Development of gold to gold interconnection flip chip bonding for chip ...
Development of Gold to Gold Interconnection Flip Chip Bonding for Chip On Suspension Assemblies C.F. Luk, Y.C. Chan and K.C. Hung Department of Electronic Engineering, City University of Hong Kong Tat Chee Avenue, Kowloon, Hong Kong Tel. no.: (852) 2788 7130, Fax no.: (852) 2788 7579 Email address: [email protected] Abstract Gold to Gold Interconnection (GGI) flip chip bonding technology has been developed to bond the driver IC chip on the integrated circuit suspension used in hard disk drives (HDDs). GGI is a lead free process where the Au bumps and Au bond pads are joined together by heat and ultrasonic power under a pressure head. I. Introduction Data storage industry has witnessed an annual Compound Growth Rate (CGR) in area density of magnetic recording of 60 % since 1991 and 100 % since 1998. Central to this phenomenal growth has been the continual breakthroughs in magnetic recording technology and periphery to it is a whole range of supporting technologies, of which the High Density Interconnect (HDI) for integrating the magnetic recording (MR) head to the drive electronics is among one of the most important. In order for the magnetic recording heads to deliver the electrical performances as can be expected in state-of-art drive products. Gold to Gold Interconnection (GGI) flip chip bonding technology has been developed to bond the driver IC chip on the integrated circuit suspension. GGI is a lead free process where the Au bumps and Au bond pads are joined together by heat and ultrasonic power under a pressure head. The use of GGI flip chip assembly process will help to eliminate some of the equipment and processing steps of the traditional flip chip C4 process and accordingly shortens the overall cycle time. The subsequent benefit of this is a lower manufacturing cost by reduced capital investment in processing equipment and work-inprogress (WIP). Head Gimbal Assembly (HGA) is the core component of a MR head where read/write exchange process takes place. Chip On Suspension (COS) is only possible after a new HGA configuration with the integrated circuit suspension was developed [1]. This new HGA configuration emerged from the conventional HGA with twist pair wires (see Figure 1). With the circuit integrated suspension design, it becomes possible to assemble the driver IC chip close just next to the MR head slider on the suspension (see Figure 2). Because of the very short distance between the drive IC chip and the MR head slider, it is possible to decrease the parasitic resistance, inductance and capacitance. With the lowest parasitic impedance or in other words a clean data transmission line, we can achieve higher data transfer rates entering the 700 Mbits/sec to 1Gbits/sec regime. This paper will describe a flip chip bonding method for joining the driver IC chip on integrated circuit suspension
0-7803-7038-4/01/$10.00 (C)2001 IEEE
using GGI. Au bumps, which are applied mechanically on the wafer or on the chip using thermosonic ball bonder enable fine pitch bumping. The process works with available chips, having peripheral bond pads of pitch down to 80 µm. In the present work the total GGI process is evaluated. Besides the bumping, the flip chip assembly process covering the pick & place, Au to Au interconnection as well as the underfill process is investigated with special emphasis on process automation. Characterization of the critical GGI process parameters such as heating temperature, bonding pressure, ultrasonic power setting and duration time for optimal bonding condition is established. The reliability evaluation is concentrating on thermo-mechanical, robustness and functional performance of the final assembly. Performed tests are ball bond shear test, high temperature storage, and temperature cycling and humidity storage. X-ray inspection for Au bumps positional precision with respect to the Au bond pads. Underfill quality is revealed by SAM inspection for void, crack and de-lamination defects.
COS Configuration Circuitintegrated suspension M R slider
C onventionalH G A using the twistwires
D riverIC chip
C urrentH G A using the circuitintegrated suspension
C hip on suspension M R H ead
Figure 1.The H G A configuration trends
HGA (COS)
Figure 2. Head Gimbal Assy with Chip On Suspension
2001 Electronic Components and Technology Conference
Figure 3. US Bonding Procedure
II. Gold to Gold Interconnection (GGI) Process The principal features of ultrasonic and thermocompression bonding are married in thermosonic bonding in GGI technique. The steps in GGI process are shown in Figure 3. Step1: Bonding tool pick up the IC chip by vacuum force and align the gold bumps with the substrate bond pads. The work stage is heated at 150 degree C. Step 2: The bonding tool lowers down and press the IC chip against the bond pads of the substrate till there is no clearance between both the interfaces of IC chip/tool and IC chip/substrate bond pad. Step 3: Ultrasonic power causes gold stub bump deformed and touch to bonding pad tightly and pad metallurgy infiltrate each other. Step 4: Bonding complete and vacuum force release from the bonding tooling and the tool raise up. Bonding parameters for good GGI bonding, includes heating temperature, bonding pressure, ultrasonic power and time duration.
