Technical challenges of stencil printing technology for ultra fine pitch flip chip bumping

Transcription

1 Technical challenges of stencil printing technology for ultra fine pitch flip chip bumping Dionysios Manessis a,*, Rainer Patzelt a, Andreas Ostmann b, Rolf Aschenbrenner b, Herbert Reichl b a Technical University of Berlin, Microperipheric Research Center, TIB 4/2-1, Gustav-Meyer-Allee 25, D Berlin b Fraunhofer Institute for Reliability and Microintegration (IZM), Gustav-Meyer-Allee 25, D Berlin Abstract Stencil printing remains the technology route of choice for flip chip bumping because of its economical advantages over traditionally costly evaporation and electroplating processes. This paper provides the first research results on stencil printing of 80 µm and 60 µm pitch peripheral array configurations with Type 7 Sn63/Pb37 solder paste. In specific, the paste particle size ranges from 2 µm to 11µm with an average particle size of 6.5 µm taken into account for aperture packing considerations. Furthermore, the present study unveils the determining role of stencil design and paste characteristics on the final bumping results. The limitations of stencil design are discussed and guidelines for printing improvement are given. Printing of Type 7 solder paste has yielded promising results. Solder bump deposits of 25 µm and 42 µm have been demonstrated on 80 µm pitch rectangular and round pads, respectively. Stencil printing challenges at 60 µm pitch peripheral arrays are also discussed. Keywords: Flip chip, Flip chip bumping, wafer bumping, ultra fine pitch printing, Type 7 solder paste. * Corresponding author. Tel.: ; Fax.: address: [email protected]

2 1. Introduction The continuous drive in electronics industry for smaller, lighter, faster and cheaper products has led to increased utilisation of advanced IC packages such as BGAs and CSPs and accelerated adoption of Wafer-Level CSP and flip chip technologies. In particular, the attractiveness of Flip chip technologies lays on the superior electrical performance, higher thermal conductivity, smaller size and higher I/O counts which are mandatory requirements for advanced semiconductor applications [1-3]. However, a substantive shift towards flip chip interconnection technologies will be witnessed only with accomplishment of cost reduction, reliability improvement and cost-efficient high density substrate technologies [3]. Low-cost flip chip bumping technology has become a reality with implementation of electroless nickel plating process for under bump metallization (UBM) in conjunction with stencil printing of solder pastes for the formation of solder bumps [4-9]. This technological approach offers significant economical advantages over the conventional C4 process which uses expensive equipment for sputtering and evaporation of high lead solder material and over established electroplating technology routes. Stencil printing of solder paste for flip chip wafer bumping has been long proved to be the cost-saving alternative for high-volume production and plausibly has been the focus of many investigations for its continuous improvement. Among its merits, it can be mentioned its compatibility with pre-existing printing equipment in a surface mount assembly line as well as its capability of producing solder bumps from a wide compositional spectrum including leadfree solders which have been the hot issue of strict environmental legislative rules. Reported studies indicate that the flip chip wafer bumping capability by solder paste printing for high volume production is in the range of 150µm and 225µm pitch for the respective peripheral and

3 area array configurations [10,11] while significant bumping achievements towards smaller pitches are continuously realised. Plausible questions arise on the limits of pitch reduction that can be reached by today s stencil printing capability especially in view of Moore s law prediction that the bits/chip grow by a factor of 4x every three years. It is a general consensus that a definite answer in this puzzling question will not be given soon as long as continuous progress is achieved on areas of vital importance for ultra fine pitch (UFP) printing such as stencil manufacturing capability, new solder paste developments, stencil printing equipment, and cost-efficient high density substrate technologies. Another economical alternative to UFP stencil printing for Flip chip bumping can be the Immersion Soldering technique which has been developed at Technical University of Berlin (TUB) in cooperation with Fraunhofer Institute (IZM), Berlin. Successful flip chip bumping of wafers for very small pitches down to 40µm has been achieved by immersion soldering [12,13]. This paper presents the results of our recent studies on stencil printing of type 7 solder paste for bumping of 4 wafers with peripheral pad configurations at 80 and 60 µm pitch. Comparisons of bumping results by stencil printing, immersion soldering and electroplating at 80 µm pitch will be derived. In addition, the present work intends to further promote the understanding of UFP printing for flip chip bumping applications. 2. Experimental procedure The main objectives of the present work were to examine the feasibility of stencil printing for flip chip bumping of wafers with ultra fine pitch peripheral pads which have been originally bumped by immersion soldering and electroplating techniques in the framework of an

