Chemistry

Physical & Theoretical Chemistry fields

Okayama University Year 1996

Shape evolution of electrodeposited

copper bumps with high peclet numbers

Kazuo Kondo Keisuke FukuiOkayama University Himeji Institute of Technology

Mitsunori Yokoyama Kunio ShinoharaUniversity of Hokkaido University of Hokkaido

This paper is posted at eScholarship@OUDIR : Okayama University Digital InformationRepository.

http://escholarship.lib.okayama-u.ac.jp/physical and theoretical chemistry/25

466 J Electrochem. Soc., Vol. 144, No. 2, February 1997 The Electrochemical Society, Inc.

2. J. H. Nordlien, S. Ono, N. Masuko, and K. Nisancioglu,This Journal, 142, 3320 (1995).

3. J. H. Nordlien, K. Nisancioglu, S. Ono, and N. Masuko,Corros. Sci., In press.

4. J. H. Nordlien, K. Nisancioglu, S. Ono, and N. Masuko,This Journal, 143, 2564 (1996).

5. 0. Lunder, J. E. Lein, P Kr. Aune, and K. Nisancioglu,Corrosion, 45, 9, 741 (1989).

6. T. Beijoudi, C. Fiaud, and L. Robbiola, ibid., 9, 738(1993).

7. 0. Lunder K. Nisancioglu, and R. S. Hansen, SAETechnical Paper No. 930755, SAE International,Warrendale, PA (1993).

8. 0. Lunder and K. Nisancioglu, in Progress in theUnderstanding and Prevention of Corrosion, Vol. 2,J. M. Costa and A. D. Mercer, Editors, p. 1249, TheInstitute of Materials, London (1993).

9. Standard Practice for Codification of Certain Nonfer-rous Metals and Alloys, Cast and Wrought, B275,American Society for Testing of Materials, Philadel-phia, PA (1980).

10. J. H. Nordlien, 0. Lunder H. Leth-Olsen, and K.

Nisancioglu, Extended Abstracts of the 47th AnnualMeeting of the Scandinavian Society of ElectronMicroscopy, P Beisvâg, P-H. Iversen, and J. K.Solberg, Editors, p. 111, SCANDEM 95, Trondheim,Norway (1995).

11. P F Malis and D. Steele, in Ultrainicrotomy for MaterialsScience in Workshop on Specimen Preparation forTEM Materials II, R. Vol. 199, R. Andersen, Edito; Ma-terials Research Symposium Proceedings, MaterialsResearch Society, Pittsburgh, PA (1990).

12. R. Furneaux, G. E. Thompson, and G. C. Wood. Corros.Sci., 18, 853 (1978).

13. 0. Lunder, T. Kr. Aune, and K. Nisancioglu, Corrosion,43, 291 (1987).

14. F. Mansfeld, S. Lin, S. Kim, and H. Shih, ibid., 45, 615(1989).

15. B. R. W Hinton, D. R. Arnott, and N. E. Ryan, Mater.Forum, 9, 162 (1986).

16. N. A. Braaten, J. K. Grepstad, and S. Raaen, Surf. Sn.,222, 499 (1989); 5. Raaen, N. A. Braaten, and J. K.Grepstad, Phys. Scr., 41, 1001 (1990).

17. N. A. Braaten and S. Raaen, ibid., 43, 430 (1991).

Shape Evolution of Electrodeposited Copper Bumps with HighPeclet Numbers

Kazuo Kondo,* Keisuke Fukui,b Mitsunori Yokoyama; and Kunio Shinohara°

"Graduate School of Engineering, University of Hokkaido, Nishi-8, Kita-13, Sapporo 060, JapanbFaculty of Engineering, Himei Institute of Technology, Shosha, 2167, Himeji 671-22, Japan

ABSTRACT

We report the shape evolution of initial copper bumps at Peclet numbers higher than a hundred. The role of vorticesand of penetration flow within the cavity was discussed with numerical fluid dynamics computation to obtain a bumpwith a single hump at the center. The current distributions, or flux profiles, were calculated at the diffusion controlledoverpotential and were compared with the electrodeposited bump shapes. For the 100 jim cavity width, the vorticesincrease at the upstream corners with Peclet numbers 1410 and 7311. The vortices are the local resistance of mass trans-fer to the cathode. These vortices cause the hollows in flux profiles at the upstream corner with these Peclet numbers. Thepenetration flow collides with the photoresist sidewall and the vortices decrease at downstream corners.These decreasedvortices cause the increase in flux profile at downstream corners. For a 30 jim cavity width a single large vortexformsfor the higher Peclet number 44,500 and a single hump in flux is achieved.

Introduction

Electrodeposited bumps are the indispensable micro-connectors for high density interconnection in the latestmicroelectronics applications. For example, interconnec-tion between microprocessors after 32 bits and randomaccess memory (RAM) demand bumps, since the number ofpads are several hundreds and the number is increasingfor every new model. The bumps may be used not only asa substitute for wire bonding {tape automated bonding(TAB) and flip chip]"2 but also as a new testing method forbare chips with high pin counts.3 Another application isthe interconnection between liquid crystal display (LCD)and driver chips, which are mostly interconnected withTAB. The bumps act as microconnectors of TAB technolo-gy and conduct digital signals from the chip to the LCD

pixels.4The bumps are electrodeposited onto dot-shaped cath—

odes several 10 to 200 jim in diameter. The cathode, orphotolithographic patterns, are patterned by photomaskand photoresist. The control of electrodeposited bumpshapes as well as the uniformity in their heights areimportant for exhibiting proper interconnection reliabili-ty.1 The bump shape with hump at the center, mushroomshape, is required for high density interconnection.

* Electrochemical Society Active MembeiiC Present address: Department of Applied Chemistry, Okayama

University, 3-1-1 Tshima-Naka, Okayama 700, Japan.