Au-plated surface finish is required for the substrate to have a good bonding with the gold stub bumped IC by Thermosonic bonding method. Integrated circuit Suspension Suspension used in this experiment is integrated circuit suspension. Bonding pads surface finish is Ni-Au plated. The material structure and thickness of different layers for the integrated suspension substrate is shown in Figure 6.
A B
Figure 4. IC Chip.
III. Components Integrated circuit Chip The IC chip used in COS evaluation is 803AC pre-amp. The specifications of dimensions is shown below: • Chip size: 1.4(L)x1.1(W)x0.125(H)mm • Bump material: Gold • Count of bumps: 8 bumps • Pad-to-Pad distance: A=0.19mm; B=0.26mm (Refer to Figure 4) • Bump size: D=0.105mm; H2=25.4µm; H1=40µm; (Refer to Figure 5)
0-7803-7038-4/01/$10.00 (C)2001 IEEE
H1 H2
D Figure 5. Au stud bump.
2001 Electronic Components and Technology Conference
Figure 6. Cross section of suspension pads. IV. Results & Discussion 1. Optimization of the thermosonic Au-Au bonding process Figure 7 shows the relationship between the shear strength of COS samples and the ultrasonic time of the thermosonic bonding. It is commonly known that thermosonic Au-Au bonding uses ultrasonic energy to soften the joint material to promote the solid diffusion between the Au bump and the Au pad [2]. Too short ultrasonic time (i.e. 0.1 s) will lead to not enough time for the softening of the joint material. 400 350
Shear strength (g)
300 250 200 150 100
Figure 9. Optical micrograph of cross-section of COS sample using 0.5 s of ultrasonic time. Figure 10 shows the plot of the shear strength of COS samples against the bonding pressure. In order to interpret the results in Figure 10, we also show the optical micrographs of the cross-section of COS samples in Figure 11. As the bonding pressure increases, the Au bump will be deformed more and more, and thus larger and larger effective contact area at the bonding interface between Au bump and Au pad will be obtained. This can also be seen from Table 1 which shows the standoff height of the COS samples. Increasing the bonding pressure leads to reduction of standoff height from 33 to 18 µm. However, too high bonding pressure may result in damage to the device as deformation by bonding tool may occur [2].
50 0
0.1
0.2
0.3
0.4
0.5
400
0.6
Time (s)
Die
300
Shear strength (g)
Figure 7. The plot of shear strength vs ultrasonic time.
350
250 200 150 100 50 0
Underfill
0.1
0.2
Au/Ni/Cu pad
Suspension
Figure 8. Optical micrograph of cross-section of COS sample using 0.1 s of ultrasonic time. Figure 8 shows the optical micrograph of the cross-section of the COS sample using 0.1 s of ultrasonic time. It can be seen that the Au bump is not deformed properly and only the central part of the Au bump is directly bonded to the Au pad of the suspension. Thus, This will reduce the effective bonding area of the Au bump to the Au pad and cause the low shear strength of the COS sample. If the ultrasonic time is enough, the Au bump can be deformed such that good Au-Au bonding will be achieved (see Figure 9). 0-7803-7038-4/01/$10.00 (C)2001 IEEE
0.3
0.4
0.5
0.6
Pressure (kg)
Au bump
Figure 10. The plot of shear strength vs bonding pressure. Table 1. Bonding parameters vs standoff height. Bonding parameters Ultrasonic time (s) 0.1 0.3 0.5 Bonding pressure (kg) 0.1 0.3 0.5 Ultrasonic power (mW) 10 18 26 Substrate temperature (°C) 120 160 200
Standoff height (µm) 26.43 ± 0.13 23.44 ± 0.29 23.28 ± 0.53 32.84 ± 1.09 22.75 ± 0.11 17.96 ± 0.43 30.53 ± 0.95 25.39 ± 0.17 23.25 ± 0.49 25.99 ± 1.89 22.10 ± 0.24 21.185 ± 0.9
2001 Electronic Components and Technology Conference
(a) 0.1 kg
movement may affect the efficient of the bonding at the interface between Au bump and Au pad. 400 350
Shear strength (g)
300 250 200 150 100 50 0
8
10
12
14
16
18
20
22
24
26
28
Power (mW)
(b) 0.3 kg
Figure 12. The plot of shear strength vs ultrasonic power.
