Thermomigration and electromigration in Sn58Bi ball grid array solder joints

Similar documents
Electromigration failure mechanisms for SnAg3.5 solder bumps on Ti/Cr-Cu/Cu and Ni P /Au metallization pads

Dissolution of electroless Ni metallization by lead-free solder alloys

Interfacial reactions of BGA Sn 3.5%Ag 0.5%Cu and Sn 3.5%Ag solders during high-temperature aging with Ni/Au metallization

Bi Layer Formation at the Anode Interface in Cu/Sn 58Bi/Cu Solder Joints with High Current Density

Electromigration Behavior of through-si-via (TSV) Interconnect for 3-D Flip Chip Packaging

Lead-Free Solder Bump Technologies for Flip-Chip Packaging Applications

Interfacial Reactions between the Sn-9Zn Solder and Au/Ni/SUS304 Multi-layer Substrate

Microelectronic Engineering

Study of the Interface Microstructure of Sn-Ag-Cu Lead-Free Solders and the Effect of Solder Volume on Intermetallic Layer Formation.

Influence of Thermomigration on Lead-Free Solder Joint Mechanical Properties

IEEE TRANSACTIONS ON COMPONENTS, PACKAGING, AND MANUFACTURING TECHNOLOGY PART B, VOL. 20, NO. 1, FEBRUARY

Failure Modes of Flip Chip Solder Joints Under High Electric Current Density

Y.C. Chan *, D.Y. Luk

Interfacial Reactions between Ni-Zn Alloy Films and Lead-free Solders

Study of anisotropic conductive adhesive joint behavior under 3-point bending

Advanced Analytical Techniques for Semiconductor Assembly Materials and Processes. Jason Chou and Sze Pei Lim Indium Corporation

1 Thin-film applications to microelectronic technology

SCV Chapter, CPMT Society, IEEE September 14, Voids at Cu / Solder Interface and Their Effects on Solder Joint Reliability

Copyright 2009 Year IEEE. Reprinted from 2009 Electronic Components and Technology Conference. Such permission of the IEEE does not in any way imply

Fabrication of Smart Card using UV Curable Anisotropic Conductive Adhesive (ACA) Part I: Optimization of the curing conditions

Suppression of Cu 3 Sn and Kirkendall voids at Cu/Sn-3.5Ag solder joints by adding a small amount of Ge

EFFECT OF Ag COMPOSITION, DWELL TIME AND COOLING RATE ON THE RELIABILITY OF Sn-Ag-Cu SOLDER JOINTS. Mulugeta Abtew

The Effect of Cu and Ni on the Structure and Properties of the IMC Formed by the Reaction of Liquid Sn-Cu Based Solders with Cu Substrate

Effects of Current Stressing on Shear Properties of Sn-3.8Ag-0.7Cu Solder Joints

WF6317. A superactive low-volatile/high heat-resistant water-soluble flux for ball soldering

Correlations between IMC thickness and three factors in Sn-3Ag-0.5Cu alloy system

IMPACT OF MICROVIA-IN-PAD DESIGN ON VOID FORMATION

Endoscopic Inspection of Solder Joint Integrity in Chip Scale Packages

Interface Reaction Between Electroless Ni Sn P Metallization and Lead-Free Sn 3.5Ag Solder with Suppressed Ni 3 P Formation

Effects of Bi Content on Mechanical Properties and Bump Interconnection Reliability of Sn-Ag Solder Alloys

Influence of Thermal Cycling on the Microstructure and Shear Strength of Sn3.5Ag0.75Cu and Sn63Pb37 Solder Joints on Au/Ni Metallization

Effects of Flux and Reflow Parameters on Lead-Free Flip Chip Assembly. Sandeep Tonapi 1 Doctoral Candidate

Composition/wt% Bal SA2 (SABI) Bal SA3 (SABI + Cu) Bal

Microelectronics Reliability

Unique Failure Modes from use of Sn-Pb and Lead-Free (mixed metallurgies) in PCB Assembly: CASE STUDY

Thermomigration in lead-free solder joints. Mohd F. Abdulhamid, Shidong Li and Cemal Basaran*

PCB Technologies for LED Applications Application note

Microelectronics Reliability

White Paper Quality and Reliability Challenges for Package on Package. By Craig Hillman and Randy Kong

