Impact of probing procedure on flip chip reliability - Semantic Scholar
Microelectronics Reliability 43 (2003) 123–130 www.elsevier.com/locate/microrel
Impact of probing procedure on flip chip reliability Kuo-Ming Chen 1, Kuo-Ning Chiang
*
Department of Power Mechanical Engineering, National Tsing Hua University, 101, Sec. 2 Kung-Fu Rd., Hsinchu 300, Taiwan, ROC Received 15 January 2002; received in revised form 16 July 2002
Abstract Probe-after-bump is the primary probing procedure for flip chip technology, since it does not directly contact the bump pad, and involves a preferred under bump metallurgy (UBM) step coverage on the bump pads. However, the probe-after-bump procedure suffers from low throughputs and high cost. It also delays the yield feedback to the fab, and makes difficult clarification of the accountability of the low yield bumped wafer between the fab and the bumping house. The probe-before-bump procedure can solve these problems, but the probing tips may over-probe or penetrate the bump pads, leading to poor UBM step coverage, due to inadequate probing conditions or poor probing cards. This work examines the impact of probing procedure on flip chip reliability, using printing and electroplating bumpings on aluminum and copper pads. Bump height, bump shear strength, die shear force, UBM step coverage, and reliability testing are used to determine the influence of probing procedure on flip chip reliability. The experimental results reveal that bump quality and reliability test in the probe-before-bump procedure, under adequate probing conditions, differ slightly from the corresponding items in the probe-after-bump procedure. UBM gives superior step coverage of probe marks in both probe-before-bump and probe-after-bump procedures, implying that UBM achieves greater adhesion and barrier function between the solder bump and the bump pad. Both printing and electroplating bump processes slightly influence all evaluated items. The heights of probe marks on the copper pads are 40–60% lower than those on the aluminum pads, indicating that the copper pad enhances UBM step coverage. This finding reveals that adequate probing conditions of the probe-before-bump procedure are suited to sort flip chip wafers and do not significantly affect bump height, bump shear strength, die shear force, or flip chip reliability. Ó 2002 Elsevier Science Ltd. All rights reserved.
1. Introduction Probing attempts to detect the electrical continuity, leakage and gross function of a wafer. Probing must precede wire-bonding packaging. The primary type of probe card used in wire-bonding packaging is the epoxy probe card. Flip chip packaging has been broadly used in microelectronics, owing to its higher density, better electrical performance, easier thermal management, and smaller footprint than other types of packaging. In the
*
Corresponding author. Tel.: +886-3-5719034; fax: +886-35722840. E-mail addresses: [email protected] (K.-M. Chen), [email protected] (K.-N. Chiang). 1 United Microelectronics Corp., Taiwan.
1970s, IBM developed Cobra vertical probing technology to probe C4 (controlled collapse chip connection) bumps [1]. Probe-after-bump is the most widespread probing procedure for flip chip technology. This procedure does not involve direct contact with the bump pad, and under bump metallurgy (UBM) yields good step coverage on the bump pad. However, the probe-afterbump procedure suffers from low throughputs and high cost. It also delays the yield feedback to the fab [2], and the cause of the low yield of the bumped wafer between the fab and the bumping house is hard to clarify. The probe-before-bump procedure can solve these problems. However, the procedure may cause inferior UBM step coverage due to the heavy probe marks, and lower the bump quality and the reliability of the device. Fig. 1 illustrates poor UBM step coverage of a bump pad, due to a heavy probe mark. Voids appear between
0026-2714/02/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 0 2 6 - 2 7 1 4 ( 0 2 ) 0 0 2 6 7 - 6
124
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
those formed using cantilever probe technology. Such a probe mark supports preferred UBM step coverage. The pointed tip with a diameter of 15–30 lm directly contacts the pad during the probe-before-bump procedure. The flat tip with a diameter of 75–125 lm directly contacts the top of the bump. The contact force of the vertical probe card is from 0.0932 to 0.128 g/lm (2.33 to 3.2 g/mil). The planarity of the tips is less than 75 lm [8].
3. Experiment
Fig. 1. Poor UBM step coverage on probe mark.
UBM and the bump pad. Besides, the probing tips may penetrate the bump pad due to inappropriate probing conditions or a poor probe card. This work examines the effect of the probing procedure on bump height, bump shear strength, die shear force, UBM step coverage, and reliability, using electroplating and printing bump processes on aluminum and copper pads.
