheet in ProductionLapping Op and Services Volume 11 • Issue 2 • 2019 shape) Opertion Time

7 Recycle Slurry To evaluate impact of Every 5 mins Data capture BMS Database PSE Op Status different slurry application Active Agent KPIV (can affect slurry Every time fresh 8 MainMain Effects Effects Plot Plot Machine Setting (SOP) for for Means Means Slurry Sheet Lapping Op Volume viscosity) slurry prepared DataData Means Means KPIV (can affect wafer Any time changed 9 Sun Gear Ratio Machine Setting (SOP) QA Records Engineer rotationSunSun Gear )Gear Ratio Ratio by engineer BottomBottom Plate Plate Speed Speed ExhaustExhaust Timer Timer KPIV (can affect wafer Any time changed 10 Bottom Plate Speed 1.21.2 Machine Setting (SOP) QA Records Engineer rotation) by engineer 1.01.0 KPIV (can affect wafer Any time changed 11 Exhaust Timer Machine Setting (SOP) QA Records Engineer 0.80.8 rotation) by engineer KPIV (can affect wafer Any time changed 12 Acceleration Timer 0.43 0.70 25 Machine Setting (SOP) 35 1 QA Records 5 Engineer s 0.43 0.70 25 35 1 5 s rotation) by engineer n n AA cceleration cceleration Timer Timer PlatePlate Temperature Temperature RecyRecy cle cle S S lurry lurry S S tatus tatus a Slurry a KPIV (can affect slurry Every time fresh e 13 e Slurry Sheet Lapping Op 1.2 Machine Setting (SOP) Temperature M 1.2

M viscosity) slurry prepared

f f

o KPIV (can affect slurry Lot Processing o 1.0 1.0 14 Plate Temperature At the start of shift Digital thermometer Lapping Op n n viscosity) Sheet a a 0.80.8 e

e M M Tab. 6. Key results from process capability charts 11 33 1818 3030 OnOn OO ff ff AA ctiv ctiv e eA A gent gent C C oncentration oncentration SS lurry lurry Temperature Temperature SlurrySlurry Mixing Mixing Time Time TIGHT 1.2FLATNESS LOOSE FLATNESS PARAMETERS 1.2 COMBINED SPEC SPEC 1.01.0

Wafer Qty 0.80.8 5,623 28,120 33,743 TTV Reject % 9.82 3.35 4.43 0.360.36 1.001.00 2525 3232 11 1010 DPMO 98,168 33,499 44,282 Process Capability Fig. 6. Main effect0.21 plots for means0.52 ‐NA‐ (Ppk) Fig. 6. Main effect plots for means

Tab. 7. Response Table for Means

Tab. 8. Displaying the impact of improvements on TTV rejects The list of experimental variables was still too was followed by factor B and an interaction effect BEFORE IMPROVEMENT AFTER IMPROVEMENT long, and only screening DOETIGHT wasFLATNESS feasible. TaguchiLOOSE between factorsTIGHT AFLATNESS and B. TheL OOSEPareto chart was also PARAMETERS COMBINED COMBINED L12 design was used to rank theS PECfactors in FtheLATNESS orderSPEC used to determineSPEC statistically FLATNESS significantSPEC factors and ofWafer Qty their impact on TTV. Taguchi L125,623 implied that by28,120 the effect33,743 of their interactions19,677 on the50,666 output. 70,343 The fac- performingTTV Reject % 12 lapping batches, nine9.82 experimental3.35 tors above4.43 the red line0.02 were significant,0.01 and the0.01 fac- DPMO 98,168 33,499 44,282 152 118 127 factors at two different values were evaluated for their tors below were non-significant. Process Capability (Ppk) 0.21 0.52 ‐NA‐ 3.87 1.30 ‐NA‐ impact on TTV, as shown in Fig. 6 and Tab. 7. T main effects plot (Fig. 8) indicated that an From Taguchi DOE, the top two ranked factors increase in the sun gear ratio, plate temperature and (the sun gear ratio and plate temperature) seemed to the bottom plate speed resulted in reduced TTV. have a significant impact on TTV. Although the fac- Further, as centre points (marked red ) for all the tor ranked third (the bottom plate speed) did not three factors were not on the line, the impact of these seem significant, it was still chosen for further analy- factors on TTV was not linear. In simple terms, the sis. As Taguchi is only a screening DOE, it is always increase in the plate temperature to 24 °C did not recommended to perform a more comprehensive result in any reduction in TTV; however, a plate DOE to confirm the results of any screening DOE. So, temperature closer to 30 °C was most likely to reduce three factors of the second level of the full factorial the TTV reject rate. design of experiment (DOE) were designed with To make things simple, Minitab’s response opti- three centre points to characterise and optimise the miser function could be used, which calculated values three experimental factors. for input variables to achieve the optimum output. DoE results in the Pareto chart represent the rela- Results are shown in Fig. 9, suggesting that to achieve tive impact of differences between different factors on the minimum TTV (0.462), inputs should be set to the process outcome. As shown in Fig. 7, factor A the following parameters (highlighted in red): (sun gear ratio) had the biggest impact on TTV. This • Sun Gear Ratio = 0.7;

Engineering Management in Production and Services Volume 11 • Issue 2 • 2019

Normal Plot of the StandardizedNormal Plot Effects of the Standardized Effects Pareto Chart of the StandardizedPareto Chart Effects of the Standardized Effects (response is Avg TTV, Alpha = (response0.05) is Avg TTV, Alpha = 0.05) (response is Avg TTV, Alpha = (response0.05) is Avg TTV, Alpha = 0.05) 99 99 2.20 2.20 Effect Type Effect Type F actor Name F actor Name Not Significant Not Significant A Sun Gear Ratio A Sun Gear Ratio 95 95 Significant ASignificant A B Plate Temperature B Plate Temperature C Bottom Plate Speed C Bottom Plate Speed 90 90 AB F actor Name AB F actor Name A Sun Gear Ratio A Sun Gear Ratio 80 80 B Plate Temperature B B Plate Temperature B AC C BottomAC Plate Speed C Bottom Plate Speed 70 70 t t

n 60 n 60 AB AB e e m m c c

50 50 r r r r e e

e e C C

40 40 T T P P 30 30 C C B B 20 20 AC AC 10 A 10 A

5 5 BC BC

1 1 -60 -50 -40 -30 -20 -60-10 -500 -4010 -3020 -2030 -10 0 10 20 30 0 10 20 30 0 4010 5020 6030 40 50 60 Standardized Effect Standardized Effect Standardized Effect Standardized Effect Fig. 7. Normal and standardisedFig. 7. Normal effects and standardisedfor Pareto charts effects for Pareto charts

Main Effects Plot for AvgMain TTV Effects Plot for Avg TTV Interaction Plot for AvgInteraction TTV Plot for Avg TTV Data Means Data Means Data Means Data Means

Sun Gear Ratio SunPlate Gear Temperature Ratio PlatePoint Temperature Ty pe Point Ty pe18 24 30 25 30 18 35 24 30 25 30 35 1.5 1.5 1.08 1.08 Corner Corner Sun Gear Sun Gear Center Center Ratio Point Ty pe Ratio Point Ty pe 0.96 0.96 0.430 Corner 0.430 Corner 1.0 1.0 Sun Gear Ratio Sun Gear Ratio 0.565 Center 0.565 Center 0.84 0.84 0.700 Corner 0.700 Corner 0.72 0.72 0.5 0.5 1.5 1.5 Plate Plate

n 0.60 n 0.60

a a Temperature Point Ty pe Temperature Point Ty pe e 0.430 0.565 e 0.700 0.43018 0.56524 0.70030 18 24 30 18 Corner 18 Corner M M 1.0 1.0 Bottom Plate Speed Bottom Plate Speed Plate Temperature Plate Temperature 24 Center 24 Center 1.08 1.08 30 Corner 30 Corner