Aging Treatment Characteristics of Shear Strength in Micro Solder Bump

Manufacturing and Reliability Modelling

IMPLEMENTATION OF A FULLY MOLDED FAN-OUT PACKAGING TECHNOLOGY

Effects of Pd Addition on Au Stud Bumps/Al Pads Interfacial Reactions and Bond Reliability

IMPACT OF LEAD-FREE COMPONENTS AND TECHNOLOGY SCALING FOR HIGH RELIABILITY APPLICATIONS

Project Proposal. Cu Wire Bonding Reliability Phase 3 Planning Webinar. Peng Su June 6, 2014

Jeong et al.: Effect of the Formation of the Intermetallic Compounds (1/7)

Mater. Res. Soc. Symp. Proc. Vol Materials Research Society

Controlling the Microstructures from the Gold-Tin Reaction

' Department of Electronic Engineering

Thermo-Mechanical FEM Analysis of Lead Free and Lead Containing Solder for Flip Chip Applications

Chips Face-up Panelization Approach For Fan-out Packaging

Calorimetric Study of the Energetics and Kinetics of Interdiffusion in Cu/Cu 6. Film Diffusion Couples. Department of Physics

New Pb-Free Solder Alloy for Demanding Applications. Presented by Karl Seelig, VP Technology, AIM

Available online at ScienceDirect. Procedia Engineering 79 (2014 )

INTERFLUX ELECTRONICS NV

System Level Effects on Solder Joint Reliability

Atmosphere Effect on Soldering of Flip Chip Assemblies. C. C. Dong Air Products and Chemicals, Inc. U.S.A.

Faculty of Mechanical and Manufacturing Engineering, Universiti Tun Hussein Onn Malaysia, Johor, Malaysia

Loading Mixity on the Interfacial Failure Mode in Lead-Free Solder Joint

Non-Conductive Adhesive (NCA) Trapping Study in Chip on Glass Joints Fabricated Using Sn Bumps and NCA

14. Designing with FineLine BGA Packages

The Morphology Evolution and Voiding of Solder Joints on QFN Central Pads with a Ni/Au Finish

Mater. Res. Soc. Symp. Proc. Vol Materials Research Society

Ag Plating and Its Impact on Void-Free Ag/Sn Bumping

Arch. Metall. Mater. 62 (2017), 2B,

Reliability And Processability Of Sn/Ag/Cu Solder Bumped Flip Chip Components On Organic High Density Substrates

Arch. Metall. Mater. 62 (2017), 2B,

LEAD FREE ALLOY DEVELOPMENT

Fast Spreading of Liquid SnPb Solder on Gold-coated Copper Wheel Pattern

Low Cycle Fatigue Testing of Ball Grid Array Solder Joints under Mixed-Mode Loading Conditions

Critical Use Conditions and their Effect on the Reliability of Soldered Interconnects in Under the Hood Application

Hybrid atomization method suitable for production of fine spherical lead-free solder powder

THIN IMMERSION TIN USING ORGANIC METALS

Lead-Free Inspection Methods. Tom Perrett Marketing Manager Soldertec & Keith Bryant European Sales Manager Dage Precision Industries

Electromigration induced Kirkendall void growth in Sn-3.5Ag/Cu solder joints

3D-WLCSP Package Technology: Processing and Reliability Characterization

Effects of Design, Structure and Material on Thermal-Mechanical Reliability of Large Array Wafer Level Packages

Interconnection Reliability of HDI Printed Wiring Boards

Welding Processes. Consumable Electrode. Non-Consumable Electrode. High Energy Beam. Fusion Welding Processes. SMAW Shielded Metal Arc Welding

Visit

YOUR Strategic TESTING ENGINEERING CONCEPT SMT FLIP CHIP PRODUCTION OPTO PACKAGING PROCESS DEVELOPMENT CHIP ON BOARD SUPPLY CHAIN MANAGEMENT

Failure Modes in Wire bonded and Flip Chip Packages

Impacts of the bulk Phosphorous content of electroless Nickel layers to Solder Joint Integrity

Silicon Wafer Processing PAKAGING AND TEST

Copyright 2008 Year IEEE. Reprinted from IEEE ECTC May 2008, Florida USA.. This material is posted here with permission of the IEEE.