2. Probing technology Over ten manufacturing methods use bumping, of which evaporation, electroplating, printing and stud bumping are the most important [3]. The popularity of the electroplating and the printing bumping process has increased because evaporation bumping is more expensive and has a lower throughput than these other two approaches. This work employs eutectic solder with UBM of Al/Ni–V/Cu, and Ti/Ni–V/Cu deposited on aluminum and copper pads using printing bumping. Ti/ Cu/Ni is also deposited on both aluminum and copper pads using electroplating bumping. All UBM materials were deposited on wafers by sputtering, except for Ni in the electroplating bumping, which was deposited by electroplating. Cobra, epoxy (cantilever), and membrane are the main probe card technologies [4–7]. The epoxy probe card was developed first; and remains the most extensively used, applied especially to peripheral pads. The Cobra probe card is used in vertical probing technology, and is used in area array pads. IBM developed the Cobra probe card to probe C4 bumps in the 1970s. The main disadvantages of Cobra are its high cost, low bandwidth, and the very large force required to deflect the probe needles. The membrane probe card has also been available for many years, but it is not widely used in industry. The advantages of vertical probing technology are that it can be applied to area array pads such that the probe marks are smaller and shallower than
The semiconductor industry has long used aluminum alloys as interconnections and pad material. Recently, lower resistance capacitance delays, higher electromigration resistance and lower cost have become very important requirements in advanced semiconductor technology. Copper meets these requirements, and has been successfully used in interconnections in some advanced devices, by applying the Damascene process [9]. However, copper is easily oxidized and thoroughly removing the oxide is difficult. Therefore, the applications of copper as a bump pad or wire-bonding pad material are limited. This work employed electroplating and printing as methods to manufacture bumps on a test vehicle. The experiment ignores probe-after-bump for copper pads since the procedure does not affect the UBM step coverage. Table 1 specifies the test vehicle. Figs. 2 and 3 display the cross-sections of the solder bumping and flip chip packages. Both the probe-before-bump and the probe-afterbump procedures use the same Cobra vertical probe card but different probing tip sizes and shapes. The Advantest T5335P tester and the TSK UF200 prober are Table 1 Description of test vehicle Items Bumping Bump pitch (lm) Bump height (lm) Bump count Bump process Bump material UBM size (lm) UBM material Packaging Packaging type Die size (mm) Substrate material Packaging size (mm) Ball diameter (mm) Ball pitch (mm)
Data 200 100 48 Printing and electroplating Sn/Pb eutectic 100 Printing: Al/Ni–V/Cu, Ti/Ni–V/Cu; electroplating: Ti/Cu/Ni FC-CSP 5:4 4:6 Organic 10 8 mm 0.35 0.75
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
125
Fig. 5. Schematic diagram of die shear test.
Fig. 2. Cross-section of bump.
test is below 0.5 mm/s. Eq. (1) gives the minimum bump shear force. Fbf ¼ Fbs A
Fig. 3. Cross-section of flip chip package.
employed to probe the test vehicle at 85 °C with onetime touchdown and a 110 lm overdrive. The bump height variation must be within 15 lm to ensure a high flip chip assembly yields, and the bump coplanarity should be under 30 lm. The bump shear test aims to evaluate the UBM adhesion to the solder bump and the pad. The bump shear strength must exceed 3.1 mg/lm2 [10], and that the fracture must appear in the solder bump. A robust UBM adhesive layer indicates that the solder bump is well able to withstand thermal fatigue stress. Fig. 4 depicts the bump shear test. The distance between blade and die, 25 lm, is approximately a quarter of the bump height. The speed of bump shear
Fig. 4. Schematic diagram of bump shear test.
ð1Þ
where Fbf is the bump shear force; Fbs denotes the bump shear strength, and A represents the UBM area and equals pð50 lmÞ2 for the test vehicle. Therefore, the minimum bump shear force can be calculated as 3:1 mg=lm2 pð50 lmÞ2 ¼ 24:35 g. The die shear test was performed on completion of die bonding and the reflowing process, but without encapsulating the underfill, to verify the bump quality and the die bonding process. The fracture should occur in the solder bumps. Eq. (2) gives the minimum die shear force. Fdf ¼ Fbf n
ð2Þ
where Fdf denotes the die shear force; Fbf is the bump shear force and equals 24.35 g for this test vehicle, and n represents the number of bumps and equals 48. Thus, the minimum die shear force is calculated as Fdf ¼ 24:35 g 48 ¼ 1170 g, implying that the minimum die shear force is 1170 or 24.35 g/bump. Fig. 5 depicts the die shear test. The distance between blade and substrate is 40–50 lm. The speed of the die shear test is 0.1 mm/s. In this work, six groups of specimens underwent the flip chip reliability test, following the JEDECs (Joint Electron Device Engineering Council) level three-moisture sensitivity (MSL 3). The reliability test includes temperature cycle test (TCT), thermal shock test (TST), temperature and humidity with bias (THB), highly accelerated stress test (HAST), and high temperature storage (HTS) test. Each test item is performed on 45 specimens. In the TCT, the dwell times at high and low temperatures were 10 min, while the ramp up and ramp down times between the high and low temperatures were 5 min: thus, each test cycle took 30 min. Table 2 summarizes the test items and conditions of the flip chip reliability test.
