Prognostication of Copper-Aluminum Wirebond Reliability under High Temperature Storage and Temperature-Humidity

Pradeep Lall [1], Shantanu Deshpande [1], Luu Nguyen [2], Masood Murtuza [2]

[1]Auburn University NSF-CAVE3 Electronic Research Center Department of Mechanical Engineering

Auburn, AL. 36849 [2]Texas Instruments, Santa Clara, CA 95052

Tele: (334) 844-3424 E-mail: lall@auburn.edu

Abstract Gold wire bonding has been widely used as first-level interconnect in semiconductor packaging. The increase in the gold price has motivated the industry search for alternative to the gold wire used in wire bonding and the transition to copper wire bonding technology. Potential advantages of transition to Cu-Al wire bond system includes low cost of copper wire, lower thermal resistivity, lower electrical resistivity, higher deformation strength, damage during ultrasonic squeeze, and stability compared to gold wire. However, the transition to the copper wire brings along some trade-offs including poor corrosion resistance, narrow process window, higher hardness, and potential for cratering. Formation of excessive Cu-Al intermetallics may increase electrical resistance and reduce the mechanical bonding strength. Current state-of-art for studying the Cu-Al system focuses on accumulation of statistically significant number of failures under accelerated testing. In this paper, a new approach has been developed to identify the occurrence of impending apparently-random defect fall-outs and pre-mature failures observed in the Cu-Al wirebond system. The use of intermetallic thickness, composition and corrosion as a leading indicator of failure for assessment of remaining useful life for Cu-al wirebond interconnects has been studied under exposure to high temperature and temperature-humidity. Damage in wire bonds has been studied using x-ray Micro-CT. Microstructure evolution was studied under isothermal aging conditions of 150°C, 175°C, and 200°C till failure. Activation energy was calculated using growth rate of intermetallic at different temperatures. Effect of temperature and humidity on Cu-Al wirebond system was studied using Parr Bomb technique at different elevated temperature and humidity conditions (110°C/100%RH, 120°C/100%RH, 130°C/100%RH) and failure mechanism was developed. The present methodology uses evolution of the IMC thickness, composition in conjunction with the Levenberg-Marquardt algorithm to identify accrued damage in wire bond subjected to thermal aging. The proposed method can be used for quick assessment of Cu-Al parts to ensure manufactured part consistency through sampling.

Introduction Gold wire has been widely used in the electronic industry for first-level interconnects because of its high electrical conductivity, low hardness and resistance to oxidation. The recent steep rise in the gold prices has forced the electronic industry to look for other low cost alternatives. Copper and

palladium coated copper are leading candidates to replace the incumbent gold wire [Breach 2010]. The transition to copper wire bond has been encouraged by not only the lower cost of copper wire but also desirable properties such as higher thermal conductivity and electrical conductivity compared with gold. In addition, copper wire also has higher mechanical strength and a low reaction rate with aluminum, which bodes well for long term reliability of the copper-aluminum system [Boettcher 2010; Kim 2003]. However, introduction of copper in to wire bonding process have also created new set of challenges. Copper is harder than gold, and thus requires more ultrasonic energy and mechanical force during the mechanical bonding process. Higher energy during the bonding process requires a narrow bonding window, increasing the potential for chip cratering beneath bonding surface. Further, copper and copper intermetallics tend to oxidize at high temperatures reducing bond strength. Substantial amount of work has been done on developing and optimizing thermosonic wire bonding process for copper and significant improvements are noted in literature [Breach 2010; Qin 2011; Shah 2009; Chen 2004; Schmitz 2011].

In spite of process advancements, reliability is still one of the major concerns for copper wire bonding. Past research has shown that formation of excessive intermetallics at Cu-Al interface is a major contributor to failure of the wire bond accompanied by resistance increase and fracture or bond lifts [Boettcher 2011; England 2011]. Past researchers have studied the IMC growth under isothermal aging. IMC has been studied using various techniques, including the crystal structure, effect of annealing time and effect of temperature on IMC growth [Na 2011]. Composition of IMC has been studied using TEM and XRD, and reported to be CuAl2 and Cu9Al4 [Wieczorek-Ciurowa 2005; Laik 2008; Lee 2005]. Tian, et.al. [2011] reported IMC growth behavior of Cu-Al wirebond system when subjected to multiple thermal stresses. Cu-Al IMCs grow much faster in thermal aging compared to thermal shock and that rate of growth is higher for higher aging temperatures. Three predominant phases of Cu-Al IMCs include CuAl2, Cu9Al4, CuAl. However, the timing of appearance of each phases is still not clear. It has also been found that IMC growth is diffusion driven [Chen 2011; Hand 2008]. Activation energy of Cu-Al IMCs reported in literature have wide range from 2.5l kcal/mol to 23.9 kcal/mol [Lee 2005; Goh 2013; Simonovic 2009; Levenson 1989]. Prior data indicates that cracking in the vicinity of the Cu-Al intermetallics is a predominant location of failure. The initial cracks may form in the copper-aluminum bond because of

978-1-4799-2407-3/14/$31.00 ©2014 IEEE 1973 2014 Electronic Components & Technology Conference

ultrasonic squeeze during the bond process, and that the cracks may propagate with aging time [Boettcher 2010]. The Cu-Al IMCs grow at a slower pace compared to the Au-Al system [Goh 2013]. Oxidation of Cu-Al intermetallics during operation at high temperature and high humidity may cause oxidation of IMC, followed by crack initiation, and eventual failure. The process of corrosion is accelerated in the presence of ionic species, such as halide, hydroxyl ions, and elevated temperature. Prior data indicates that under 85%RH/85°C conditions, the aluminum pad gets corroded faster and causes accelerated failure compared to the Au-Al wirebond system [Breach 2010]. Different corrosion reactions have been proposed for Cu rich and Al rich phases of IMC. Copper rich interface (Cu9Al4) undergoes preferential corrosion compared to the aluminum rich phase. Reaction rate is greatly affected by concentration of hydroxyl, halide ions in surrounding. Corrosion of IMC causes sudden increase in resistance and reduction in ball shear strength [Osenbach 2013; Liu 2011; Su 2013; Zeng 2013]. Understanding this abnormal behavior is important step in troubleshooting reliability issues related to Cu-Al wire bonding.