Impact of Intermetallic Growth on the Mechanical Strength of Pb-Free BGA Assemblies

Effect of Magnesium Addition on Microstructure and Mechanical Properties of Lead-Free Zinc-Silver Solder Alloys

IBM Research Report. Yoon-Chul Sohn, Jin Yu KAIST 373-1, Guseong-Dong, Yuseong-Gu Daejeon Korea

A LOW TEMPERATURE CO-FIRED

Ultralow Residue Semiconductor Grade Fluxes for Copper Pillar Flip-Chip

Field Condition Reliability Assessment for SnPb and SnAgCu Solder Joints in Power Cycling Including Mini Cycles

COMPUTER SIMULATION AND EXPERIMENTAL RESEARCH OF CAST PISTON POROSITY

Y. Zhou, X. Zhou, Q. Teng, Q.S. Wei, Y.S. Shi

UBM (Under Bump Metallization) Study for Pb-Free Electroplating Bumping : Interface Reaction and Electromigration

Fluxless soldering using Electron Attachment (EA) Technology

Automotive Electronic Material Challenges. Anitha Sinkfield, Delphi

Challenges and Future Directions of Laser Fuse Processing in Memory Repair

Dynamic strength of anisotropic conductive joints in flip chip on glass and flip chip on flex packages

Intermetallic Phase Growth and Reliability of Sn-Ag-Soldered Solar Cell Joints

Transcription:

J Mater Sci: Mater Electron (2010) 21:1090 1098 DOI 10.1007/s10854-009-9992-2 Thermomigration and electromigration in Sn58Bi ball grid array solder joints X. Gu K. C. Yung Y. C. Chan Received: 18 August 2009 / Accepted: 30 September 2009 / Published online: 13 October 2009 Ó Springer Science+Business Media, LLC 2009 Abstract In the present study, individual effect of thermomigration (TM) and combined effects of TM and electromigration (EM) in Sn58Bi ball grid array (BGA) solder joints were investigated using a particular designed daisy chain supplied with 2.5 A direct current (DC) at 110 C. Driven by the electric current, Bi atoms migrated towards the anode side and formed a Bi-rich layer therein. With a thermal gradient, Bi atoms tended to accumulated at the low temperature side. When the effects of TM and EM were in same direction, TM assisted EM in the migration of Bi, otherwise it counteracted the effect of EM. The effect of electron charge swirling were detected when the electric current passed by the Cu trace on the top of the solder bump instead of entering into it. For the joint without current passing by or passing through, only TM induced the migration of the Bi atoms. 1 Introduction Electromigration (EM) of solder joints has become a most persistent reliability issue in interconnects of microelectronic devices due to continuous miniaturization and the X. Gu Y. C. Chan (&) Department of Electronic Engineering, City University of Hong Kong, Tat Chee Avenue, Kowloon Tong, Hong Kong, People s Republic of China e-mail: eeycchan@cityu.edu.hk K. C. Yung Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong, People s Republic of China requirements for enhanced performance of devices [1]. Recently, thermomigration (TM) has been regarded as another reliability concern as it has been found to accompany EM in flip chip solder joints [2]. TM may assist or counteract EM depending on the direction of the thermal gradient. Most of the studies are focused on the TM of Sn Pb solder joints [2 6]. The results showed that with a substantial thermal gradient, Pb atoms in the solder migrated towards the lower temperature side. Due to the increasing awareness of the toxicity of Pb, Pb-free solders are being widely used in the electronic industry to replace Pb-containing solders in recent years. Among all the Pb-free solders, the Sn58Bi solder alloy is a promising candidate for low-temperature applications due to its low melting point (139 C) [7], so it is important to study the reliability of the Sn58Bi solder joints. Especially for the similar bi-phase structure of Sn58Bi alloy to that of Sn37Pb alloy, the investigation of TM and EM in Sn58Bi solder joint is very necessary. Many studies have reported the EM behavior of Sn58Bi solder joints [8 11]. Generally, these investigations have found that a continuous Bi-rich layer formed near the anode side, while a Sn-rich band formed at the cathode side after the current stressing. However, very few studies have yet been published to explore the TM behavior of Sn58Bi solder. In our previous study [12], the combined effects of TM and EM in Sn58Bi solder joints were investigated by using a line-type structure supplied with direct current (DC) in opposite directions. The results showed that Bi migrated to the lower temperature side driven by the thermal gradient. This study is intended to study the individual effect of TM and the combined effects of EM and TM in Sn58Bi ball grid array (BGA) solder joints using a particular designed daisy chain structure supplied with DC current. In addition, to help understand the mechanism of TM, a three-dimensional