126
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
Table 2 Flip chip packaging reliability test items and condition
1
2
3
4
5
Test items and condition
Sample size
Readouts
Precondition level 3, followed by: TCT, condition B ()55 °C/þ125 °C) Precondition level 3, followed by: TST, condition B ()55 °C/þ125 °C) Precondition level 3, followed by: THB, 85 °C/ 85%RH, 1.1Vcc Precondition level 3, followed by: HAST, 130 °C/ 85%RH, 1.1Vcc HTS, 150 °C
45
500, 1000 cycles
45
200, 500 cycles
45
168, 500, 1000 h
45
100 h
45
500, 1000 h
Note: Precondition level 3: 30 °C/60%RH @ 192 h, Reflow 3X @ 220þ5/)0 °C.
4. Results and discussion 4.1. Probe marks Fig. 6(a) and (b) display the probe marks on aluminum and copper pads, following the probe-before-bump procedure, while Fig. 7 displays the probe mark on the top of the bump under the probe-after-bump procedure. Fig. 6(a) and (b) show that the height of the probe marks on the aluminum and cooper pads is 0.30–0.50 and 0.15– 0.25 lm, respectively. The probe marks on the copper pads are lower and shallower since copper has a higher modulus and hardness. The material with the higher modulus and hardness exhibits less plastic deformation, such that the height of the probe mark is lower. Normally, the shallower probe mark indicates greater UBM step coverage because the target material is more effectively deposited on the smooth pad during sputtering. Copper is the preferred pad material if the oxidized copper can be removed completely. Fig. 8(a)–(c) show UBM step coverage on higher and lower probe marks, and multiple touchdown on the bump pad, respectively, indicating that good UBM step coverage is inversely proportional to the height of the probe mark and the number of touchdowns. Fig. 9 shows the excellent UBM step coverage on the probe mark and pad, indicating that UBM achieves good adhesion and functions well as a barrier between the solder bump and pad.
Fig. 6. Probe mark on pad (probe before bump).
4.2. Bump height The bump height sample size is 35 for each group. Fig. 10 compares the bump heights of the six groups of specimens, for different bumping processes, pad materials, probing procedures; it shows standard deviations
Fig. 7. Probe mark on top of bump (probe after bump).
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
127
Fig. 9. UBM step coverage condition.
from 3.5 to 6.4 lm. The average bump height deviation of the six groups is below 2%. The probing procedure, bumping process, and pad material slightly influence the bump height. 4.3. Bump shear test
Fig. 8. UBM step coverage on different height of probe marks.
The bump shear test sample size is 25 for each group. Fig. 11 displays the fracture located on the solder bump after the bump shear test. Fig. 12 compares the bump shear strength across the specimens in the six groups, with standard deviations from 0.33 to 0.46 mg/lm2 . The minimum shear strength of the eutectic solder is 3.1 mg/ lm2 . The bump shear strength always exceeds 3.1 mg/ lm2 by 95–115%. All fractures appear on the solder bump, showing that UBM achieves superior adhesion between the pad metal and the solder bump. The bump shear test results yield the following information.
Fig. 10. Comparison of bump height.
128
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
material properties of the solder, and is unrelated to the pad materials.
Fig. 11. Fractured location of bump shear test.
The difference in bump shear strength between the probe-before-bump and the probe-after-bump procedures, performed on aluminum pads, is 2.9% by the electroplating process, and 0.2% by printing. The results indicate that UBM achieves good adhesion. The bump shear test relates only to the material properties of the solder, and is independent of the probing procedure. The bump shear strength difference between the aluminum and the copper pads is 3.7% by the electroplating process and 1.2% by printing when the probe-beforebump procedure is applied. This result implies that UBM gives excellent step coverage and adhesion to both aluminum and copper pads. Although the copper pads always have superior UBM step coverage due to light probe marks, the bump shear strength relates only to the
Fig. 13. Solder bump wetting status after die bonding process. Excellent solder bump wetting for prob before (a) and after (b) bump procedure.
Fig. 12. Comparison of bump shear strength.
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
The difference in bump shear strength between electroplating and printing is 9.3% on aluminum pads and 3.7% on copper pads when the probe-before-bump procedure is used. This large difference for aluminum pads follows from the heavy probe marks thereon. The bump shear strength of the electroplating bumping exceeds that of the printing bumping by 7–9%, because the electroplating bumping is more compact and has fewer voids than the printing bumping. 4.4. Die shear test Four groups of specimens were examined by the die shear test using different bumping and probing proce-
129
dures, since the copper pad negligibly affects the die shear force. The sample size for the die shear test is five. Fig. 13(a) and (b) reveal that the solder bump exhibits excellent wetting with UBM and the substrate pad. Fig. 14 compares the die shear forces across the four groupsÕ specimens, showing standard deviations from 3.7 to 5.4 g/bump. The minimum die shear force should exceed 1.17 kg or 24.35 g/bump. Fig. 14 shows that the die shear force always exceeds 24.35 g/bump by 45–59%, indicating that UBM has good step coverage and adhesion. The die shear test supports the following claims. The difference in the die shear force between the probe-before-bump and the probe-after-bump procedures is 3.0% with electroplating and 2.1% with printing.