The role of mass transport in etching rectangular cavi-ties formed by photoresist was discussed extensively bythe following authors. Alkire and Deligianni developed atwo-dimensional numerical fluid dynamics computationto study the electrolytic etching of copper.5'6 The flow pat-terns were classified into two modes for the rectangularcavity of aspect ratios of 1.0 and 0.25, and their etch ratedistributions were compared. Georgiadou and Alkire re-ported the experimental results on chemical etching ofcopper.78 They also developed numerical fluid dynamicscomputation to study the presence of sparingly solublesurface film and compared the average etch rate with ex-periment for cavity aspect ratios of 1.0 and 0.20. Shin andEconomou studied electrolytic etching by numerical fluiddynamics of moving boundary for forced and natural con-vection.'0 They found that natural convection is effectivefor rinsing the dissolution products out of deep cavities.Jaw et at. studied jet electrochemical micromachining ofthroughholes. They found that changing nozzle diameter,cell voltage, nozzle-to-sample distance, and electrolyteflow is effective in shape control of throughholes."12

Bumping is electrodeposition onto dots and rectangularcavities formed by photoresist. Dukovic developed a two-dimensional numerical computation of tertiary current dis-tribution. The moving boundary problem was discussed onshape evolution of copper electrodeposit and the effect ofresist wall angle and leveling agent was reported.'3 Simonand Reichel developed a high speed gold bumping processand reported their bump shapes with a fountain-type plat-

466 J. Electrochem. Soc., Vol. 144, No. 2, February 1997 The Electrochemical Society, Inc.

2. J. H. Nordlien, S. Ono, N. Masuko, and K. Nisancioglu,This Journal, 142, 3320 (1995).

3. J. H. Nordlien, K. Nisancioglu, S. Ono, and N. Masuko,Corros. Sci., In press.

4. J. H. Nordlien, K. Nisancioglu, S. Ono, and N. Masuko,This Journal, 143, 2564 (1996).

5. 0. Lunder, J. E. Lein, T. Kr. Aune, and K. Nisancioglu,Corrosion, 45, 9, 741 (1989).

6. T. Beijoudi, C. Fiaud, and L. Robbiola, ibid., 9, 738(1993).

7. 0. Lunder, K. Nisancioglu, and R. S. Hansen, SAETechnical Paper No. 930755, SAE International,Warrendale, PA (1993).

8. 0. Lunder and K. Nisancioglu, in Progress in theUnderstanding and Prevention of Corrosion, Vol. 2,J. M. Costa and A. D. Mercer, Editors, p. 1249, TheInstitute of Materials, London (1993).

9. Standard Practice for Codification of Certain Nonfer-rous Metals and Alloys, Cast and Wrought, B275,American Society for Testing of Materials, Philadel-phia, PA (1980).

10. J. H. Nordlien, 0. Lunder, H. Leth-Olsen, and K.

Nisancioglu, Extended Abstracts of the 47th AnnualMeeting of the Scandinavian Society of ElectronMicroscopy, T. Beisvàg, 11-H. Iversen, and J. K.Solberg, Editors, p. 111, SCANDEM 95, Trondheim,Norway (1995).

11. 11 F Malis and D. Steele, in Ultramicrotomy for MaterialsScience in Workshop on Specimen Preparation forTEM Materials II, R. Vol. 199, R. Andersen, Editor, Ma-terials Research Symposium Proceedings, MaterialsResearch Society, Pittsburgh, PA (1990).

12. R. Furneaux, G. E. Thompson, and G. C. Wood. Corros.Sci., 18, 853 (1978).

13. 0. Lunder, 11 Kr. Aune, and K. Nisancioglu, Corrosion,43, 291 (1987).

14. F Mansfeld, S. Lin, S. Kim, and H. Shih, ibid., 45, 615(1989).

15. B. R. W. Hinton, D. R. Arnott, and N. E. Ryan, Mater.Forum, 9, 162 (1986).

16. N. A. Braaten, J. K. Grepstad, and S. Raaen, Surf. Sci.,222, 499 (1989); S. Raaen, N. A. Braaten, and J. K.Grepstad, Phys. 5cr., 41, 1001 (1990).

17. N. A. Braaten and S. Raaen, ibid., 43, 430 (1991).

Shape Evolution of Electrodeposited Copper Bumps with HighPeclet Numbers

Kazuo Kondo,*0c Keisuke Fukui,b Mitsunori Yokoyama? and Kunio Shinohara°

"Graduate School of Engineering, University of Hokkaido, Nishi-8, Kita-13, Sapporo 060, Japan'Faculty of Engineering, Himeji Institute of Technology, Shosha, 2167, Himeji 671-22, Japan

ABSTRACT

We report the shape evolution of initial copper bumps at Peclet numbers higher than a hundred. The role of vorticesand of penetration flow within the cavity was discussed with numerical fluid dynamics computation to obtain a bumpwith a single hump at the center. The current distributions, or flux profiles, were calculated at the diffusion controlledoverpotential and were compared with the electrodeposited bump shapes. For the 100 p.m cavity width, the vorticesincrease at the upstream corners with Peclet numbers 1410 and 7311. The vortices are the local resistance of mass trans-fer to the cathode. These vortices cause the hollows in flux profiles at the upstream corner with these Peclet numbers.Thepenetration flow collides with the photoresist sidewall and the vortices decrease at downstream corners.These decreasedvortices cause the increase in flux profile at downstream corners. For a 30 p.m cavity width a single large vortexformsfor the higher Peclet number 44,500 and a single hump in flux is achieved.