J Mater Sci: Mater Electron (2010) 21:1090 1098 1091 thermal electrical finite element method (FEM) will be conducted to simulate the current density distribution and temperature distribution in the structure. 2 Experimental Both of the small substrates and large substrates used to fabricate the BGA solder joints were made of flame retardant 4 (FR4) materials. The thickness of each substrate was 1.5 mm. The BGA pads were a solder-mask-defined type, and the opening diameters of the pads on the small substrate and large substrate were 280 lm and 250 lm, respectively. The thickness of the Cu pads was 35 lm with organic solderability preservative (OSP) surface finishing. The pads were linked with designed Cu traces with a thickness of 35 lm for electrical continuity. The small substrate was attached to the large substrate using flip chip bonding technology by a Flip-Chip bonder (SUSS, FCM). The assembly process was similar to that of a conventional surface mount technology (SMT) procedure, including flux printing, ball placement, reflow soldering and cleaning. The aligned samples were reflowed using a hot air convection reflow oven (BTU PYRAMAX 100 N). The peak temperature in the temperature profile was 176 C and the duration above the melting point of Sn58Bi alloy was about 82 s. Figure 1a gives images of a small substrate and large substrate before the assembly. Figure 1b gives the image of a prepared structure, and four designed daisy chains formed at the edges of the small substrate. Solder joints 1 6 were prepared for each chain. Only one chain was supplied with a DC current during the testing. In order to create the thermal gradient across the solder joints, two Cu pads with dimensions of (35 lm 9 12000 lm 9 24000 lm) were designed on the other side of the large substrates. Each Cu pad was connected with the BGA pads of joints 2 and 5 through the plating through holes (PTH) (as shown in Fig. 1a). The prepared structure with one chain supplied with DC current and was put into the furnace set at 110 C. Figure 1c shows the electron flow through the solder joint chain. For the solder joints 1, 3, 4 and 6, the electric current passes through the solder bumps. While the electric current passes by the top of the solder bump of joint 2 instead of entering it. No current passes by or passes through the solder bump of joint 5. After predetermined periods of time, samples were taken out of the furnaces and quenched in air. Each one sample was mounted and cross sectioned carefully towards the center of the solder joint. Microstructural observations were performed using a Philips XL 40 FEG Scanning Electron Microscope (SEM) equipped with an energy dispersive x-ray (EDX). Three-dimensional thermal electric FEM simulation was used to calculate the current density distribution and temperature distribution in the test structure. In the simulation model, the height of the solder bump between the two substrates is 150 lm. The physical parameters are listed in Table 1 [3, 13 15]. TCR is the abbreviation of temperature coefficient of resistance. 3 Results and discussion 3.1 Results Figure 2 shows the typical microstructure of one Sn58Bi BGA solder joint aged at 110 C without current stressing for 480 h. It is clear that the two interfaces exhibited nearly the same microstructure. Figure 3 shows the microstructures of six solder joints in one daisy chain supplied with a 2.5 A DC current at 110 C for 480 h. The measured temperatures of the small substrate surface and the large substrate surface were about 112 and 132 C, which showed that a thermal gradient was created between the small substrate and the large substrate during the test. According to the microstructures of the solder joints shown in Fig. 3, substantial mass migration had occurred in the solder joints during the current stressing. In particular for joints 1 and 6, serious separation of Bi atoms had occurred. It is clear that phase separation in Sn58Bi solder is especially prominent for solder joints 1 and 6. Figure 4 shows the detailed microstructures of joint 1 and joint 6 with local magnified images. The Bi atoms accumulated at the small substrate side in solder joint 1, while a large Sn-rich area formed at the large substrate side. By contrast, Bi atoms accumulated at the large substrate side of joint 6. It is well known that Bi atoms migrate in the direction of the electron flow. As the direction of the electron flow was from the large substrate to the small substrate for joint 1, while it was from the small substrate to the large substrate for joint 6, it is reasonable to find the phenomena as the above illustrated. For solder joint 1, as shown in Fig. 4b, the IMC layer at the small substrate side is composed of a Cu 3 Sn IMC layer (near the Cu pad side) about 2.4 lm thick and a Cu 6 Sn 5 IMC layer (near the solder side) about 4.2 lm thick. By contrast, the IMC layer at the large substrate side (as shown in Fig. 4c) is mainly the Cu 6 Sn 5 phase with a thickness of more than 20 lm. About 12 lm of the Cu pad at the large substrate side had been consumed during the test. The IMC layer at the small substrate side is thinner than that at the large substrate side. The enhanced effect of the electric current and the barrier effect of the Bi-rich layer played important roles in the growth of the Cu-Sn IMC layer in