Electronic components operating in harsh environment are often subjected to high temperature and/or humidity conditions. It is often not feasible to capture the continuous time temperature operational history over the use life of the product. Current health management systems provide nearly zero visibility into health of electronics and packaging for prediction of impending failures [McCann 2005; Marko 1996; Schauz 1996; Shiroishi 1997]. The built-in-self tests (BIST) are generally used to give electronic assemblies the ability to test and diagnose themselves with minimal interaction from external test equipment. BIST has a built-in circuit capable of providing error detection and correction [Zorian 1994; Chandramouli 1996; Drees 2004; Hassan 1992]. However, several studies conducted [Drees 2004; Gao 2002; Daniel 1990] have shown that BIST can be prone to false alarms and can result in unnecessary costly replacement, requalification, delayed shipping, and loss of system availability. There is need to develop scientific methodology to predict damage in copper-aluminum electronic packages, without knowing there prior aging history. Prognostic health management (PHM) is a method for assuring the reliability of a system by monitoring the system in real time as it is used in the field. PHM system if implemented for copper-aluminum wire bond interconnects could provide information of the electronic system’s state and the possibility of impending failure. Leading-information on near failure systems could allow enough time for repair or replacement of the damaged modules. In addition, it is envisioned that the leading indicators for the copper-aluminum system could be used for defect detection.

Previously, leading indicators based prognostic and health management methodologies for residual life computation of electronic solder joints subjected to single, multiple and superimposed thermal environments comprising of isothermal aging and thermal cycling have been developed [Lall 2008a,b; Lall 2007c, 2009c]. Examples of second-level interconnect damage precursors include micro-structural evolution, inter-metallic compound growth, stress and stress gradients. Previously, damage pre-cursors have been developed for

various lead-free alloy compositions on a variety of area-array architectures. PHM methodology for second-level interconnects was successfully developed for microstructure evolution of damage based leading indicator for estimating prior accrued damage [Lall 2012a,c]. Researchers have also implemented PHM technique for assessment of damage and remaining useful life in electronic system, which are subjected to thermo-mechanical load, sequential multiple thermal environment [Lall 2012b; 2010b]. In this paper; PHM framework has been introduced using Levenberg-Marquardt (LM) algorithm based on physics based state-vectors for assessment of accrued damage and future degradation prediction, using IMC layer development when Cu-Al system is subjected to thermal aging, as a leading indicator.

Test Vehicle The test vehicle used for the study is the 32-pin chip scale package. The package is 4.5 mm in length, 5.5 mm in width, 0.7 mm in height. Each pin has a length of 0.45 mm and width of 0.3 mm. The package interconnects has an I/O pitch of 0.5 mm. The package has 30m diameter copper wires and aluminum pads. The packages used for the study were not daisy chained. Package dimensions are listed in Table 1.

Table 1: Dimensions of Test Vehicle Parameter Dimensions (mm)

Width of Package 4.5 Length of Package 5.5 Height of Package 0.7

Length of Pin 0.45 Width of Pin 0.3

Pitch 0.5 Figure 1 shows the optical microscopic images of the package top and bottom. Figure 2 shows X-ray images of the package taken using the YXLON Cougar CT System. Figure 3 shows the 3D μ-CT reconstruction showing copper-aluminum wirebonds. The chip and the electronic mold compound have been deselected for better visibility of the copper-aluminum interconnects. Figure 4 shows magnified view of single wire-bond.

(a) (b)

Figure 1 – Optical images (a) bottom view, (b) top view.

1974

Figure 2 -X-ray image (Top View)

Figure 3– 3D µCT reconstruction of package

Figure 4– Close-up of copper wirebond

Test Matrix Test packages were subjected to 150°C, 175°C, and 200°C isothermal aging environment. For aging at 150°C and 175°C, IMC growth was monitored at the interval of 1 week. For aging temperature of 200°C, reading interval was 2 days.

To study effect of temperature-humidity bias, set of packages were subjected to three different conditions, 130°C/100%RH, 120°C/100%RH, 110°C/100%RH. Reading interval for these tests was 24 hours. Packages were potted into resin, polished, and then polished surface was sputter coated with gold, at 25µA, for 45 seconds. IMC growth was observed using scanning electron microscopy. IMC thickness at multiple locations was recorded in each image, and mean value of readings was considered as final IMC thickness. This ensured accuracy in measurement. Different modes of EDS scans (e.g. line scan, point scans) were performed for material characterization. Experimental Results and Analysis High Temperature Thermal Aging Figure 5 shows development of IMC layer when packages were subjected to 150°C isothermal aging. IMC growth over period of temperature can be observed. Similarly, images of IMC growth have been captured at thermal aging conditions of 175°C and 200°C as shown in Figure 6 and Figure 7.

