Current Trend of Thermosonic Bonding

The wire material used in thermosonic bonding has been dominated by the use of gold (Au), be it 99.99 % (4N) purity or lesser purity of 99.9 % (3N) and 99 % (2N) until recent years. Almost all thermosonic ball bonders designed and manufactured for mass production of integrated chip (IC) device assembly before mid- to late 1990 were for Au wire only. The majority of size used in device assembly has been around 25–30 μm (1–1.2 mil) in diameter. The decision on wire diameter followed the ITRS wafer technology node road map for ICs and the associated wire bond interconnect pad pitch trend. As wafer technology node evolved into smaller dimension, the size of the wire also reduced accordingly to ensure that more wires can be packed into the device. Before 2008, the industry had been working on reducing Au wire diameter for fine technology node as well as for cost reduction. Figure 10 shows how much cost can be saved with reducing wire diameter. Just by reducing the wire diameter from, say 25 to 23 μm(1 to 0.9 mil), a 16%cost savings can be realized. Compared to, however, the cost of using copper (Cu), which is just a tiny fraction of today’s Au price, the cost savings by wire diameter reduction is insignificant.

With the continued increase of Au price over the last 10 years, semiconductor package assembly houses have to consider lower cost of bonding wire materials in order to remain competitive in the market, especially high-density devices that use more than a hundred or even a thousand wires. Gold price increased from around US$300 per troy ounce in the beginning of 2002 to well over US$1700 10 years later, and at one point of time, it even reached close to US$1900 in September 2011 (Fig. 11). Using a lower-cost material, such as Cu, can help the assembly houses save millions of dollars due to the huge difference in unit cost of the material (less than one US dollar per oz).
Copper wire had been used for many years in power devices where thick wires in the dimension of 50 μm or larger have been used and the number of wires per device was few. These wires are mainly used with ultrasonic bonding process in which no ball formation is required. However, the economic tsunami in 2008–2009 period had driven the use of fine diameter Cu wire to be used to replace Au wire for major cost savings in IC packages. The size of Cu wire started at 25 μm (1 mil) to replace the more expensive Au material of the same diameter used in low-pin count IC devices such as SOP, TSOP, etc., but the trend did not stop there and more conversion from Au to Cu happened to all wire diameters down to 20 μm (0.8 mil) as well as 18 μm (0.7 mil) in recent years.
When processing Cu bonding wire, changes to the design of ball bonder must be carried out because Cu being an easily oxidized material when heated in the presence of oxygen, a protection environment must be provided during the melting of the wire to form a ball before bonding onto the IC pad. Figure 12 shows an EFO (electric flame off) wand design for Cu wire bonder. For Au wire, there is no need for oxidation protection due to its inertness to environment. EFO wand is visible underneath the tip of the wire. Hence the kit, which can be of glass or ceramic, is not needed and the Au wire is exposed in air during processing. However, for Cu, depending on the machine maker and design model, an EFO kit must be used so as to allow the protection gas to be purged to protect the Cu ball during formation to be oxide-free. The common protection gas is a mixture of nitrogen and hydrogen in a ratio of 95:5 (forming gas), although for Pd-coated Cu wire, nitrogen alone can be used. The kit can be an enclosure type or open concept as long as the tip of the wire is inside the gas region during ball formation. The EFO assembly shown in Fig. 12 is in a Kulicke & Soffa (K&S) Cu wire bonder. The rectangular white object is the EFO tube which has an enclosed feature allowing better protection to the tip of the wire which is located inside the flow path of forming gas.

Properties of Thermosonic Bonding Wires

The elements, Au, Cu, Ag, and Al, in wire form with low electrical resistivity – 1.63 to 2.7 μΩ.cm – are commonly used for interconnections (Harman 1997). These four elements are face centered cubic in crystal structure with different atomic properties (Cullity 1978). Moreover, for a thermosonic ball-wedge bonding, a wire has to satisfy the essential factors: (a) free air ball (FAB) formation, (b) deformation of the FAB and easy welding to Al bond pad, (c) loop formation, (d) deformation of wire and easy welding to lead fingers, and (e) reliability of ball and wedge bonds (first and second bond) to attain good electrical interconnections. Among these four elements, the inert Au satisfies all the five vital factors and has been used for three decades in the field (Harman 1997; Prasad 2004). Next to Au, Cu wire bonding technology has matured, especially in the formation of FAB under preventive environment. For niche applications, Ag wire bonding is widely used in the industries, while Al wire bonding is still in developmental stage to find a suitable method for the formation of a good FAB. The following sections mainly discuss the salient engineering advancement in the field of Au and Cu wire bonding.

Composition of Bonding Wires

High pure Au (5N, 99.999 %) is extremely soft, and for a successful wire drawing and bonding, it is usually doped (Qi and Zhang 1997; Chew et al. 2004) with Ca, Cu, Be, Pd, Pt, Ag, Ce, etc., in parts per million (ppm). In addition, the annealing process design of 4N and 3N purity Au wire produces a stiff and soft wire suitable for bonding, particularly for looping. Later, 2N purity Au wire, a high-reliability product without any Kirkendall voids, has been used. Electrical grade oxygen-free high conductivity (OFHC) 4N Cu has been used for bonding. Fine wires made using 5N and 6N purity Cu are examined in the view of developing a softer Cu wire than 4N Cu.Alloyed Cu with1N purity (>90 %) with certain elemental addition and Pd-coated Cu (PCC) wire with 4N purity Cu core and 4Npurity-plated Pd are examined for better bondability and reliability (specifically biased HAST/THB) performances.