Fig. 14. Comparison of die shear force.
Table 3 Flip chip packaging reliability test results before and after probing procudure Electroplating Before
1 2 3 4 5
Precondition level 3, followed by: TCT, condition B ()55 °C/þ125 °C) Precondition level 3, followed by: TST, condition B ()55 °C/þ125 °C) Precondition level 3, followed by: THB, 85 °C/ 85%RH, 1.1Vcc Precondition level 3, followed by: HAST, 130 °C/ 85%RH, 1.1Vcc HTS, 150 °C
Printing After
Before
After
Al
Cu
Al
Al
Cu
Al
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
Bumping process: electroplating, printing; pad material: Al, Cu.
130
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
Fig. 15. Acceptance of probe mark angle.
The difference is due to the variation in the solder bumping process. This result indicates that the UBM gives excellent step coverage, and that the die shear force relates to the material properties of the solder bump. The die shear force of the electroplating bumping exceeds that of the printing bumping by 3.6% with the probe-before-bump procedure, and by 4.4% with the probe-after-bump procedure, because the electroplating bumping is more compact and includes fewer voids than the printing bumping.
6.1%, respectively, indicating that the bump process negligibly influences bump quality and flip chip reliability. The probing procedure impacts neither the bump quality nor flip chip reliability. Furthermore, copper can be used as the bump pad material if its oxide can be thoroughly removed. The probe mark on the copper pad is 40–60% shallower than that on the aluminum pad, implying that the copper pad enhances UBM step coverage for the probe-before-bump procedure. This finding verifies that the adequate probing conditions for the probe-before-bump procedure described here, can be employed in flip chip wafer sorting.
Acknowledgements The authors would like to thank UST (Unitive Semiconductor Taiwan Corp.) and SPIL (Siliconware Precision Industrial Co., Ltd.) for their support in bump manufacturing and flip chip assembly of test vehicle.
4.5. Reliability test References Table 3 presents the flip chip reliability results for six groupÕs specimens with various bumping processes, pad materials, and probing procedures. All specimens passed the reliability test, implying that the probe-before-bump procedure with adequate probing conditions can be used in wafer sorting for flip chip devices. The evaluation indicates that copper can be directly employed as the bump pad material when the copper oxide is completely removed. Accordingly, the acceptable probe mark angle should exceed 90° to guarantee that UBM has preferred step coverage. Fig. 15 shows the acceptable probe mark angle.
5. Conclusion This work examines the probe-before-bump procedure applied to flip chip wafer sorting. The procedure does not influence bump height, bump shear strength, die shear force, or flip chip reliability. Experimental results demonstrate that the bump shear strength and the die shear force in the probe-before-bump and probeafter-bump procedures exceed the specifications by 95– 115% and 45–59%, respectively. The UBM yields excellent step coverage on probe marks in the probebefore-bump procedure, indicating that UBM achieves superior adhesion and barrier function. The difference in bump shear strengths and die shear forces between the electroplating and printing processes are below 9.3% and
[1] Genin D, Wurster M. Probing consideration in C-4 testing of IC wafers. Int J Microcirc Electron Pack 1992;12(4), ISSN 1063-1674. [2] Tan Q et al. Reliability evaluation of probe-before-bump technology. In: IEEE/CPMT International Electronics Manufacturing Technology Symposium, Austin, TX, 1999. p. 320–4. [3] Lau JH. Low cost flip chip technologies for DCA, WLCSP, and PBGA assemblies. McGraw-Hill; 2000. [4] Zhang Y et al. Thermally actuated microprobes for a new wafer probe card. IEEE J Microelectromech Syst 1999;8(1):43–9. [5] Zargari M et al. A BiCMOS active substrate probe-card technology for digital testing. IEEE J Solid-state Circ 1999;34(8):1118–34. [6] Enrique R et al. Wafer probe process verification tools. In: IEEE/SEMI Advanced Semiconductor Manufacturing Conference, Boston, MA, 2000. p. 207–12. [7] Akashi T et al. Eutectic solder bumped flip chip development. In: IEEE/CPMT International Electronics Manufacturing Technology Symposium, Austin, TX, 1997. p. 319–25. [8] Cobra Probe, Probe cards information, 2001. Available from <www.kns.com>. [9] Tyu C et al. Microstructure and reliability of copper interconnects. IEEE Trans Electron Dev 1999;46(6):1113– 20. [10] Elenius P. Choosing a flip chip bumping supplier. Adv Pack 1998;(March/April):70–3; In: IEEE/CPMT International Electronics Manufacturing Technology Symposium, Austin, TX, 1998. p. 185–8.