Figure 5 – IMC Development at 150°C  

1975

 Figure 6 - IMC Development

at 175°C

Figure 7 - IMC Development

at 200°C Their IMC growth is plotted in Figure 13 (b) and (c) respectively. Even after aging for 20 weeks, at 150°C, IMC corrosion was not found. IMC growth is very slow. Initially only Cu-rich phase of IMC was visible. However, after aging for 840 hours, Al-rich phase was distinctly visible in the Cu-Al intermetallics. In case of thermal aging at 175°C, the Al-rich phase was detected after aging after only 336 hours of thermal aging, much sooner than the 840 hours required at 150°C. Initial cracking of wirebond was observed after 1008 hours of aging at 175°C. Initial cracks were observed at edges of ball bond, in copper rich interface, and then they propagated towards center, resulting into complete cracking in most of the wirebonds after 1176. Even after complete cracking it was observed that the IMC growth continued resulting in formation of singular IMC phase below the crack. The aluminum from the Al-pad continued to diffuse into the IMC, but diffusion of copper from ball bond got restricted due to crack. IMC growth after crack formation resulted in the development of a continuous layer of aluminum rich phase

below the crack. Similar trend was observed in third thermal aging condition at 200°C. IMC growth at 200°C was the fastest of the three thermal conditions measured in this study. Crack initiation and propagation was faster at higher thermal aging temperature with the crack initiation observed after 432 hours and near complete cracking in majority of the wirebonds after 528 hours of 200°C aging. Crack initiation and propagation is shown in Figure 8.

Figure 8 – Crack initiation and propagation

EDX analyses were performed on IMC phases, as well as region of cracking. Point scans and line scanning techniques were used. EDX analysis on IMC phases to determine its composition is shown in Figure 10. It was found that two phases found are Cu-rich and Al-rich phases. They preliminary consist Cu9Al4 (Spectrum-4; Figure 10) and CuAl2 (Spectrum-3; Figure 10) respectively, which is in good agreement with previous reports [Wieczorek-Ciurowa 2005; Liak 2008; Lee 2005]. EDX spot scan also shows that cracking takes place in copper rich interface as shown in Figure 11. Location of the crack has also been confirmed using the EDX elemental line scan in Figure 12. IMC

1976

thickness at each reading interval was measured using a RMS average of the IC thickness at 20-points. The average value of the IMC thickness was plotted as shown in Figure 9.

Figure 9 – Method for measurement of the IMC growth.

Figure 10 – EDX sport analysis to determine IMC Composition

Figure 11 – EDX Spot analysis of crack.

Figure 12 – EDX Line scan through crack.

In line scan image shown in Figure 12, the area of crack and IMC is marked by black box, projected as circle in scan result. Image analysis shows that Cu rich zone has undergone oxidation, and cracking. Source of this oxygen is hypothesized to be outgassing from the molding compound under thermal aging, or broken polymer chain in molding compound.

Figure 13 – IMC growth for thermal aging at 150°C

Figure 14 – IMC growth for thermal aging at 175°C

Figure 15 – IMC growth for thermal aging at 200°C

Accelerated Test for Corrosion-Susceptibility of Cu-Al In this test, two sets of 10-packages each were subjected to three different conditions of 110°C/100%RH, 120°C/100%RH, and 130°C/100%RH. To achieve the test conditions of 110°C/100%RH, 120°C/100%RH and 130°C/100%RH, a Parr bomb test apparatus was used. Parr

Ln of Aging Duration (Hours)

4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5

Ln

of

No

rma

lize

d IM

C (

Dim

ens

ion

les

s)

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

Thermal aging at 150oCRegression Line

Ln of Aging Duration (Hours)

5.0 5.5 6.0 6.5 7.0 7.5

Ln

of

No

rma

lized

IM

C (

Dim

ensi

on

less

)

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

2.2

Thermal aging at 175oCRegression Line

Ln of Aging Duration (Hours)

3.5 4.0 4.5 5.0 5.5 6.0 6.5

Ln

of

No

rma

lized

IM

C (

Dim

ensi

on

less

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

2.2

Thermal aging at 200oCRegression Line

1977

bomb test apparatus shown in Figure 16 is a closed pressure vessel in which 100%RH can be maintained at high ambient temperatures. PTFE cup inside the Parr bomb was filled with water, and packages were placed in cup such that they will not be immersed into water but rather only in contact with the water vapors.

 Figure 16 – Parr Bomb (Courtesy of Parr Instruments

Company)

 Figure 17 - Crack Development and Propagation at

110°C/100%RH Microstructure analysis using scanning electron microscope was performed on parts subjected to the highly-accelerated stress test. EDS line scan, point scan and area scanning techniques were used. Figure 20 shows the point scan analysis at the Cu-Al wirebond interface. Analysis indicates that oxygen was present at the Cu-Al interface. Percentage atomic weight of oxygen was higher at edges, showing that initial corrosion took place at edges, and then it propagated towards center. For all three cases, oxidation was predominantly found in the Cu-rich phase.

Figure 18 – Crack Development and Propagation at

120°C/100%RH

Figure 19 - Crack Development and Propagation at

130°C/100%RH

Figure 21 shows EDS line scan from Cu ball bond through Al pad to the Si-Chip. Analysis results indicate that cracking of wirebond took place in Cu-rich zone, and the Al-rich zone was still intact. In order to ascertain that the cracking is due to

1978

corrosion; but a visual polishing artifact, area mapping was done. Figure 22 shows area mapping performed to find oxygen content of selected area. Intensity of red dots indicates concentration of oxygen at the location of the red dot. Figure 22 shows that dark red spots were spotted in the region of cracking, and area near it. Density of the oxygen rich region indicated by the red-dots shows that the cracks shown are due to oxidation.

