1_pvi8-waferingarticle

8/6/2019 1_PVI8-waferingarticle

http://slidepdf.com/reader/full/1pvi8-waferingarticle 1/6

Photovol ta ics In ternat ional 1

PVI8-17_2

Market Watch

PowerGeneration

CellProcessing

PVModules

Materials

ThinFilm

Fab &FacilitiesCost reduction and productivity

improvement strategies formulticrystalline wafering processesMark Osborne , Senior News Editor, Photovoltaics International

Int odu tionPolysilicon, wafer and solar module pricesall declined severely in 2009. According toa recent report from iSuppli Corp., averagecrystalline module prices dropped by 37.8%, solar wafer prices fell by 50%, andpolysilicon prices declined by 80%.

The market research firm expects furtherprice declines in 2010, though not at thesteep levels seen last year. In 2010, iSuppliis forecasting price declines for crystalline

modules of 20%, solar wafers to declineby 18.2%, and polysilicon prices falling by a further 56.3% (see Fig. 1). Many industry observers now believe that with thecontinued growth in polysilicon productionby both major and new entrant producers,declining prices along the supply chain arenow set in stone for the industry.

Of course, this is a key prerequisite foran industry chasing the elusive grid parity (and beyond), which will benefit the playersin becoming increasingly competitive withother renewable energy forms.

“The erosion in pricing is bound tochange the face of the solar industry,”noted Henning Wicht, Senior Directorand Principal Analyst for PhotovoltaicSystems at iSuppli. “The freefall of PV prices represents a permanentratcheting down of price structures thatwill transform the industry into a morecompetitive marketplace.”

iSuppli expects PV manufacturers tocontinue to focus on cost reductions inorder to keep up with the price declinesand to repair compressed profit marginsexperienced in 2009. When the cost-down programs eventually catch upwith the rate of price declines, an overallimprovement in the profit picture can beexpected, Wicht said.

This is of particular importance topolysilicon and wafer producers, who haveexperienced the most aggressive pricedeclines, yet have significantly higher capitalcosts compared to others downstream, suchas pure-play module manufacturers.

Challenges also exist for waferproducers. With the strong wafer demandcreating a very tight supply environment,further capital expenditures will benecessary to meet customer demand.Traditionally in periods of strong demand,equipment lead times extend to over ninemonths as key production equipmentsuch as furnaces and wire saws are slowly churned out. These tools are large in sizeand/or complex in construction makingthem inherently difficult to manufacturein high quantities at the short lead times.

The ole of poly ili onDisconnects between points in the supply chain as seen recently in polysiliconshortages are in danger of transferring tothe wafering sector. This results in greaterpressure to boost productivity whilenew capacity is planned and eventually implemented.

As can be seen in Fig. 2, iSuppli hassurveyed wafer production expansion

AbsTrAcT

Multicrystalline wafers are the workhorse of the PV industry, with approximately 60% of crystalline silicon solar cellsmade from the substrate. They offer cost advantages in the form of good conversion efficiencies, which should continueto improve as cell technology advances continue. However, wafer prices were acutely impacted by the fall in PV marketdemand in late 2008, which continued through most of 2009. With relatively high capital costs, continued pricingpressures and calls for greater quality and control, wafer producers are now set on a course that requires rigorous andsustainable production cost-reduction strategies to meet customer requirements. This paper focuses on strategies thatcan be adopted to address this need for tighter quality specifications that reduce manufacturing costs downstream andboost cell conversion efficiencies.

Figu e 1. Ave age elling p i e fo e a t 2008-2012 in Us$/MW. C o u r t e s y o

f i S u p p

l i C o r p .



2

PVI8-17_2

www.pv-tech.org

Market Watch

Cellrocessing

Fab &Facilities

ThinFilm

Materials

Powereneration

PVModules

plans of producers that show only agradual expansion (left column) in wafercapacity based on actual planned capacity additions. The middle column shows thepotential additional plans that could beundertaken but have yet to materialize.Only in 2012 and beyond will we see thoseplans add meaningful production capacity.

It could be argued that such a forecastonly reinforces the pressures on waferproducers to optimize current capacity throughput, yield and overall productivity.

