450 mm silicon wafers challenges wafer thickness scaling

Size: px

Start display at page:

Download "450 mm silicon wafers challenges wafer thickness scaling"

Lilian Roberts
6 years ago
Views:

10.1149/1.98088 The Elecrochemical Sociey 450 mm silicon wafers challenges wafer hickness scaling M. Goldsein Inel, Sana Clara CA, USA and M.

1 / The Elecrochemical Sociey 450 mm silicon wafers challenges wafer hickness scaling M. Goldsein Inel, Sana Clara CA, USA and M. Waanabe NuFlare Technology, Numazu, Shizuoka, Japan Pilo line using 450 mm wafer will sar in 01 according o ITRS. One of he mos imporan discussions is wafer hickness scaling from mm wafer o 450 mm wafer. I is discussed in erms of empirical scaling, graviaional sag, hermal sress and fracure srengh. INTRODUCTION Today, we ake our favorie music and movies o anywhere. I is a resul of neverending reducion of bi cos brough by he developmen of semiconducor devices and heir producion echnology. There is he well-known Moore s Law. I is a core of he Inernaional Technology Roadmap of Semiconducors (ITRS) (1) ha serves as a echnology arge for he semiconducor indusry. An increase of silicon wafer size is embedded in i for more efficien and economical producion of devices. Silicon as a key maerial of he elecronic indusry keeps progress in defec engineering and producion echnology. An increase in he silicon wafer size every 4-5 echnology generaion is required for more efficien and economical manufacuring of IC devices. Pilo line producion on nex generaion wafer size, wih a diameer of 450 mm, is expeced in 01. R&D of 450 mm is underway as shown in Figure 1. In he 450 mm era, a silicon wafer will be 335g and consequenly graviy sag is larger han ever before. Large wafers will be more fragile. Discussions abou his provided useful informaion beyond mm echnology. How he 450 mm crysal will come o producion and wheher i is economically feasible is now a ho opic in he silicon indusry. A new indusry sandard () of 450 mm handling wafer is approved in 008 July a SEMICON WEST. Figure mm single crysal of is R&D sage. 3

2 450 mm WAFER THICKNESS ISSUES There are many echnical challenges facing 450 mm crysal growing and wafer manufacuring (3). One of hem is he wafer hickness scaling. Many facors have o be considered in defining he righ wafer hickness: Device manufacuring requiremens slip generaion due o hermal sress, elasic deformaion due o film sress, breakage during he processes Silicon suppliers manufacurabiliy surface polishing o igh flaness requiremens, breakage due o wafer increase fragiliy Wafer shipping and handling large graviaional sag, process induced bow and vibraion in he wafer carrier. Cos Wafer hickness should be considered carefully. Once i is decided, i is raher difficul o change aferwards. Many device producion echnologies and equipmens are wafer hickness dependen and a change of wafer hickness has a major impac. However, hickness can be revisied during he developmen of specificaion for 450 mm Process Wafers. Because acual daa available are no enough o decide on wafer hickness, herefore any revision based on he daa mus be considered a various sages of sandards developmen. Wafer hickness should be considered carefully. Many device producion echnologies and equipmens are wafer hickness dependen and a change of wafer hickness has a major impac. Lessons learned from pas wafer size ransiions (5) showed ha once he wafer hickness is specified i is difficul o change. Based on he simulaion daa (4,5), some daa using sinered polysilicon 450 mm wafer and suppliers inpu, an iniial hickness of 95 ± 5 µm was recommended as a good hickness for handling wafers. This is hicker han he exrapolaed hickness rend from 15 mm o mm wafers, as shown in Figure. Handling wafer specificaion sandards have been deermined as given in Table 1 (). Handling wafers used for roboics and equipmen developmen are manufacured o his spec. Table 1. Specificaion of 450 mm handling wafer Propery Specificaion DIAMETER ±0. mm THICKNESS, CENTER POINT 95 ± 5 µm NINE-POINT THICKNESS VARIATION, less han NOTCH DIMENSIONS (see Figure 1) Deph Angle FRONT SURFACE BACK SURFACE FINISH (GLOSS) EDGE PROFILE SURFACE FINISH 10 µm 1.00 mm+ 0.5 mm 0.00 mm POLISHED POLISHED POLISHED 4