Processing of Bonding Wires

Wire processing (Zompi et al. 1991; Camenschi and Sandru 1980) is an established technique since the 1980s. In general, the processing of bonding wire sequences is as follows: melting the raw material under preventive atmosphere/vacuum, continuous casting to thick rods, heavy wire drawing at room temperature, intermediate annealing, final wire drawing at room temperature with stringent die management, and final annealing and spooling in a clean room environment.
At present, bonding wires using Au, Cu, Ag, and Al are wire drawn at room temperature until 10 μm diameter. The manufacturing yield loss is high on drawing the wires in the diameter range of 10–15 μm. Compared to electrical grade Cu wires in the market, the bonding wires have to be processed with high quality and tight tolerances, as it is guided via a capillary during wire bonding processes.

Properties of Bonding Wires

Physical, thermal, electrical, and mechanical properties of the bonding wires are tabulated in Table 1. The tensile curves of Au, Cu, PCC, Ag, and Al wires are shown in Fig. 13. The tensile test was performed using Instron 5,400 series model. As Au is malleable, the fine wires are annealed between 2%and 6%elongation (EL) to suit for wire bonding. Fine Cu wire (standard (Std)) is harder and stiffer than Au and generally annealed to a higher EL of 12–16 %. Soft Cu wire developed by special processing technique requires low load to deform under tensile (Fig. 13).Also, the modulus of soft Cu wire remains the same as that of Au. Thus, the soft Cu wire gains popularity and is being used nearly 50 % of market shares among the bare Cu wires available in the bonding wire manufacturing industries. PCC wire is annealed similar to Cu, and Ag wire is annealed similar toAuwire. The low melting point of element Al, which forms a stable surface oxide film, has to be annealed in batches and not continuous strand annealed like the other bonding wires with a high-temperature melting point.

Table 1 provides the measured resistivity of bonding wires, calculated using the basic formula ρ = R.A/L, where R is the measured resistance in ohms using an ohmmeter connected to a fixture, A is the cross-section area of the wire, and L is the measured length of the wire. Fusing current is the maximum current carrying capacity of a wire, above which the wire fuses (melt). The fusing current is measured at room temperature and atmospheric conditions (Table 1). Factors such as wire diameter, wire length, wire type (Au, Cu, Ag, Al), and input pulse current strongly influence the fusing current. For instance, increase in wire diameter or decrease in wire length increases the fusing current (Fig. 14). Literature (Krabbenborg 1999; Mertol 1995; Loh 1983) reports few methods to calculate fusing current of fine metal wires. Trend of the predicted and measured reading is similar for varying wire diameter, length, and type, yet the measured value is 10 % lower than the predicted one.

Floor and Shelf Life of Bonding Wires

End users consider floor and shelf life of bonding wires as an important factor. Gold is inert, and if it is stored inside the plastic holder in a dirt-free environment, the floor and shelf life is generally practiced as 3 months and 1 year, respectively. No rigorous floor and shelf life studies are being conducted for Au wires. Copper is prone for oxidation and is usually sealed under vacuum. It is recommended to be stored away from corrosive industrial atmosphere, such as humid and chlorine-rich coastal atmosphere. Similarly, Ag is prone to corrode if the atmosphere is rich in sulfur. Moreover, Al wire with stable oxide layer is much safer than Cu and Ag wire, yet it is possible to corrode if the environment is corrosive.
Floor life of a wire is defined as how long the wire can be used after opening the vacuum seal, and shelf life is defined as the period that the vacuum-sealed wire spools can be stored and used later for bonding. The annealed wires are stable and expected to have no significant changes in microstructures and tensile properties. Therefore, wire surface oxidation and bonding performance are studied in detail. For instance, wire surface oxidation and stitch pull evaluation of 0.8 mil soft Cu wire tested for 90 days after opening the vacuum seal and 6 months vacuum-sealed conditions are shown in Fig. 15. The recommended floor and shelf life of soft Cu wire is 7 days and 6 months, respectively. Similarly, the floor and shelf life of Ag and Al wires are recommended by the manufacturers.

Surface Characterization of Bonding Wire, Bond Pad, and Lead Finger

The roughness and oxide layer of the two joining surfaces are the two factors that influence a good bonding of a wire to bond pad or lead finger to establish an interconnection. This section provides data on the surface characterization of wire/pad/finger for a good bonding.

Surface Oxidation of Bonding Wire

Except Au, almost all metals readily get oxidized when exposed to air and so as Cu. ECI Technology has designed and developed an equipment, QC200, which measures oxide layer thickness by sequential electrochemical reduction analysis (SERA) technique (Tench 1994). The measurement is based on Faraday’s principle of electrolytic reduction of a cell. Surface Cu oxides are reduced with hydrogen ions in a cell consisting of reference and auxiliary electrodes, borate electrolyte, and specimen. The cathode potential is constant for each type of surface film (CuO, Cu2O). Oxide film reduction and its thickness are proportionate to surface area of Cu wire sample, current flow, potential difference, electrolyte, etc. The SERA technique is accurate enough to measure the copper oxide layer thickness in A. Reduction of Cu2O or CuO occurs at a unique voltage range, and the time taken for the reaction is used to calculate the thickness of the layer as per Faraday’s equation:

where I is current in ampere, t is reduction time measured in sec, M is molecular weight in g, F is Faraday’s constant (96,498 coulombs), n is number of electrons, S is surface area of immersed Cu section in cm2, and d is density of film in g/cm3.
The annealed 4N Cu wires in normal production and laboratory environment showed neither presence of Cu2S nor detectable. On the other hand, CuO is detected when Cu wire is exposed to harsh oxidizing atmosphere (Fig. 16). The system is also capable to measure AgO or Ag2S on Ag wire surface.
As a quality control measure, every batch of final annealed Cu wire is examined for dirt and loose wires under low power scope (1–10) as well as surface oxidation using SERA. For a good Cu wire bonding, the surface oxide (Cu2O) layer thickness is observed to be ~20 A°. A typical voltage versus time plot of a reduction of 0.8 mil Cu wire surface is shown in Fig. 16. Observation of the wire surface in scanning electron microscopy (SEM-EDX analysis) at high magnification (200–10,000x) revealed a clean, contamination-free surface with no traces of oxygen.