0.96 0.96 0.5 0.5

0.84 0.84

0.72 0.72 Bottom Plate Speed Bottom Plate Speed 0.60 0.60 25 30 35 25 30 35 Fig. 8. Main effects andFig. interaction8. Main effects plots and for interactionavg TTV plots for avg TTV

Optimal OptimalSun Gear Sun GearPlate Te Plate BottomTe P Bottom P High 0.70 High 0.70 30.0 30.0 35.0 35.0 D Cur D0.70 Cur 0.70 30.0 30.0 35.0 35.0 0.76000 Low 0.760000.430 Low 0.430 18.0 18.0 25.0 25.0

Composite Composite Desirability Desirability 0.76000 0.76000

Avg TTV Avg TTV Minimum Minimum y = 0.4620 y = 0.4620 d = 0.76000 d = 0.76000

Fig. 9. Response optimiserFig. 9. resultResponse optimiser result

• Plate Temperature = 30 °C; Using the recommended input values, Response • Bottom Plate Speed = 35 rpm. Optimiser predicted Avg TTV to be around 0.462 To validate the theoretical predictive DOE model, μm, which is exactly the same as the confirmation a confirmation run should be conducted using the run result of 0.460 μm (see Fig. 10). The confirmation recommended input values. A confirmation run was test, thus, confirmed the validity of the theoretical carried out with six lapping batches, and the results model. are presented in Fig. 10.

98 Engineering Management in Production and Services Volume 11 • Issue 2 • 2019 Summary for Confimation Test TTV A nderson-Darling Normality Test A -S quared 0.50 Summary for ConfimationP -V alue Test 0.207TTV Mean 0.46084 A nderson-Darling Normality Test StDev 0.10751 A -S quared 0.50 V ariance 0.01156 P -V alue 0.207 Skew ness -0.011959 Kurtosis -0.157437 Mean 0.46084 N 210 StDev 0.10751 V ariance 0.01156 Minimum 0.17696 Skew ness -0.011959 1st Q uartile 0.38360 Kurtosis -0.157437 Median 0.46912 N 210 3rd Q uartile 0.53777 0.2 0.3 0.4 0.5 0.6 0.7 M aximum 0.74813 Minimum 0.17696 95% C onfidence Interv al for Mean 1st Q uartile 0.38360 Median 0.46912 0.44622 0.47547 3rd Q uartile 0.53777 0.2 0.3 0.4 0.5 0.695% C onfidence0.7 Interv al for Median M aximum 0.74813 0.43928 0.49011 95% C onfidence Interv al for Mean 95% C onfidence Interv al for StDev 0.44622 0.47547 95% Confidence Intervals 0.09812 0.11891 95% C onfidence Interv al for Median Mean 0.43928 0.49011 95% C onfidence Interv al for StDev Median 95% Confidence Intervals 0.09812 0.11891 0.44 0.45 Mean 0.46 0.47 0.48 0.49 Median

Fig. 10. Graphical Summary0.44 of Confirmation0.45 0.46Test Run0.47 Results0.48 0.49

Fig. 10. Graphical Summary of Confirmation Test Run Results

Boxplot of TTV for Tight Flatness Spec by Improvement Stages 8 Before ImprovementBoxplot of TTV forPilot TightStage -Flatness After Improvement Spec by Improvement Stages 7 8 Before Improvement Pilot Stage - After Improvement 6 7

5 6

V 4 T 5 T

3 V 4 T T 2 3 2

1 2 2

0 1

6 5 1 6 0 4 3 2 9 7 8 5 5 7 9 6 9 0 9 8 7 0 8 7 6 0 7 4 4 4 6 6 5 0 1 8 2 5 5 9 2 8 8 8 4 7 9 4 9 0 8 6 6 5 1 3 3 0 1 9 1 5 6 3 1 7 1 0 9 4 7 8 0 9 8 1 0 8 8 9 1 9 2 1 3 1 6 5 3 1 7 9 1 :5 :1 :5 :2 :5 :1 :3 :3 :1 :4 :2 :5 :4 :1 :2 :1 :1 :3 :0 :2 :3 :4 :4 :2 :5 :1 :4 :0 :1 :0 :2 :4 :2 :4 :5 :2 :0 :2 :3 :4 :3 :1 :0 :1 :4 :0 :4 :1 :3 :4 :2 :4 :1 :5 :4 :4 :3 :3 :0 :2 :1 :2 :1 :2 :4 :0 :0 :5 :3 :0 :3 :1 :5 :4 :5 :2 :5 :0 :5 :2 :5 :0 :1 :2 :5 :2 :4 :1 :2 :4 :0 :1 :0 3 9 0 2 3 1 2 4 6 4 6 9 2 3 0 4 5 3 0 3 3 4 3 0 2 2 2 3 0 2 3 2 3 4 1 3 0 2 3 4 3 4 4 3 3 3 9 2 3 7 2 0 0 1 2 2 2 0 0 0 0 0 0 0 1 14 20 23 02 02 05 05 0 6 09 11 23 23 03 0 0 0 0 2 0 0 0 0 2 0 0 0 1 15 17 00 00 03 05 1 6 17 21 23 14 15 17 22 2 0 0 2 0 0 0 2 2 0 0 0 0 04 23 00 02 04 0 4 23 23 03 03 05 17 23 23 0 0 0 2 0 0 0 1 1 1 0 0 2 2 1 2 12 12 12 12 12 12 12 12 1 2 12 12 1 2 2 2 2 2 2 2 2 2 2 2 2 2 1 2 12 12 12 12 12 12 12 12 1 2 12 12 12 1 2 2 2 2 2 2 2 2 2 2 2 2 2 12 12 12 12 12 12 12 12 1 2 12 12 12 12 2 2 2 2 2 2 2 2 2 2 2 2 2 2 12 12 12 12 12 12 1 2 12 12 12 12 12 12 0 0 0 0 0 0 0 0 0 0 0 0 0 01 01 01 01 01 01 0 1 01 01 01 01 01 01 0 0 0 0 0 0 0 0 0 0 0 0 0 0 01 01 01 01 0 1 01 01 01 01 01 01 01 01 0 0 0 0 0 0 0 0 0 0 0 0 0 01 01 01 01 0 1 01 01 01 01 01 01 01 01 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2/2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2/ 2/2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2/2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 0 2 2 2 2 3 3 3 3 4 4 14 14 4 4 4 5 5 5 5 505 5 3 3 4 4 6 6 6 7 8 8 8 8 9 0 0 0 3 3 3 5 5 5 5 5 5 5 5 8 8 8 8 8 9 9 3 4 4 4 5 5 6 6 6 6 6 6 7 7 7 7 0 0 1 1 1 1 7 7 8 8 8 2 3 3 7 7 7 7 0 2 /1 /1 /1 /1 /1 /1 /1 /1 /1 /1 /1 / / 1 1 1 1 1 1 1 1 1 1 0 0 /0 /0 /0 /0 /0 /0 /0 /0 /0 /0 /0 /1 /1 /1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 /1 /1 /1 /1 /2 /2 /2 /2 /2 /2 /2 /2 /2 2 2 2 2 2 2 2 3 3 3 3 3 3 0 /0 /0 /0 /0 /1 /1 /1 /1 /1 /1 /1 /2 /2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 / 3/ 3/ 3/ 3/ 3/ 3/ 3/ 3 / 3/ 5/ 5/ 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 / 5/ 5/ 5/ 5/ 5/ 5/ 5/ 5 / 5/ 5/ 5/ 5/ 5/ 5 5 5 5 5 5 5 5 5 5 5 5 5 5/ 5/ 5/ 5/ 5/ 5/ 5/ 5 / 5/ 5/ 5/ 5/ 5/ 6/ 6 6 6 6 6 6 6 6 6 6 6 6 6