Introduction

Electrodeposited bumps are the indispensable micro-connectors for high density interconnection in the latestmicroelectronics applications. For example, interconnec-tion between microprocessors after 32 bits and randomaccess memory (RAM) demand bumps, since the number ofpads are several hundreds and the number is increasingfor every new model. The bumps may be used not only asa substitute for wire bonding {tape automated bonding(TAB) and flip chip]"2 but also as a new testing method forbare chips with high pin counts.' Another application isthe interconnection between liquid crystal display (LCD)and driver chips, which are mostly interconnected withTAB. The bumps act as microconnectors of TAB technolo-gy and conduct digital signals from the chip to the LCDpixels.4

The bumps are electrodeposited onto dot-shaped cath-odes several 10 to 200 p.m in diameter. The cathode, orphotolithographic patterns, are patterned by photomaskand photoresist. The control of electrodeposited bumpshapes as well as the uniformity in their heights areimportant for exhibiting proper interconnection reliabili-ty.' The bump shape with hump at the center, mushroomshape, is required for high density interconnection.

* Electrochemical Society Active MemberPresent address: Department of Applied Chemistry, Okayama

University, 3-1-1 Tshima-Naka, Okayama 700, Japan.

The role of mass transport in etching rectangular cavi-ties formed by photoresist was discussed extensively bythe following authors. Alkire and Deligianni developed atwo-dimensional numerical fluid dynamics computationto study the electrolytic etching of copper." The flow pat-terns were classified into two modes for the rectangularcavity of aspect ratios of 1.0 and 0.25, and their etch ratedistributions were compared. Georgiadou and Alkire re-ported the experimental results on chemical etching ofcoppet7" They also developed numerical fluid dynamicscomputation to study the presence of sparingly solublesurface film and compared the average etch rate with ex-periment for cavity aspect ratios of 1.0 and 0.20.' Shin andEconomou studied electrolytic etching by numerical fluiddynamics of moving boundary for forced and natural con-vection." They found that natural convection is effectivefor rinsing the dissolution products out of deep cavities.Jaw et al. studied jet electrochemical micromachining ofthroughholes. They found that changing nozzle diameter,cell voltage, nozzle-to-sample distance, and electrolyteflow is effective in shape control of throughholes."

Bumping is electrodeposition onto dots and rectangularcavities formed by photoresist. Dukovic developed a two-dimensional numerical computation of tertiary current dis-tribution. The moving boundary problem was discussed onshape evolution of copper electrodeposit and the effect ofresist wall angle and leveling agent was reported." Simonand Reichel developed a high speed gold bumping processand reported their bump shapes with a fountain-type plat-

j. iecrrocnem. soc., voi. 144, NO. , t-eoruary 1 / g I lie Izlectrocflemical Society, Inc.

ing cell.'4 Detailed studies on shape evolution of goldbumps were discussed by Kondo et aL" We studied theeffects of dot diameter, electrolyte flow, and additive onbump shapes. The numerical fluid dynamics computationstudy on initial stage of gold and copper bumps at lowerPeclet number of less than a hundred were also reported.'6"7

The present investigation discusses the shape evolutionof copper bumps at Peclet numbers higher than a hundred.The role of penetration and vortex flows within the cavi-ties are discussed to obtain a bump with a single hump atthe center which is necessary for high density intercon-nection. Further, the initial bump shapes electrodepositedat diffusion controlled overpotentials are compared withthe current distribution calculated by the numerical fluiddynamics.

Numerical Analysis and ExperimentalTwo-dimensional cross sections of dot patterns were

analyzed numerically by solving the equation of continu-ity, Navier-Stokes equations, and mass-transfer equation.The computational area and the boundary conditions areillustrated in Fig. 1 and the details are given in Ref. 17.The numerical method proposed by Patankar" is adoptedfor discretization and pressure equation calculation. Thedimensionless Peclet number is defined as follows

Pe = hu62h/D

Electrodeposition and photolithography conditions arepresented in Ref. 17. The photolithographed substrateshave dot patterns of 100 p.m in diam and these patterns arelocated 8 mm from the center of the rotating disk elec-trode. The substrates were attached to the rotating diskelectrode and rotational speed varied from 200 to3000 rpm. The bumps were electrodeposited potentiostat-ically at the diffusion controlled overpotential. The bumpswere observed by scanning electron microscopy (SEM)with sample stage tilting of 1.22 rad.

Results and DiscussionStream functions and isoconcentration contours.—

Figure 2 shows the normalized stream functions for the100 p.m cavity width calculated by the numerical fluiddynamics. Cathodes are located at the cavity bottom andare illustrated with slant lines. Photoresists 10 p.m inheight are located at both sides of the cathodes. The elec-trolyte flows from left to right as indicated by arrows. ThePeclet numbers are 126, 1410, and 7311 for Fig. 2 a, b, andc, respectively.

The vortices occur at the corners of cathodes both at up-and downstream. The streamline 6 is the penetration flowwhich starts from the upstream side bulk solution to thecathode centers at cavity bottoms and to the downstreamside bulk solution. The vortices at upstream side cornersincrease and grow toward the downstream side with theincrease in Peclet numbers. The vortices at the down-

Fig. 1. Illustration of two-dimensional cross section of photoresistand cathode; x- and y-coordinates and the boundary conditionsare shown.

U / / / , / 17150! ff1/Ill /100x, jim

Fig. 2. Effect of Peclet numbers on stream lines. The cavity widthwas 100 p.m. (a) Pe = 126, (b) Pc = 1410, and (c) Pe = 7311.

stream side corners decrease with the increase in thePeclet numbers.