New capacity additions could also benefitfrom new technology introductions thatalso reduce production costs. Those thatsuccessfully tackle these challenges willnot only benefit from lower costs andimproved margins, but also attract morecustomers downstream requiring lowerprices to remain competitive. They willalso gain advantage in a continuing overallprice decline environment.

“Wafer producers havebenefited from an overallimprovement in material

quality due to betteravailability of virgin

polysilicon. ”There should be little doubt that

po lys i l i con and so la r wafe r s a reintrinsically linked. Wafer price declineshave primarily been a result of the evengreater decline in virgin polysiliconprices since mid-2008. Numerous market

research reports point to continuingcapacity expansions that should lead tofurther price declines. A recent reportfrom Bernreuter Research estimates thatglobal polysilicon production will reach

Figu e 2. Glo al ola wafe p odu tion apa ity plan (MW).

C o u r t e s y o

f i S u p p

l i C o r p .

Figu e 3. be n eute re ea h: Poly ili on upply demand p i ing dynami .

Figu e 4. c u i le at ce adyne Tianjin Te hni al ce ami .




PVI8-17_2

Market Watch

PowerGeneration

CellProcessing

PVModules

Materials

ThinFilm

Fab &Facilities

250,000MT in 2012, with approximately 80,000MT produced in China alone,making up about one-third of globalproduction.

Johannes Bernreuter, founder of the research firm, told Photovoltaics

International that polysilicon supply/demand dynamics this year wouldtranslate into spot pricing in the range of US$45 to US$50/kg.

“In 2011, I expect that a significant volume in China will probably beproduced at manufacturing costs belowUS$35/kg, and that the spot price will fallto this level – whether by the end of the year or earlier,” commented Bernreuter.

Wafer prices will therefore continue todecline, but as seen in 2009, prices fell muchmore than manufacturing costs, leading tosignificant margin declines. However, waferproducers have actually benefited from theplentiful supply of polysilicon in a differentway. According to Nick Sarno, the formerVP of Manufacturing at LDK Solar andnow currently consulting for the industry,wafer producers have benefited from anoverall improvement in material quality dueto better availability of virgin polysiliconand the declining dependence for someusing scrap silicon to complement ingotproduction.

“This translates into cost reductionsthroughout the wafer manufacturingprocess,” noted Sarno. “Less man hourson procurement of scrap silicon, sortingand storing has a small impact but betterquality supply of virgin material producesbetter ingots, blocks and wafers, reducingscrap, consumables and delivering a betterproduct all round.”

Fu na e : i ize the olution?In an effort to reduce ingot costs, there hasbeen a gradual but growing trend towardslarger ingot sizes. The larger the ingot,the more blocks and consequently waferscan be produced, resulting in increasedoverall yield increases (see Table 1). Thistherefore requires larger ingot growthfurnaces. Much of the current ingotproduction is performed using DirectionalSolidification System (DSS) furnaces

that cast multicrystalline ingots using450kg furnaces for volume production.Developments continue on charge sizesabove 500kg and 800kg, and there areefforts underway to increase ingot sizes to1,000kg in the not-too-distant future.

In today’s market, a key limitation is thetechnological capability to manufacturecrucibles of ever-increasing size.Common industry practice has been totackle this obstacle by adopting largerfurnaces but using multiple crucibles at atime within the furnace for optimizationand greater throughput.

“From a capital expenditure point of view this strategy is a plus,” noted Sarno.“You are future-proofing investments inanticipation of the larger crucibles beingavailable, while not having additional

investments for going bigger just because

of crucible size limitations.”Crucible manufacturers are attemptingto address several challenges at once,answering increased demand for larger-sized crucibles – in particular 450kgproducts – while re-tooling for evenbigger sizes in the future.

“The two key metricsfor cost reduction at thefurnace are throughput

and yield,

”Ceradyne, a leading fused silica cruciblemanufacturer, is expanding manufacturingin the first half of 2010 in its existingfacility, Ceradyne Tianjin TechnicalCeramics (see Fig. 4), and has brokenground on a second facility, CeradyneTianjin Advanced Materials, which willbe operational during early 2011, to meetgrowing demand.

A leading DSS furnace supplier is GTSolar, a company that has certainly seenthe growing demand for larger systems inthe last few years. During fiscal year 2009and 2010, GT Solar sold approximately US$1 bil l ion of capital equipmentinto the market to support the build-out of installed end-user capacity of approximately 8GW to 10GW.