6 5 1 6 0 4 3 2 9 7 8 5 5 7 9 6 9 0 9 8 7 0 8 7 6 0 7 4 4 4 6 6 5 0 1 8 2 5 5 9 2 8 8 8 4 7 9 4 9 0 8 6 6 5 1 3 3 0 1 9 1 5 6 3 1 7 1 0 9 4 7 8 0 9 8 1 0 8 8 9 1 9 2 1 3 1 6 5 3 1 7 9 1 :5 :1 :5 :2 :5 :1 :3 :3 :1 :4 :2 :5 :4 :1 :2 :1 :1 :3 :0 :2 :3 :4 :4 :2 :5 :1 :4 :0 :1 :0 :2 :4 :2 :4 :5 :2 :0 :2 :3 :4 :3 :1 :0 :1 :4 :0 :4 :1 :3 :4 :2 :4 :1 :5 :4 :4 :3 :3 :0 :2 :1 :2 :1 :2 :4 :0 :0 :5 :3 :0 :3 :1 :5 :4 :5 :2 :5 :0 :5 :2 :5 :0 :1 :2 :5 :2 :4 :1 :2 :4 :0 :1 :0 3 9 0 2 3 1 2 4 6 4 6 9 2 3 0 4 5 3 0 3 3 4 3 0 2 2 2 3 0 2 3 2 3 4 1 3 0 2 3 4 3 4 4 3 3 3 9 2 3 7 2 0 0 1 2 2 2 0 0 0 0 0 0 0 1 14 20 23 02 02 05 05 0 6 09 11 23 23 03 0 0 0 0 2 0 0 0 0 2 0 0 0 1 15 17 00 00 03 05 1 6 17 21 23 14 15 17 22 2 0 0 2 0 0 0 2 2 0 0 0 0 04 23 00 02 04 0 4 23 23 03 03 05 17 23 23 0 0 0 2 0 0 0 1 1 1 0 0 2 2 1 2 12 12 12 12 12 12 12 12 1 2 12 12 1 2 2 2 2 2 2 2 2 2 2 2 2 2 1 2 12 12 12 12 12 12 12 12 1 2 12 12 12 1 2 2 2 2 2 2 2 2 2 2 2 2 2 12 12 12 12 12 12 12 12 1 2 12 12 12 12 2 2 2 2 2 2 2 2 2 2 2 2 2 2 12 12 12 12 12 12 1 2 12 12 12 12 12 12 0 0 0 0 0 0 0 0 0 0 0 0 0 01 01 01 01 01 01 0Lapping1 01 01 01 01 01 01 0 0 0 0 Date0 0 0 0 0 0 0 Out0 0 0 01 01 01 01 0 1 01 01 01 01 01 01 01 01 0 0 0 0 0 0 0 0 0 0 0 0 0 01 01 01 01 0 1 01 01 01 01 01 01 01 01 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2/2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2/ 2/2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2/2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 0 2 2 2 2 3 3 3 3 4 4 14 14 4 4 4 5 5 5 5 5 5 5 3 3 4 4 6 6 6 7 8 8 8 8 9 0 0 0 3 3 3 5 5 5 5 5 5 5 5 8 8 8 8 8 9 9 3 4 4 4 5 5 6 6 6 6 6 6 7 7 7 7 0 0 1 1 1 1 7 7 8 8 8 2 3 3 7 7 7 7 0 2 /1 /1 /1 /1 /1 /1 /1 /1 /1 /1 /1 / / 1 1 1 1 1 1 1 1 1 1 0 0 /0 /0 /0 /0 /0 /0 /0 /0 /0 /0 /0 /1 /1 /1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 /1 /1 /1 /1 /2 /2 /2 /2 /2 /2 /2 /2 /2 2 2 2 2 2 2 2 3 3 3 3 3 3 0 /0 /0 /0 /0 /1 /1 /1 /1 /1 /1 /1 /2 /2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 / 3/ 3/ 3/ 3/ 3/ 3/ 3/ 3 / 3/ 5/ 5/ 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 / 5/ 5/ 5/ 5/ 5/ 5/ 5/ 5 / 5/ 5/ 5/ 5/ 5/ 5 5 5 5 5 5 5 5 5 5 5 5 5 5/ 5/ 5/ 5/ 5/ 5/ 5/ 5 / 5/ 5/ 5/ 5/ 5/ 6/ 6 6 6 6 6 6 6 6 6 6 6 6 6 Fig. 11. Boxplot of TTV for tight flatness spec by improvement stagesLapping Date Out