Surface Contaminants and Oxidation of Bond Pad

Metals exposed to atmosphere at room temperature form an oxide film thickness of 20–100 A° . Auger analysis of Al-metallized bond pad surfaces clearly revealed an oxygen peak. Depth profile analysis using secondary ion mass spectroscopy (SIMS) on bare Cu lead frame surface revealed presence of oxygen until 2.4 nm thickness (Murali et al. 2007). Both Al and Cu surfaces possess good bondability. This indicates that the aluminum/copper oxide films do exist on the surfaces before joining. However, they are broken in the course of wire bonding for a successful welding. Hard oxide layers such as aluminum oxide (alumina, 1,800 VHN) seen on Al are comparatively more difficult to break than soft metallic oxide layers (Table 2) (Murali et al. 2007). During wire bonding, supplied ultrasonic energy displaces the capillary by ~8 μm; hence, the FAB compressed by capillary first scrubs the bond pad surface as well as breaks the oxide layers (aluminum oxide, copper oxide, etc.) of the contact surfaces. The oxide film on Cu FAB and wire surface have to be broken in order to weld. In addition, contaminants such as deposits of carbon, fine dirt, epoxy outgassing, substrate outgassing, fluorine, sulfur, etc., are generally found on the metallized bond pad (Al, Au) and plated lead frame finger (Ag, Cu, Pd, Au) surfaces (Murali et al. 2007). Deposition of the outgassed compound may cause nonstick on lead (NSOL).

Surface Roughness and Plastic Deformation of Asperities of Bondable Surfaces

Surface roughness (Ra) of Al- and Au-metallized bond pad surfaces is within the range of 0.02–0.09 μm, and bonding to this smooth surface does not cause a problem with either Au or Cu ball bonds. Ra measured for plated lead frame finger surfaces varies between 0.02 and 1.4 μm. The surface profile of Al bond pad and Ag-plated lead frame measured using 3-D laser microscope are shown in Fig. 17. Au-, Cu-, Ag-, and Pd-plated surfaces with Ra of 0.2–1 μm revealed good bonding. For better grip of epoxy molding compound to lead finger surface, rough lead finger surface is preferred to avoid delamination. But increase in Ra greater than 1 μm may cause nonstick on lead (NSOL). Considering bare Cu lead frame, Ra greater than 0.15 μm causes NSOL. Ra measured by noncontact method on Cu FAB spherical surface and wire surface revealed around 0.17–0.55 μm Ra (Murali et al. 2007). The reported data are measured for the devices used in the industries.

Surfaces assumed to have hills and valleys and the asperity contact are important for joining. Deformation at the asperity contact can be elastic or elastic–plastic. Welding is not possible below the threshold deformation of asperity. Table 2 consolidates the hardness and tensile strength of some pure metals, threshold deformation of certain metals, and hardness of some oxides. The threshold deformation for Al is 40 %, while to crack the aluminum oxide film, 1 % or less deformation is required (Murali et al. 2007). Hence, oxide film can be easily broken by the supplied energies when two surfaces contact with each other. Grain orientation and surface microstructure may also affect the bondability, especially for bare Cu surface.

Thermosonic Bonding Process

Wire bonding process involves four important steps:

(a) Axisymmetric spherical and consistent FAB formation without off-center
(b) First bond (ball bond) bondability, FAB to deform, weld, and possess intact contact with the bond pad without any irregularity
(c) Looping – bonded wire between ball and stitch bonds to be stiff and not to sag during post processing to avoid shorting with adjacent or upper/lower tier wires. It also has to satisfy the dimensional tolerances for long loop, low loop, high loop, reverse bonding, leaning, swaying, loop height, fine-pitch bonding, etc.
(d) Second bond (“stitch” or “wedge” bond) bondability, the wire to deform, weld, and possess integral contact with the lead finger. Create a consistent tail end on breaking the wire at stitch.

For the last three decades, Au wire has satisfied the abovementioned steps for the diameter from 15 to 75 μm (0.6–3 mil). With hike in Au price, industries are forced to find an alternate to Au wire, which must satisfy the above steps to enter the family of bonding wires. Undoubtedly, bare Cu, alloyed Cu, and PCC wires started replacing Au wire (Breach 2010; QiJia et al. 2010; Zhong 2009). Industries have started to investigate Ag wire for niche applications. This section discusses the advances in the engineering of wire bonding.

FAB Formation

Electric spark to the tail end of the wire melts a controlled length of it, as the temperature of the wire reaches higher than the melting point. Due to the surface tension, the molten pool rolls up into spherical ball shape and then solidifies (Huang et al. 1995). This is termed as “free air ball (FAB)” (Fig. 18a). In principle, balance between heat and mass transfer creates the FAB. Heat transfer is predominantly by thermal conduction at the neck between the ball and wire regions. Secondary mode of heat transfer may be due to thermal convection and radiation by air cooling. The shape-up of spherical molten pool satisfies the minimum surface area criteria. Rapidly solidified FAB in the atmosphere is axisymmetrical to the wire axis (Fig. 18a). The ratio of FAB to wire diameter is usually maintained between 1.6 and 2.2. The processing parameters that govern the formation of FAB are:

(a) Electric flame off (EFO) firing current
(b) EFO time
(c) EFO voltage
(d) EFO wand gap (gap between the wire tip and electrode tip)
(e) Morphology of the wire tip
f) Gas flow rate, for bonding wires other than Au
(g) Plating thickness, for coated bonding wires

The factors that influence the FAB formation are:

(1) Reaction of the molten FAB surface to its environment
(2) Balance of surface tensional force on the molten FAB surface
(3) Gravitational force
(4) Composition of the wire
(5) Density of the wire
(6) Thermal conductivity of the wire
(7) Electrical conductivity of the wire

Au wire is inert; on melting no sign of oxidation is observed on the FAB surface, thus balancing the gravitational force (weight of FAB) and the surface tension at the solid–liquid interface of the FAB. Certain elemental addition in weight ppm can cause apple bite structure in Au FAB (Fig. 18b). During alloy design of bonding wires, study of FAB formation has to be considered as an utmost factor to qualify the wire.