(a) (b) (c) Figure 20 – EDS spot analysis (a) Spectrum 1, (b) Spectrum

2, (c) Spectrum 3

Table 2 –Percentage atomic Weight of Elements. Elements Spectrum 1 Spectrum 2 Spectrum 3

O 33.36 17.37 21.56 Al 3.08 2.74 2.63 Si 13.00 5.25 7.80 Cl 0.19 0.03 0.03 Cu 50.37 74.61 67.98

Figure 21 – EDS Line Scan of crack in wirebond

Figure 22 – Area Mapping for Oxygen in the region of

cracking. The corrosion mechanism has been analyzed using the Pourbaix Diagram or the potential/pH diagram which maps out the potential stable phases of the aqueous electrochemical

system. The vertical axis is labeled for the voltage potential with respect to a standard hydrogen electrode (SHE) as calculated by the Nernst Equation. The lines in the Pourbaix diagram show the equilibrium condition where the activities for each of the species are equal on either side of the line. On either side of the line one of the ionic species is said to be predominant. The Pourbaix Diagram has been used to identify the regions of immunity, corrosion, and passivity for both copper and aluminum. The vertical lines indicate the species that are in acid-alkali equilibrium. The non-vertical lines separate the species at redox equilibrium. Specifically, the horizontal lines separate redox equilibrium species not involving hydrogen or hydroxide ions. The diagonal lines separate redox equilibrium species involving hydrogen or hydroxide ions. The dashed lines enclose the practical region of stability of the aqueous solvent to oxidation or reduction and thus the region of interest in aqueous systems. Outside the dashed region, water breaks down and not the metal.

Figure 23 – Pourbaix Diagram of Copper at 100°C

[Beverskog 1997]

Figure 24 - Pourbaix Diagram of Aluminum [McCafferty

2010, Pg. 96]

1979

In general, the metal is not attacked and forms stable unreacted metal species in the region of immunity, metal forms a stable oxide or stable hydroxide in the region of passivity, and metal is susceptible to corrosion in the region labeled as corrosion. Low E (or pE) values represent a reducing environment. High E values represent an oxidizing environment. Electrochemical potential represents the driving force for oxidation or reduction. The electrochemical potential is generally cited as standard reduction potential and quantifies the tendency of the chemical species to be reduced. More positive values of reduction potential indicate higher ease with which that the chemical species will be reduced. Corrosion of metal usually occurs at the anode. Standard electrochemical potential of pure Cu is +0.34V. Electrochemical potential of aluminum is -1.67V. Thus in the copper-aluminum system, copper is easier to reduce and aluminum is easier to oxidize. Exact value of electrochemical potential of IMC components, i.e. Cu9Al4 and CuAl2 is unknown and it reported that it should be in between potential values of pure Cu and Al [Benedeitit 1995]. So theoretically based on electrochemical potential values, being most active metal in the system, Al would act as sacrificial anode, and undergo following reaction – Al + 3OH- Al(OH)3 + 3e- (1) Figure 24 shows Pourbaix diagram for Al, for 1e-6 concentration, 373°K. In regions where Al+++ is stable, corrosion is possible. In region where aluminum oxide is stable, resistance or passivity is possible. If the pH is between 4 and 8.3, Al2O3 is stable and thus protects the aluminum. Aluminum hydroxide forms stable passivation layer around bare Al pad. The pH values of most of the commercial electronic molding compounds falls in this range of 4 to 8.3. Passivation layer stability window might get narrow at in the presence of ionic contaminations, such as halide ions. With the industry migration to green molding compound, which are halide ion free or having very less halide concentration the possibility of narrowing of the stability window and the possibility corrosion of Al pad are lower.

During testing of samples in the current study, corrosion of wirebond was found during highly accelerated stress test in the Parr Bomb apparatus. Two main phases found in the IMC development study were Cu9Al4 and CuAl2. Cu9Al4 is on copper ball side, while CuAl2 was found on Al pad side of IMC. A higher fraction of Aluminum in Al rich phase forms strong passivation layer which stops further reaction, even at low pH values [Birbilis 2008, Osenbach 2013]. On the other hand, Cu9Al4 which is Cu rich IMC layer has lower Al fraction – which means it will have weaker passivation layer than CuAl2 and Al pad. This layer can be easily attacked by moisture and even small concentration of ionic contamination can trigger the corrosion reaction, making it corrosion prone. Detailed microstructure analysis of the aged samples revealed that IMC formed during bonding process was corroded. IMC layer formed near copper ball region (Cu rich interface) found to be corrosion prone than other IMC layer. Line scans and point scans showed that higher oxygen content i.e. corrosion followed by cracking was found in Cu-rich interface (Figure 21, Figure 22). There are different possibilities in which reaction may occur: Cu9Al4 + Cl2 4AlCl3 + 9Cu (2)

Cu9Al4 + 6H2O 2Al2O3 + 9Cu +6H2 (3) 2AlCl3 + 3H2O Al2O3 + 6HCl (4) Main byproducts of reaction are Al2O3 and HCl. Al2O3 is brittle oxide and can cause cracks in oxidized region. Hydrochloric acid further gets dissolved into moisture and breaks into chlorine ions. These chlorine ions again attack Cu9Al4 and continue corrosion process. The process continues and can eat away Cu-rich IMC phase, making it mechanically and electrically unstable, causing excessive stresses, and mostly results into crack formation, and propagation. The observations noted in this experiment are in good agreement with previous literature [Boettcher 2010; Breach 2010; Osenbach 2013; Liu 2011; Zeng 2013; Gan 2013]. Prior studies were limited to maximum 120 hours of aging. However, for 120°C and 130°C Parr bomb testing after 168+ hours of aging, it was found that along with Cu rich IMC, Al rich IMC and Al pad were also oxidized. This might be due to extreme conditions, and prolonged exposure to such environment, which might have caused degradation of molding compound, releasing some byproduct which can make reaction more aggressive. Initially Al will react with moisture to produce passivation layer of Al(OH)3. This can be attacked by chlorine ions and broken down as follows. 2Al(OH)3+Cl- 2Al(OH)2Cl + OH- (5)

Al + 4Cl- AlCl4- + 3e (6)

2 AlCl4- + 3H2O 2Al(OH)3 + 8Cl- (7)

One of the products of reaction is chlorine ion. So once the reaction initiates, it will keep on going and Al pad will undergo pitting corrosion. Typical corrosion reaction of CuAl2 can be described as following reaction [Osenbach 2013] CuAl2 + 6Cl- 2AlCl3 + Cu (8)