4 on-going European project. Experimental details on test wafers, stencil design and solder paste materials used in this study are cited below. 2.1 Wafer design Two different wafer types A and B with a diameter of 4 were employed in this study. Both types had peripheral pad arrangements at 80, 60, and 40µm pitch with rectangular and circular UBM pads for type A and type B, respectively. The electroless Ni/Au plating technology of TUB was used to deposit 5 µm Ni /50 nm flash (Au) under bump metallization layers (UBM) (over the chip passivation layer) on the (Al) metallization pads. The lateral overlappment of the Ni/Au UBM pads on the chip passivation layer was also 5 µm. Detailed description of TUB s Electroless technology approach can be found in literature [4,5]. In addition, a well-established 6 IZM 21 test wafer with peripheral 80 µm square Ni/Au pads at 200 and 300 µm pitch was used to test the printability and reflow characteristics of the new solder pastes developed for the present study and to provide useful feedback to solder paste vendors for tailoring the paste properties. Table 1 summarizes the main features of the wafer types A and B. 2.2 Stencil design The stencil design in this study targeted primarily on exploring the challenges of stencil printing the part of the wafer with 80µm pitch and in second priority the part with pads at 60µm pitch. The part of the wafer with 40µm pitch was not accommodated in the stencil design because its proper design would be in conflict with the design at 80µm and 60µm pitch. A laser-cut stencil was used in order to take advantage of its inherent aperture wall tapering which

5 admittedly promotes paste release [6,7,9]. The variation in aperture shape and size intends more to uncover the challenges and limits of solder paste printing with respect to issues such as paste filling characteristics, paste release, minimal separation distance without bridging rather than deposition of maximum solder paste volume. Basic guidelines found in literature for stencil design regarding aspect ratio and area ratio have been taken into consideration [6-9]. On the other hand, stencil manufacturing limitations and minimum separation distances between apertures for stencil integrity greatly affect the stencil design strategy. The stencil thickness was 30µm and the stencil design was rotated 45 o to the printing direction in order to avoid notuniform aperture fillings due to orientation phenomena. The dimensions of apertures asmanufactured are cited in Table 2. It is worth mentioning that the present study explored the possibility of using an electroformed stencil of 30µm thickness but current manufacturing constraints for aperture separation distances larger or equal to 50 µm were not compatible with the proposed design of Table Solder paste materials and process sequence In this study, Type 7 water-soluble Sn63/Pb37 solder pastes from two vendors - A and B- were used. The metal content in both pastes was 90%. The solder powder size in Type 7 was in the range of 2-11µm in diameter (min 90% of powder). An average particle size of 6.5 µm was used for aperture filling considerations. The use of Type 7 solder paste (2-11µm) is deemed necessary so that an aperture width/average particle size ratio for the smallest aperture size is maintained at least (5 to 6) for good aperture filling, good paste release and longer stencil life. Commercially available printing equipment was employed. Stencil printing was conducted using vacuum handling of the wafers and manual cleaning of the apertures at the end of each

6 print in the case of wafers types A and B. Metal squeegees at 60 o angle were used whereas the printing pressure and speed were selected in order to ensure clear stencil print wipe and aperture filling, respectively. Reflow of wafers was done in a forced convection oven with nitrogen reflow atmosphere. The oxygen level was maintained less than 20 ppm. Cleaning of flux residues was performed in VIGON solution at 65 o C for 1 hour. The solder bump heights were measured with a Cyberscan Vantage 3D System using a DRS-300 sensor with a vertical resolution of µm. 3. Results and discussion 3.1 General characteristics of type 7 paste One SEM perspective of the Type 7 solder powder is given in Fig. 1. The paste was printed using the stencil described in Table 2 on a monitor wafer. Subsequently, it was heated at 100 o C for 0.5 hr to allow evaporation of the flux vehicle and to be suitable for SEM examination. The powder particle size range is in accordance with the classification provided by the vendor (min 90%, 2 to 11µm). The printing behavior of Type 7 solder paste was initially tested on a 6 test wafer IZM 21 with peripheral pad arrangements at 200 and 300µm pitch in order to gain valuable experience and examine the paste s overall performance. It should be noted that the IZM 21 wafer is originally printed with Type 6 paste (5-15µm). An 80 µm thick laser-cut stencil was used with oblong apertures for better paste filling characteristics. Both pastes A and B have exhibited good rolling behavior, filling efficiency, good printing results, and stencil life. Fig. 2 shows the 200µm pitch part of the wafer printed with Type 7. As expected, the Type 7 paste has shown for the specific application worse aperture release behavior than Type 6 pastes since more paste