Interestingly, the issue of larger

furnace sizes was not the first point of conversation that Henry Chou, ProductMarketing Manager, PV Equipmentfor GT Solar International wanted tomention. Rather, he noted that as theindustry becomes more mature, GT Solarhas brought to the table what he describesas the ‘value metric.’ Using simplified variables to highlight manufacturing costs,the company targets cost reductions viaimproved throughput of the tool coupledto higher yields and better overall ingotquality. This then translates into lesswastage from the ingot to the wafersawing steps, also lowering costs.

“Being able to get better quality wafers as a result does indeed enable cellproducers to achieve higher conversionefficiencies,” commented Chou. “What welook for is good kilograms produced perhour.”

The two key metrics for cost reductionat the furnace are throughput and yield,according to Chou. With respect tothroughput, this relates to the charge sizedivided by the process time.

“Interestingly, if you start to add moremass to the charge, your process time willstart to increase. Based on the processtime, the quality of the segregation of impurities and growth of the crystalschanges the optimum balance. So if yougrow bigger ingots it doesn’t necessarily mean that you are going to get higher

Figu e 5. GT sola ’ Dss450HP ingot g owth fu na e.

Model Name JZ-270 JZ-450 JZ-520 JZ-800

Initial Charge of Poly-silicon (kg) 270 450 520 800Estimated Dimension of the L & W 680 840 840 996Solidified Ingot (mm) Thickness 250 270 315 345Estimated Dimension of L & W 156 156 156 156Cropped Blocks (mm) Thickness 200 220 265 295

Estimated Weight of Cropped Ingot (kg) 181.4 311.8 375.7 602.2Estimated Yield for Ingot (%) 67.2 69.3 72.3 75.3

Ta le 1. JYT la ge fu na e enefit analy i .



4

PVI8-17_2

www.pv-tech.org

Market Watch

Cellrocessing

Fab &Facilities

ThinFilm

Materials

Powereneration

PVModules

throughput, due to the growing cycleneeding to be longer,” added Chou.

He also believes that ingot sizes above600kg could lead to other equipmentchanges that actually add to the cost of production. Consensus on whether chargesizes above 450kg are inevitable wouldseem to be premature. The company hasrecently been tapped for its ‘DSS450’furnaces and ancillary equipment andservices to the value of US$200 million.The orders come from a range of waferproducers, including Tianwei New Energy,Phoenix Photovoltaic Technology, YingliGreen Energy, JA Solar and Sino-AmericanSilicon Products (SAS).

In early May 2010, the company launched the ‘DSS450HP,’ a highperformance ingot growth furnace with athermally optimized, second-generationhot zone that is said to improve throughputwhile delivering industry-leading crystalquality. Current users of the DSS450and DSS240 models can migrate to thenew DSS450HP with a field upgrade kit,which means that in the near term, waferproducers are regarding proven technology offerings as preferable to pushing theboundaries of even larger ingot/furnacesizes.

slu y olutionSlu r ry i s a t t he hea r t o f wafe rmanufacturing. It is used with wire sawsto cut the wafers out of silicon ingots,normally comprising an abrasive mixtureof silicon carbide and ethylene glycolthat erodes, rather than cuts, the ingot.Generally, a conversation centring onwhere key costs are located across the various wafering processes will leadstraight to the topic of slurry.

As the slurry is fed along a guide wire,the abrasive silicon carbide in the slurry (or

‘grit’) cuts through the ingots to producethe wafers. Slurry should be regularly replaced and reprocessed in order to keepthe number and quality of the wafers highwhile maintaining acceptable unit cost.

Although slurry is vital to the process,it also represents a significant proportionof the wafer production cost; slurry costsare often second only to the cost of thepolysilicon itself. Despite the relatively low price of US$3.50 per kilogram fornew slurry, the overall cost can quickly escalate. In a scenario where 10 wiresaws are being used in a process, theoperating cost can be as high as US$16million a year, not including disposal feesor labour charges.

“A do-it-yourself scenario

may impact quality controland productivity. ”For the vast majority of manufacturers,

slurry reprocessing ranks in the top fivecontributors to operating cost. Optimizingthis aspect of the business can havedramatic and lasting effects on waferproduction costs and overall quality.