Levenberg-Marquardt Algorithm The relationship between the IMC growth parameter and time is nonlinear. Inverse solution for interrogation of system-state is challenging for damage evolution in such systems. Levenberg-Marquardt (LM) algorithm n iterative technique that computes the minimum of a non-linear function in multidimensional variable space has been used for identifying the solution in the prognostication neighborhood [Madsen 2004, Lourakis 2005, Nielsen 1999]. Let “f” be an assumed functional relation between a measurement vector referred to as prior damage and the damage parameter vector, p, referred to as predictor variables. The measurement vector is the current values of the leading-indicator of failure and the parameter vector includes the prior system state, and accumulated damage and the damage evolution parameters. An initial parameter estimate p0 and a measured vector x are provided and it is desired to find the parameter vector p, that best satisfies the functional relation “f” i.e. minimizes the

squared distance or squared-error, t. The minimize

parameter vector p, given by

)p(g)p(g2

1)))p(g(

2

1)p(F T

m

1i

2i

(9)

)p(g)p(J)p(F T' (10)

1980

m

1i

''ii

T'' )x(g)x(g)p(J)p(J)p(F (11)

Where )p(F represents the objective function for the squared

error term t, )p(J is the Jacobian, and )p(F' is the

gradient, and )p(F '' is the Hessian. The variation of an F-

value starting at “ p ” and with direction “ h ” is expressed as

a Taylor expansion, as follows: 2'T o)p(Fh)p(F)hp(F (12)

Where is the step-length from point “p” in the descent direction, “h”. Mathematically, “h” is the descent direction of

)p(F if 0)p(Fh 'T . If no such “h” exists, then

0)(' pF , showing that in this case the function is

stationary. Since the condition for the stationary value of the objective function is that the gradient is zero, i.e.

0)h(L)hp(F '' (13) The descent direction can be computed from the equation,

gJh)IJJ( TT (14)

The term is called as the damping parameter, 0ensures that coefficient matrix is positive definite, and this ensures descent direction. When the value of is very small,

then the step size for LM and Gauss-Newton are identical. Algorithm has been modified to take the equations of inter-metallic growth under isothermal aging to calculate the unknowns.

Calculation of Activation Energy for Cu-Al System Consistent with the acquired data on the Cu-Al IMC system, a square dependence of the IMC thickness on the aging duration has been assumed to calculate the activation energy.

CKtX2 (15) Where, X is the IMC thickness (μm), t is the aging time (s), K is the reaction rate of IMC formation (μm2/s), C is the constant related to initial IMC thickness (μm2) and, the reaction rate is represented as function of temperature:

RT

Q

0eKK

(16)

Where, 0K is the Multiplication Factor (μm2/s), R is the Gas

constant (1.99 cal/mol K), T is the Aging Temperature (K), Q is the Activation Energy (Kcal/mol). Using equation

(15), data on square of IMC thickness for the temperatures of 150°C, 175°C and 200°C has been fit versus versus aging time in seconds (Figure 25). Three values of growth rate i.e. k were obtained, and plotted against 1/T as shown in Figure 26. Activation energy calculated from the plot is 12.893 Kcal/mol, i.e. 0.559eV. Table 3 and Table 4 shows comparison of energy of activation and IMC growth rate with data previously published in literature. Activation energy obtained in current experiment is in good agreement of activation energy reported in [Goh 2013; Na 2011]. In all the papers, growth rate of IMC because of temperature bias is in

the range of 10-6 µm2/s to 10-8 µm2/s, which is also in good agreement with current study (10-7 µm2/s).

 Figure 25 – Intermetallic growth rates at different

temperatures.

 Figure 26 – ln(k) vs 1/T

Table 3 – Comparison of Activation Energies for Cu-Al

WB Type Reference Test Temperatures

(°C)

Q

(Kcal/mol)

Copper

[Goh 2013] 175, 200, 225 6.1 [Kim 2003] 150, 250, 300 26 [Na 2011] 150, 200, 250 10.71

Current Study 150, 175, 200 12.89

Table 4 – Comparison of IMC Growth Rate for Cu-Al System WB Type Reference

Temperature Temperature

(°C) Growth Rate

(µm2/s) Copper 150°C [Kim 2003] 1.88*10-8

[Na 2011] 2.15*10-8 Current Study 1.21*10-7

175°C [Goh 2013] 3.57*10-7 Current Study 3.25*10--7

200°C [Na 2011] 2.56*10-8 [Goh 2013] 6.26*10-7

Current Study 7.02*10-7

1981

Prognostication Approach for HAST In order to assess the accrued damage in the Cu-Al package, IMC growth has used as leading indicator of failure for interrogation of state of system and remaining useful life calculations. Measurements of IMC thickness growth have been fit into equation (17).

n1

0

01 kty

yy

(17)

Prior damage in each case has been prognosticated based on the IMC evolution. Prognostication involves withdrawal of three samples at three periodic intervals. The samples were then cross-sectioned, potted and polished to measure intermetallic thickness. Prior damage accrued was prognosticated. IMC growth parameter was used to compute life consumed due to exposure to thermal aging:

n1

0

01 kty

yy

(18)

n1

0

02 )tt(ky

yy

 

(19)

)t2t(ky

yy1

0

03

 (20)

Where, 1y is IMC thickness at time 1t , t is the time interval

at which future IMC thickness i.e. 32 y,y . The parameter 0y

is IMC thickness before initiation of thermal aging. The solution requires three equation and three unknowns. Levenberg- Marquardt (LM) algorithm was used to solve these three nonlinear equations (18), (19), and (20) and optimization of three unknowns. Figure 27 gives an overview of the methodology used for prognostication of Cu-Al wire bond for prior accrued damage using IMC layer as leading indicator of failure. Consider a field deployed electronic

package, which has been used for certain time, say 1t , which

is unknown. For prognostication, samples will be taken from field at uniform-interval of time t . IMC layer of all packages can be measured and then using equation (18), (19) and (20) providing measurements of y1, y2, y3. The LM algorithm is then used to solve for prior unknown damage.