Copper FAB

In the last few years, industries have moved to Cu wire bonding on a large scale of manufacturing. Electrical firing of Cu wire at the atmosphere leads to severe surface oxidation (Fig. 18c) and mostly absence of FAB formation. On purging with reactive forming gas (N2:H2, ratio of 95:5), a good Cu FAB is formed. An axisymmetrical Cu FAB formation depends strongly on gas flow rate besides  EFO processing parameters. Nominal forming gas flow rate required to form axisymmetrical FAB is 0.5–0.6 lpm. Gas flow rate greater than 1 lpm causes pointed FAB (Fig. 18d) due to rapid solidification, and gas purging lesser than 0.3 lpm creates surface oxidation and pointed FAB (Fig. 18e). The rapid solidification of high volume of purging forming gas exhibits forced convection of heat transfer of Cu FAB. Now and then, at the neck region, few ring patterns are observed (Fig. 18f), perhaps molten ball roll-down in steps. Though reason is unknown, it could be due to a slight fluctuation in gas flow velocity in the Cu kit causing the imbalance between gravitational force and surface tension.

From the cross-sectioned microstructure, pore-free FAB is apparent (Fig. 19a).  SEM observation by backscattered mode of etched FAB is shown in Fig. 19b, where surface grain boundaries are clear. Columnar grain growth inside the FAB and wire-deformed structures are observed for both 4N Au and 4N Cu wires, but no distinct heat-affected zone (HAZ) is observed for Cu wire (Fig. 19c, d). A typical grain orientation of 4N Cu FAB is shown in Fig. 19e with mostly {101} plane of orientation. Texture analysis of FAB formed using 3N, 4N, and 5N Cu wires exhibited variation in plane of crystallographic orientation of grains (Srikanth et al. 2007). High-purity 5N Cu wire showed FAB grains formed with mixed modes of {111} and {001} plane of orientation, while 3N and 4N Cu wires revealed majority of the FAB grains formed in the mode of {001} plane of orientation (Srikanth et al. 2007). Published literature on the texture analysis of wire and FAB is minimal. Nucleation and growth of grains in the FAB depends on the presence of impurity elements and its concentration. Residual stress, density of dislocation pileup, and type of atomic defects in the deformed wire due to the presence of impurities are the source of grain nucleation and hence resulted in the change of FAB texture. Differences in the FAB hardness are expected when the FAB texture varies.

Pd-Coated Cu (PCC) FAB

Considering the FAB microstructure of PCC wire, conveniently PCC wire can be categorized as a next-generation bonding wire. PCC wire can form an axisymmetrical spherical FAB with the purging gases, inert N2 or reactive forming. A good FAB is attained by purging 0.3–0.4 lpm of N2 or forming gas. On electric sparking, the core Cu melts and rolls up to form FAB; it is expected that the coated Pd be dissolved into the molten Cu as Cu–Pd has a good dissolution (Cu–Pd is solid–solution mixture and forms a homogeneous alloy (ASM Handbook 1995)). Surprisingly, Pd with high melting point (1,554o C) than Cu (1,083o C) has neither dissolved into the molten Cu FAB nor wetted and covered the FAB. Instead, Pd moved up to the top and then segregated within the neck region of FAB (Fig. 20a). This indicates that within a few milliseconds of time of FAB formation, nanometer thick coated Pd layer has poor dissolution into the molten Cu. Furthermore, coated Pd has melted, non-wetted with molten Cu, shaped along with the FAB and accumulated along the top of FAB periphery (near to wire neck region).
The macro observation of the PCC FAB using an optical scope showed bluishsilvery color for Pd areas and reddish-brown color for exposed Cu regions (Fig. 20a). The cross-sectioned FAB also revealed segregation of Pd within the FAB closer to wire neck region (Fig. 20b). The optical color contrast images with bluish (bluish-silvery) color on EDX exhibited rich in Pd. Some FAB revealed swirl in bluish color, thus indicating Pd segregates in a swirl form (Fig. 20c–e). This swirl reveals a possibility of the presence of two immiscible liquid pools (i.e., high volume of molten Cu non-wetted with low volume of molten Pd) leading to the composite microstructure of swirl Pd in a Cu FAB on solidification. Interestingly, the unequal density of molten Pd and molten Cu causes different speeds of motion during roll-up into a ball due to gravitational effects, resulting in swirl form. The balance of gravitational force and surface tension by the two unequal density immiscible molten pools generate an axisymmetrical spherical PCC FAB with composite microstructure.