Figure 3 shows the isoconcentration contours for the100 p.m cavity width calculated by numerical fluiddynamics. The location of cathodes and photoresists arethe same as in Fig. 2. The copper ions are consumed oncathodes at the cavity bottoms and the concentrationboundary layer develops. The Peclet numbers are 126,1410, and 7311 for Fig. 3a, b, and c, respectively. The iso-

aI: 0.005: 0.2

10: 0.45015: 0.70020: 0.950

b

C..x,pm

Fig. 3. Effect of Peclet numbers on isoconcentration contours. Thecavity width was 100 p.m. (a) Pc = 126, (b) Pe = 1410, and (c)Pc = 7311.

aI :-2.O3XIG-4

- - — --__________________5:0,00 - ______________6:4.44Xl03 -

10: 0.111 _______

____ ________

b

15

Df/fIIf/J/fII!If/C

15

_____ _________ 10

CCbuIk

CC bUlkand u at Y5OLLm

hototX=L+5Oi mf f f I I 'I

Cathode

L= 100p.m

J. Electrochem. Soc., Vol. 144, No. 2, February 1997 The Electrochemical Society, Inc. 467

ing cell.'4 Detailed studies on shape evolution of goldbumps were discussed by Kondo et al.1' We studied theeffects of dot diameter, electrolyte flow, and additive onbump shapes. The numerical fluid dynamics computationstudy on initial stage of gold and copper bumps at lowerPeclet number of less than a hundred were also reported."7

The present investigation discusses the shape evolutionof copper bumps at Peclet numbers higher than a hundred.The role of penetration and vortex flows within the cavi-ties are discussed to obtain a bump with a single hump atthe center which is necessary for high density intercon-nection. Further, the initial bump shapes electrodepositedat diffusion controlled overpotentials are compared withthe current distribution calculated by the numerical fluiddynamics.

Numerical Analysis and ExperimentalTwo-dimensional cross sections of dot patterns were

analyzed numerically by solving the equation of continu-ity, Navier-Stokes equations, and mass-transfer equation.The computational area and the boundary conditions areillustrated in Fig. 1 and the details are given in Ref. 17.The numerical method proposed by Patankar19 is adoptedfor discretization and pressure equation calculation. Thedimensionless Peclet number is defined as follows

Fe =

Electrodeposition and photolithography conditions arepresented in Ref. 17. The photolithographed substrateshave dot patterns of 100 p.m in diam and these patterns arelocated 8 mm from the center of the rotating disk elec-trode. The substrates were attached to the rotating diskelectrode and rotational speed varied from 200 to3000 rpm. The bumps were electrodeposited potentiostat-ically at the diffusion controlled overpotential. The bumpswere observed by scanning electron microscopy (SEM)with sample stage tilting of 1.22 rad.

Results and DiscussionStream functions and isoconcent ration contours.—

Figure 2 shows the normalized stream functions for the100 p.m cavity width calculated by the numerical fluiddynamics. Cathodes are located at the cavity bottom andare illustrated with slant lines. Photoresists 10 p.m inheight are located at both sides of the cathodes. The elec-trolyte flows from left to right as indicated by arrows. ThePeclet numbers are 126, 1410, and 7311 for Fig. 2 a, b, andc, respectively.

The vortices occur at the corners of cathodes both at up-and downstream. The streamline 6 is the penetration flowwhich starts from the upstream side bulk solution to thecathode centers at cavity bottoms and to the downstreamside bulk solution. The vortices at upstream side cornersincrease and grow toward the downstream side with theincrease in Peclet numbers. The vortices at the down-

C"Cand u, stY'5O p.m

Fig. 1. Illustration of two-dimensional cross section of photores;stand cathode; x- and y-coordinates and the boundary conditionsare shown.

a1:-2.O3X10 ______________________________

10:0.111 ______15: 0.444

If/Ill/fl/f/fl / 1/I/IlC

U///ffff//501/Jff//f/100X,pm

Fig. 2. Effect of Peclet numbers on stream lines. The cavity widthwas 100 p.m. (a) Pe 126, (b) Pc = 1410, and (c) Pc 7311.

stream side corners decrease with the increase in thePeclet numbers.

Figure 3 shows the isoconcentration contours for the100 p.m cavity width calculated by numerical fluiddynamics. The location of cathodes and photoresists arethe same as in Fig. 2. The copper ions are consumed oncathodes at the cavity bottoms and the concentrationboundary layer develops. The Peclet numbers are 126,1410, and 7311 for Fig. 3a, b, and c, respectively. The iso-

Fig. 3. Effect of Pedet numbers on isoconcentration contours. Thecavity width was 100 p.m. (a) Pe 126, (b) Pe 1410, and (c)Pc = 7311.

____ — ii______ e

_________ _______________ ieHT'

Csthode

L*iIOtim

X'L+50p.m

Fig. 4. Effect of Peclet num-bers on flux (normalizedSherwood number) profiles. Thecavity width was 100 sun. (a) Pc= 126, (b) Pc = 1410, and (c)Pc = 7311.

X,.tm

concentration contours converge and come close to thecathodes with the increase in the Peclet numbers. The iso-concentration contours form distortions both at up- anddownstream corners for Fig. 3b and c because of the largeconvection within the vortices.

Flux profiles on cathodes—Figure 4 shows the currentdistributions, or flux profiles, for the 100 sm cavity widthcalculated from the isoconcentration contours. The flux isexpressed by the ratio of local and maximum dimension-less Sherwood numbers, Sh/Shmax.The Peclet numbers are126, 1410, and 7311 for Fig. 4a, b, and c, respectively. Forthe Peclet number of 126, Fig. 4a, the Sh/Shmax is constantat the upstream side of x = 0 to 10 sm. This ratio decreas-es toward the downstream and also becomes constant atthe downstream side of x = 90 to 100 sm. For Peclet num-ber 1410 (Fig. 4b), the ratio is low at x = 0 and increasesrapidly from x = 0 to 10 sm. The ratio decreases towardthe downstream side and slightly increases again at thedownstream side of x = 90 to 100 sm. For Peclet number7311 (Fig. 4c), the ratio is also' low at x = 0 and increasesrapidly from x = 0 to 10 m. The ratio decreases towardthe downstream side and increases at the downstream sideof x = 95 m.