 Figure 27 - Prognostication of thermally aged samples

In case of isothermal aging, in which IMC development is primarily driven by diffusion, value of “n” from equation (17)

is theoretically known to be 0.5 so, equation has been updated to,

5.01

0

01 kty

yy

(21)

5.01

0

02 )168t(ky

yy

(22)

5.01

0

03 )336t(ky

yy

(23)

Equation (17) was used to plot graph of normalized IMC against square root of aging time, for all three cases, shown in Figure 28. Linear plot implies that value of n in equation (17) is 0.5. Levenberg- Marquardt (LM) algorithm was used to solve these three nonlinear equations, and prediction of remaining useful life.

 Figure 28 – Normalized IMC vs Square root of aging time

Table 5 : Damage accrual Relationship Using IMC as Leading

Indicators Aging

Condition 150°C 175°C 200°C

Equation Kn= 0.023(t)0.482

Kn= 0.427(t)0.498

Kn= 0.045(t)0.513

Table 6 - Comparison of experimental and prognosticated

Results; for isothermal aging at 150°C Experimental

(Hours) Prognosticated

(Hours) % Error

1t 1176 1286 8.5%

2t 1344 1298 3.5%

3t 1512 1578 4.1%

Levenberg-Marquardt algorithm was developed based on equations (21), (22) and (23) for prediction of remaining useful life, when packages were subjected to isothermal aging conditions. Table 5 represents damage accrual relationships, where Kn is IMC thickness at any time t. The equations were derived from experimental data. Exponent of time “t” in all three cases is in the vicinity of 0.5.

1982

Figure 29 - 3D plot of error versus aging duration in hours versus normalized IMC thickness, for isothermal aging at 150°C, for time duration (a) 1176 hours, (b) 1344 hours, (c) 1512 hours.

Table 7 - Comparison of experimental and prognosticated results; for isothermal aging at 175°C.

Experimental (Hours)

Prognosticated (Hours)

% Error

1t 504 539 6.5%

2t 672 656 2.5%

3t 840 897 6.3%

Figure 30 - 3D plot of error versus aging duration in hours versus normalized IMC thickness, for isothermal aging at 175°C, for time duration (a) 504 hours, (b) 672 hours, (c) 840 hours.

It shows that IMC growth in Cu-Al wire bond system is driven by Fickian diffusion. For thermal aging at 150°C, prognostication was done at 1176, 1344, 1512 hours of thermal aging. Results of the prognostication are shown in Table 6. 3D plots of algorithm output are shown in Figure 29. Percentage error in between experimental aging duration and prognosticated aging duration was calculated. The error bound was in the range of 3.5% to 8.5%. The solution in Figure 29 corresponds to the point at minimum error. The minimum error point represents prognosticated aging duration. Similar approach was used to prognosticate prior accrued damage when packages were subjected to thermal aging of 175°C. Results are tabulated in Table 7 and 3D plot of output are shown in Figure 30. Percentage error bound for this model was in between 2.5% to 6.5%. The approach was extended to packages subjected to thermal aging at 200°C. Results are shown in Table 8 and 3D plots are shown in Figure 31. Percentage error in models prediction was in between 2.8% to 8.7%.

Table 8 - Comparison of experimental and prognosticated

Results; for isothermal aging at 200°C. Experimental

(Hours) Prognosticated

(Hours) % Error

1t 288 265 8.7%

2t 336 361 7.0%

3t 384 395 2.8%

Figure 31 - 3D plot of error versus aging duration in hours versus normalized IMC thickness, for isothermal aging at 200°C, for time duration (a) 288 hours, (b) 336 hours, (c) 384 hours.

Conclusions Microstructure evolution of Cu-Al intermetallics when subjected to thermal aging has been studied in this paper. Two distinct phases were found, namely Cu-rich interface (Cu9Al4) and Al rich interface (CuAl2). Complete cracking of wirebond was found after prolonged thermal aging at 175°C and 200°C.

1983

Crack initiates from edges of ball bond, and propagates towards center. Cu-rich IMCs were found to be corrosion at the Cu-Al interface. A method has been developed for prognostication of accrued prior damage and remaining useful life after exposure to thermal aging. The presented approach uses the Levenberg-Marquardt Algorithm in conjunction with development of damage based leading indicator for estimating prior accrued damage. Specific damage proxies examined is the intermetallic thickness in Cu-Al wire bond. The viability of the approach has been demonstrated 32 pin chip-scale package without any prior knowledge aging duration, subjected to thermal aging at 150°C, 175°C, and 200°C. The prognosticated values have been validated versus experimental data. Correlation between the prognosticated damage and the actual accrued damage demonstrates that the proposed approach can be used to assess prior damage accrued because of aging.

Acknowledgments The work presented here in this paper has been supported by a research grant from the Semiconductor Research Corporation (SRC), Research ID 2284.

References Appelt, Bernd K., Andy Tsenga, Chun-Hsiung Chenb, Yi-

Shao Lai. “Fine pitch copper wire bonding in high volume production”, Microelectronics Reliability 51, pp.13- 20, 2011

Benedeitit, A.V., P.T.A. Sumodjo, K. Nobe, P. L. Cabot and W. G. Proud, “Electrochemical Studies of Copper, Copper-Aluminium and Copper-Aluminium-Silver Alloys: Impedance in 0.5M NaCl”, Electrochimica Acta 1995. Vol. 40, pp. 2657-2668.

Birbilis, N., and R.G. Buchheit, “Investigation and Discussion of Characteristics for Intermetallic Phases Common to Aluminum Alloys as a Function of Solution pH”, Journal of The Electrochemical Society 155, C117- C126 (2008).