A progressive sectioning of PCC FAB showed segregation of Pd (bluish area) in the outer area of the FAB and is absent towards the inner core of the FAB (Fig. 21). Varying FAB processing parameters can improve Pd distribution especially on the lower hemisphere of the FAB surface. But a complete (100 %) coverage of Pd distribution cannot be achieved (Lin et al. 2012; Tang et al. 2010). With an increase in EFO current, bluish layer at the neck is reduced and hardness of the PCC FAB is increased, indicating that Pd dissolution into Cu is much improved leading to PCC FAB microstructure with less in composite structure.
On etching the PCC FAB using Cu etchant showed unetched layer along the periphery of FAB indicating the dissolution of Pd into Cu to a few micron thickness (Fig. 21 E-E). However, this Pd-enriched Cu peripheral layer (solid solution of Cu–Pd) of FAB cannot be considered as Pd-coated Cu FAB, which is the demand of wire bonding industries. Globally, wire bonding end users aim to get a uniformly coated Cu FAB with Pd, which would form a Pd line at the interface between the Cu wire bond and Al bond pad. But in reality, Pd segregates at the neck of PCC FAB. This constitutes a major disadvantage of PCC FAB leading to a nonuniform distribution of Pd on the FAB surface. The bluish segregated Pd-rich area is unidentified, whether it is a pure Pd or a solid solution Cu–Pd or a CuPd-ordered phase.

Silver and Aluminum FAB

Since the 1970s, researchers have worked on Al wire and reported that the formation of a good FAB is difficult to be attained even if purged using inert (N2, Ar) and reactive forming gases (Onuki et al. 1986). The reduction of surface aluminum oxide layer on Al wire by hydrogen (purging forming gas) may not be effective like in the case of copper oxide reduction on Cu wire surface using forming gas. Continuous forming of aluminum oxide film and its stability finds it difficult to remove in an Al wire to form a good FAB. Furthermore, axisymmetrical Ag FAB formation is observed on purging the forming gas. Alloyed Cu, alloyed Ag, and Pd-coated Ag wires (Tanna et al. 2012) are also examined to replace Au wires.

Ball Bonding Process (First Bond)

The steps involved in first bond process are the following: (a) FAB tip comes into contact with Al bond pad with certain contact velocity without damaging the bond pad, (b) deformation of FAB, (c) mass displacement of Al on bond pad surface, and (d) diffusion bonding of deformed FAB and Al bond pad. During the thermosonic bonding, the wire bonds undergo up to 8 μm displacement, which are bound to cause heating due to the friction between two surfaces in contact.
Microstructures of Au and Cu wire bonds reveal no sign of melting at the bond interface (Murali 2006). An increase in temperature around 200o C during wire bonding is reported. Perhaps, thermosonic wire bonding of these common metals undergoes solid-state reaction. The bonded unit is soaked in 20 % NaOH or KOH solution to dissolve the remaining aluminum, and then gold aluminide distribution is observed by placing the detached bonded ball in a topsy–turvy position (Fig. 22a). It is well known that a uniform distribution of gold aluminide (intermetallic coverage (IMC)) of about 85%is recommended. On thermal aging, a nonuniform IMC would lead to an erratic intermetallic growth as shown in Fig. 22b, for Au ball bonds. The apparent diffusion of atoms is always a combined effect of lattice diffusion and grain boundary diffusion. However, the interface reveals mostly the lattice diffusion with intermetallic nucleation along the horizontal direction. In addition to IMC study, a good ball bond needs to have the following performance characteristics:

(a) Ball shear to be greater than 5.5 g/mil2
(b) Ball pull with neck fracture, without bond pad metal lift or bond pad cratering
(c) Pass dimensional specification of ball bond concentricity
(d) Minimal Al splash
(e) Through cross-sectional analysis, maintains at least 30 % of the remaining Al thickness underneath the bonded ball

The intermetallic nucleation at the Cu wire bond interface is of few nanometers, revealed clearly in transmission electron microscopy (TEM) studies (Xu et al. 2009). While the technique of dissolving bond pad and topsy–turvy Cu wire bond to observe the intermetallic coverage at the interface does not provide the information as good as Au wire bond because of the formation of copper aluminide in nanometer level. Instead, for Cu wire bond intermetallic coverage observations, it is practiced to heat-treat the wire bonds at 175o C to 250o C for 30 min to 5 h and then dissolve the Cu wires using nitric acid and observe the intermetallic coverage on the bond pad (Fig. 22c). Though the intermetallic coverage can be seen on the bond pad, strictly the area of coverage cannot be considered for representing the nucleation of intermetallic (distribution of copper aluminide at t ¼ 0 condition) on bonding without any heat treatment. While the intermetallic coverage measures the nucleation (on bonding) and growth of copper aluminide on heat treatment, in contrast to gold aluminide distribution on nucleation.

Phonon Interaction Mechanism

Phonon is a quantized mode of vibration of atomic lattice of a solid, and it plays a major role on the physical properties such as thermal, electrical, and sound of solid materials. It is a special type of lattice oscillation with the same frequency as vibration movements. Quantized behavior results in wave characteristics, that is, lattice vibrations (thermal energy) move as a strain wave (elastic wave) at the speed of sound. Phonon generation is a dislocation movement and the frequency required for dislocation movement is about 108 Hz (megahertz range). Ultrasonic energy is in the frequency range of 105–106 Hz, the predominant source of supplied energy during wire bonding. Hence, fundamentally, the FAB and bond interface materials may experience phonon interactions or effects close to it: softening the FAB for an easy deformation on compression, consequently influencing a good interdiffusion at the interface and welding the FAB to bond pad. Gold ismalleable; hence, it is expected to be deformed and flow much more easily than Cu on thermosonic bonding.
Slips, deformation bands, and grain substructures are evident in Cu wire bonds (Murali et al. 2003a). Also, modeling of the wire bonds revealed Cu wire has a higher stress to flow than Au wire. Sensitive devices with complex structures underneath the pad especially designed and fabricated with fragile low-k materials suffer from cratering. Hence, the development of soft Cu wire to produce FAB with lower work-hardening rate on bonding is being investigated in order to reduce bond pad cratering.