1092 J Mater Sci: Mater Electron (2010) 21:1090 1098 Fig. 1 a Optical images of a small substrate (chip) and large substrate (substrate) before the assembly, b optical image of a sample prepared for subsequent tests, and c schematic diagram of electrified daisy chain Table 1 Thermal conductivities and electrical resistivities for the materials used in the simulation Material Thermal conductivity (W/m C) Resistivity (lx cm) TCR (10-3 / C) Cu 398 1.7 4.3 FR4 0.7 Sn58Bi 19 30 4.4 Sn58Bi/Cu solder joints, which has been reported in our previous study [11]. It is worth noting that big voids formed at the interface between the Bi-rich phase and the Sn-rich phase. However, the solder joints chain did not open after the test. According to Fig. 4a, the solder joint is still partly connected. The formation of the crack can be ascribed to substantial Cu-Sn IMC growth in the solder joint which consumed a large amount of Sn atoms. Fig. 2 SEM image of one solder joint aged at 110 C without current stressing for 480 h

J Mater Sci: Mater Electron (2010) 21:1090 1098 1093 Fig. 3 SEM images of Sn58Bi solder joints in one daisy chain under 2.5A DC current stressing at 110 C for 480 h, the red arrows indicate the electron current Fig. 4 SEM images of solder joints 1 and 6 under 2.5 A current stressing at 110 C for 480 h: a cross section of joint 1, b local magnified image of joint 1 at the small substrate side, c local magnified image of joint 1 at the large substrate side, d cross section of joint 6, e local magnified image of joint 6 at the small substrate side, and f local magnified image of joint 6 at the small substrate side For solder joint 6, as the direction of the electron flow was from the small substrate (cathode) to the large substrate (anode), most of the Bi atoms accumulated at the large substrate side. The IMC layer at the cathode is mainly Cu 6 Sn 5 with a thickness of about 8.3 lm. By contrast, the IMC layer at the anode is mainly Cu 3 Sn with a thickness of 3.1 lm, and only a very thin (about 0.5 lm) Cu 6 Sn 5 formed at the solder side. It is worth noting that more than 90% of the Bi atoms had accumulated at the anode side for joint 1 and joint 6, which was due to the rapid transfer of the Bi atoms driven by the electron flow. Also, the IMC formed at the anode