Boettcher, T., Michael Rother Stefan Liedtke, Mandy Ullrich, Marc Bollmann, Andreas Pinkernelle, Daniel Gruber, Hans-Juergen Funke, Michael Kaiser, Kan Lee, Martin Li, Karina Leung, Tina Li, Mark Luke Farrugia, Orla O’Halloran, Matthias Petzold, Benjamin März, Robert Klengel “On the Intermetallic Corrosion of Cu-Al wire bonds”, Proceedings of 12th Electronics Packaging Technology Conference, pp 585-590 2010.

Breach, C. D., Ng Hun Shen ; Tee Wai Mun ; Teck Kheng Lee. “Effects of Moisture on Reliability of Gold and Copper Ball Bonds”, Proceedings of 12th Electronics Packaging Technology Conference, pp 44-51, 2010.

Breach, C. D., What is the future of bonding wire? Will copper entirely replace gold?, Gold Bulletin, Volume 43, No 3, pp- 150-167, 2010.

Chandramouli, R., Pateras, S., “Testing Systems on a Chip”, IEEE Spectrum, Vol. 33, No. 11, pp. 42-47, Nov. 1996.

Chen, J., Dominiek Degryse, Petar Ratchev, Ingrid De Wolf, “Mechanical Issues of Cu-to-Cu Wire Bonding”, IEEE Transactions on Components and Packaging Technologies, Vol. 27, No. 3, pp.539-545, September 2004.

Chen, Jiunn, Yi-Shao Lai, Yi-Wun Wang, C.R. Kao. “Investigation of growth behavior of Al–Cu intermetallic

compounds in Cu wire bonding”, Journal of Microelectronics Reliability, Vol 51, pp 125–129, 2011.

Chong L.G.., Francis Classe, Bak Lee chan, Hasjhim Uda, “Extended Reliability of gold and copper ball bonds in microelectronic packaging”, gold Bulletin, June 2013, Volume 46, Issue 2, pp 103-115.

Drees, R., and Young, N., Role of BIT in Support System Maintenance and Availability, IEEE A&E Systems Magazine, pp. 3-7, August 2004.

England, L. Siew Tze Eng,, Chris Liew, Hock Heng Lim. “Cu wire bond parameter optimization on various bond pad metallization, and barrier layer material schemes”, Microelectronics Reliability, Volume 51, No.1, pp.81-87, 2011

Facility for the Analysis of Chemical Thermodynamics. http://www.crct.polymtl.ca/factweb.php

Gao, R. X., Suryavanshi, A., BIT for Intelligent System Design and Condition Monitoring, IEEE Transactions on Instrumentation and Measurement, Vol. 51, Issue: 5, pp.1061-1067, October 2002

Goh, Chwee Sim, Wee Ling Eddy Chong, Teck Kheng Lee and Christopher Breach. “Corrosion Study and Intermetallics Formation in Gold and Copper Wire Bonding in Microelectronics Packaging”, Crystals Journal, Vol 3, pp 391-404, 2013.

Hang, C.J., C.Q. Wang, M. Mayer, Y.H. Tian, Y. Zhou, H.H. Wang “Growth behavior of Cu/Al intermetallic compounds and cracks in copper ball bonds during isothermal aging”, Journal of Microelectronics Reliability, Vol 48, pp 416–424 2008.

Hassan, A., Agarwal, V. K., Nadeau-Dostie, B., Rajski, J., BST of PCB Interconnects Using Boundary- Scan Architecture, IEEE Transactions on Computer-Aided Design, Vol. 11, No. 10, pp. 1278-1288, October 1992

Kim, Hyoung-Joon, Joo Yeon Lee, Kyung-Wook Paik, Kwang-Won Koh, Jinhee Won, Sihyun Choe, Jin Lee, Jung-Tak Moon, and Yong-Jin Park “Effects of Cu/Al Intermetallic Compound (IMC) on Copper Wire and Aluminum Pad Bondability”, IEEE Transactions On Components And Packaging Technologies, Vol. 26, No. 2, June 2003

Laik, A., K. Bhanumurthy, G.B Kale. “Diffusion in Cu(Al) Solid Solution”, Defect and Diffusion Forum Vol. 279, pp 63-69, 2008.

Lall, P., Bhat, C., Hande, M., More, V., Vaidya, R, Pandher, R., Suhling, J., Goebel, K., Interrogation of System State for Damage Assessment in Lead-free Electronics Subjected to Thermo-Mechanical Loads, Proceedings of 58th ECTC conference pp. 918-929, 2008.

Lall, P., Hande, M., Bhat, C., More, V., Vaidya, R., Suhling, J., Algorithms for Prognostication of Prior Damage and Residual Life in Lead-Free Electronics Subjected to Thermo-Mechanical Loads, Proceedings of the 10th ITherm, pp. 638-651, May 28-31, 2008.

Lall, P., M. Hande, C. Bhat, J. Suhling, Jay Lee, Prognostics Health Monitoring (PHM) for Prior-Damage Assessment in Electronics Equipment under Thermo-Mechanical Loads,Proceedings of 57th ECTC conference, pp. 1097-1111, 2007c.

1984

Lall, P., Mahendra Harsha, Jeff Suhling, Kai Goebel. “Sustained damage and remaining useful life assessment in lead-free electronics subjected to sequential multiple thermal environments”, electronics components and technology conference 2012, pp. 1-14, 2012a

Lall, P., Mahendra Harsha, Kai Goebel, “Level of Damage and Remaining Useful Life Assessment in Lead-free Electronics Subjected to Multiple Thermo-mechanical Environments”, IEEE International Conference on Prognostics and Health Management, pp.1-14, 2012b

Lall, P., Mahendra Harsha, Kai Goebel, Jim Jones. “Interrogation of Thermo-Mechanical Damage in Field-Deployed Electronics”, IEEE International Conference on Thermal, Mechanical and Multi-Physics Simulation and Experiments in Microelectronics and Microsystems, pp.1-16, 2012c

Lall. P., More, V., Vaidya, R., Goebel, K., “Assessment of Residual Damage in Lead-free Electronics Subjected to Multiple Thermal Environments of Thermal Aging and Thermal Cycling”, Proceedings of 60th ECTC, pp. 206- 218,June 2-5, 2010b.