At the upstream side corner, the vortices increase andgrow toward the downstream side with the increase inPeclet numbers (Fig. 2). The vortices are the resistance ofmass transfer to the cathode because the vortices recircu-late within themselves and are not able to capture the cop-per ion.16'17 These resistances of mass transfer flatten theflux profile at the upstream side corner for Peclet number126 (Fig. 4a). Further increase in the Peclet numbers to1410 and 7311 decrease the flux and cause hollows in theflux profiles at upstream side corners (Fig. 4b, c). At thecenter of the cavity, the flux decreases gradually from theup- to the downstream side because of the penetrationflows (Fig. 2). These penetration flows transport fresh bulkelectrolyte of high copper ion concentration from theupstream side to the bottom surface and to the down-stream side. As this bulk electrolyte moves from the up- tothe downstream side along the cathode surface, the copperions are consumed and the flux decreases. At the down-stream side corners, the vortices decrease with an increasein the Peclet numbers of 1410 and 7311 (Fig. 2b, c). Theseincrease the flux at the downstream side corners (Fig. 4b,c). These decreases in vortices are due to the collisionbetween the penetration flows and the photoresist side-walls at the downstream side corners.

Bump shape evolution at diffusion controlled overpoten-fiat—Figure 5 shows the cathodic polarization curves forPeclet numbers 126, 498, and 1410. The current densityincreases as the Peclet numbers increase. The potential of

—450 mV vs. 3.33 mol KC1-AgCl is diffusion control for allPeclet numbers and this potential is used to electrodepositthe bumps.

Figure 6 shows SEM micrographs of initial bump shapeselectrodeposited on the dot cathode 100 jsm in diameter atthe potential —450 my vs. 3.33 mol KC1-AgC1. Total num-ber of coulombs was 4.0 C/cm2. The larger arrow in Fig. 6ashows the direction of electrolyte flow. The bump shape isflat at the upstream side corner and this is indicated witha smaller arrow in Fig. 6a. The bump height decreasestoward the downstream and also becomes flat at thedownstream side corner. For Fig. 6b, the bump heightincreases rapidly at the upstream side corner and theheight decreases toward the downstream side and slightlyincreases again at the downstream side corner. For Fig. 6c,the bump height also increases rapidly at the upstreamside corner and the height decreases toward the down-stream side and increases again at the downstream sidecorner (indicated with an arrow). These observed initialbump shapes coincide well with the calculated flux profileshown in Fig. 4.

Shape evolution at higher Peclet numbers—The twovortices exist separately at the up- and downstream sidecorners for 100 sm cavity width (Fig. 2). These vortices are

0

EU

C.—-

.WW

00C

—2

Potential, mV vs. 3.33 moIJl KCI-AgCI

fig. 5. Cathodic polarization curves. (a) Pc 126, (1,1 Pc = 498,and jc) Pc = 1410.

468 J. Electrochem. Soc., Vol. 144, No.2, February 1997 The Electrochemical Society, Inc.

in _________ -S0.5

0

1.0

1.00.50

a..e. S • • •

—tilI,..SSSS. S • •

S • • •

S S • • • • •S••

• • • S • • •S. S

0

• S •

50

• S S S •

100

Fig. 4. Effect of Peclet num-bers on flux (normalizedSherwood number) profiles. TheccMty width was 100 run. (a) Pc= 126, (b) Re = 1410, and (c)Re = 7311.

concentration contours converge and come close to thecathodes with the increase in the Peclet numbers. The iso-concentration contours form distortions both at up- anddownstream corners for Fig. 3b and c because of the largeconvection within the vortices.

Flux profiles on cathodes.—Figure 4 shows the currentdistributions, or flux profiles, for the 100 p.m cavity widthcalculated from the isoconcentration contours. The flux isexpressed by the ratio of local and maximum dimension-less Sherwood numbers, Sh/Shmax. The Peclet numbers are126, 1410, and 7311 for Fig. 4a, b, and c, respectively. Forthe Peclet number of 126, Fig. 4a, the Sh/Shmax is constantat the upstream side of x = 0 to 10 p.m. This ratio decreas-es toward the downstream and also becomes constant atthe downstream side of x = 90 to 100 p.m. For Peclet num-ber 1410 (Fig. 4b), the ratio is low at x = 0 and increasesrapidly from x = 0 to 10 p.m. The ratio decreases towardthe downstream side and slightly increases again at thedownstream side of x = 90 to 100 p.m. For Peclet number7311 (Fig. 4c), the ratio is also. low at x = 0 and increasesrapidly from x = 0 to 10 p.m. The ratio decreases towardthe downstream side and increases at the downstream sideof x = 95 p.m.

At the upstream side corner, the vortices increase andgrow toward the downstream side with the increase inPeclet numbers (Fig. 2). The vortices are the resistance ofmass transfer to the cathode because the vortices recircu-late within themselves and are not able to capture the cop-per ion.16'17 These resistances of mass transfer flatten theflux profile at the upstream side corner for Peclet number126 (Fig. 4a). Further increase in the Peclet numbers to1410 and 7311 decrease the flux and cause hollows in theflux profiles at upstream side corners (Fig. 4b, c). At thecenter of the cavity, the flux decreases gradually from theup- to the downstream side because of the penetrationflows (Fig. 2). These penetration flows transport fresh bulkelectrolyte of high copper ion concentration from theupstream side to the bottom surface and to the down-stream side. As this bulk electrolyte moves from the up- tothe downstream side along the cathode surface, the copperions are consumed and the flux decreases. At the down-stream side corners, the vortices decrease with an increasein the Peclet numbers of 1410 and 7311 (Fig. 2b, c). Theseincrease the flux at the downstream side corners (Fig. 4b,c). These decreases in vortices are due to the collisionbetween the penetration flows and the photoresist side-walls at the downstream side corners.