Pilot testing was used Figfor. 11 the. Boxplot validation of TTV for of tight the flatness TTV spec had by improvement become stable; stages and few wafers were above results found duringProcess Capabilitythe improvement of TTV for Tight Spec_After stage Improvementby the spec limit.Process This Capability improvement of TTV for Loose Spec_Afterwas statistically Improvement implementing it on largerLB sample size.USL The incremen- significant (W=1367172420.0LB and p< 0.01).USL This data Process Data Process Capability of TTV for TightWithin Spec_After ImprovementProcess Data Process Capability of TTV for LooseWithin Spec_After Improvement LB 0 Ov erall LB 0 Ov erall tal implementationTarget * with a continuous review of provided Targetsufficient* evidence in support of the USL 2 Potential (Within) C apability USL 3.5 Potential (Within) C apability LB USL LB USL Sample Mean 0.351407 Cp * Sample Mean 1.68514 Cp * results was conductedSample N 19677 to reduceProcess Data the risk of large CPL improved* Sample condition NWithin 50666 for looseProcess Data spec material. TheseCPL * Within StDev (Within) 0.125247 LB 0 CPU 4.39 StDev (Within)Ov erall 0.371802 LB 0 CPU 1.63 Ov erall StDev(Overall) 0.141981 Target * C pk 4.39 StDev(Overall) 0.464149 Target * C pk 1.63 Potential (Within) C apability Potential (Within) C apability rejects. Loose TTV spec materialsUSL 2 were run on one O v erall Cconditions apability were then transferredUSL 3.5 to tight spec mateO v erall- C apability Sample Mean 0.351407 Cp * Sample Mean 1.68514 Cp * Pp * Pp * Sample N 19677 CPL * Sample N 50666 CPL * PPL * PPL * lapping machine for a limitedStDev (Within) period0.125247 of one week, rial, and the improvementCPU 4.39 StDev (Within)in TTV0.371802 was again statisti- CPU 1.63 PPU 3.87 PPU 1.30 StDev(Overall) 0.141981 C pk 4.39 StDev(Overall) 0.464149 C pk 1.63 Ppk 3.87 Ppk 1.30 O v erall C apability O v erall C apability and then, the results were awaited. Since results from C pm cally* significant according to the Mann Whitney testC pm * Pp * Pp * PPL * PPL * this initial phase were as expected, this optimum (W=124139220.0PPU 3.87 and p= 0.00). PPU 1.30 Ppk 3.87 Ppk 1.30 condition was expanded-0.00 to 0.65other1.30 machines,1.95 2.60 3.25 3.90 one4.55 at a The EstablishingC pm * 0.00 Process0.55 1.10 1.65capability2.20 2.75 3.30 and3.85 DPMO C pm * O bserv ed P erformance Exp. Within Performance E xp. O v erall P erformance O bserv ed P erformance Exp. Within Performance E xp. O v erall P erformance time. BoxplotPPM results < LB 0.00 fromPPM < LBthe * pilotPPM stage < LB * for loose were the PPMnext < LB step,0.00 aimingPPM < LB *to quantifyPPM < LB * and validate PPM > USL 152.46 PPM > USL 0.00 PPM > USL 0.00 PPM > USL 118.42 PPM > USL 0.53 PPM > USL 46.13 spec material PPMare Total shown152.46 inPPM TotalFig.0.00 11. PPM Total-0.000.000.65 1.30 1.95 2.60 3.25the3.90 new4.55 improvedPPM Total 118.42 results.PPM Total At0.53 thePPM start Total0.00 46.13of0.55 the1.10 process,1.65 2.20 2.75 3.30 3.85 O bserv ed P erformance Exp. Within Performance E xp. O v erall P erformance O bserv ed P erformance Exp. Within Performance E xp. O v erall P erformance PPM < LB 0.00 PPM < LB * PPM < LB * PPM < LB 0.00 PPM < LB * PPM < LB * There were rather a fewPPM > wafersUSL 152.46 withPPM > USLTTV0.00 > 3.5PPM > USL 0.00the baseline for the lappingPPM > USL process118.42 PPM was > USL 0.53establishedPPM > USL 46.13 Fig. 12. Process capabilityPPM Total of152.46 TTV forPPM tight Total and0.00 loosePPM specTotal 0.00 after improvement PPM Total 118.42 PPM Total 0.53 PPM Total 46.13 microns (spec limit) before improvement. However, using DPMO, which was now repeated for the a noticeable difference wasFig visible. 12. Process after capability the introduc of TTV- for tightimproved and loose process. spec after The improvement process capability charts are tion of the improved condition. It was found that

99 Summary for Confimation Test TTV A nderson-Darling Normality Test A -S quared 0.50 P -V alue 0.207 Mean 0.46084 StDev 0.10751 V ariance 0.01156 Skew ness -0.011959 Kurtosis -0.157437 N 210 Minimum 0.17696 Tab. 4. Critical to Quality metrics of the lapping process 1st Q uartile 0.38360 E Median 0.46912 TYP NAME VOC 3rd Q uartile 0.53777 CTQ External 0.2 Polishing 0.3 0.4 No sharp roll‐off at the edge of wafer 0.5 0.6 0.7 M aximum 0.74813 95% C onfidence Interv al for MeanFor Loose Spec TTV < 3.5 µm 0.44622 0.47547 For Tight Spec TTV < 2.0 µm 95% C onfidence Interv al for Median Internal Lapping Comparable yield to old active agent 0.43928 0.49011 TTV reject < 0.10 % 95% C onfidence Interv al for StDev 95% Confidence Intervals Good Flatness 0.09812 0.11891 For Tight Spec TTV < 2.0 µm Mean

Tab. 5. Data collection plan Median ITEM NO. WHAT WHY WHEN HOW WHERE WHO 0.44 0.45 0.46 0.47 0.48 0.49 1 Wafer Traceability To correlate poor TTV Every Lot Data capture Central Database ‐ Lapping Op Fig. 10. Graphical Summary of Confirmationwith different Test Run Results application Browser Lapping ADE, Data 2 TTV KPOV, CTQ Every wafer Central Database CW Insp. Op upload Every time fresh 3 Slurry Mixing Time KPIV Machine Setting (SOP) Slurry Sheet Lapping Op slurry prepared 4 Slurry Density Boxplot of TTVKPIV for Tight FlatnessEvery time fresh Spec by ImprovementManual Measurement StagesControl Charts Lapping Op slurry prepared 8 5 Machine flowrate Before ImprovementKPIV PilotEvery time loop Stage - After ImprovementManual Measurement Control Charts Lapping Op was changed 7 KPIV (can affect wafer Every 40 hr of 6 Plate Shape Using dial gauge Plate Shape Sheet Lapping Op shape) Opertion Time 6 7 Recycle Slurry To evaluate impact of Every 5 mins Data capture BMS Database PSE Op Status different slurry application 5 Active Agent KPIV (can affect slurry Every time fresh 8 Machine Setting (SOP) Slurry Sheet Lapping Op Volume viscosity) slurry prepared V 4 T KPIV (can affect wafer Any time changed 9 T Sun Gear Ratio Machine Setting (SOP) QA Records Engineer rotation) by engineer 3 KPIV (can affect wafer Any time changed 10 Bottom Plate Speed Machine Setting (SOP) QA Records Engineer rotation) by engineer 2 2 KPIV (can affect wafer Any time changed 11 Exhaust Timer Machine Setting (SOP) QA Records Engineer 1 rotation) by engineer KPIV (can affect wafer Any time changed 12 Acceleration Timer Machine Setting (SOP) QA Records Engineer 0 rotation) by engineer Slurry KPIV (can affect slurry Every time fresh 6 5 1 6 0 4 3 2 9 7 8 5 5 7 9 6 9 0 9 8 7 0 8 7 6 0 7 4 4 4 6 6 5 0 1 8 2 5 5 9 2 8 8 8 4 7 9 4 9 0 8 6 6 5 1 3 3 0 1 9 1 5 6 3 1 7 1 0 9 4 7 8 0 9 8 1 0 8 8 9 1 9 2 1 3 1 6 5 3 1 7 9 1 13 :5 :1 :5 :2 :5 :1 :3 :3 :1 :4 :2 :5 :4 :1 :2 :1 :1 :3 :0 :2 :3 :4 :4 :2 :5 :1 :4 :0 :1 :0 :2 :4 :2 :4 :5 :2 :0 :2 :3 :4 :3 :1 :0 :1 :4 :0 :4 :1 :3 :4 :2 :4 :1 :5 :4 :4 :3 :3 :0 :2 :1Machine Setting (SOP) :2 :1 :2 :4 :0 :0 :5 :3 :0 :3 :1 :5 :4 :5 :2 :5 :0 :5 :2 :5 :0 :1 :2 :5 :2Slurry Sheet :4 :1 :2 :4 :0 :1 :0 Lapping Op 3 9 0 2 3 1 2 4 6 4 6 9 2 3 0 4 5 3 0 3 3 4 3 0 2 2 2 3 0 2 3 2 3 4 1 3 0 2 3 4 3 4 4 3 3 3 9 2 3 7 2 0 0 1 2 2 2 0 0 0 0 0 0 0 1 14 20 23 02 02 05 05 0 6 09 11 23 23 03 0 0 0 0 2 0 0 0 0 2 0 0 0 1 15 17 00 00 03 05 1 6 17 21 23 14 15 17 22 2 0 0 2 0 0 0 2 2 0 0 0 0 04 23 00 02 04 0 4 23 23 03 03 05 17 23 23 0 0 0 2 0 0 0 1 1 1 0 0 Temperature 2 viscosity) 2 slurry prepared 1 2 12 12 12 12 12 12 12 12 1 2 12 12 1 2 2 2 2 2 2 2 2 2 2 2 2 2 1 2 12 12 12 12 12 12 12 12 1 2 12 12 12 1 2 2 2 2 2 2 2 2 2 2 2 2 2 12 12 12 12 12 12 12 12 1 2 12 12 12 12 2 2 2 2 2 2 2 2 2 2 2 2 2 2 12 12 12 12 12 12 1 2 12 12 12 12 12 12 0 0 0 0 0 0 0 0 0 0 0 0 0 01 01 01 01 01 01 0 1 01 01 01 01 01 01 0 0 0 0 0 0 0 0 0 0 0 0 0 0 01 01 01 01 0 1 01 01 01 01 01 01 01 01 0 0 0 0 0 0 0 0 0 0 0 0 0 01 01 01 01 0 1 01 01 01 01 01 01 01 01 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 / / / / / / / / / / / 4/ 4/ / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / 1 0 12 12 12 12 13 13 13 1 3 14 14 1 1 4 4 4 5 5 5 5 5 5 5 3 3 0 4 04 06 06 06 07 08 08 08 0 8 09 10 10 10 3 3 3 5 5 5 5 5 5 5 5 8 8 18 18 18 19 19 23 24 24 2 4 25 25 26 26 26 6 6 6 7 7 7 7 0 0 1 1 1 1 7 07 08 08 08 12 1 31 3 17 17 17 17 20 22 / / / / / / / / / / / / / /1 /1 /1 /1 /1 /1 /1 /1KPIV (can affect slurry /1 /1 /0 /0 / / / / / / / / / / / / / / /1 /1 /1 /1 /1 /1 /1 /1 /1 /1 /1 /1 /1 / / / / / / / / / / / / / / /2 /2 /2 /2 /2 /2 /2 /3 /3 /3 /3 /3 /3 /0 / / / / / / / / / / Lot Processing / / / 14 Plate Temperature 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 At the start of shift 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5Digital thermometer 5 5 5 5 5 5 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6 6 6 6 6 6 Lapping Op viscosity) Sheet Lapping Date Out