Au pad
bonded area
bonded area
(c) 0.5 kg
bonded area
Figure 13. SEM picture of the sheared surface of the COS sample using ultrasonic power of 26 mW.
Figure 12 shows the relationship between the shear strength of COS samples and the ultrasonic power of the thermosonic bonding. When the ultrasonic power is low, the ultrasonic energy applied is not enough for softening and bonding of Au bump to Au pad. As the ultrasonic power increases, the bonding should become stronger and stronger. However, when the ultrasonic power is further increased to 26 mW, the shear strength is unexpectedly reduced. Actually, the larger the ultrasonic power, the larger the amplitude of the Horn tip vibration of the thermosonic bonder [3]. When the amplitude is large, the fresh bond formed in a cycle may be broken by the next cycle. As a result, this large amplitude
0-7803-7038-4/01/$10.00 (C)2001 IEEE
400 350 300
Shear strength (g)
Figure 11. Optical micrograph of cross-section of COS sample using different bonding pressure.
250 200 150 100 50 0
120
140
160
180
200
Temperature (°C)
Figure 14. The plot of shear strength vs substrate temperature.
2001 Electronic Components and Technology Conference
2. Reliability of the optimal setting of the thermosonic Au-Au bonding After the optimization of the bonding process, some of the standard reliability testing for the optimal setting have been performed. The reliability testing includes high temperature and high humidity storage (85°C and 85% RH) for 1000 hours, high temperature storage (125 °C) for 1000 hours, low temperature storage (- 40 °C) for 1000 hours, and thermal shock (-40 to 125 °C) for 1000 cycles. Figures 15 – 18 are the plot of measured current of the COS samples against the reliability testing time. All these figures show a good reliability of the GGI COS samples after those standard reliability testing.
Bias Current
Bias Current Change 40 35 30 25 20 15 10 5 0 0
24
100
250
500
750
1000
Hours
Figure 15. Bias current change under high temperature and high humidity storage (85°C and 85% RH).
Bias Current
Bias Current Change
Bias Current
Bias Current Change 40 35 30 25 20 15 10 5 0 0
24
100
250
500
750
1000
Hours
Figure 17. Bias current change under low temperature storage (- 40 °C). Bias current change 40
Bias current
Figure 13 is the SEM picture of the sheared surface of the COS sample using ultrasonic power of 26 mW, which shows the evidence of the above postulation. It can be seen that some positions are not bonded well as pointed by arrows. Therefore, too large ultrasonic amplitude will cause bad thermosonic bonding of the COS samples even the effective bonding area of Au bump to Au pad is large. Figure 14 shows the plot of the shear strength against the temperature provided to the substrate during thermosonic bonding. In fact, a higher substrate temperature leads to a better plastic deformation of the Au bump [4]. This makes it possible to get a better bonding at a constant bonding force since a larger effective bonding area of Au bump to Au pad is achieved. Regarding the detail evaluation of the above experiment, we can obtain a quite optimal setting of the thermosonic AuAu bonding of COS samples, which shows as follows: • Ultrasonic time = 0.2 s; • Bonding pressure = 0.45 kg; • Ultrasonic power = 20 mW; • Substrate temperature = 160 °C.