1094 J Mater Sci: Mater Electron (2010) 21:1090 1098 side is composed of the Cu 3 Sn phase and Cu 6 Sn 5 phase, it is different from the results reported in previous study [11], which showed the main phase of the IMC at the interface of the Cu/Sn58Bi solder is Cu 5 Sn 6. Two causes are responsible for this. One is the barrier effect of the Bi-rich layer. After long periods of current stressing, the Bi-rich layer became thick enough to block the transfer of Sn atoms and further IMC development was limited due to the lack of Sn atoms at the anode side, while more Cu atoms could be supplied by the Cu pad and more Cu 6 Sn 5 IMC turned into Cu 3 Sn. As the Bi-rich layer developed in the joints illustrated in the previous report was thinner than that in solder joint 1 and joint 6 considered here, substantial amounts of Sn atoms in the bulk solder could diffuse to the interface and react with the Cu atoms to form Cu 6 Sn 5. The other reason is that the actual annealing temperature in the solder joints in the previous report was lower than that in the case considered here. According to Laurila et al. [16], the high temperature will be helpful for the growth of the Cu 3 Sn. It was also reported that Kirkendall voids always tend to form inside Cu 3 Sn during the anneling of the Cu/solder [16, 17] interconnections. But no Kirkendall voids could be observed in our results, which maybe due to the shorter annealing time in this test than that cited in the above literature [16, 17]. For the cathode side, there were enough Sn atoms supplied for the Cu-Sn IMC formation and Cu 6 Sn 5 IMC was the main product. Figure 5 gives the magnified images of solder joint 3 and joint 4. The upper side of the image is the small substrate while the lower side is the large substrate side. There is no big difference in the microstructure between the two solder joint, except the Bi separation occurred at opposite substrate sides. The average thickness of the Bi-rich layer in solder joint 3 is about 24 lm, and that in solder joint 4 is about 22 lm. A layer of Cu 3 Sn IMC with a thickness of about 1 lm was detected at the interface of the anode side for each joint, but the main phase of the IMC layer is Cu 6 Sn 5. The IMC at the cathode interface is mainly Cu 6 Sn 5 for each solder joint. Figure 6 shows the detailed microstructure of solder joint 2. The electric current passed by on the top of the solder bump instead of entering into it. Despite the current did not pass through the solder joint, a Bi-rich phase also formed at the larger substrate side during the test. A major difference of the microstructure from the joints those shown in Figs. 4 and 5 is that two Bi-rich phase regions can be observed in Fig. 6. One Bi-rich phase region formed at the interface of the large substrate (as shown in Fig. 6d). The other one with small area formed at one side of the contact between the solder and the Cu trace on the small substrate (as shown in Fig. 6c), while the Sn-rich phase formed at the other side of the contact between the solder and the Cu trace (as shown in Fig. 6b). The mechanism of formation of the Bi-rich areas in solder joint 2 may be ascribed to the TM and the electron charge swirling, and this will be discussed in a following section. The average thickness of the Bi-rich layer at the large substrate side is about 15.5 lm. The IMC layer at the small substrate is about 4.9 lm, while that at the large substrate is about 4.3 lm which is only a slightly thinner than that at the small substrate side. Figure 7 shows the microstructure of solder joint 5. During the test, there was no current passing through this solder joint or pass by the Cu pads those contacted with the solder bump. It is clear that a Bi-rich layer with an average thickness about 5.9 lm formed at the large substrate side, which is thinner than that in solder joint 2, which also indicates that the migration of Bi in this joint is not prominent. No obvious Sn-rich phase can be observed at the small substrate side. As the Sn-rich phase is induced by the migration of Bi, the small amount of Bi atoms migration towards the large substrate could not leave a visible Sn-rich phase at the small substrate side. IMC layers at the two interfaces are nearly the same thickness, with the thickness of about 3.4 lm. For the Sn58Bi solder joint with current stressing, the polarity effect of the electric current and the barrier effect of the Bi-rich layer played critical roles in the growth of the IMC layers, so the thickness of Fig. 5 SEM images of the cross section of solder joints under 2.5 A DC current stressing at 110 C for 480 h: a solder joint 3 and b solder 4