Tab. 6. Key results from process capability charts Fig. 11. Boxplot of TTV for tight flatness spec by improvement stages Engineering Management in Production and Services TIGHT FLATNESS LOOSE FLATNESS Volume 11 • Issue 2 • 2019 P ARAMETERS COMBINED SPEC SPEC

Wafer Qty Process Capability of5,623 TTV for Tight Spec_After28,120 Improvement33,743 Process Capability of TTV for Loose Spec_After Improvement

TTV Reject % LB 9.82USL 3.35 4.43 LB USL Process Data Within Process Data Within LB 0 98,168 33,499 Ov44,282 erall LB 0 Ov erall DPMOTarget * Target * USL 2 Potential (Within) C apability USL 3.5 Potential (Within) C apability Process Capability Sample Mean 0.351407 Cp * Sample Mean 1.68514 Cp * Sample N 19677 0.21 0.52 CPL ‐NA‐* Sample N 50666 CPL * (Ppk) StDev (Within) 0.125247 CPU 4.39 StDev (Within) 0.371802 CPU 1.63 StDev(Overall) 0.141981 C pk 4.39 StDev(Overall) 0.464149 C pk 1.63 O v erall C apability O v erall C apability Tab. 7. Response Table for Means Pp * Pp * PPL * PPL * PPU 3.87 PPU 1.30 Ppk 3.87 Ppk 1.30 SUN GEAR BOTTOM PLATE EXHAUST ACCELERATIONC pm * PLATE RECYCLE SLURRY ACTIVE AGENT SLURRY C pmSLURRY* LEVEL RATIO SPEED TIMER TIMER TEMP. STATUS CONCENTRATION TEMP. MIXING TIME

1 1.2893 -0.00 0.65 1.30 1.951.0338 2.60 3.25 3.901.0053 4.55 1.0092 1.1997 0.000.9940 0.55 1.10 1.65 2.201.0120 2.75 3.300.9957 3.85 1.0062 O bserv ed P erformance Exp. Within Performance E xp. O v erall P erformance O bserv ed P erformance Exp. Within Performance E xp. O v erall P erformance 2 PPM < LB 0.00 0.7123 PPM < LB * 0.9678 PPM < LB * 0.9963 0.9925 0.8020 PPM < LB 0.00 PPM1.0077 < LB * PPM < LB0.9897 * 1.0060 0.9955 PPM > USL 152.46 PPM > USL 0.00 PPM > USL 0.00 PPM > USL 118.42 PPM > USL 0.53 PPM > USL 46.13 Delta PPM Total 152.46 0.5770 PPM Total 0.00 0.0660 PPM Total 0.00 0.0090 0.0167 0.3977 PPM Total 118.42 PPM0.0137 Total 0.53 PPM Total0.0223 46.13 0.0103 0.0107 Rank Fig. 12. Process capability1 of TTV for3 tight and loose9 spec after improvement5 2 6 4 8 7

shown in Fig. 12. The key results from Fig. 12 and method was exercised on the lapping machine, which from the start of the project are summarised in Tab. 8. would automatically prevent the machine from run- As shown in Tab. 8, there was a massive reduc- ning if any of these two parameters were set incor- tion in TTV rejects from 9.82% to 0.02% for tight rectly. This was achieved using continuous, automatic flatness spec; from 3.35% to 0.01% for loose flatness measurement of these parameters by the machine, spec; and from 4.43% to 0.01% for the combined and as soon as incorrect values were identified, an materials. This is also reflected in the DPMO values. alarm went-off and the machine stopped running. The initial target of the project was to reduce TTV The Statistical Process Control (SPC) Chart: the rejects to less than 0.1%; and after the improvements, underlying principle of Six Sigma is to reduce the TTV reject for both specs was found to be less than process output variation by controlling key process 0.03%. Process capability charts clearly demonstrated input variables (KPIV’s), which is done with the help that the mean and variation for loose spec material of SPC charts. As discussed earlier, the second most was comparatively higher than that of the tight spec critical parameter for good TTV is the plate tempera- material. Loose spec material went through other ture, where poka yoke cannot be applied. However, it manufacturing processes that increased its TTV is imperative to control it using SPC charts. The most value, whereas tight spec material went straight to the appropriate SPC chart for this purpose is the Indi- TTV measurement stage. Hence, a true reflection of vidual – Moving Range (I-MR) control chart. the lapping process capability could be evaluated by Statistical Quality Control Chart is used to con- tight flatness spec material only. The process capabil- trol the key process output variables (KPOV). In this ity index increased from 0.21 to 3.87, suggesting project, the KPOV was TTV, which could be con- a significant improvement in the output quality of the trolled using the Mean and Std. Deviation (Xbar-S) process, as well as sustainable long-term stability of chart. the lapping process, in terms of controlling TTV of Control Plan is the master document, which lists wafers. all the parameters of the process that should be Stage 5 — the Control phase: poka yoke. This monitored and controlled to achieve the desired term, commonly referred to as “mistake proofing”, results. It is of utmost importance to communicate originated from Japanese words “poka” meaning the control procedure amongst various stakeholders. “avoid” and “yokeru” meaning “mistakes”. In the cur- Maintaining timely updates to the lapping process rent scenario, the sun gear ratio and the bottom plate control plan was also important. Proper control and speed were identified as two key input variables for vigilance mechanisms for the key input parameters TTV control in the lapping process. So, the poka yoke impacting the identified output and placing adequate