30 20 10 0 0
24
100
250 Hours
500
750
1000
Figure 18. Bias current change under high temperature storage (125 °C). V. Conclusion Chip on suspension design option will bring exciting new products to market and help keep the hard disk drive industry on its current projected growth rate well into the decade. Since GGI flip chip bonding for COS application is still relatively new and have yet to achieve volume use, work is still being done to establish and improve the limits of the technology with regard to reliability. Acknowledgments The authors would like to acknowledge the financial support provided by the Innovation and Technology Fund (Project no. 9440007) of the Innovation and Technology Commission. References 1. B. Mclnerney and M. Sheperek, “Chip-On-Suspension Electronics Close Preamp Head Gap”, Data storage, Vol. 7 No. 12, Page(s): 34 – 38, Dec 2000. 2. P.H. Lawyer, D. Choudhury; M.D. Wetzel; D.B. Rensch, Electronics Manufacturing Technology Symposium, Twenty-Third IEEE/CPMT, Page(s): 390 –393, 1998. 3. Sa-Yoon Kang; Teh-Hua Ju; Y.C. Lee, Electronic Components and Technology Conference, Proceedings, 43rd, Page(s): 877 –882, 1993. 4. S. Weiss; E. Zakel; H. Reichl, Electronic Components and Technology Conference, Proceedings, 44th, Page(s): 929 – 937, 1994.
40 35 30 25 20 15 10 5 0 0
24
100
250
500
750
1000
Hours
Figure 16. Bias current change under thermal shock condition (-40 to 125 °C).
0-7803-7038-4/01/$10.00 (C)2001 IEEE
2001 Electronic Components and Technology Conference
0-7803-7038-4/01/$10.00 (C)2001 IEEE
using GGI. Au bumps, which are applied mechanically on the wafer or on the chip using thermosonic ball bonder enable fine pitch bumping. The process works with available chips, having peripheral bond pads of pitch down to 80 µm. In the present work the total GGI process is evaluated. Besides the bumping, the flip chip assembly process covering the pick & place, Au to Au interconnection as well as the underfill process is investigated with special emphasis on process automation. Characterization of the critical GGI process parameters such as heating temperature, bonding pressure, ultrasonic power setting and duration time for optimal bonding condition is established. The reliability evaluation is concentrating on thermo-mechanical, robustness and functional performance of the final assembly. Performed tests are ball bond shear test, high temperature storage, and temperature cycling and humidity storage. X-ray inspection for Au bumps positional precision with respect to the Au bond pads. Underfill quality is revealed by SAM inspection for void, crack and de-lamination defects.
COS Configuration Circuitintegrated suspension M R slider
C onventionalH G A using the twistwires
D riverIC chip
C urrentH G A using the circuitintegrated suspension
C hip on suspension M R H ead
Figure 1.The H G A configuration trends
HGA (COS)
Figure 2. Head Gimbal Assy with Chip On Suspension
2001 Electronic Components and Technology Conference
Figure 3. US Bonding Procedure
II. Gold to Gold Interconnection (GGI) Process The principal features of ultrasonic and thermocompression bonding are married in thermosonic bonding in GGI technique. The steps in GGI process are shown in Figure 3. Step1: Bonding tool pick up the IC chip by vacuum force and align the gold bumps with the substrate bond pads. The work stage is heated at 150 degree C. Step 2: The bonding tool lowers down and press the IC chip against the bond pads of the substrate till there is no clearance between both the interfaces of IC chip/tool and IC chip/substrate bond pad. Step 3: Ultrasonic power causes gold stub bump deformed and touch to bonding pad tightly and pad metallurgy infiltrate each other. Step 4: Bonding complete and vacuum force release from the bonding tooling and the tool raise up. Bonding parameters for good GGI bonding, includes heating temperature, bonding pressure, ultrasonic power and time duration.
Au-plated surface finish is required for the substrate to have a good bonding with the gold stub bumped IC by Thermosonic bonding method. Integrated circuit Suspension Suspension used in this experiment is integrated circuit suspension. Bonding pads surface finish is Ni-Au plated. The material structure and thickness of different layers for the integrated suspension substrate is shown in Figure 6.
A B
Figure 4. IC Chip.
III. Components Integrated circuit Chip The IC chip used in COS evaluation is 803AC pre-amp. The specifications of dimensions is shown below: • Chip size: 1.4(L)x1.1(W)x0.125(H)mm • Bump material: Gold • Count of bumps: 8 bumps • Pad-to-Pad distance: A=0.19mm; B=0.26mm (Refer to Figure 4) • Bump size: D=0.105mm; H2=25.4µm; H1=40µm; (Refer to Figure 5)
0-7803-7038-4/01/$10.00 (C)2001 IEEE
H1 H2
D Figure 5. Au stud bump.