Stitch Bonding Process (Second Bond)

After first bond, wire is looped following the trajectory of capillary and the last thermosonic sequence is the second bond process. The second bond is also termed as “stitch bond” or “wedge bond.” It follows the following steps: (a) the wire scrubs the plated finger surface the moment it has been contacted, (b) deformation of wire, (c) mass displacement of plated material but less significant than Al bond pad displacement, (d) diffusion bonding of deformed wire to plated finger, and (e) cutting the wire such that it possesses a consistent tail end leading to a consistent FAB diameter of the next cycle. A good stitch bonding needs to have the following performance characteristics:

(a) Stitch pull higher than the specification.
(b) Fracture of heal area on stitch pull.
(c) Low in mass displacement of the plating material.
(d) Through cross-sectional analysis, confirm the intact stitch bonding.

Stitch bonding using Au wires (2N to 4N purity) to Ag-/Au-/Pd-/Cu-plated surfaces is normally good with higher stitch pull strength (Harman 1997; Fan et al. 1999; Johnson et al. 1999). The process parameters used for stitch bonding may vary with different plated surfaces. For example, Cu wire bonding can be bonded using soft 4N Cu wire than standard Cu wire (Fig. 23). The stitch pull and second bond process window are wider for soft Cu wire than standard Cu wire. PCC wire bonding shows a significant benefit on second bonding than any bare Cu wire (soft or standard) with high throughput (number of device bonded in an hour also referred to as unit per hour (UPH)). The enhancement of diffusion bonding between the plated Pd of the wire to the lead finger material is attributed for the second bond benefit of PCC wire than bare Cu wire. The consistency of tail end and capillary contamination may also be added factors for the betterment of PCC wire performance in the second bonding process step. For this reason, PCC wire has taken a place in the bonding wire market since 2010.

The 4N purity Ag, 4N purity Al, and alloyed Ag wires can be bonded to plated fingers. Wedge to wedge ultrasonic bonding of Al wire for automotive application is another versatile topic which is not covered in thermosonic bonding.
The supply of ultrasonic energy and compressive force work hardens the stitch bond that is similar to ball bond. The behavior is obvious in Cu bonds than Au bonds. Hardness of soft Cu wire used in the industries is in the range of 85–95 HV. After bonding, the stitch bond possesses 110–140 HV. The wires from 0.6 to 6 mil diameter reveal the same behavior. Observation of slip bands, micro bands, and deformation cells (sub-grains) probably created by the application of ultrasonic energy is attributed to the work hardening of ball and stitch bonds (Murali et al. 2007; Murali et al. 2003a).

Reliability of Wire Bonds

Basically, reliability of wire bonds is tested in two ways:

(a) Thermal aging (also referred to as high-temperature storage (HTS)) of the unmolded device
(b) Rigorous accelerated stress test of molded device using epoxy molding compound (EMC)

The molded device undergoes reliability tests such as thermal aging, thermal cycling, thermal shock, highly accelerated stress test (HAST), temperature–humidity test (TH/THB), pressure cooker test, reflow, preconditioning, moisture-sensitive level, tested biasing 1 to +5 V supply or unbiased, etc. First bond is more sensitive for these tests than second bond. Development of any new wire has to be first tested for thermal aging, and once it passed this reliability test, further detailed reliability investigations have to be carried out.
The high-temperature stored unmolded device is evaluated for bond pad cratering failure. Moreover, the molded device is tested for electrical resistance and I–V characteristic curve. Also, the epoxy mold is de-encapsulated (decapped) using fuming sulfuric acid and/or concentrated nitric acid. Epoxy molded device with Au wire bonds is decapped manually or by using semiautomated unit with a mixture of these two acids, whereas for epoxy molded device with Cu, wire bonds are decapped only by using semiautomated unit (Murali and Srikanth 2006). But to some extent, the Cu wire and its bond can be dissolved. Currently, combination of laser and acid decapping approach has proved that epoxy molds can be decapped without damaging the Cu wire bonds. Ag wire bonds and Al wire bonds are also examined by this technique.
Thermal aging of Au ball bonds has been investigated for the void formation at the interface causing the failure, whereas the reliability tests such as HAST, THB, and PCT have been studied for the failure of Cu ball bonds due to galvanic corrosion. These two topics are associated with the interfacial reaction occurring at the ball bond on reliability tests and are discussed below.

Microstructures and Interfacial Reactions of Ball Bonds

In the 1970s, voids observed at the Au wire bond interface were considered as a major research project. Later it was explained that the voids are Kirkendall type created due to atomic diffusion at the wire bond interface. From 1990 to 2000, 2N alloyed Au wire (99 % purity) was developed by wire suppliers which could avoid Kirkendall voids at the Au wire interface. Studies on the interfacial reaction in a wire bond on high-temperature storage are vital to understand the stability and reliability of Au wire bonds, especially first bond. The interfacial reaction strongly depends on the following factors:

(a) Properties of the joining metals (similar or dissimilar)
(b) Testing temperature
(c) Testing time
(d) Environment
(e) Area of welding

Wire bond interface has different combinations of dissimilar metal joints. For first bond, more than 90%of the bond pad materials are Al bond pad. In general, the Al bond pad compositions are Al-1wt%Si-0.5wt%Cu, Al-0.5wt%Cu, and pure Al. On bonding the four types of wire Au, Cu, Ag, and Al to Al bond pad, one can expect the Al–Al wire bond to be the most stable one, but it is not in use (because of the difficulty in the formation of FAB). On thermal aging, the other three types of wire bonds form intermetallics at the interface (Au–Al, Cu–Al, Ag–Al). From the phase diagram, the type of second phases that are formed at the interface for different two-phase mixtures is provided in Table 3.