Lall. P., More, V., Vaidya, R., Goebel, K., “Prognostication of Latent Damage and Residual Life in Lead-free Electronics Subjected to Multiple Thermal-Environments”, 59th ECTC,pp. 1381-1392, 2009c

Lee, Won-Bae, Kuek-Saeng Bang, Seung-Boo Jung. “Effects of intermetallic compound on the electrical and mechanical properties of friction welded Cu/Al bimetallic joints during annealing”, Journal of Alloys and Compounds, Vol 390, pp 212–219 2005.

Levenson, L. L., “Grain boundary diffusion activation energy derived from surface roughness measurements of aluminum thin films”, Applied Physics Letters, Vol 55, pp 2617-2619, 1989.

Liu, Hai, Zhenqing Zhao, Qiang Chen, Jianwei Zhou, Maohua Du, Senyun Kim, Jonghyun Chae, Myungkee Chung “Reliability of Copper Wire Bonding in Humidity Environment”, Proceedings of 13th Electronics Packaging Technology Conference, pp 53-58, 2011.

Marko, K.A., J.V. James, T.M. Feldkamp, C.V. Puskorius, J.A. Feldkamp, and D. Roller, Applications of Neural Networks to the Construction of "Virtual" Sensors and Model-Based Diagnostics, Proceedings of ISATA 29th International Symposium on Automotive Technology and Automation, pp.133-138, June 3-6, 1996.

McCann, R. S., L. Spirkovska, “Human Factors of Integrated Systems Health Management on Next-Generation Spacecraft, First International Forum on Integrated System Health Engineering and Management in Aerospace”, Napa, CA, pp. 1-18, November 7-10, 2005.

Na, Seok Ho, TaeKyeong Hwang, JungSoo Park, JinYoung Kim, HeeYeoul Yoo and ChoonHeung Lee “Characterization of Intermetallic Compound (IMC) growth in Cu wire ball bonding on Al pad metallization”, Proceedings of 2011 IEEE Electronic Components and Technology Conference, pp 1740-1745, 2011.

Osenbach, John, Wang, B.Q. ; Emerich, S. ; DeLucca, J. “Corrosion of the Cu/Al Interface in Cu-Wire-Bonded Integrated Circuits”, Proceedings of IEE Electronic

Components & Technology Conference, pp 1574-1586, 2013.

Qin, I., A., Shah, C. Huynh M. Meyer, M. Mayer, Y. Zhou. “Role of process parameters on bond ability and pad damage indicators in copper ball bonding”, Microelectronics Reliability 51, pp.60-66, 2011.

Rosenthal, Daniel, Brian C. Wadell, “Predicting and Eliminating Built-In Test False Alarms”, IEEE Transactions On Reliability, Vol. 39, No. 4, pp 500-505, 1990.

Schauz, J. R., Wavelet Neural Networks for EEG Modeling and Classification, PhD Thesis, Georgia Institute of Technology, 1996.

Schmitz, S., M. Schneider-Ramelow, S. Schröder, “Influence of bonding process parameters on chip cratering and phase formation of Cu ball bonds on AlSiCu during storage at 200C”, Microelectronics Reliability 51, pp.107-112, 2011

Shah A, Mayer M, Zhou Y, Hong SJ, Moon JT. “Low-stress thermo sonic copper ball bonding”, IEEE Trans Electronics Packaging Manufacturing (32) 2009.

Shiroishi, J., Y. Li, S. Liang, T. Kurfess, and S. Danyluk, “Bearing Condition Diagnostics via Vibration and Acoustic Emission Measurements”, Mechanical Systems and Signal Processing, Vol.11, No.5, pp.693-705, Sept. 1997.

Simonovic, Darko, Marcel H. F. Sluiter, “Impurity diffusion activation energies in Al from first principles”, Journal of The American Physical Society- Physics Review, pp 1-12, 2009.

Su, Peng, Hidetoshi Seki, Chen Ping, Shingo Itoh Louie Huang, Nicholas Liao, Bill Liu, Curtis Chen, Winnie Tai, and Andy Tseng. “Effects of Reliability Testing Methods on Microstructure and Strength At the Cu Wire-Al Pad Interface”, Proceedings of Electronic Components & Technology Conference, pp 179-185, 2013.

Tian, Y.H., C.J. Hang, C.Q. Wanga, G.Q. Ouyanga, D.S.Yanga, J.P. Zhao. “Reliability and failure analysis of fine copper wire bonds encapsulated with commercial epoxy molding compound”, Microelectronics Reliability, volume 51, pp 157–165, 2011.

Wieczorek-Ciurowa, K., K. Gamrat, Z. Sawlowicz “Characteristics Of Cual2–Cu9al4/Al2O3 Nanocomposites Synthesized By Mechanical Treatment”, Journal of Thermal Analysis and Calorimetry, Vol. 80, pp 619–623 2005.

Zeng, Yingzhi, Kewu Bai, Hongmei Jin “Thermodynamic study on the corrosion mechanism of copper wire bonding”, Microelectron Reliab (2013), http://dx.doi.org/10.1016/j.microrel.2013.03.006.

Zorian, Y., Hakim Bederr, A Structured Testability Approach for Multi Chip Boards Based on BIST and Boundary Scan, IEEE Transactions on Components, Packaging, and Manufacturing Technology-Part B, Vol. 17, No. 3, pp. 283- 290, August 1994.

1985