2001 Electronic Components and Technology Conference
Figure 6. Cross section of suspension pads. IV. Results & Discussion 1. Optimization of the thermosonic Au-Au bonding process Figure 7 shows the relationship between the shear strength of COS samples and the ultrasonic time of the thermosonic bonding. It is commonly known that thermosonic Au-Au bonding uses ultrasonic energy to soften the joint material to promote the solid diffusion between the Au bump and the Au pad [2]. Too short ultrasonic time (i.e. 0.1 s) will lead to not enough time for the softening of the joint material. 400 350
Shear strength (g)
300 250 200 150 100
Figure 9. Optical micrograph of cross-section of COS sample using 0.5 s of ultrasonic time. Figure 10 shows the plot of the shear strength of COS samples against the bonding pressure. In order to interpret the results in Figure 10, we also show the optical micrographs of the cross-section of COS samples in Figure 11. As the bonding pressure increases, the Au bump will be deformed more and more, and thus larger and larger effective contact area at the bonding interface between Au bump and Au pad will be obtained. This can also be seen from Table 1 which shows the standoff height of the COS samples. Increasing the bonding pressure leads to reduction of standoff height from 33 to 18 µm. However, too high bonding pressure may result in damage to the device as deformation by bonding tool may occur [2].
50 0
0.1
0.2
0.3
0.4
0.5
400
0.6
Time (s)
Die
300
Shear strength (g)
Figure 7. The plot of shear strength vs ultrasonic time.
350
250 200 150 100 50 0
Underfill
0.1
0.2
Au/Ni/Cu pad
Suspension
Figure 8. Optical micrograph of cross-section of COS sample using 0.1 s of ultrasonic time. Figure 8 shows the optical micrograph of the cross-section of the COS sample using 0.1 s of ultrasonic time. It can be seen that the Au bump is not deformed properly and only the central part of the Au bump is directly bonded to the Au pad of the suspension. Thus, This will reduce the effective bonding area of the Au bump to the Au pad and cause the low shear strength of the COS sample. If the ultrasonic time is enough, the Au bump can be deformed such that good Au-Au bonding will be achieved (see Figure 9). 0-7803-7038-4/01/$10.00 (C)2001 IEEE
0.3
0.4
0.5
0.6
Pressure (kg)
Au bump
Figure 10. The plot of shear strength vs bonding pressure. Table 1. Bonding parameters vs standoff height. Bonding parameters Ultrasonic time (s) 0.1 0.3 0.5 Bonding pressure (kg) 0.1 0.3 0.5 Ultrasonic power (mW) 10 18 26 Substrate temperature (°C) 120 160 200
Standoff height (µm) 26.43 ± 0.13 23.44 ± 0.29 23.28 ± 0.53 32.84 ± 1.09 22.75 ± 0.11 17.96 ± 0.43 30.53 ± 0.95 25.39 ± 0.17 23.25 ± 0.49 25.99 ± 1.89 22.10 ± 0.24 21.185 ± 0.9
2001 Electronic Components and Technology Conference
(a) 0.1 kg
movement may affect the efficient of the bonding at the interface between Au bump and Au pad. 400 350
Shear strength (g)
300 250 200 150 100 50 0
8
10
12
14
16
18
20
22
24
26
28
Power (mW)
(b) 0.3 kg
Figure 12. The plot of shear strength vs ultrasonic power.
Au pad
bonded area
bonded area
(c) 0.5 kg
bonded area
Figure 13. SEM picture of the sheared surface of the COS sample using ultrasonic power of 26 mW.
Figure 12 shows the relationship between the shear strength of COS samples and the ultrasonic power of the thermosonic bonding. When the ultrasonic power is low, the ultrasonic energy applied is not enough for softening and bonding of Au bump to Au pad. As the ultrasonic power increases, the bonding should become stronger and stronger. However, when the ultrasonic power is further increased to 26 mW, the shear strength is unexpectedly reduced. Actually, the larger the ultrasonic power, the larger the amplitude of the Horn tip vibration of the thermosonic bonder [3]. When the amplitude is large, the fresh bond formed in a cycle may be broken by the next cycle. As a result, this large amplitude
0-7803-7038-4/01/$10.00 (C)2001 IEEE
400 350 300
Shear strength (g)
Figure 11. Optical micrograph of cross-section of COS sample using different bonding pressure.