Abundant studies are available on gold aluminide and void formation at Au–Al wire bond interface (Harman 1997; Murali et al. 2004). The wire bonds bonded with 4N Au wires that are in the range of 20–75 μm diameter revealed void growth for a storage longer than 2,000 h at 175o C. The finer the wire diameter, the earlier the bond failure occurs; the failure is bond pad peeling/cratering on ball pull due to Kirkendall void growth (Murali et al. 2004). Kirkendall voids observed in Au ball bond increase the junction resistance (Maiocco et al. 1990). Lattice interdiffusion is predominant over grain boundary diffusion leading to a uniform growth of intermetallics along the entire bond interface (horizontal direction). Voids are always observed at the gold–gold aluminide interfacial layer suggesting the interdiffusion couple between Au and Al: diffusion of Au is faster than Al.
Nowadays, on thermal aging, wire bonds using 2N Au showed no bond pad cratering on ball pull until 10,000 h of storage at 175o C. Cross-sectioned analysis also showed no significant growth of Kirkendall voids (Fig. 24a). The diffusion rate at the interface is slowed down by an appropriate addition of dopants in the 2N Au wire. Thermal aging wire bonds of ultrafine wire in the diameter of 10–15 μm with 2N Au also showed no ball pull lift up to 5,000 h of storage at 175o C. According to JEDEC standard, a reliable bond has to pass without any bond pad lift failure for 1,000 h of storage at 150o C. Industries demand the reliability of wire bond for longer storage period at higher temperature than 150o C such as to test at 175o C or 200o C.

Wire bonds with 4N Cu are stable on thermal aging unlike Au wire bonds, and a very slow growth rate of copper aluminide is formed at the interface. The thickness of the intermetallics observed at the interface for gold aluminide and copper aluminide is shown in Fig. 24b. For the same thermal aging conditions, growth of intermetallics is exponential. For 4N Au, within a few hours, thick gold aluminide is formed, while for 4N Cu, a micron thick copper aluminide is formed after 1,000 h for the same temperature of storage. The growth of copper aluminide behavior of PCC wire bond is similar to 4N Cu wire bond.
Wire bonds with alloyed Ag wire have passed 1,000 h of thermal aging at 175o C, and the growth rate of silver aluminide is faster than copper aluminide formation in 4N Cu wire bonds (Fig. 24b). Interestingly, of all the tested wires, 4N Au, 2N Au, 4N Cu, PCC, alloyed Cu, and alloyed Ag, the alloyed Cu showed the slowest growth rate of intermetallic (copper aluminide) formation at the interface (Fig. 24b). Even at 250o C of thermal aging alloyed Cu wire bonds, the growth rate is much slower than 4N Cu wire bonds.
In Pd-doped Au wire (2N Au) bonded to Al bond pad on high temperature, Pd segregates at the interface between Au and Au–Al intermetallic layers. Industries are expected to have a similar behavior with PCC wire bonds or alloyed Cu wire bonds, but no distinct segregation of secondary elemental addition is observed at the interface between Cu and Cu–Al intermetallic layers.

Mechanism of Interfacial Reaction (Intermetallics/Kirkendall Void Formation)

Interdiffusion of atoms plays a crucial role on the formation of intermetallics. The physical, thermal, and atomic properties of the joining metals, especially the vacancy–solute binding energy and the atomic radii, have a significant influence in the interdiffusion of atoms (Murali 2006; Murali et al. 2003b). Mobility of atoms in the substitutional solid solution strongly depends on the atomic radii of the joining elements. With regard to Au–Al, Au–Ag, and Ag–Al systems, all the three elements have the same atomic radii revealing an absence of lattice mismatch and thus providing an unstrained crystal lattice. Therefore, there may be a free flow of atoms without any hindrance assisted by vacancy mechanism (Murali 2006; Murali et al. 2003b). In other combinations such as Cu–Al, Au–Pd, Cu–Pd, and Au–Cu of metallic bonding, in spite of similar prevailing vacancy mechanism, difference in atomic radii leads to striving of mobility of atoms, which is expected to hinder the growth of intermetallics on thermal aging. The following points are discussed on the interdiffusion of atoms of different bonding systems (Murali 2006):

• In Au–Al system, higher vacancy–solute binding energy of gold-vacancy pair (Murali 2006; Murali et al. 2003b) and strain-free lattice aid boundless atomic mobility, causing a thorough mixing of atoms. In addition, difference in electronegativity between Au and Al atoms is higher than other systems. Hence, the ionic character of the bond is high, favoring the formation of intermetallics.
• Effortless interdiffusion can occur due to strain-free lattice of Ag–Al system.
• During bonding Cu wire to Al bond pad, lower degree of vacancy–solute binding energy and higher percentage of lattice misfit hinder the mobility of atoms, thus, resulting in the absence of formation of intermetallics (t = 0).
• Similarly in Cu–Au system, due to severe lattice misfit, formation of ordered phases may be difficult.
• In Cu–Pd system, the vacancy mobility is equivalent to Au–Al system. Difference in atomic radii between Cu and Pd causes lattice misfit and reduces the mobility. In addition, both elements are stiff and more ultrasonic power is required to deform and bond.
• High lattice misfit, low vacancy–solute binding energy, same valence, and electronegativity of Cu and Ag lead to the formation of diffusion bonding with eutectic mixture.
• Same atomic radii and valence of Au and Ag and its closer electronegativity values lead to the formation of solid solution on mixing and so in the wire bond.

The concepts of vacancy mechanism and atomic strain-free lattice with effortless mobility of atoms/vacancies are attributed to the growth of the intermetallics and voids. Therefore, in order to slow down the growth of intermetallics and formation of Kirkendall voids, the rate of interdiffusion has to be reduced (Murali 2006). This can be achieved by adding elements with a large difference of atomic radii to the bonding metals. For example, Y, Sn, Zr, In, Cd, Ni, Cr, Fe, Mn, and Si can be added to bond pad Al metallization or bonding wire. Addition of the mentioned elements either individually or combined may effectively reduce the rate of interdiffusion at the interface.