According to CRS ReprocessingServices, a leading slurry recycling firm,wafer manufacturers that are planning aslurry reprocessing program or starting

from scratch in a greenfield operationcan benefit from significant margin-enhancing improvements. For a waferingoperation running 10 wire saws with a volume of 385 metric tons per month,slurry reprocessing can result in a savingof anwhere in the region of US$8-10million per year, net of recycling costs,utilities, and infrastructure investment.

“At CRS Reprocessing Services we areable recover between 80% and 90% of both the grit and the carrier,” noted BillLawrence, founder and president. “Theactual amount of recovery is a functionof the wafer size, wafer thickness, saw set-up and the silicon kerf loading in the usedslurry. For oil-based slurries, the recovery of the cleaning solvents used to rinsewafers and equipment is typically greater

than 90%. In short, the higher the recovery rate, the more material is effectively recovered and reused, and the less thatmust be spent on new material.”

The benefits of slurry recycling were

echoed by Nick Sarno, who recalled ahuge move toward recovery strategiesduring his time at LDK Solar. Sarno notedthat LDK has a 30,000MT recyclingsystem designed in-house (see Fig. 7),with a similar system being installed asthe company ramped wafer productionbeyond 2GW per annum. Sarno reiteratedthat cost avoidance was a key aspect asrecycling significantly reduces the use of new material.

Perhaps not surprisingly, CRS’sLawrence cautioned against the tendency for wafer producers attempting to design

in-house systems instead of using fully supplied and managed systems from expertthird parties.

“The downside is that do-it-yourself options tend to have much lower recovery rates for both grit and for the carrier.These processes typically recover only asmall portion, say, 20-30%, of the carrier.In addition, a do-it-yourself scenario may impact quality control and productivity resulting from lower specifications,limited lab verification tools and increaseddowntime.

“To put it in perspective, an optimal

reprocessing solution that increased therecovery of both grit and carrier couldeasily decrease the overall costs for a10-wire-saw operation by $275,000 to$375,000 a month. These factors, alongwith the need to fund resources that arenot central to the business, reduce thepotential savings that are expected by taking the process in-house.”

In general, there are four options whenit comes to reprocessing slurry: do-it- yourself, off-site reprocessing, on-sitereprocessing, or do no reprocessing at all.

Lawrence went on to point out some

of the cost issues associated with off-sitereprocessing, which is popular in Europe.He noted that the service fee for off-sitereprocessing can be competitive, perhapseven lower than the common alternative of

Figu e 6. crs rep o e ing se vi e ’ typi al lu y e y ling y tem.

Figu e 7. In-hou e lu y e y lingy tem at LDK sola .




PVI8-17_2

Market Watch

PowerGeneration

CellProcessing

PVModules

Materials

ThinFilm

Fab &Facilities

on-site reprocessing. There are a numberof other costs, however, that factor into anoff-site processing scenario that should beconsidered.

Chief among these are freight costs– both trucking and shipping, whichcan vary depending on the amountof slurry and transportation distance.In some cases, the expense can beconsiderable. Rates range from $0.03 perkilogram within a country to up to $0.40per kilogram around the world. For acustomer processing 800MT per month,this could total over $200,000 per month.Off-site reprocessing also requires that amanufacturer have large amounts of slurry ‘in play.’

Many of these issues have longpervaded in parallel industries such assemiconductor manufacturing, wheremost of the required recycling is done onsite, bringing with it other benefits.

From a quality control standpoint,on - s i t e r ep rocess ing b r ings fu l ltransparency to the manufacturer, whogains the advantage of immediate and verifiable slurry. Since the slurry doesnot leave the facility, it is reprocessedin a closed loop that eliminates the riskof outside contaminants entering theproduction stream. On-site testing,conducted by experts who can easily andquickly adjust levels to achieve consistentand optimized slurry, helps ensure wafer yields are high with minimal waste.

Once again, the issue of quality is akey factor that should not be overlookedin the pursuit of straightforward costanalysis. Optimized yield and highquality products are often greater costsaving pursuits.

chemi al ont i utionOf course, slurry is not the only chemicalused in wafer processing. Cleaning andtexturizing steps use HF acids in baths.Simply shopping around for lower costbulk chemicals may prove effective inthe short term, but once again, usingthe right chemistries for yield and waferquality considerations could prove to be agreater contribution to cost reduction andoptimization strategies.