250 200 150 100 50 0
120
140
160
180
200
Temperature (°C)
Figure 14. The plot of shear strength vs substrate temperature.
2001 Electronic Components and Technology Conference
2. Reliability of the optimal setting of the thermosonic Au-Au bonding After the optimization of the bonding process, some of the standard reliability testing for the optimal setting have been performed. The reliability testing includes high temperature and high humidity storage (85°C and 85% RH) for 1000 hours, high temperature storage (125 °C) for 1000 hours, low temperature storage (- 40 °C) for 1000 hours, and thermal shock (-40 to 125 °C) for 1000 cycles. Figures 15 – 18 are the plot of measured current of the COS samples against the reliability testing time. All these figures show a good reliability of the GGI COS samples after those standard reliability testing.
Bias Current
Bias Current Change 40 35 30 25 20 15 10 5 0 0
24
100
250
500
750
1000
Hours
Figure 15. Bias current change under high temperature and high humidity storage (85°C and 85% RH).
Bias Current
Bias Current Change
Bias Current
Bias Current Change 40 35 30 25 20 15 10 5 0 0
24
100
250
500
750
1000
Hours
Figure 17. Bias current change under low temperature storage (- 40 °C). Bias current change 40
Bias current
Figure 13 is the SEM picture of the sheared surface of the COS sample using ultrasonic power of 26 mW, which shows the evidence of the above postulation. It can be seen that some positions are not bonded well as pointed by arrows. Therefore, too large ultrasonic amplitude will cause bad thermosonic bonding of the COS samples even the effective bonding area of Au bump to Au pad is large. Figure 14 shows the plot of the shear strength against the temperature provided to the substrate during thermosonic bonding. In fact, a higher substrate temperature leads to a better plastic deformation of the Au bump [4]. This makes it possible to get a better bonding at a constant bonding force since a larger effective bonding area of Au bump to Au pad is achieved. Regarding the detail evaluation of the above experiment, we can obtain a quite optimal setting of the thermosonic AuAu bonding of COS samples, which shows as follows: • Ultrasonic time = 0.2 s; • Bonding pressure = 0.45 kg; • Ultrasonic power = 20 mW; • Substrate temperature = 160 °C.
30 20 10 0 0
24
100
250 Hours
500
750
1000
Figure 18. Bias current change under high temperature storage (125 °C). V. Conclusion Chip on suspension design option will bring exciting new products to market and help keep the hard disk drive industry on its current projected growth rate well into the decade. Since GGI flip chip bonding for COS application is still relatively new and have yet to achieve volume use, work is still being done to establish and improve the limits of the technology with regard to reliability. Acknowledgments The authors would like to acknowledge the financial support provided by the Innovation and Technology Fund (Project no. 9440007) of the Innovation and Technology Commission. References 1. B. Mclnerney and M. Sheperek, “Chip-On-Suspension Electronics Close Preamp Head Gap”, Data storage, Vol. 7 No. 12, Page(s): 34 – 38, Dec 2000. 2. P.H. Lawyer, D. Choudhury; M.D. Wetzel; D.B. Rensch, Electronics Manufacturing Technology Symposium, Twenty-Third IEEE/CPMT, Page(s): 390 –393, 1998. 3. Sa-Yoon Kang; Teh-Hua Ju; Y.C. Lee, Electronic Components and Technology Conference, Proceedings, 43rd, Page(s): 877 –882, 1993. 4. S. Weiss; E. Zakel; H. Reichl, Electronic Components and Technology Conference, Proceedings, 44th, Page(s): 929 – 937, 1994.
40 35 30 25 20 15 10 5 0 0
24
100
250
500
750
1000
Hours
Figure 16. Bias current change under thermal shock condition (-40 to 125 °C).
0-7803-7038-4/01/$10.00 (C)2001 IEEE
2001 Electronic Components and Technology Conference