3 EMPIRICAL SCALING Hisorically, wafer hickness has been increased by some facor as he wafer diameer seps up o he nex diameer since hin large wafers may be fragile and hard o handle. As shown in Figure, exrapolaion of he raher recen 15 mm, 00 mm and mm diameers rend o 450 mm indicaes ha 450 mm wafer hickness would be abou 85 µm. Third order polynomial curve fiing o semilog plo of he diameer rend indicaes ha he 450 mm wafer hickness would be abou 800 µm. This is a simple empirical exension of he hisorical diameer rend and 5 µm o 50 µm hicker han curren 775 µm hick mm wafer. SSi esimaed 85 µm for 400 mm and 875 µm for 450 mm (6). Taking ino accoun commens from wafer suppliers, anoher 50 µm is added. Specificaion of SEMI 450 mm handling wafer hickness is 95 µm as shown in Table 1. Thickness (µm) Exracion of empirical hickness rend Diameer (mm) Figure. Empirical scaling of 450 mm wafer hickness. WAFER SCALING IMPACT ON STRESS Wafer hickness has a minimal effec on wihin wafer sress. Sress due o graviy modeling in a verical furnace boa ype suppor is depiced in Figure 3. S ress due o Graviy 4.00 Sress (MPa) mm 775um 450mm 775um 450mm 875um 450mm 975um 450mm 1500um Figure 3. Wafer sress due o graviy wih verical furnace boa ype suppor. 5

4 Changing he wafer hickness from 775 µm o 900 µm changes he sress by ~5%. The higher diameer o hickness aspec raio increases he risk o slip during hermal processing due o sress from emperaure gradiens and deposied film o subsrae coefficien of hermal expansion. Modeling shows ha sresses in 450 mm wafer during ramp, from 100ºC 1100ºC a 10ºC/min are subsanially higher han in mm wafer. Figure 4 shows he sress evoluion. For a given furnace boa pich he emperaure gradien and he sress across he wafer increases when he diameer or he hickness are increased, as shown in Figure 4a Graviy + Ramp: Sress mm/775um 450mm/775um 450mm/875um 450mm/1500um Sress (MPa) Pich (mm) (a) (b) Figure 4. Simulaion of hermal effec on wafer sress wih verical furnace boa ype suppor. a. during 10ºC/min ramp from 100ºC o 1100ºC b. during fas push-in a 700ºC Fas boa push-in a a speed of ~00 mm/min from room emperaure ino a 700ºC ho furnace creaes high emperaure gradiens ha lead o very high sresses in he wafer during push-in, as shown in Figure 4b. 6

5 As expeced sress rends show srong dependence on he boa pich. Sresses in deposied film induce curvaure of he wafer. The radius of curvaure is proporional o he square of he wafer hickness. I can be calculaed using he Soney equaion WAFER SCALING IMPACT ON GRAVITY SAG Wafer graviaional sag is generaed by he wafer weigh under differen suppor posiions. Graviaional sag plays an imporan role in wafer handling and ransfer design. A good undersanding of wafer sag is required o enable wafer roboic ransfer and wafer carrier designs and wafer graviaional sag versus for differen suppor sysems was calculaed and a preliminary specificaion was issued. Modeling daa of graviaional sag of 450 mm wafer wih side rail suppor (FOSB ype) is presened in Figure 5a. When peripheral suppor is applied 450 mm wafer graviaional sag is expeced o be ~5x higher han ha of a mm wafer wih he same hickness (Figure 5b) Bow (um) (a) Graviy Bow in FOSB mm 450mm 450mm 450mm 775 um 85um 900um 1000um (b) Graviy Bow wih peripheral suppor Bow (um) mm 450mm 450mm 450mm 775 um 85um 900um 1000um Figure 5. Graviy sag modeling. 7

6 Early in he roboics and wafer carrier developmen work real experimenal daa was required. Mechanical grade, sinered silicon wafers wih differen hicknesses have been manufacured enabling wafer carriers, auomaion and OEM developmen work by providing wafers having he righ mechanical and opical characerisics. Some daa is given in Table. Table. Sinered Wafers Mechanical Properies Bending Srengh (MPa) Vickers Hardness (Hv) Young s Modulus (GPa) Poisson s Raio Sinered Si wafer Graviaional sag experimenal daa have been colleced wih differen wafer suppor. Experimenal daa measured by 450mm wafer wih peripheral suppor is presened in Figure mm wafer graviaional sag wih peripheral suppor Wafer sag (um ) um 900 um 1000 um Wafer hickness CMM Capaciance Figure mm sinered wafer graviaional sag wih peripheral suppor. The measuremen daa seem o be impaced by he measuremen procedure. CMM measuremen affec he sag due o addiional weigh of he measuring arm and capaciance measuremens do no supply poin daa bu an area average. In a FOSB ype of rail suppor he 900 µm wafer CMM measuremens show a bow of 950 µm. Repeiive measuremens daa of sinered wafers wih hickness of 95+/- 5 mm show graviaional sag of 940 +/- 0 mm. The wafer shape is impaced by he wafer suppor. Wih peripheral suppor he wafer has a bowl ype shape. When hold in a FOSB ype suppor he wafer ake a saddle shape ype (Figure 7). 8