Galvanic Corrosion in Cu Ball Bonds

In the last 2 years, rigorous research work has been carried out on reliability of bare Cu wire bond in semiconductor packaging industries. So far the key observations are Cu wire bonded to Al bond pad (Al, Al-0.5Cu, Al-1Si-0.5Cu) using BGA substrate and tested for HAST (130 C, 85 %RH, biased and unbiased), and THB (85 C, 85 %RH, biased and unbiased) tests revealed failure in short period (less than 100 h). Industries require a stringent criterion to pass as long as 500 h of BHAST and THB tests, for instance, biased at +3 to +5 V. The epoxy mold compound manufacturers reported (Su et al. 2011) Cl content and pH of the EMC are the reason for de-bonding of Cu ball bond from the Al bond pad causing open circuit. They also mentioned that Cl ions could be diffused from BGA substrates, especially containing high Cl ions of about 60 ppm compared to a green mold compound containing less than 20 ppm.

It is well known (Fontana 1987) that when two dissimilar metals are welded together and exposed to corrosive environment (Cl-, Br-, F-, SOx-, NOx-, OH-, etc.), an electrochemical galvanic corrosion is unavoidable. The rate of corrosion strongly depends on:

(a) Welded dissimilar metals
(b) Area of welding
(c) Source of the corrosive medium and its strength
(d) External bias and its current density and polarity

Among the interfacial layers, the Cu and Al interface is expected to corrode under galvanic conditions than the rest. Any corrosion reaction is electrochemical in nature and needs a path for electron to flow, since EMC is an insulator and its interface in contact with Al bond pad or Cu wire may not corrode without an electron flow path. The second bond interface, welding Cu wire to noble metals (Ag/Au/Pd) or Cu, may have less significance on galvanic corrosion. Where Au/Ag/Pd/Cu belongs to cathodic EMF series (unlike Al in anodic series), therefore, the EMF potential difference is negligible between Cu and noble metals (Fontana 1987).
In this concern, the PCC wire which was extensively examined with an anticipation to solve the galvanic corrosion problem of Cu wire bond failed due to inhomogeneity of Pd distribution on the FAB surface (bond interface). The confidence level of PCC wire bond on BHAST is still in the evaluation stage. Moreover, the alloyed Cu wire is examined on these grounds. Table 4 provides BHAST and HTS performance of a molded device with alloyed Cu wire bond. The test was conducted in collaboration with Sumitomo Bakelite Co., Ltd. using 16pSOP device, molded with G-700 series EMC, and biasing +20 V. Several industries have keen interest on the alloyed Cu wire, since the corrosion potential of the wire is much better than the bare Cu wire. High ball shear 11–13 g/mil2 for 0.8 mil bare and PCC Cu wires reveal survival of wire bonds for longer periods on BHAST compared to bonded with 9–10 g/mil2. This factor of Cu wire bonds with high ball shear is now examined for alloyed Cu wire.
In addition to Cl ions, SOx ions can corrode Ag wire bonds dramatically if they exist in the surrounding environment. EMC manufacturer claim that the presence of sulfur (S) is less than 10 ppm in green mold compound. Studies on galvanic corrosion of Ag wire bonds due to Cl ions and SOx ions are utmost important to qualify Ag wire; the topic is under evaluation with high priority.

Future Trends

Future thermosonic wire bonding will see continual decrease in the use of Au wire due to the high precious metal price, and industry is continuously converting this material to lower-cost alternatives like Cu, be it bare or coated for oxidation protection. There will be increasing use of Ag or Ag alloy wires also for applications that Cu wire is unable to replace Au such as in LED which saw application in low-end device. For IC devices like memory chip, there has been evaluation and qualification ongoing with Ag alloy wires in areas where cost is less critical. Based on industry prediction, the demand for Au wire will reduce to less than 50 % of total bonding wires used (Fig. 25). Majority of wire material will still be Cu based and with complex configuration in composition for bare Cu and plating technology for coated wire. Silver alloy wire will see higher growth in low-end LED devices in the next few years but not for high-end devices until more confidence is built up on its reliability. Industry will continue to evaluate the use of Ag wire for high-speed memory device such as DDR4 in which Cu wire is not able to meet the application requirement. The composition of Ag alloy wire will continue to evolve for different application needs if cost is not a critical requirement.

Size of wire is also getting finer for high-density packages moving from 20 μm (0.8 mil) in 2008/2009 to 18 μm (0.7 mil) in 2010/2011 and even 15 μm (0.6 mil) beyond 2012. Even finer diameters of 12 and 10 μm are being evaluated for future product and applications.
The bond pad and substrate materials are also moving more and more to nonaluminum type such as Au, Pd–Au, Ni–Pd, and Ni–Pd–Au for better first bond reliability while that of metal substrate for second bond will see more usage of pre-plated lead frame (PPF) which has Au on Pd and Ni layers. For major cost savings, bare Cu lead frame is already been evaluated but not with great challenges. On organic substrates, continual development of Au plating technology has been ongoing to achieve better and more robust stitch pull strength and reliability. ENEPIG, or electroless Ni, electroless phosphorus (P) immersion Au plating, is the key to future BGA substrate.
With more variety of wire and bond surface material, the interfacial reaction between wire and bond surface is becoming more complicated, and studies needed to be carried out to understand the intermetallic compound formation and growth between different alloying composition of Cu and Ag wire with bond pad material (i.e., first bond) and how their reliability performance can be compared with conventional high-purity Au and Cu. The second bond reliability must also be studied due to less well-known interfacial reaction between new wire material and Ag or Au on bond fingers.