Bump shape evolution at diffusion controlled overpoten-tial.—Figure 5 shows the cathodic polarization curves forPeclet numbers 126, 498, and 1410. The current densityincreases as the Peclet numbers increase. The potentialof

—450 mV vs. 3.33 mol KC1-AgCl is diffusion control for allPeclet numbers and this potential is used to electrodepositthe bumps.

Figure 6 shows SEM micrographs of initial bump shapeselectrodeposited on the dot cathode 100 p.m in diameter atthe potential —450 my vs. 3.33 mol KC1-AgC1. Total num-ber of coulombs was 4.0 C/cm2. The larger arrow in Fig. 6ashows the direction of electrolyte flow. The bump shape isflat at the upstream side corner and this is indicated witha smaller arrow in Fig. 6a. The bump height decreasestoward the downstream and also becomes flat at thedownstream side corner. For Fig. 6b, the bump heightincreases rapidly at the upstream side corner and theheight decreases toward the downstream side and slightlyincreases again at the downstream side corner. For Fig. 6c,the bump height also increases rapidly at the upstreamside corner and the height decreases toward the down-stream side and increases again at the downstream sidecorner (indicated with an arrow). These observed initialbump shapes coincide well with the calculated flux profileshown in Fig. 4.

Shape evolution at higher Peclet numbers—The twovortices exist separately at the up- and downstream sidecorners for 100 p.m cavity width (Fig. 2). These vortices are

U

C.—-.—

, —

00C

Potential, mY vs. 3.33 moIJl KC1-AgCI

Fig. 5. Cathodic polorizotion curves. (a) Pe = 126, (b) Pe = 498,

and jc) Pe = 1410.

468 J. Electrochem. Soc., Vol. 144, No.2, February 1997 The Electrochemical Society, Inc.

. S • •1.00.50

1.02 0.5CE

CE

1.00.50

S • • •

• S • • • . S

S 5 • • • •

• S • • • •— 55S5S5 • •p0 50

X,jtm100

Fig. 6. Effect of Peclet numbers on bump shapes for 100 m dotdiameter. The bumps were formed by potentiostatic electrolysis atthe diffusion controlled region of —450 mV vs. 3.33 mol KCI-AgCI.(a) Pe = 126, (b) Pe = 1410, and (c) Pe = 7311.

15

10

1/75

. .

Fig. 7. Effect of Peclet numbers on stream lines and isoconcenfration contours. The cavity width was 30 p.m. Stream lines for (a) Pe = 7311and (b) Pe = 44,500; isoconcenfration contours for (c) Pe = 7311 and (d) Pe = 44,500.

J. biectroctiem. SOC., Vol. 144, NO.2, 1-ebruary 1991 () The Electrochemical Society, Inc. 4b1

expected to merge into one for shorter cavity width athigher Peclet numbers. Numerical fluid dynamics resultsare shown below.

Figure 7 shows the stream functions and the isoconcen-tration contours. For the Peclet number of 7311, the twovortices grow toward the center and start to merge togeth-

C1:0.005: 0.2

10: 0.45015: 0.70020: 0.950—

0/ / 15/ / 30x, pm

LcI2 __/ / // 30x,pm

Fig. 6. Effect of Peclet numbers on bump shapes for 100 m dotdiameter. The bumps were formed by potentiostatic electrolysis atthe diffusion controlled region of —450 mV vs. 3.33 mol KCI-AgCI.(a) Pe = 126, (b) Pe = 1410, and (c) Pe = 7311.

15

10

1/75

. .

Fig. 7. Effect of Peclet numbers on stream lines and isoconcenfration contours. The cavity width was 30 p.m. Stream lines for (a) Pe = 7311and (b) Pe = 44,500; isoconcenfration contours for (c) Pe = 7311 and (d) Pe = 44,500.

J. biectroctiem. SOC., Vol. 144, NO.2, 1-ebruary 1991 () The Electrochemical Society, Inc. 4b1

expected to merge into one for shorter cavity width athigher Peclet numbers. Numerical fluid dynamics resultsare shown below.

Figure 7 shows the stream functions and the isoconcen-tration contours. For the Peclet number of 7311, the twovortices grow toward the center and start to merge togeth-

C1:0.005: 0.2

10: 0.45015: 0.70020: 0.950—

0/ / 15/ / 30x, pm

LcI2 __/ / // 30x,pm

4/U J. Electrochem. Soc., Vol. 144, No.2, February 1997 The Electrochemical Society, Inc.

1.0

0.5

'0

1.0

0.5

010 20 30

X,RmFig. 8. Effect of Peclet numbers on flux (normalized Sherwood

number) profiles. The cavity width was 30 jim. (a) Pc = 7311 and(bJ Pc = 44,500.

er at x = 18 jim. For Peclet number 44,500, two vorticesmerge into one and form a single large vortex. The isocon-centration contours form distortions because of the largeconvection within the vortices. The vortices circulateclockwise (Fig. 2, 7), Flow direction of these vortices alongthe cathode surface are from the down- to the upstreamside. Because of this flow direction, the copper ions areconsumed and the flux decreases toward the down- toupstream side.

Figure 8 shows the flux profiles on cathodes. Figure 8ahas two peaks in its profile which correspond to the exis-tence of two vortices. One peak is at x = 10 jim at theupstream side and the other peak is at x = 28 jim at thedownstream side. In Fig. 8b, single hump at x = 25 jimexists which corresponds to the formation of a single largevortex at the cavity center.