7 sticks to the aperture walls due to its higher surface energy. The extent of solder balling after reflow is not severe but definitely more profound than with Type 6 paste. Perspective of the wafer after reflow and cleaning is given in Fig. 3. All the above mentioned reasons can potentially account for the smaller bump height of 95µm obtained by Type 7 paste than the bump height of 105µm achieved by Type 6 paste using the same stencil thickness of 80 µm. 3.2 Printing on wafer type A Many prints have been carried out on monitor and the real wafer in order to find out the optimum printing parameters for elimination of flux bleeding and more importantly paste bridging. In this context, the minimum pressure which yields a clean swipe is used. A pressure of 3 Kgr was found adequate for satisfactory printing. A metal squeegee of 12 cm length at 60 o angle was used. Typical speeds for wafer bumping in the range of 7 to 25 mm/sec can be used as indicated also in the literature [6,9]. Printing with a snap-off is generally not advisable for this study since it can severely harm the very thin stencil due to repeatable bending. However, snap-off printing with relatively thicker stencils can alleviate flux bleeding under the stencil with a significant compromise in printing definition and coplanarity achievement. Schake has also verified similar effects of snap-off printing on wafer bumping [6]. Obviously, the alignment of the stencil apertures with the pad configuration plays a paramount role for the printing and reflow results. In this respect, printing equipment position accuracy and repeatability come into play, especially for UFP stencil printing. On the other hand, paste properties such as slumping behavior and paste release can affect greatly the printing and can be adjusted step-by-step by the vendor for process optimisation. Printing with apertures 1 and 2 yields insufficient deposits and poor uniformity. Paste release characteristics of these apertures

8 can not be considered satisfactory because the aspect and area ratios of the apertures are below the critical point of good paste performance. In addition, aperture 1 with a separation distance between deposits around 33µm has shown considerable bridging. Aperture 3 yields considerable better paste release compared to apertures 1 and 2 as well as print uniformity although its aspect ratio is smaller than aperture 1. This can be attributed to the relatively better laser cutting quality of rectangular apertures (aperture 3) than square apertures (aperture 1) and also in the larger area ratio of aperture 3 than aperture 1. The extent of bridging observed has been minimal. Indicative pictures of aperture 3 printing with paste A are shown in Fig. 4. It should be noted that the print sharpness and definition of the deposits in Fig. 4 have been greatly affected by some irregularities of the as-manufactured aperture walls. Laser-cutting of very small apertures should be extremely well tuned in order to produce apertures with good dimensional stability and smoothness. In this study, two stencils were manufactured until an acceptable quality could be achieved. In this context, it is believed that the superior aperture quality of electroformed stencils compared to laser stencils can help printing UFP designs provided that manufacturing limitations for aperture separation distances of 50µm can be lifted. Apertures 4 and 5 have yielded the best paste release, deposit definition and sufficiency. However, considerably more bridging has occurred especially on the left and right edges of the wafer. The results for apertures 4 and 5 indicate that an aperture separation distance in the range of 30 to 33µm may be the lower limit for printing acceptance at 80 µm pitch provided that printing equipment capabilities and paste performance have been optimised as much as possible. The above findings very vividly raise the dilemma for stencil design; on one hand the aspect and area ratios should be respectively greater than 1.5 and 0.55 for best printing results, as witnessed also in other studies [6,7,9], but on the other hand real estate constraints for not bridging along

9 with stencil manufacturing difficulties become the predominant factors for the final stencil design. As an evidence of this discrepancy, apertures 4 and 5 with ratios larger or the same as the printing guidelines account for some bridging whereas aperture 3 with ratios of 1.37 and 0.47 yields almost no bridging. This interesting finding may imply that type 7 paste can define new lower limits of aspect and area ratios for good printing performance, decreasing simultaneously the stencil manufacturing challenges. Obviously, more investigations are needed to support adequately the above scenario. As it was discussed above, aperture 3 has yielded the best results for printing at 80µm particularly with paste A. Optical and SEM micrographs of aperture 3 solder deposits after reflow and cleaning are shown in Figs. 5 and 6, respectively. Printing at 60µm pitch has been attempted with apertures 6 and 7. Both pastes have completely clogged in aperture 6. The aperture 7 has shown better paste release than aperture 6. These results indicate that for apertures with approximately the same aspect and area ratios, rectangular shape is more preferable than round with respect to paste release. Paste filling performance of the apertures is not a major consideration for printing with very thin stencils. Figure 7 indicatively shows solder deposits at 60µm after reflow and cleaning of paste s A flux residues. Bridging between deposits is also shown. This bridging in Fig. 7. is characterized as solder tunnelling and is believed to stem from the slight misregistration of the solder deposits with respect to the pad position. The solder bump height (UBM pad height + solder cap height) at 80 µm pitch with rectangular aperture 3 (41µmx92µm) (Fig. 5) is measured to be 25±1.4 µm whereas the corresponding bump height at 60 µm pitch is 19 ± 1.6 µm. The UBM pad height accounted for