Frank Haunert, Product Manager atBASF, commented that ‘solar-grade’ bulkchemicals such as HF acids that are notspecific to semiconductor purity levelshave been developed, which means thatthey can therefore be offered at lowerprices, while maintaining the quality andconsistency requirements that optimizeprocesses. With respect to BASF’s‘Seluris’ range of etching and texturizingchemicals, Haunert noted that they enable saw damage to be rectified andthe wafer surfaces to be structured from‘drop-in’ solutions, reducing quality inconsistencies and delivering a morehomogenous wafer result.

With respect to slurry, Haunert notedthey are developing slurries that have

tackled issues such as wire pairing: “Thetypical wire saw can experience surfacetensions that mean the wires cannot getthrough the slurry solution and startcoming together creating one thin and onethicker wafer. This has a negative impactof yield and overall productivity.”

Diamond wi e olutionThe new kid on the block for sawingbricks and wafers is diamond wiretechnology (DWT). A key bonus of diamond wires is that slurries are notrequired, therefore dispensing withthe expensive material altogether.This of course means that there are norecycling costs nor any of the other costsassociated with the use of slurry. DWTis also claimed to provide 2.5-3x thethroughput compared to conventionalwire/slurry combinations and is saidto offer improved cutting accuracy,reducing kerf loss and providing stableprocessing partly due to its simplicity relative to slurry-based processing.Although still in its infancy from anadoption perspective, the technology hassome potentially compelling aspects.

“We believe we can drive down thewafering costs by 10 to 15% initially withdiamond wire technology and furtherimprovements later,” commented PeterPauli, CEO of Meyer Burger Technology AG. “The important part of that is yieldimprovement. The industry as a wholeonly has an 80 to 90% total yield. If weonly deliver a 5% yield gain this would bea dramatic improvement for the industry.”

“A key bonus of diamondwires is that slurries arenot required, therefore

dispensing with the expensivematerial altogether.

”However, concerns have been raised

over the use of DWT, not only because of the inherent cost of diamond wire, but dueto the physical difference of the resultingwafers compared to conventional wire/slurry-produced wafers.

Nick Sarno cautioned that diamond

Figu e 8. Meye be ge ’ diamond wi e ompa i on ha t.

Figu e 9. Fewe si lo e a e po i le th ough the optimal po itioning of the lo k within the opping aw.



6

PVI8-17_2

www.pv-tech.org

Market Watch

Cellrocessing

Fab &Facilities

ThinFilm

Materials

Powereneration

PVModules

wire leaves score marks on the surfaceof the wafer, making the wafers lookphysically different from non-DWTproduced wafers. This has returnedsome negative feedback from customers,reporting that they are ‘a little afraid of using’ this type of wafer.

“Significant efforts are

ongoing to provide the yieldand cost reductions required

for the industry across thecomplete wafering

process flow. ”Meyer Burger’s CEO was well versed on

this issue: “The sub-surface damage is actually less than with slurry,” Pauli pointed out. “Thesurface may look a little bit different, butchanging to another technology takes time.We don’t expect a dramatic uptake of thetechnology right now.”

“Changing to another technology takes time” – the same can be saidabout changing perceptions regardingDWT. To that end, the company hasmade great strides in developing andimportantly demonstrating the cost and yield benefits of DWT. Meyer Berger hasinvested positively in the complete systemtechnology and infrastructure that wouldenable the industry to migrate to DWT atthe high volume commercial level. Paulinoted that the actual diamond wire was

only one aspect of the complete packageneeded to make adoption successful.

“The key is that DWT changeswafering, internally and dramatically,”he added. “It gives us the opportunity to change and improve the whole line– simplifying and reducing steps andequipment needed in the value chain.With overall complexity of the waferingprocess reduced, we have the ability toimprove the overall yield and bring theindustry into the industrialization phase.This, to me, is the future.”