7 Graviy sag Bowl wafer shape wih peripheral suppor Saddle wafer shape in FOSB Figure 7. wafer shape under differen suppor condiions. FRACTURE STRENGTH SCALING Thinner wafers are ofen supposed o be more fragile. If he wafer is no hick enough, i is expeced ha larger diameer wafers will be easy o break, a leas, according o one s experience, during device or wafer processes. Fracure srengh of brile maerial has been sudied for a long ime. Calculaion of he fracure srengh from measured daa of brile ceramic plaes, recangular or circular, is well esablished. Tha is, σ F G = where σ is fracure srengh, F is applied force a he fracure, is hickness and G is geomery fracure. G depends on es fixure, such as, four poin bending, hree poin bending, pison on ring, ball on ring and ring on ring. In case of pison on hree balls, ASTM F394 (8) is G = B C 1 ν B C A C ( 1+ ν ) ln + ( 1+ ν ) 1+ ln ( 1 ν ) A C where ν is Poisson s raio, A is radius of suppor circle on which hree balls are locaed, B is radius of pison o apply force o specimen and C is radius of specimen. Fracure srengh is supposed o be he same if wafers are processed in he same way. Applied force F is ambiguous. I would be some force ha causes wafer breakage during he wafer process. Geomery facor G can be calculaed assuming proper combinaion of A, B and C. Bawa (9) made choice ofσ, A, B and F and calculaed for various C. Alhough calculaed for 15, and mm wafers looks reasonable, i depends on he parameers. If a choice of parameers of any fracure srengh measuremen mehod, i will become a reasonable scaling mehod. Oherwise, his will be unrealisic. Fracure probabiliy of silicon wafer as brile maerial is P = A P + A P + A fracure bulk bulk surface surface edge P edge where P fracure is wafer fracure probabiliy. A bulk P bulk, A surface P surface, A edge P edge are bulk, surface and edge conribuion where A bulk, A surface and A edge are weighing facors respecively. Bulk conribuion will come from sress induced by oxygen precipiaes bu 9

8 i can be negleced since he fracure srengh of a plain wafer is considered as a bench mark. Edge conribuion will be imporan if edge conac or edge polishing are considered. However, daa ha show how edge flaw conribues o fracure srengh are no available in he public domain. Therefore, his facor canno be considered for he momen. Only surface facor is considered here. Fracure of brile maerials is described by Weibull saisics (10) like P where P(σ,V) is fracure probabiliy, in general, V is specimen volume, V eff is effecive volume of specimen under consideraion, V 0 is volume of reference specimen, σ is fracure srengh of specimen under consideraion, σ 0 is fracure srengh of reference specimen and m is specimen size scaling facor. Large m means ha fracure srengh scaers more due o wide size disribuion of flaw ha causes fracure. Wider he disribuion is more chance o include large flaw ha causes low fracure srengh in he specimen. If he disribuion is very sharp, fracure srengh does no depend on he specimen volume, since fracure srengh is a funcion of he size of he larges flaw in he specimen. More precisely (11), surface area dependen fracure probabiliy is P ( σ, V ) ( σ, S) V = 1 exp V eff 0 S 1 exp S m σ σ 0 m 1 σ Γ 1 + m σ 0 m eff = 0 where S eff is effecive surface area of specimen under consideraion and S 0 is surface area of reference specimen. To ge same probabiliy, S eff σ m = S 0 σ 0 m or σ = (S 0 / S eff ) 1/m σ 0. To compare fracure srengh of 450 mm wafer o ha of mm wafer, 450 mm wafer is (S 0 / S eff ) 1/m =1/.5 1/m weaker han mm wafer. To compensae his weak fracure srengh, 450 mm wafer needs o be hicker han mm wafer. Thickness dependence of fracure srengh of he measuremen heory canno be applied for his since i includes measuremen fixure geomery. When force is applied on a fron surface region where subhreshold flaw exiss on is back surface, fracure srengh is F σ = log B E E s where F is applied force, B=1.35, E is Young s modulus of specimen and E s is Young s modulus of supporing plae (1). Then, σ F/, his is equivalen o calculae fracure srengh from measured daa wihou geomery facor. When wafer is expeced o encouner he same force, σ will be consan for wafers of any size. Tha is 10