ConclusionThe current distributions, or flux profiles, were calcu-

lated by the numerical fluid dynamics and compared withthe initial bumps electrodeposited at Peclet numbers high-er than a hundred. The role of penetration and vortexflows within the cavities were discussed in order to obtainbumps with single humps at the centers.

1. The vortices form separately both at up- and down-stream side corners for 100 jim cavity width. These vor-tices, however, merge into one and form a single large vortexfor the 30 jim cavity width of high Peclet number 44,500.

2. For the 100 jim cavity width of Peclet number 126, thecurrent distribution, or the flux profile, is constant at theupstream side corner. The flux profile decreases towardthe downstream and also becomes constant at the down-stream side corners. For Peclet number 1410, the fluxdecreases and increases rapidly at the upstream side cor-ner and increases slightly at the downstream side cornetFor Peclet number 7311, the flux increases significantly atthe downstream side. These flux profiles coincide wellwith the initial bump shapes electrodeposited on the 100jim diameter patterns at the diffusion controlled overpo-tential.

For the 30 jim cavity width of the Peclet number of44,500, however, the flux profile has a single hump at thecentet

3. The vortices are the resistance of mass transfer to thecathode. The vortices increase with the increase of thePeclet numbers 1410 and 7311 at the upstream side cometThese vortices decrease in flux profile at the upstream sidecorners. The vortices decrease with the increase of the

Peclet numbers at the downstream side cornet Thisdecrease in vortices causes an increase in flux profiles.

For the 30 jim cavity width of Peclet number 7311, thetwo vortices at the up- and downstream side start to mergetogether. For Peclet number 44,500, the vortices merge intoone and form a single large vortex. The flux profile of thesingle hump at the center is achieved because of this sin-gle large vortex.

AcknowledgmentThe authors thank Professor R. C. Alkire and his

research group for instructive discussions. This work wassupported by TEPCO Research Foundation.

Manuscript submitted May 20, 1996; revised manuscriptreceived Oct. 10, 1996.

Hokkaido University assisted in meeting the publicationcosts of this article.

LIST OF SYMBOLSc concentration, mol m3D diffusion coefficient, m2 s'h resist height, mk [(D ac/ay150)/Ac] mass transfer coefficient, m sL cathode length, mP static pressure, PaPe Peclet number, Pe = huY2h/Dr radius of photomask pattern, mSh Sherwood number, Sh = kL/DShma* maximum value of the Sherwood numberu velocity in the x-direction, m sv velocity in y-direction, m sx coordinate in the streamwise direction, my coordinate normal to the cathode surface, mv kinematics viscositj, m2 sji viscosity, kg mt sp density, kg m3fi number of rotation, s

REFERENCES1. R. R. Tummala and E. J. Rymaszewski, Microelec-

tronics Packaging Handbook, p. 361, Van NostrandReinhold, New York (1989).

2. M. Sage and D. Gross, Hardware Developments inElectronic Systems, BPA (Technology & Manage-ment) Ltd., Surrey, England (1992).

3. Y. Yamamoto, M. Sugimoto, and K. Miyake, in Pro-ceedings of ISHM '93, p. 370, Dallas, TX (1993).

4. K. Hatada, Introduction to TAB Technology (inJapanese), Kougiyouchiyousakai, Tokyo (1989).

5. R. C. Alkire and H. Deligianni, This Journal, 135, 1093(1988),

6. R. C. Alkire, H. Deligianni, and J-B Ju, ibid., 137, 818(1990).

7. M. Georgiadou and R. C. Alkire, ibid., 140, 1340 (1993).8. M. Georgiadou and R. C. Alkire, ibid., 140, 1348 (1993).9. M. Georgiadou and R. C. Alkire, ibid., 141, 679 (1994).

10. C. B. Shin and D. J. Economou, ibid., 138, 527 (1991).11. S.-J. Jaw, J. M. Fenton, and M. Datta, in Electrochem-

ical Microfabrication II, M. Datta, K. Sheppard, andJ. 0. Dukovic, Editors, PV 94-32, p. 217, The Elec-trochemical Society Proceedings Series, Pennington,NJ (1994).

12. S.-J. Jaw, J. M. Fenton, and M. Datta, in High RateMetal Dissolution Process, M. Datta, B. R. Mac-Dougall, and J. M. Fenton, Editors, PV 95-19, p. 236,The Electrochemical Society Proceedings Series,Pennington, NJ (1995).

13. J. 0. Dukovic, IBM J. Res. Develop., 37, 125 (1993).14. J. Simon and H. Reichel, in Proceedings of the 9th

International Microelectronics Conference, p. 265,Ohmiya, Japan (1996).

15. K. Kondo, T. Miyazaki, and Y. Tamura, This Journal,141, 1644 (1994).

16. K. Kondo and K. Fukui, Kagaku-kougaku Ronbun-shu, 22 (1996).

17. K. Kondo, K. Fukui, K. lJno, and K. Shinohara, ThisJournal, 144, 1880 (1996).

18. H. Schlichting, Boundary Layer Theory, p. 93,McGraw-Hill, New York (1968).

19. 5. V. Patankar, Numerical Heat Transfer and FluidFlow, Hemisphere Publishing Corp., New York(1980).

• ..••l.. S S •

55

S • • •••••••.

ss55555

•

0

470 J. Electrochem. Soc., Vol. 144, No. 2, February 1997 The Electrochemical Society, Inc.

Fig. 8. Effect of Peclet numbers on flux (normalized Sherwoodnumber) profiles. The cavity width was 30 pm. (a) Pc = 7311 and(b) Pc = 44,500.

er at x = 18 p.m. For Peclet number 44,500, two vorticesmerge into one and form a single large vortex. The isocon-centration contours form distortions because of the largeconvection within the vortices. The vortices circulateclockwise (Fig. 2, 7). Flow direction of these vortices alongthe cathode surface are from the down- to the upstreamside. Because of this flow direction, the copper ions areconsumed and the flux decreases toward the down- toupstream side.