100 Engineering Management in Production and Services Volume 11 • Issue 2 • 2019 controls for achieving quantifiable improvements ambiguity or confusion. Following which, a detailed were crucial. breakdown of the components of the problem facili- tated a productive brainstorming session, resulting in a list of possible causes and solutions. For example, 4. Discussion of the results initially, a long list of over 60 potential causes was generated. It was practically impossible to evaluate all The current implementation of the Six Sigma of them at the same time. However, it did not take methodology to reduce flatness defects in lapping very long for the Six Sigma experts in the team to was successful. The silicon wafer flatness was found to suggest solutions for dealing with such situations, be dependent on several and their interaction. This which was only possible due to a clear and detailed project presented an interesting revelation regarding written statement of the problem. Six Sigma tools of the key input variables that can affect lapping TTV. project framing, cause and effect diagram and Y=f(x) Based on the controlled design of experiments, it can cascade helped to shorten the long list of identified be concluded that the sun gear ratio and the plate potential causes, thus making it easier to reduce the temperature were the two most critical input variables scope of the project to a manageable extent and pro- affecting TTV. The interaction between these two viding the necessary focus to the evaluation. factors was also statistically significant. The sun gear A seamless, systematic roadmap of statistical ratio of 0.7 and the plate temperature of 30 ºC were thinking was ensured. Six Sigma is a step-by-step the optimum setting to minimise TTV. However, care procedure for problem solving or process improve- must be taken as these were optimum settings for one ment. Previously, at this organisation, all the efforts to manufacturing plant location and were completely identify the root cause for high TTV rejects had different from other locations using a similar manu- failed. Whereas, with the help of Six Sigma methodol- facturing set-up. ogy, wherein every step was clearly laid out in suffi- The bottom plate speed was a statistically signifi- cient detail, the problem-solving process was much cant input factor. However, its impact on TTV was more straightforward. Whenever the project met not as high as the other two significant factors (the a roadblock, a range of statistical tools was available sun gear ratio and the plate temperature). It was to facilitate the problem-solving process. For example, found that the bottom plate speed should be opti- after using several statistical and graphical analysis mised at 35 rpm. The wafer rotation was critical for techniques, such as Multi-Vari charts, there were still achieving the optimum TTV: Lapping is a closed nine potential factors whose impact on wafer flatness process, which makes it difficult to see what is going was unknown. At that stage, Taguchi screening DOE inside the machine. As per the results of the test, helped reveal that causal relationship. The learning which were discussed earlier, the increasing sun gear point here is that since there are a wide number of ratio and the bottom plate speed helped with wafer tools and techniques listed at every step of the five rotation. Further, the increasing plate temperature phases of the Six Sigma process, at least one of them reduced the viscosity of lapping slurry, thereby aiding is likely to work for the concerned process. the wafer rotation as well. This suggested that if sili- Statistical tools were used to validate the engi- con wafer does not receive enough rotation in the neering models. Statistical knowledge forms the core lapping carrier, it can result in uneven lapping of the Six Sigma methodology, which is yet another removal, thereby high TTV. important learning point gained from the current Detailed scoping of the problem at early stages project. The implementation of the Six Sigma meth- helped to generate the following solutions. As part of odology in the current project can be argued to be the Six Sigma approach, a clear statement of the a creative interplay between three components of problem was drafted at the start of the project. Con- managerial objectives, the expansion of engineering sistent with evidence of previous research (Mishra, knowledge and the application of statistical tech- 2018; Pyzdek, 2003), the scoping exercise largely niques. Statistical analyses were used to validate the consisted of ensuring that team members understand theoretical engineering models. the nature of the problem and what is expected of The full factorial DOE technique gives the Six them. A clear and shared understanding of the prob- Sigma methodology an edge over traditional problem and the process at the early stages enabled work- lem-solving techniques. Full factorial DOE tool, ing towards a common goal and channelling of team which also belongs to Six-Sigma, was particularly members’ efforts in the right direction, avoiding any useful in identifying key factors that were responsible

101 Engineering Management in Production and Services Volume 11 • Issue 2 • 2019 for causing TTV rejects. Previous studies have also original state. Nevertheless, another measure of long- shown that the application of Taguchi and full facto- term stability of an improved process achieved rial DOE technique helps to identify sources of varia- through the implementation of Six Sigma methodolo- tion in the process (Ghosh & Rao, 1996). Previous gies is the process capability (Ppk) index. The Ppk attempts at this organisation involved the use of One index of the current process was found to be 3.87, Factor at a Time (OFAT) technique. This technique which indicates that a sustainable solution to the had repeatedly failed to identify the underlying cause problem has been identified and implemented. It also in the current scenario. In hindsight, it is now logical ensures that correct factors responsible for poor wafer to understand why traditional problem-solving tech- flatness have been identified and that adequate con- niques had failed. In the current project, an interac- trol has been exerted to maintain a long-term process tion between more than one factors (plate temperature stability. Any decline in the process capability or and sun gear ratio) was responsible for high TTV quality over long term should, therefore, be unlikely. rejects. Such an evaluation of the interaction between more than one input factor would have been impossible to perform through traditional problem-solving Conclusions techniques, such as OFAT. The use of Full factorial DOE technique revealed not only the impact of all This study aimed to reduce flatness rejects in the three significant factors but also imparted new lapping process of silicon wafer manufacturing using insights on the interaction between these factors. It the Six Sigma methodology. It is a novel application was found that unless these factors are set at their of Six Sigma as previously it has been implemented in respective optimum levels simultaneously, inputting several industries, such as finance, service, manufac- of individual optimum values of these factors will still turing and non-manufacturing but not for the lapping result in poor TTV values. process of silicon wafer manufacturing in the semi- Implementation and control were also carefully conductor industry. A significant reduction in rejects considered. It had been argued in previous research from 4.83% to 0.02% was achieved through the (Sreedharan et al., 2019; Raman & Basavaraj, 2019) implementation of the Six Sigma methodology, that after the desired results through the Six Sigma resulting in the savings of £57.5k/month. A signifi- methodology are achieved, if appropriate monitoring cant increase in the capability index (Ppk) of the lap- and control procedures are not established, then most ping process also occurred, which is indicative of the likely the process would regress back to its original, enhanced product quality and efficiency, thereby flawed state (Antony & Coronado, 2002; Muraliraj et increasing customer satisfaction. In the current proj- al., 2018). A similar phenomenon had been observed ect, three factors (the sun gear ratio, the plate tem- in this organisation, where an effective solution had perature and the bottom plate speed) and the temporarily fixed a problem, but a lack of appropriate interaction between the sun gear ratio and the plate control systems caused the re-appearance of the same temperature were statistically significant. The optimi- issue. For such reasons, in the current project, much sation and tighter control of these variables were key emphasis was put on the control stage of the Six for the successful reduction in TTV rejects. Sigma methodology, which was a new experience for On a large scale, dealing with such complex pro- the entire team and, probably, the organisation too. cesses as lapping, it is often difficult to agree on a starting point and contain the scope to a manageable extent. The Six Sigma techniques of define and Limitations measure provided a much-needed direction and structure to the problem-solving process, at an early How stable are the improvements in the lapping stage. Previous attempts using traditional problem- process? It has been shown that despite noticeable solving techniques had failed since several potential improvements achieved by a project in the short- factors and their complex interactions were respon- term, its sigma level and performance can decline sible for high TTV rejects. On the other hand, the Six over time by 1.5 sigma. The standardised six-sigma Sigma technique could quantify and rank order sev- level accounts for such variations in the long-term. eral factors with their interactions in terms of their Since the current process did not achieve the full six- relative effect on the process outcome. Other Six sigma level, it is likely that in case of a 1.5 sigma shift Sigma pointers, such as the commitment of the top occurring, the process capability might regress to its management, the interim publication of success sto-