J Mater Sci: Mater Electron (2010) 21:1090 1098 1095 Fig. 6 SEM images of solder joint 2: a cross section of solder joint 2 with the electric current passing by on top of it, b, c and d local magnified micrographs of each indicated parts IMC layer developed at the anode side is not the same as that at the cathode side. For solder joint 5, no electric current affected the IMC growth, and the barrier effect is not obvious for a Bi-rich layer with a thickness of 5.9 lm, so the IMC layers at the two sides had similar thicknesses. As a thermal gradient was created across the small substrate and the large substrate, the accumulation of the Bi atoms at the large substrate side was due to the TM effect in the solder joint. This will be discussed later. Figure 8 gives the FEM simulation results of the current density distribution in solder joints 1 6 with a 2.5 A DC current applied. The calculated average current density in the powered solder joints is 2.2 9 10 3 A/cm 2. According to the simulation results, the current density in most parts of solder joints 1, 3, 4, and 6 reached above 2.3 9 10 3 A/cm 2. Current crowding was existed at the contact areas between the Cu traces and the solder bumps. The current density at the entry location was about 10 times of the average current density. For the joint 2, the current passing by the Cu trace on the top of the bump, it is clear that current crowding also existed at the entry location and exit location. The current density in most part of solder joint 2 was above 500 A/cm 2, which was due to the effect of electron charge swirling. The effect of electron charge swirling has been reported by Lai et al. [18], their results showed that for a solder joint with the electric current passing by on top of it, the induced vertical current density field leads to the initiation of a void around the UBM. As no current passed by solder joint 5, no current density existed in solder joint 5. In order to understand the effect of electron charge swirling, a vector display of the current density distribution in solder joint 2 is shown in Fig. 9. It is clear that a swirling current density field was created in the solder bump. The magnitude of the current density in the part near the Cu trace on the top of the solder bump was comparable to that on other solder joints having the electric current going through them. Figure 10 shows the FEM simulation results of temperature distribution in the test module supplied with a 2.5 A DC current at an ambient temperature of 110 C. It is obvious that a temperature differences about 40 C existed between the small substrate and the large substrate. The calculated local temperatures are larger than those measured by thermocouple. This may be because the measured locations are the outside surfaces of the test module. Figure 11 shows the detailed temperature distributions in solder joints 1 6. Thermal gradient and the maximum temperature in each solder joint is shown in Table 2. Itis clear that the largest thermal gradient existed in solder joint 2 during the test. 3.2 Discussion For solder joint 1 and solder joint 6, the maximum temperatures in the solder are 131.2 and 136.8 C, respectively.

1096 J Mater Sci: Mater Electron (2010) 21:1090 1098 Fig. 8 Current density distribution in solder joints (a) solder joints 1 3 and (b) solder joints 4 6 (units: A/m 2 ) Fig. 7 SEM images of solder joint 5: a cross section of solder joint 5 (without current stressing), b local magnified micrograph of the interface at the small substrate side, and c local magnified micrograph of the interface at the large substrate side The thermal gradient across joint 1 is 119.2 C/cm, while that across solder joint 6 is about twice of this. As Bi atoms migrate towards the lower temperature side when the solder joint undergoes the thermal gradient [12], the TM and EM of Bi in solder joint 1 are in the opposite directions, Fig. 9 Vector show of the current density distribution in solder joints 2 (units: A/m 2 ) while those are in the same direction for the solder joint 6. However, EM played the overwhelming effect on the migration of Bi atoms for solder joints 1 and 6, it is rational

J Mater Sci: Mater Electron (2010) 21:1090 1098 1097 Table 2 Thermal gradient and the maximum temperature in solder joints Solder joint Thermal gradient ( C/cm) Maximum temperature ( C) 1 119.2 131.2 2 403.5 130.0 3 0 136.8 4 317.3 142.1 5 226.9 124.7 6 239.9 136.8 Fig. 10 Temperature distribution in the test module (units: C) to observe that most of the Bi atoms in the solder accumulated at the small substrate side for joint 1. For solder joint 6, as the thermal gradient is in the same direction of Fig. 11 Temperature distribution in solder joints 1 6 (units: C)