9 σ = = σ σ σ m =.5 where 450 and are hickness and σ 450 and σ are fracure srengh of 450 mm and mm wafers. Weibull modulus m of some silicon plaes are repored as shown Table 3. Table 3. Weibull modulus m from public domain daa. (Mpa) m size (mm ) 15x15 15x15 15x15 15x15 15x15 100x x x100 Surface Sawn Sawn eched eched eched Sawn Eched Eched (fine) (Coase (A) (B) (B) Alkalin Acid Thickness( m) Reference : Towards he correlaion of mechanical properies and sawing parameersof silicon wafer by A Bidiville, K. Wasmer, J. Michler, C. Ballif, M. Van der Meer and M. Nasch : Mechanical srengh of silicon wafers depending on wafer hickness and surface reamen G. Colei, N.J.C.M. Van der Borg, S. de Iullis, C.J.J. Tool and L.J. Geerligs, Proc. 1s European Solar Conf. 006 Dresden Germany 3: Mechanical effecs of chemical echings on monocrysalline silicon for phoovolaic use M. Sefancich, M. Buuri, D. Vincenzi, G. Marinelli, Solar Energy Maerials and Solar Cells 69, , (001) Togeher wih hese repored m, scaled 450 mm wafer hickness is shown in Figure mm Wafer Thickness (µm) Weibull modules m Scaling model Repored m Figure 8, Wafer hickness scaling based on Weibull size scaling. 11

10 450 mm scaled hickness scaers around 850 µm excep for he large m case. I is fairly smaller han he hickness of he handling wafer. This underesimaion will be caused by Edge conribuion o fracure srengh is no aken ino calculaion due o he lack of daa in he public domain. Calculaion assumes he same fracure probabiliy for 450 mm wafer and mm wafer. If 450 mm wafer needs smaller fracure probabiliy because of is high value, is fracure srengh used in he calculaion should be smaller. This will resul in increase of wafer hickness o compensae he decrease of fracure srengh. Daa shown in Table 3 come from wafers for solar cells. I may be differen from ha of mm wafers for LSI devices. These should be improved in he nex fracure scaling model. CONCLUSIONS We used sinered mechanical wafers o measure he graviaional sag and he wafer shape as a funcion of wafer hickness for various suppor sysems. Sinered mechanical wafers daa have been validaed wih single crysal measuremens. We found a good correlaion beween he simulaion and he real measuremen daa. The daa colleced show a very srong correlaion beween he wafer suppor design he graviaional sag and wafer shape and will be very useful for designing wafer carriers and handling equipmen. The daa colleced on he mechanical grade wafers will be used o fine-une he 450 mm process wafer hickness recommendaion for he SEMI 450 mm process wafers spec draf. Wafer hickness scaling based on he Weibull size scaling of he fracure srengh is a quaniaive physical model bu i ends o underesimae he 450 mm wafer hickness. Refining he model and use of realisic daa will improve he underesimaion once more pracical 450 mm wafer hickness will become available. REFERENCES 1. ITRS 007 ediion; hp// SEMI New Sandard: Specificaion for 450 mm Diameer Mechanical Handling Polished Wafers 3. M. Waanabe and S. Kramer, Inerface 15, No.4, p8, (006) 4. M. Waanabe, T. Fukuda, A. Ogura, K. Kirino and M. Kohno Discussion on Issues oward 450mm Wafer, ECS Transacions () (006) 5. J. Draina, D. Fandel, J. Ferrell, and S. Kramer, ECS Transacions (), 135 (006). 6. K. Takada, H. Yamagishi, and M. Imai, in Semiconducor Silicon/1998, H. R. Huff, H. Tsuya, and U. Gösele, Ediors. The Elecrochemical Sociey, PV 98-1, 376, (1998) 7. M. Goldsein, presened a 450 mm handling wafer ask force, SEMICON WEST 007 1

11 8. ASTM F (Reapproved 1996) Sandard Tes Mehod for Biaxial Flexure Srengh (Modulus of Rapure) of Ceramic Subsraes 9. M. S. Bawa, E. F. Pero and H. M. Grimes, Semiconducor Inernaional, p. 115, Nov R. Danzer, T. Lube, P. Supancic and R. Damani, Adv. Eng. Maerials 10, 75, (008) 11. C. A. Klein, J. Appl. Phys. 96, 317, (004) 1. Y. Jung, A. Pajares, R. Banerjee and B. R. Lawn, Aca Maerial. 5, 3459, (004) 13

P. Bruschi: Project guidelines PSM Project guidelines.

P. Bruschi: Project guidelines PSM Project guidelines. Projec guidelines. 1. Rules for he execuion of he projecs Projecs are opional. Their aim is o improve he sudens knowledge of he basic full-cusom design flow. The final score of he exam is no affeced by