102 Engineering Management in Production and Services Volume 11 • Issue 2 • 2019 ries and open communication, also helped to main- Bothe, D. (2001). Statistical reason for the 1.5s shift. Quality Engi- tain the motivation and focus throughout the project. neering, 14(3), 479-488. Brewer, P., & Bagranoff, N. A. (2004). Near zero-defect accounting Overall, Six Sigma was shown to be an effective with six-sigma. Journal of Corporate Accounting & Finance problem solving and quality improvement statistical (Wiley), 15(2), 67-43. technique for the semiconductor industry, which lead Breyfogle, F. W. III. (1999). Implementing six-sigma: smarter so- to a series of other benefits (such as the increase in the lutions using statistical methods. New York, United States: John Wiley & Sons. process efficiency, the reduction in time and financial Breyfogle, F. W., Cupello, J. M., & Meadows, B. (2001). Managing costs, the enhanced satisfaction of customers and six-sigma: a practical guide to understanding, assessing, and employees as well as the provision of innovative team implementing the strategy that yields bottom-line success. building and networking opportunities), in addition New York, United States: Wiley. to the defect rate reduction in the lapping process. Brown, M. G., Hitchcock, D. E., & Willard, M. A. (1994). Why TQM fails and what to do about it. Boston, United States: The relevance of Six Sigma in the manufacturing Irwin Professional. industry is irrefutable, but there may be a need to Bunce, M., Wang, L., & Bidanda, B. (2008). Leveraging six-sigma conduct further rigorous research on its utility across with industrial engineering tools in crateless retort produc- non-conventional sectors and the practicalities in tion. International Journal of Production Research, 46(23), 6701-6719. terms of incurring costs and resources. Byrne, J. A. (1998). How Jack Welch runs GE. Business Week, Charles, 8 June 1997, 90-104. Cao, G., Clarke, S., & Lehaney, B. (2000). A systematic view of or- Literature ganisational change and TQM. The TQM Magazine, 12(3), 186-193.

Agarwal, R., & Bajaj, N. (2008). Managing outsourcing process: Chappell, A., & Peck, H. (2006). Risk management in military sup- applying six-sigma. Business Process Management Journal, ply chains: is there a role for six-sigma? International Jour- 14(6), 829-837. nal of Logistics, 9(3), 253-67. Al-Aomar, R. (2006). A simulation-based DFSS for a lean ser- Cole, R. (1998). Learning from the quality movement. California vice system. International Journal of Product Development, Management Review, 41, 43-73. 3(3/4), 349-368. Craven, E., Clark, J., Cramer, M., Corwin, S., & Cooper, M. (2006). Alexander, P., Antony, J., & Rodgers, B. (2019). Lean Six Sigma New York-Presbyterian hospital uses six-sigma to build for small- and medium-sized manufacturing enterpris- a culture of quality and innovation. Journal of Organiza- es: a systematic review. International Journal of Quality tional Excellence, 25(4), 11-19. & Reliability Management, 36(3), 378-397. doi: 10.1108/ Crosby, P. B. (1979). Quality is free. New York, United States: IJQRM-03-2018-0074 McGraw-Hill. Andersson, R., Eriksson, H., & Torstensson, H. (2006) Similari- Crosby, P. B. (1984). Quality without tears. New York, United Sta- ties and differences between TQM, six-sigma and lean. The tes: McGraw-Hill. TQM Magazine, 18(3), 282-296. Cupryk, M., Takahata, D., & Morusca, D. (2007). Crashing the Antony, J. (2004). Some pros and cons of six-sigma: an academic schedule’ in DCS validation pharmaceutical projects with perspective. The TQM Magazine, 16(4), 303-306. lean Six-sigma and project management techniques: case Antony, J. (2006). Six-sigma for service processes. Business Process study and discussion. Journal of Validation Technology, Management Journal, 12(2), 234-248. 13(3), 222-233. Antony, J., & Coronado, R. B. (2002). Design for six-sigma. IEE Da Silva, I. B., Filho, M. G., Luiz, A. O., & Lima junior, O. F. (2019). Manufacturing Engineer, 24-26. A new Lean Six Sigma framework for improving competi- tiveness. Acta Scientiarum Technology, Maringa, 41, e37327. Antony, J., Snee, R., & Hoerl, R. (2017). Lean Six Sigma: yester- doi:10.4025/actascitechnol.v41i2.37327 day, today and tomorrow. International Journal of Quality & Reliability Management 34(7),1073-1093. doi: 10.1108/ Dimitri, K. (1991). Reliability engineering handbook. New Jersey, IJQRM-03-2016-0035 United States: Prentice-Hall, Englewood Cliffs. Arul, K., & Kohli, H. (2004). Six-sigma for software application of Doble, M. (2005). Six-sigma and chemical process safety. Interna- hypothesis tests to software Data. Software Quality Journal, tional Journal of Six-sigma and Competitive Advantage, 1(2), 12(1), 29-42. 229-244. Bandyopadhyay, J., & Lichtman, R. (2007). Six-sigma approach to Edgeman, R., Bigio, D., & Ferleman, T. (2005). Six-sigma and busi- quality and productivity improvement in an institution for ness excellence: strategic and tactical examination of IT ser- higher education in the United States. International Journal vice level management at the office of the chief technology of Management, 24(4), 802-808. officer of Washington, DC. Quality and Reliability Engineer- ing International, 21(3), 257-273. Benedetto, A. (2003). Adapting manufacturing-based six-sigma methodology to the service environment of a radiology Esaki, K. (2016). Common Management Process Model of new film library. Journal of Healthcare Management, 48, 263-80. TQM based on the situation analysis. Intelligent Informa- tion Management, 8(6), 181-193. Boaden, R. J. (1997). What is total quality management… and does it matter? Total Quality Management, 8(4), 153-171. Eskildson, L. (1994). Improving the odds of TQM’s success. Quality Progress, 27(4), 61-63. Bolt, C., Keim, E., Kim, S., & Palser,th L. (2000). Service quality six- sigma case studies. ASQ’s 54 Annual Quality Congress Pro- Evans, J. R., & Dean, J. W. Jr. (2002). Total Quality: Management, ceedings, Anais, 225-231. organisation, strategy. Mason, United States: South Western