Impact of probing procedure on flip chip reliability Kuo-Ming Chen 1, Kuo-Ning Chiang
*
Department of Power Mechanical Engineering, National Tsing Hua University, 101, Sec. 2 Kung-Fu Rd., Hsinchu 300, Taiwan, ROC Received 15 January 2002; received in revised form 16 July 2002
Abstract Probe-after-bump is the primary probing procedure for flip chip technology, since it does not directly contact the bump pad, and involves a preferred under bump metallurgy (UBM) step coverage on the bump pads. However, the probe-after-bump procedure suffers from low throughputs and high cost. It also delays the yield feedback to the fab, and makes difficult clarification of the accountability of the low yield bumped wafer between the fab and the bumping house. The probe-before-bump procedure can solve these problems, but the probing tips may over-probe or penetrate the bump pads, leading to poor UBM step coverage, due to inadequate probing conditions or poor probing cards. This work examines the impact of probing procedure on flip chip reliability, using printing and electroplating bumpings on aluminum and copper pads. Bump height, bump shear strength, die shear force, UBM step coverage, and reliability testing are used to determine the influence of probing procedure on flip chip reliability. The experimental results reveal that bump quality and reliability test in the probe-before-bump procedure, under adequate probing conditions, differ slightly from the corresponding items in the probe-after-bump procedure. UBM gives superior step coverage of probe marks in both probe-before-bump and probe-after-bump procedures, implying that UBM achieves greater adhesion and barrier function between the solder bump and the bump pad. Both printing and electroplating bump processes slightly influence all evaluated items. The heights of probe marks on the copper pads are 40–60% lower than those on the aluminum pads, indicating that the copper pad enhances UBM step coverage. This finding reveals that adequate probing conditions of the probe-before-bump procedure are suited to sort flip chip wafers and do not significantly affect bump height, bump shear strength, die shear force, or flip chip reliability. Ó 2002 Elsevier Science Ltd. All rights reserved.
1. Introduction Probing attempts to detect the electrical continuity, leakage and gross function of a wafer. Probing must precede wire-bonding packaging. The primary type of probe card used in wire-bonding packaging is the epoxy probe card. Flip chip packaging has been broadly used in microelectronics, owing to its higher density, better electrical performance, easier thermal management, and smaller footprint than other types of packaging. In the
*
Corresponding author. Tel.: +886-3-5719034; fax: +886-35722840. E-mail addresses: [email protected] (K.-M. Chen), [email protected] (K.-N. Chiang). 1 United Microelectronics Corp., Taiwan.
1970s, IBM developed Cobra vertical probing technology to probe C4 (controlled collapse chip connection) bumps [1]. Probe-after-bump is the most widespread probing procedure for flip chip technology. This procedure does not involve direct contact with the bump pad, and under bump metallurgy (UBM) yields good step coverage on the bump pad. However, the probe-afterbump procedure suffers from low throughputs and high cost. It also delays the yield feedback to the fab [2], and the cause of the low yield of the bumped wafer between the fab and the bumping house is hard to clarify. The probe-before-bump procedure can solve these problems. However, the procedure may cause inferior UBM step coverage due to the heavy probe marks, and lower the bump quality and the reliability of the device. Fig. 1 illustrates poor UBM step coverage of a bump pad, due to a heavy probe mark. Voids appear between
0026-2714/02/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 0 2 6 - 2 7 1 4 ( 0 2 ) 0 0 2 6 7 - 6
124
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
those formed using cantilever probe technology. Such a probe mark supports preferred UBM step coverage. The pointed tip with a diameter of 15–30 lm directly contacts the pad during the probe-before-bump procedure. The flat tip with a diameter of 75–125 lm directly contacts the top of the bump. The contact force of the vertical probe card is from 0.0932 to 0.128 g/lm (2.33 to 3.2 g/mil). The planarity of the tips is less than 75 lm [8].
3. Experiment
Fig. 1. Poor UBM step coverage on probe mark.
UBM and the bump pad. Besides, the probing tips may penetrate the bump pad due to inappropriate probing conditions or a poor probe card. This work examines the effect of the probing procedure on bump height, bump shear strength, die shear force, UBM step coverage, and reliability, using electroplating and printing bump processes on aluminum and copper pads.