1098 J Mater Sci: Mater Electron (2010) 21:1090 1098 the electron flow, TM assisted the effects of EM on the migration of Bi. For solder joint 2, three factors played important effects on the migration of Bi. One is the effect of the electron charge swirling, which induced a current density above 500 A/cm 2 in the solder bump. The other is that a thermal gradient with magnitude of 403.5 C/cm was created across the solder bump during the test, TM was in the direction from the small substrate to the large substrate side, which drove Bi atoms moving towards the large substrate. The last factor is that the maximum temperature in this joint is about 136.7 C, and this temperature is very near the melting point of Sn58Bi solder alloy. As the migration of Bi is diffusion-controlled process, the temperature played an important effect on the diffusion of Bi atoms. For the above reasons, a Bi-rich layer about 15.5 lm thick formed at the large substrate in solder joint 2. The Bi-rich phase which formed at the contact of Cu trace with the solder bump can be ascribed to the electron charge swirling. For the joint 5, a thermal gradient with magnitude of 226.9 C/cm existed in it, TM is the only driving force for Bi migration. According to the temperature simulation results, the thermal gradient and the maximum temperature are smaller than those for solder joint 2, so a thinner Bi-rich layer with thickness of 5.9 lm formed during the test, which is rational. For solder joint 3, there was no temperature difference in the solder bump, only EM induced the migration of Bi. For the solder joint 4, a thermal gradient about 317.3 C/cm existed in the bump, but it is in the opposite direction of EM. Despite a maximum temperature of 142 C, since the TM counteracted with EM, the migration of Bi was not dramatic and an even thinner layer formed than that in solder joint 3. It is worth noting that the maximum simulation temperature in the solder joint is above the melting point of Sn58Bi alloy, but no obvious melting of the solder joint can be observed. This may be due to the counteract affect of the TM in the solder joint. 4 Conclusions TM and EM in Sn58Bi BGA solder joints were investigated using a particular designed daisy chain supplied with a 2.5 A DC current at 110 C. The individual effects of TM and EM, and combined effects of TM and EM were detected. Driven by the electric current, the Bi migrated towards the anode side and formed a Bi-rich layer therein. With a thermal gradient, the Bi atoms tended to migrate towards the lower temperature side. When the effects of TM and EM were in same direction, TM assisted EM in the migration of Bi, otherwise TM counteracted the effect of EM. The effects of electron charge swirling were detected when the electric current passed by the Cu trace on the top of the solder bump instead of entering into it. After stressed for 480 h, two Bi-rich areas formed in the solder bump. One Bi-rich area formed at one side of the contact between the Cu trace and the solder bump. The other formed at the large substrate side (low temperature side). The electron charge swirling and TM played combined effects on the accumulation of Bi. For the joint without current passing by or passing through, only TM induced the migration of the Bi atoms. A Bi-rich layer with a thickness about 5.9 lm formed at the low temperature side, which was thinner than that with the current passing by the Cu trace on the top of the solder bump. Acknowledgments This project has been supported by RGC General Research Fund (GRF) of Hong Kong (Project No. 9041486). References 1. K.N. Tu, J. Appl. Phys. 94, 5451 (2003) 2. H. Ye, C. Basaran, D.C. Hopkins, Appl. Phys. Lett. 82, 1045 (2003) 3. A.T. Huang, A.M. Gusak, Y.S. Lai, K.N. Tu, Appl. Phys. Lett. 88, 141911 (2006) 4. F.Y. Ouyang, K.N. Tu, Y.-S. Lai, A.M. Gusak, Appl. Phys. Lett. 89, 221906 (2006) 5. A.T. Huang, K.N. Tu, Y.S. Lai, J. Appl. Phys. 100, 033512 (2006) 6. D. Yang, B.Y. Wu, Y.C. Chan, K.N. Tu, J. Appl. Phys. 102, 043502 (2007) 7. K.J. Puttlitz, K.A. Stalter, Handbook of Lead-Free Solder Technology for Microelectronic Assemblies, vol. 9 (Marcel Dekker, Inc., New York, 2004), p. 282 8. Q.L. Yang, J.K. Shang, J. Electron. Mater. 34, 1363 (2005) 9. L.-T. Chen, C.-M. Chen, J. Mater. Res. 21, 962 (2006) 10. C.-M. Chen, L.-T. Chen, Y.-S. Lin, J. Electron. Mater. 36, 168 (2007) 11. X. Gu, D. Yang, Y.C. Chan, B.Y. Wu, J. Mater. Res. 23, 2591 (2008) 12. X. Gu, Y.C. Chan, J. Appl. Phys. 105, 093537 (2009) 13. D.R. Frear, S.N. Burchett, H.S. Morgan, J.H. Lau, The Mechanics of Solder Alloy Interconnects, vol. 3 (Van Nostrand Reinhold, New York, 1994), p. 61 14. Indium Corporation product data sheet, http://www.indium.com/ products/alloy_sorted_by_temperature.pdf 15. S.W. Liang, Y.W. Chang, C. Chen, J. Electron. Mater. 36, 159 (2007) 16. T. Laurila, V. Vuorinen, J.K. Kivilahti, Mater. Sci. Eng. R 49, 1 (2005) 17. K. Zeng, R. Stierman, T.-C. Chiu, D. Edwards, K. Ano, K.N. Tu, J. Appl. Phys. 97, 024508 (2005) 18. Y.S. Lai, C.W. Lee, C.L. Kao, J. Electron. Pack. 129, 56 (2007)