10 5µm. The printing results at 80µm pitch with aperture 3 look promising and with further improvement of aperture quality the current die yield of 80% can be significantly raised. In comparison, the solder bump height at 80µm pitch obtained by immersion soldering is 12 ± 1.5µm [12]. Stencil printing of solder paste may alternatively offer more flexibility for bump height adjustment than the immersion soldering technique especially for applications where good control of stand-off distance during assembly of the chips is very important. The solder cap height at 80 and 60µm pitch produced by stencil printing is in good agreement with the requirements for the assembly of wafer Type A flip chips on flex substrates using Thermode Bonding (~20µm, flat solder caps) [13]. Fig. 8. shows a cross sectional view of one bumped chip of wafer type A assembled on flex substrate. 3.3 Printing on Wafer Type B Stencil printing on wafer Type B with circular UBM pads has yielded the best results with aperture 3 at 80 µm pitch in agreement with findings for wafer Type A. Apertures 4 and 5 exhibit better paste release than aperture 3 but the extent of solder bridging and solder stealing is more profound. Cross sectional view of the solder bumps at 80µm pitch is shown in Fig. 9. Results indicate that a relative misalignment of the stencil apertures with respect to wafer pads is more critical for circular pads than for the rectangular pads of wafer A. Depending on the extent of the misalignment (deviation from symmetrical position over the pad) and the area of the UBM pad covered by the aperture, competitive phenomena develop between adjacent deposits resulting in solder stealing and solder tunnelling (Fig. 7.) and consequently remarkable variation in solder bump height. In this respect, the rectangular pads of type A wafer provide more allowance to possible misalignment contingency than the circular pads. This turns out to

11 be very essential, taken into account the difficulties encountered in wafer-stencil alignment for UFP printing. The average solder bump height for aperture 3 at 80 µm pitch is 42 ±5 µm included the UBM pad thickness of 5µm. The respective solder bump heights produced by electroplating are 23±3.2µm, 13.5±1.8µm, and 8.5±1.85 µm at 80, 60, and 40µm pitch. The above results indicate the capabilities of cost-efficient stencil printing at 80µm pitch to produce almost double bump height than electroplating techniques with approximately the same standard deviation. 4. Conclusions The printing behavior of Type 7 Sn63/Pb37 solder paste (2-11µm) for UFP printing has been revealed. Paste exhibits good rolling and filling characteristics as well as stencil life. Solder balling becomes a more important issue than with Type 6 solder paste, necessitating tighter reflow oxygen levels and environmental stability. Future studies will explore Type 7 lead-free paste compositions due to their increased environmental significance. Printing at 80 µm pitch is feasible with very careful stencil design. Many challenging issues arise such as maintenance of the right aspect and area ratios for good paste release and print definition and on the other hand minimum separation distance for avoidance of bridging and stencil integrity. Stencil aperture dimensional stability and quality play a very significant role in UFP printing. An aperture separation distance of 40µm and rectangular shape of apertures have yielded the best results for UFP printing at 80µm. Rectangular UBM pads have shown better results than circular pads with respect to solder stealing between adjacent deposits. Solder caps of 25µm at 80 µm pitch have been produced on rectangular pads and the flip chips are assembled on flex substrates. Furthermore, solder bumps of 42 µm have been produced on 80µm pitch round