2. Probing technology Over ten manufacturing methods use bumping, of which evaporation, electroplating, printing and stud bumping are the most important [3]. The popularity of the electroplating and the printing bumping process has increased because evaporation bumping is more expensive and has a lower throughput than these other two approaches. This work employs eutectic solder with UBM of Al/Ni–V/Cu, and Ti/Ni–V/Cu deposited on aluminum and copper pads using printing bumping. Ti/ Cu/Ni is also deposited on both aluminum and copper pads using electroplating bumping. All UBM materials were deposited on wafers by sputtering, except for Ni in the electroplating bumping, which was deposited by electroplating. Cobra, epoxy (cantilever), and membrane are the main probe card technologies [4–7]. The epoxy probe card was developed first; and remains the most extensively used, applied especially to peripheral pads. The Cobra probe card is used in vertical probing technology, and is used in area array pads. IBM developed the Cobra probe card to probe C4 bumps in the 1970s. The main disadvantages of Cobra are its high cost, low bandwidth, and the very large force required to deflect the probe needles. The membrane probe card has also been available for many years, but it is not widely used in industry. The advantages of vertical probing technology are that it can be applied to area array pads such that the probe marks are smaller and shallower than
The semiconductor industry has long used aluminum alloys as interconnections and pad material. Recently, lower resistance capacitance delays, higher electromigration resistance and lower cost have become very important requirements in advanced semiconductor technology. Copper meets these requirements, and has been successfully used in interconnections in some advanced devices, by applying the Damascene process [9]. However, copper is easily oxidized and thoroughly removing the oxide is difficult. Therefore, the applications of copper as a bump pad or wire-bonding pad material are limited. This work employed electroplating and printing as methods to manufacture bumps on a test vehicle. The experiment ignores probe-after-bump for copper pads since the procedure does not affect the UBM step coverage. Table 1 specifies the test vehicle. Figs. 2 and 3 display the cross-sections of the solder bumping and flip chip packages. Both the probe-before-bump and the probe-afterbump procedures use the same Cobra vertical probe card but different probing tip sizes and shapes. The Advantest T5335P tester and the TSK UF200 prober are Table 1 Description of test vehicle Items Bumping Bump pitch (lm) Bump height (lm) Bump count Bump process Bump material UBM size (lm) UBM material Packaging Packaging type Die size (mm) Substrate material Packaging size (mm) Ball diameter (mm) Ball pitch (mm)
Data 200 100 48 Printing and electroplating Sn/Pb eutectic 100 Printing: Al/Ni–V/Cu, Ti/Ni–V/Cu; electroplating: Ti/Cu/Ni FC-CSP 5:4 4:6 Organic 10 8 mm 0.35 0.75
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
125
Fig. 5. Schematic diagram of die shear test.
Fig. 2. Cross-section of bump.
test is below 0.5 mm/s. Eq. (1) gives the minimum bump shear force. Fbf ¼ Fbs A
Fig. 3. Cross-section of flip chip package.
employed to probe the test vehicle at 85 °C with onetime touchdown and a 110 lm overdrive. The bump height variation must be within 15 lm to ensure a high flip chip assembly yields, and the bump coplanarity should be under 30 lm. The bump shear test aims to evaluate the UBM adhesion to the solder bump and the pad. The bump shear strength must exceed 3.1 mg/lm2 [10], and that the fracture must appear in the solder bump. A robust UBM adhesive layer indicates that the solder bump is well able to withstand thermal fatigue stress. Fig. 4 depicts the bump shear test. The distance between blade and die, 25 lm, is approximately a quarter of the bump height. The speed of bump shear
Fig. 4. Schematic diagram of bump shear test.
ð1Þ
where Fbf is the bump shear force; Fbs denotes the bump shear strength, and A represents the UBM area and equals pð50 lmÞ2 for the test vehicle. Therefore, the minimum bump shear force can be calculated as 3:1 mg=lm2 pð50 lmÞ2 ¼ 24:35 g. The die shear test was performed on completion of die bonding and the reflowing process, but without encapsulating the underfill, to verify the bump quality and the die bonding process. The fracture should occur in the solder bumps. Eq. (2) gives the minimum die shear force. Fdf ¼ Fbf n
ð2Þ
where Fdf denotes the die shear force; Fbf is the bump shear force and equals 24.35 g for this test vehicle, and n represents the number of bumps and equals 48. Thus, the minimum die shear force is calculated as Fdf ¼ 24:35 g 48 ¼ 1170 g, implying that the minimum die shear force is 1170 or 24.35 g/bump. Fig. 5 depicts the die shear test. The distance between blade and substrate is 40–50 lm. The speed of the die shear test is 0.1 mm/s. In this work, six groups of specimens underwent the flip chip reliability test, following the JEDECs (Joint Electron Device Engineering Council) level three-moisture sensitivity (MSL 3). The reliability test includes temperature cycle test (TCT), thermal shock test (TST), temperature and humidity with bias (THB), highly accelerated stress test (HAST), and high temperature storage (HTS) test. Each test item is performed on 45 specimens. In the TCT, the dwell times at high and low temperatures were 10 min, while the ramp up and ramp down times between the high and low temperatures were 5 min: thus, each test cycle took 30 min. Table 2 summarizes the test items and conditions of the flip chip reliability test.