Thinking thinWith the return of a plentiful supply of polysilicon, it would seem that there is lesstalk focused on migrating to thin wafers inorder to reduce cost. Although companieslike LDK Solar have continued to migrategradually to thinner wafers, there is lessemphasis on taking wafers thinner than170 microns, due to concerns at the cellprocessing level regarding higher rates of wafer breakage. Although developmentscontinue at the R&D level, productionefforts remain focused on kerf loss.

P o e optimizationHowever, conventional wire/slurry technology is not stagnant. Significantefforts are ongoing to provide the yieldand cost reductions required for the

industry across the complete waferingprocess flow.

Herbert Arnold GmbH & Co. KG isa leading manufacturer of productionsystems for processing mono- andmulticrystall ine sil icon bricks thatincorporate cutting, grinding and polishing.

Speaking with Wolfgang Schürgers,Sales Director at Herbert Arnold, itwas apparent that there are significantopportunities in regard to the optimizationof equipment and processes that not only shorten cycle times but also provide greaterunderstanding of the overall processes thatin return improve productivity, quality and yield. The introduction of advancedtechnology manufacturing systemsimprove process control and an integrated,process data analysis creates transparentand efficient production.

“What I can see is a lot of effort tooptimize Gen5 ingots to produce 25 bricksof higher quality and improve the height of ingots by 300 to 400mm,” noted Schürgers.“I don’t see the trend towards the one-toningot. Kerf loss is still an important issueand we are able to provide a cut with thesmallest kerf loss possible.”

Schürgers highlighted that croppingwith blade saws rather than wire sawsprovides accurate straight cutting andlower cost. The development of highprecision (1.5mm) blades with theoptimization of the thickness of thesmallest of blades significantly reduceskerf loss and boosts machine uptime,increasing throughput.

With the move towards larger waferproduction lines in excess of 500MWcapacity, there will be a growing trendtowards fully automated lines, accordingto Schürgers.

Herbert Arnold have calculatedthat automation of the entire processflow would generate at least a 10%overall productivity improvement.This is achieved by optimized 24-houroperations, reduced waiting times fortool availability and reduced machine idletime as a consequence. Importantly, highly optimized positioning of the brick and sawresults in consistently improved yield.

Schürgers believes that high-volumefacilities of 200MW and above that arestill dependent on manual handling willhave to deal with high error rates as aresult of human operations, which have aconsiderable impact on the line’s ability tobe fully optimized.

Optimization of the complete line wasrepeated as a necessity by Greg Knight,Director of Process Engineering Turnkey Services for GT Solar.

“Equipment is one piece in operatinga factory,” remarked Knight. “How youutilize that facility is actually a bigger

piece. We have done a lot of analysis onstreamlining operations such as howwe process material straight out of thefurnace. There would seem to be asignificant loss of revenue occurring in a

lot of wafering facilities because they aremaking the decision to cut on the markedbrick lifetime lines… while this is a usefulmetric, using metrology to determinewhat is good and bad enables only the badmaterial to be turned into dust.”

According to Knight, better or lessconventional cutting approaches can take yields to the 90% range when typically many wafer producers are operating in thesub-70% yield range.

Significant effort has been made withGT Solar’s turnkey lines to enable thefeedback of data from sawing to ingots.The multiple issues that can occursuch as rejects and saw marks can betraced back to the specific furnace andcorrection procedures implemented inorder to limit prolonged yield impacts.Detailed monitoring of the operationsenables tighter process tolerances thatreduce reactive changes and so boostsproductivity and reduces cost.

“Better or less conventionalcutting approaches can take yields to the 90% range. ”Chemical usage optimization was also

a factor covered by BASF’s Haunert. Henoted that efforts are ongoing to maximizethe use of a given amount of material forthe highest possible number of wafers.Aspects such as bath temperatures andflow rates are fine-tuned to enhance thechemical characteristics for processes suchas wafer cleaning.

con lu ionAs wafer producers recover from rapidly declining prices in 2009 and focus on anew wave of capacity expansions to meetstrong demand in 2010, cost reductioncontinues to be a core focus. Thecontinuing abundance of polysilicon may well continue to place margin pressureon wafer prices, despite tight supply anda short-term recovery in wafer prices.

Wafer producers that are vigilant inregard to cost and process optimizationstrategies will not only weather economicconditions better, but will likely flourishahead of those competitors that fail totackle these challenges in a coherentmanner..

1_pvi8-waferingarticle

Documents