12 pads. Printing at 60µm pitch proves to be even more challenging, increasing the possibilities of deposit bridging due to pad density constraints. However, chips with bumps of 20µm have been made and are also assembled on flex. Electroformed stencils may arise as a good alternative option to laser-cut stencils with respect to manufacturing cost and aperture quality especially for UFP wafer designs. The interesting results of the present study spur further work in the area of UFP flip chip bumping through continuous collaboration with solder paste vendors and stencil manufacturers. Acknowledgements The authors would like to thank the European Commission for financial support of this work in the framework of CIRRUS Project (IST ). The provision of the wafers by the Centre Nacional de Microelectrònica (CNM), Spain is greatly appreciated. References [1] E. Vardaman. Growing demand for Flip Chip. Advancing Microelectronics. January/February 2003, pp [2] P. Elenius. Wafer-level packaging gains momentum. Solid State Technology. April 1999, pp [3] G. Meyer-Berg. KGD Roadmapping. Good-Die & Europractice-HDP newsletter. 2001;10:4-7. [4] A. Ostmann et al. Electroless Metal Deposition for Back-End Wafer Processes. Advancing Microelectronics. May/June 1999, pp [5] A. Ostmann et al. Development of an Electroless Redistribution Process. In: Proc. IMAPS

13 Europe Conference, Harrogate, June [6] J. Schake. Stencil Printing for Wafer Bumping. Semiconductor International. October 2000, pp [7] P. Coskina et al. Wafer Bumping for wafer-level CSP s and flip chips using Stencil printing technology. In: Proc. IMAPS Europe Conference, Harrogate, June [8] J. Kloeser et al. A low cost bumping process for flip chip and CSP applications. In: Proc. IMAPS Europe Conference, Harrogate, June 1999, pp [9] B. Huang et al. Solder Bumping via Paste Reflow for Array packages. Journal of Surface Mount Technology. 2002;15(1):16-31 [10] P. Elenius et al. Recent Advances in Flip Chip Wafer Bumping using Solder paste Technology. In: Proc. 49th ECTC, San Diego, 1999, pp [11] S. Nangalia et al. Issues with Fine Pitch Bumping and Assembly. In: Proc. Intern. Symp. on Advanced packaging Materials, Georgia, March 2000, pp [12] S. Nieland et al. Immersion Soldering-a new Way for Ultra Fine Pitch Bumping. In: Proc. Electronics Goes Green 2000+, Berlin, September 2000, pp [13] B. Pahl et al. A thermode Bonding Process for Fine Pitch Flip Chip Applications down to 40µm. In: Proc. 3rd Intern. Symp. on Electronics Materials and packaging (EMAP), Jeju Island, Korea, September 2001, pp

14 FIGURE CAPTIONS Fig. 1. SEM picture of Type 7 Sn63/Pb37 powder. Fig. 2. Typical print of type 7 paste at 200 µm pitch using an 80µm thick laser-cut stencil. Fig. 3. SEM perspective of bumped IZM 21 wafer at 200µm pitch using type 7 solder paste. Bump height is 95 µm. Fig. 4. Print of type 7 paste at 80 µm pitch with rectangular aperture 3 (41µmx92µm). (wafer type A -rectangular pads).

15 Fig. 5. Solder bump deposits at 80 µm pitch (wafer type A -rectangular pads). The bump height achieved was 25±1.4µm with rectangular aperture 3 (41µmx92µm). Fig. 6. SEM micrograph of solder bump deposits at 80µm pitch using rectangular aperture 3 (41µmx92µm). (wafer type A -rectangular pads). Fig. 7. Optical micrograph of solder bump deposits and bridges at 60µm pitch (wafer type A -rectangular pads). The bump height achieved was 19±1.6µm with rectangular aperture 3 (41µmx92µm). Fig. 8. Assembly of chip (wafer type A -rectangular pads) with UBM pad size of 50µmx90µm and bump height of 25µm on flex substrate.

16 Fig. 9. Cross section view of solder bumps at 80µm pitch (wafer type B -circular pads). The bump height achieved was 42±5µm on 30µm pads with rectangular aperture 3 (41µmx92µm).

17 TABLE CAPTIONS Table 1 Wafer characteristic features Table 2 Stencil design considerations

18 Fig. 1. FIGURES

19 Fig. 2.

20 Fig. 3.

21 Fig. 4.

22 Fig. 5.

23 Fig. 6.

24 Fig. 7.

25 Chip Flex Fig. 8.

26 Fig. 9.

27 TABLES Table 1 Wafer Size 4 Wafer Type A B Pitch (µm) Pad Dimensions- 50x90 35x90 20x90 Type A (µm) Pad diameter Type B (µm) Number of pads per wafer 87040

28 Table 2 Aperture Stencil Aperture Aspect Ratio Area Ratio Number/Shape Thickness Dimensions (Aperture width /t ) (Open area/aperture (t) (µm) (µm) wall area) Pitch: 80 µm 1 /Square /Rectangular 33x /Rectangular 30 41x /Rectangular 51x /Rectangular 51x Pitch: 60 µm 6/Round /Rectangular 30 28x