126
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
Table 2 Flip chip packaging reliability test items and condition
1
2
3
4
5
Test items and condition
Sample size
Readouts
Precondition level 3, followed by: TCT, condition B ()55 °C/þ125 °C) Precondition level 3, followed by: TST, condition B ()55 °C/þ125 °C) Precondition level 3, followed by: THB, 85 °C/ 85%RH, 1.1Vcc Precondition level 3, followed by: HAST, 130 °C/ 85%RH, 1.1Vcc HTS, 150 °C
45
500, 1000 cycles
45
200, 500 cycles
45
168, 500, 1000 h
45
100 h
45
500, 1000 h
Note: Precondition level 3: 30 °C/60%RH @ 192 h, Reflow 3X @ 220þ5/)0 °C.
4. Results and discussion 4.1. Probe marks Fig. 6(a) and (b) display the probe marks on aluminum and copper pads, following the probe-before-bump procedure, while Fig. 7 displays the probe mark on the top of the bump under the probe-after-bump procedure. Fig. 6(a) and (b) show that the height of the probe marks on the aluminum and cooper pads is 0.30–0.50 and 0.15– 0.25 lm, respectively. The probe marks on the copper pads are lower and shallower since copper has a higher modulus and hardness. The material with the higher modulus and hardness exhibits less plastic deformation, such that the height of the probe mark is lower. Normally, the shallower probe mark indicates greater UBM step coverage because the target material is more effectively deposited on the smooth pad during sputtering. Copper is the preferred pad material if the oxidized copper can be removed completely. Fig. 8(a)–(c) show UBM step coverage on higher and lower probe marks, and multiple touchdown on the bump pad, respectively, indicating that good UBM step coverage is inversely proportional to the height of the probe mark and the number of touchdowns. Fig. 9 shows the excellent UBM step coverage on the probe mark and pad, indicating that UBM achieves good adhesion and functions well as a barrier between the solder bump and pad.
Fig. 6. Probe mark on pad (probe before bump).
4.2. Bump height The bump height sample size is 35 for each group. Fig. 10 compares the bump heights of the six groups of specimens, for different bumping processes, pad materials, probing procedures; it shows standard deviations
Fig. 7. Probe mark on top of bump (probe after bump).
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
127
Fig. 9. UBM step coverage condition.
from 3.5 to 6.4 lm. The average bump height deviation of the six groups is below 2%. The probing procedure, bumping process, and pad material slightly influence the bump height. 4.3. Bump shear test
Fig. 8. UBM step coverage on different height of probe marks.
The bump shear test sample size is 25 for each group. Fig. 11 displays the fracture located on the solder bump after the bump shear test. Fig. 12 compares the bump shear strength across the specimens in the six groups, with standard deviations from 0.33 to 0.46 mg/lm2 . The minimum shear strength of the eutectic solder is 3.1 mg/ lm2 . The bump shear strength always exceeds 3.1 mg/ lm2 by 95–115%. All fractures appear on the solder bump, showing that UBM achieves superior adhesion between the pad metal and the solder bump. The bump shear test results yield the following information.
Fig. 10. Comparison of bump height.
128
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
material properties of the solder, and is unrelated to the pad materials.
Fig. 11. Fractured location of bump shear test.
The difference in bump shear strength between the probe-before-bump and the probe-after-bump procedures, performed on aluminum pads, is 2.9% by the electroplating process, and 0.2% by printing. The results indicate that UBM achieves good adhesion. The bump shear test relates only to the material properties of the solder, and is independent of the probing procedure. The bump shear strength difference between the aluminum and the copper pads is 3.7% by the electroplating process and 1.2% by printing when the probe-beforebump procedure is applied. This result implies that UBM gives excellent step coverage and adhesion to both aluminum and copper pads. Although the copper pads always have superior UBM step coverage due to light probe marks, the bump shear strength relates only to the
Fig. 13. Solder bump wetting status after die bonding process. Excellent solder bump wetting for prob before (a) and after (b) bump procedure.
Fig. 12. Comparison of bump shear strength.
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
The difference in bump shear strength between electroplating and printing is 9.3% on aluminum pads and 3.7% on copper pads when the probe-before-bump procedure is used. This large difference for aluminum pads follows from the heavy probe marks thereon. The bump shear strength of the electroplating bumping exceeds that of the printing bumping by 7–9%, because the electroplating bumping is more compact and has fewer voids than the printing bumping. 4.4. Die shear test Four groups of specimens were examined by the die shear test using different bumping and probing proce-
129
dures, since the copper pad negligibly affects the die shear force. The sample size for the die shear test is five. Fig. 13(a) and (b) reveal that the solder bump exhibits excellent wetting with UBM and the substrate pad. Fig. 14 compares the die shear forces across the four groupsÕ specimens, showing standard deviations from 3.7 to 5.4 g/bump. The minimum die shear force should exceed 1.17 kg or 24.35 g/bump. Fig. 14 shows that the die shear force always exceeds 24.35 g/bump by 45–59%, indicating that UBM has good step coverage and adhesion. The die shear test supports the following claims. The difference in the die shear force between the probe-before-bump and the probe-after-bump procedures is 3.0% with electroplating and 2.1% with printing.