103 Engineering Management in Production and Services Volume 11 • Issue 2 • 2019

Publishing. Kume, H. (1995). Management by quality. Tokyo, Japan: 3A Cor- Foley, K. (2004). Five essays on quality management. Sydney, Aus- poration. tralia: SAI Global Ltd. Kume, H. (Ed.) (1985). Statistical methods for quality improve- Furterer, S., & Elshennawy, A. (2005). Implementation of TQM ment. Tokyo, Japan: The Association for Overseas Technical and lean six-sigma tools in local government: a framework Scholarship. and a case study. Total Quality Management and Business Magnusson, K., Kroslid, D., & Bergman, B. (2003). Six-sigma. The Excellence, 16(10), 1179-1191. pragmatic approach. 2nd ed. Lund, Sweden: Student litera- Gangidi, P. R. (2019). Application of Six Sigma in Semiconductor ture. Manufacturing: A Case Study in Yield Improvement, Appli- McClusky, R. (2000). The rise, fall and revival of six-sigma. Mea- cations of Design for Manufacturing and Assembly. Ancuţa suring Business Excellence, 4(2), 6-17. Păcurar, IntechOpen. doi: 10.5772/intechopen.81058. Mishra, M. N. (2018). Identify critical success factors to Retrieved from https://www.intechopen.com/books/ap- implement integrated green and Lean Six Sigma. In- plications-of-design-for-manufacturing-and-assembly/ap- ternational Journal of Lean Six Sigma. doi: 10.1108/ plication-of-six-sigma-in-semiconductor-manufacturing- IJLSS-07-2017-0076 a-case-study-in-yield-improvement Muraliraj, J., Zailani, S., Kuppuswamy, S., & Santha, C. Garvin, D. A. (1988). Managing Quality: The strategic and competi- (2018). Annotated methodological review of Lean tive edge. New York, United States: The Free Press. Six Sigma. International Journal of Lean Six Sigma, 9(1), Gerhorst, F., Grömping, U., Lloyd-Thomas, D., & Khalaf, F. (2006). 2-49. doi: 10.1108/IJLSS-04-2017-0028 Design for six-sigma in product development at Ford Mo- Nwabueze, U. (2001). An industry betrayed: the case of total qual- tor Company in a case study on robust exhaust manifold ity management in manufacturing. The TQM Magazine, design. International Journal of Product Development, 13(6), 400-408. 3(3/4), 278-91. Padhy, R. K., & Sahu, S. (2011). A Real Option based Six-sigma Ghosh, S., & Rao, C. R. (Eds.) (1996). Design and analysis of experi- project evaluation and selection model. International Jour- ments. Handbook of Statistics. 13. North-Holland. nal of Project Management, 29, 1091-1102. Harari, O. (1997). Ten reasons why TQM doesn’t work. Manage- Pandey, A. (2007). Strategically focused training in six-sigma way: ment Review, 86(1), 38-44. A case study. Journal of European Industrial Training, 31(2), Harry, M. J. (1988). The nature of six-sigma quality. Rolling Me- 145-162. adows, United States: Motorola University Press. Pyzdek, T. (2019). The six-sigma handbook: A complete guide for Hoerl, R. W. (1998). Six-sigma and the future of the quality profes- black belts, green belts and managers at all levels (5th edn.). sion. Quality Progress, 31(6), 35-42. New York, United States: McGraw-Hill. Hossain, M. Z., Tasnim, M., & Hasan, M. R. (2017). Is Quality En- Raman, R., & Basavaraj, Y. (2019). Defect reduction in a capacitor suring to Get Competitive Advantages in Auto Manufactur- manufacturing process through Six Sigma concept: A case ing Industries? – A Study of Volvo Group. American Journal study. Management Science Letters, 9(2), 253-260. of Industrial and Business Management, 7(1), 48-68. Reeves, C. A., & Bednar, D. A. (1994). Defining quality: alterna- Ishikawa, K. (1968). Guide to quality control (Japanese): Gemba No tives and implications. Academy of Management Review, 13, QC Shuho. Tokyo, Japan: JUSE Press Ltd. 419-445. Ishikawa, K. (1985). What is Total Quality Control? The Japanese Roberts, C. M. (2004). Six-sigma signals. Credit Union Magazine, Way. New Jersey, United States: Englewood Cliffs, Prentice- 70(1), 40-43. Hall. Rucker, R. (2000). Citibank increases customer loyalty with defect- Jimoh, R., Oyewobi, L., Isa, R., & Waziri, I. (2018). Total Quality free processes. The Journal for Quality and Participation, 23, Management Practices and organisational performance: the 32-36. mediating role of strategies for continuous improvements. Schmidt, M., & Aschkenase, S. (2004). The building blocks of ser- International Journal of Construction Management, 19(2), vice. Supply Chain Management Review, 07-08, 34-40. 162-177. Sreedharan, R. V., Raju, R., Vijaya Sunder, M., & Jiju, A. (2019). Juran, J. (1995b). Summary, trends, and prognosis. In Juran, J. Assessment of Lean Six Sigma Readiness (LESIRE) for man- M. (Ed.). A History of managing for quality: The evolution, ufacturing industries using fuzzy logic. International Jour- trends and future directions of managing for quality (pp. 603- nal of Quality & Reliability Management, 36(2), 137- 657). Milwaukee, United States: ASQC Quality Press. 161. doi: https://doi.org/10.1108/IJQRM-09-2017-0181 Kapur, K. C., & Lamberson, L. R. (1977). Reliability in engineering Su, C. T., & Chou, C. J. (2008). A systematic methodology for the design. New York, United States: John Wiley & Sons. creation of six-sigma projects: A case study of semiconduc- Kazmierski, C. (2012). Semiconductor Industry Posts Record- tor foundry. Expert Systems with Applications, 34, 2693- Breaking Revenues Despite 2011 Challenges. Retrieved from 2703. http://www.sia-online.org/news/2012/02/06/global-sales- Talapatra, S., Uddin, M. K., Rahman, M. H. (2018). Development report-2012/semiconductor-industry-posts-record-break- of an Implementation Framework for Integrated Manage- ing-revenues-despite-2011-challenges/ ment System Based on the Philosophy of Total Quality Kedar, A. P., Lakhe, R. R., Washimkar, V. S., & Wakhare, M. V. Management. American Journal of Industrial Business Man- (2008). A comparative review of TQM, TPM and related or- agement, 8(6), 1507-1516. ganisational performance improvement programs. ICETET Tobin, L. M. (1990). The new quality landscape: Total Quality ‘08: First International Conference on Emerging Trends in Management. Journal of System Management, 41, 10-14. Engineering and Technology, India. Topalovic, S. (2015). The Implementation of Total Quality Man- Kotz, S. (1993). Process capability indices. New York, United States: agement in order to improve production performance and Chapman and Hall. enhancing the level of customer satisfaction. Procedia Tech- nology, 19, 1016-1022.

104 Engineering Management in Production and Services Volume 11 • Issue 2 • 2019

Valles, A., Sanchez, J., Morales, Noriega S., & Gomez Nunez, B. (2009). Implementation of Six Sigma in a Manufacturing Process: A Case Study. The International Journal of Indus- trial Engineering: Theory, Applications and Practice, 16, 171- 181. Wright, A. (1997). Public service quality: Lessons not learned. Quality Management, 8(5), 313-318. Wyper, B., & Harrison, A. (2000). Deployment of six-sigma methodology in human resource function: a case study. Total Quality Management, 11(4-6), S720-S727. Yang, T., & Hsieh, C. H. (2009). Six-sigma project selection using national quality award criteria and Delphi fuzzy multiple decision-making method. Expert Systems with Applications, 36, 7594-7603. Yong, J., & Wilkinson, A. (2002). The long and winding road. The evolution of quality management. Total Quality Manage- ment, 13(1), 101-121.

105