Figure 8 shows the flux profiles on cathodes. Figure 8ahas two peaks in its profile which correspond to the exis-tence of two vortices. One peak is at x = 10 p.m at theupstream side and the other peak is at x = 28 p.m at thedownstream side. In Fig. 8b, single hump at x = 25 p.mexists which corresponds to the formation of a single largevortex at the cavity center.

ConclusionThe current distributions, or flux profiles, were calcu-

lated by the numerical fluid dynamics and compared withthe initial bumps electrodeposited at Peclet numbers high-er than a hundred. The role of penetration and vortexflows within the cavities were discussed in order to obtainbumps with single humps at the centers.

1. The vortices form separately both at up- and down-stream side corners for 100 p.m cavity width. These vor-tices, however, merge into one and form a single large vortexfor the 30 p.m cavity width of high Peclet number 44,500.

2. For the 100 p.m cavity width of Peclet number 126, thecurrent distribution, or the flux profile, is constant at theupstream side corner. The flux profile decreases towardthe downstream and also becomes constant at the down-stream side corners. For Peclet number 1410, the fluxdecreases and increases rapidly at the upstream side cor-ner and increases slightly at the downstream side corner.For Peclet number 7311, the flux increases significantly atthe downstream side. These flux profiles coincide wellwith the initial bump shapes electrodeposited on the 100p.m diameter patterns at the diffusion controlled overpo-tential.

For the 30 p.m cavity width of the Peclet number of44,500, however, the flux profile has a single hump at thecenter.

3. The vortices are the resistance of mass transfer to thecathode. The vortices increase with the increase of thePeclet numbers 1410 and 7311 at the upstream side corner.These vortices decrease in flux profile at the upstream sidecorners. The vortices decrease with the increase of the

Peclet numbers at the downstream side corner. Thisdecrease in vortices causes an increase in flux profiles.

For the 30 p.m cavity width of Peclet number 7311, thetwo vortices at the up- and downstream side start to mergetogether. For Peclet number 44,500, the vortices merge intoone and form a single large vortex. The flux profile of thesingle hump at the center is achieved because of this sin-gle large vortex.

AcknowledgmentThe authors thank Professor R. C. Alkire and his

research group for instructive discussions. This work wassupported by TEPCO Research Foundation.

Manuscript submitted May 20, 1996; revised manuscriptreceived Oct. 10, 1996.

Hokkaido University assisted in meeting the publicationcosts of this article.

LIST OF SYMBOLSc concentration, mol m30 diffusion coefficient, m2h resist height, mk RD acThyI0)/Ac] mass transfer coefficient, mL cathode length, mP static pressure, PaPe Peclet number Pe =r radius of photomask pattern, mSh Sherwood number, Sh = kL/DShmax maximum value of the Sherwood numberu velocity in the x-direction, mv velocityin y-direction, m sx coordinate in the streamwise direction, my coordinate normal to the cathode surface, mv kinematics viscositj, m2 sp. viscosity, kg m1 sp density, kg m3fi number of rotation, s

REFERENCES1. R. R. Thmmala and E. J. Rymaszewski, Microelec-

tronics Packaging Handbook, p. 361, Van NostrandReinhold, New York (1989).

2. M. Sage and D. Gross, Hardware Developments inElectronic Systems, BPA (Technology & Manage-ment) Ltd., Surrey, England (1992).

3. V. Yamamoto, M. Sugimoto, and K. Miyake, in Pro-ceedings of ISHM '93, p. 370, Dallas, TX (1993).

4. K. Hatada, Introduction to TAB Technology (inJapanese), Kougiyouchiyousakai, Tokyo (1989).

5. B. C. Alkire and H. Deligianni, ThisJournal, 135, 1093

(1988).6. B. C. Alkire, H. Deligianni, and J-B Ju, ibid., 137, 818

(1990).7. M. Georgiadou and R. C. Alkire, ibid., 140, 1340 (1993).8. M. Georgiadou and B. C. Alkire, ibid., 140, 1348 (1993).9. M. Georgiadou and R. C. Alkire, ibid., 141, 679 (1994).

10. C. B. Shin and 0. J. Economou, ibid., 138, 527 (1991).11. S.-J. Jaw, J. M. Fenton, and M. Datta, in Elect rochem-

ical Microfabrication ff, M. Datta, K. Sheppard, andJ. 0. Dukovic, Editors, PV 94-32, p. 217, The Elec-trochemical Society Proceedings Series, Pennington,NJ (1994).

12. S.-J. Jaw, J. M. Fenton, and M. Datta, in High RateMetal Dissolution Process, M. Datta, B. B. Mac-Dougall, andJ. M. Fenton, Editors, PV 95-19, p. 236,The Electrochemical Society Proceedings Series,Pennington, NJ (1995).

13. J. 0. Dukovic, fBM J. Res. Develop., 37, 125 (1993).14. J. Simon and H. Reichel, in Proceedings of the 9th

International Microelectronics Conference, p. 265,Ohmiya, Japan (1996).

15. K. Kondo, T Miyazaki, and Y. Tamura, ThisJournal,141, 1644 (1994).

16. K. Kondo and K. Fukui, Kagaku-kougaku Ronbun-shu, 22 (1996).

17. K. Kondo, K. Fukui, K. TJno, and K. Shinohara, ThisJournal, 144, 1880 (1996).

18. H. Schlichting, Boundary Layer Theory, p. 93,McGraw-Hill, New York (1968).

19. S. V. Patankar, Numerical Heat Transfer and FluidFlow, Hemisphere Publishing Corp., New York(1980).

Lu

I.e •I.. .S0.5 a .•

.10

1.0 — •iSS•,5555 S••55

I I

0o 10 20 30