Fig. 14. Comparison of die shear force.
Table 3 Flip chip packaging reliability test results before and after probing procudure Electroplating Before
1 2 3 4 5
Precondition level 3, followed by: TCT, condition B ()55 °C/þ125 °C) Precondition level 3, followed by: TST, condition B ()55 °C/þ125 °C) Precondition level 3, followed by: THB, 85 °C/ 85%RH, 1.1Vcc Precondition level 3, followed by: HAST, 130 °C/ 85%RH, 1.1Vcc HTS, 150 °C
Printing After
Before
After
Al
Cu
Al
Al
Cu
Al
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
0/45
Bumping process: electroplating, printing; pad material: Al, Cu.
130
K.-M. Chen, K.-N. Chiang / Microelectronics Reliability 43 (2003) 123–130
Fig. 15. Acceptance of probe mark angle.
The difference is due to the variation in the solder bumping process. This result indicates that the UBM gives excellent step coverage, and that the die shear force relates to the material properties of the solder bump. The die shear force of the electroplating bumping exceeds that of the printing bumping by 3.6% with the probe-before-bump procedure, and by 4.4% with the probe-after-bump procedure, because the electroplating bumping is more compact and includes fewer voids than the printing bumping.
6.1%, respectively, indicating that the bump process negligibly influences bump quality and flip chip reliability. The probing procedure impacts neither the bump quality nor flip chip reliability. Furthermore, copper can be used as the bump pad material if its oxide can be thoroughly removed. The probe mark on the copper pad is 40–60% shallower than that on the aluminum pad, implying that the copper pad enhances UBM step coverage for the probe-before-bump procedure. This finding verifies that the adequate probing conditions for the probe-before-bump procedure described here, can be employed in flip chip wafer sorting.
Acknowledgements The authors would like to thank UST (Unitive Semiconductor Taiwan Corp.) and SPIL (Siliconware Precision Industrial Co., Ltd.) for their support in bump manufacturing and flip chip assembly of test vehicle.
4.5. Reliability test References Table 3 presents the flip chip reliability results for six groupÕs specimens with various bumping processes, pad materials, and probing procedures. All specimens passed the reliability test, implying that the probe-before-bump procedure with adequate probing conditions can be used in wafer sorting for flip chip devices. The evaluation indicates that copper can be directly employed as the bump pad material when the copper oxide is completely removed. Accordingly, the acceptable probe mark angle should exceed 90° to guarantee that UBM has preferred step coverage. Fig. 15 shows the acceptable probe mark angle.
5. Conclusion This work examines the probe-before-bump procedure applied to flip chip wafer sorting. The procedure does not influence bump height, bump shear strength, die shear force, or flip chip reliability. Experimental results demonstrate that the bump shear strength and the die shear force in the probe-before-bump and probeafter-bump procedures exceed the specifications by 95– 115% and 45–59%, respectively. The UBM yields excellent step coverage on probe marks in the probebefore-bump procedure, indicating that UBM achieves superior adhesion and barrier function. The difference in bump shear strengths and die shear forces between the electroplating and printing processes are below 9.3% and
[1] Genin D, Wurster M. Probing consideration in C-4 testing of IC wafers. Int J Microcirc Electron Pack 1992;12(4), ISSN 1063-1674. [2] Tan Q et al. Reliability evaluation of probe-before-bump technology. In: IEEE/CPMT International Electronics Manufacturing Technology Symposium, Austin, TX, 1999. p. 320–4. [3] Lau JH. Low cost flip chip technologies for DCA, WLCSP, and PBGA assemblies. McGraw-Hill; 2000. [4] Zhang Y et al. Thermally actuated microprobes for a new wafer probe card. IEEE J Microelectromech Syst 1999;8(1):43–9. [5] Zargari M et al. A BiCMOS active substrate probe-card technology for digital testing. IEEE J Solid-state Circ 1999;34(8):1118–34. [6] Enrique R et al. Wafer probe process verification tools. In: IEEE/SEMI Advanced Semiconductor Manufacturing Conference, Boston, MA, 2000. p. 207–12. [7] Akashi T et al. Eutectic solder bumped flip chip development. In: IEEE/CPMT International Electronics Manufacturing Technology Symposium, Austin, TX, 1997. p. 319–25. [8] Cobra Probe, Probe cards information, 2001. Available from <www.kns.com>. [9] Tyu C et al. Microstructure and reliability of copper interconnects. IEEE Trans Electron Dev 1999;46(6):1113– 20. [10] Elenius P. Choosing a flip chip bumping supplier. Adv Pack 1998;(March/April):70–3; In: IEEE/CPMT International Electronics Manufacturing Technology Symposium, Austin, TX, 1998. p. 185–8.
