A Review on Wire Bonding Process in Electronic Manufacturing

ZONG Fei,WANG Zhijie,XU Yanbo,YE Dehong,SHU Aipeng,CHEN Quan

（Freescale Semiconductor (China) Limited,Tianjin 300385,China）

1 Introduction

In the semiconductor industry,wire bonding,flip chip and tape automatic bonding（TAB for short）were 3 major methods to achieve IC internal connections.Due to the advantages in feasibility,cost,reliability quality and compatibility with other processes,wire bonding was the most popular method and accounted for more than 90% of IC counts.While the actual procedure of wire bonding was very complex and finished in a short time（less than 100 ms）,there were still many inextricable problems although wire bonding had been applied from 1960s.So a good understanding to the process of wire bonding,including bonding procedure,bonding parameters,bonding materials and tools,as well as bonding environment,would bring a great benefit to the solving of whether actual operation issues or basic bonding principles.

In this paper,dozens of journal and conference papers published recently about wire bonding process would were reviewed and numbers of experiment results were presented.Firstly the cycle of wire bonding was divided into 5 main phases and 19 sub-phases: free air ball forming（electric flam off,EFO-Open detection）,1stbonding（searching-1,impacting-1,contacting-1,bonding-1）,looping（loop reversing,loop ascending,non stick on pad detection,loop tracing）,2ndbonding（searching-2,impacting-2,contacting-2,bonding-2）and tailing（tailing ascending,short tail detection,tail tearing,on stick on lead detection,resetting）; the main 5 phases were sketched in figure-1.And the procedures of each phase,including the motion of capillary and bonder table,were introduced detailedly.Effects of key bonding parameters,installations of bonder parts,properties of bonding wire as well as lead frame or substrate,capillary design and fixtures on the quality of wire bonding were discussed; at the same time the affecting mechanism was also proposed.

Fig.2 Illumination of relative parts during wire bonding

Numbers of pictures would be used for a better introduction while if the same part was marked in all pictures,it would be fussy and confused.So here a uniform illumination was made in figure-2 and the same part wouldn’t be marked anymore if no special explanations.

As below,the process and relative information of wire bonding would be introduced phase by phase.

2 Phase I - FAB Forming

2.1 EFO（Electric flame off）

In most cases,the cycle began with a wire tail threaded through the capillary and a ball formed on the tail; this phase was known as electric flame off（EFO for short）.The motion of tail tearing（which would be addressed in the tail tearing phase）broke the wire and left a wire tail.During the EFO phase,the rod discharged an electric arc between the metal electrode rod and the wire tail,which melted the wire tail into a drop; after the discharging the drop solidified rapidly into a ball which was known as free air ball（FAB for short）.This EFO phase was shown in figure-3.

Fig.3 Sketch of EFO phase

The metal electrode rod was commonly called EFO wand,which was fixed or mobile on the wire bonder.During the electric discharging,the fixed EFO wand didn’t move and the discharging path between the wand and the wire tail was oblique; while the mobile EFO wand moved towards the wire tail horizontally and the discharging path was vertical.Figure-4 showed the motion of mobile EFO wand during the electric discharging; and due to the vertical discharging path,the symmetry and consistency of FAB was better while the horizontal motion decreased the UPH（unit per hour）.

The installation of fixed EFO wand was required strictly and figure-5 showed the configuration.The horizontal distance between the wand and the capillary outline,gap2,was about 76.2 μm generally; this distance could prevent the wand from being touched by the descending capillary as well as didn’t reduce the efficiency of electric discharging.For a certain tail length,the vertical distance between the wand and the wire tail,gap 1,should be adjusted to make the angle between the line from the wand to the capillary tip and the horizontal line,θ,within the range from 30 to 45 degrees.Improper installation might introduce some abnormal issues,such as EFO-Open and golf bond.

Fig.4 Mobile EFO wand

Fig.5 Installation configuration of fixed EFO wand

For copper wire bonding the forming gas,a mixture of nitrogen（N2）and hydrogen（H2）,was used as the environment where FAB was formed far from being oxidized by oxygen（O2）in air.The forming gas was delivered by an additional ceramic system,known as copper kit in figure-6[1],which was installed around the capillary tip and the wire tail.

Fig.6 KNS patented Cu kit

The flow rate of forming gas affected the shield of FAB forming.Shown in figure-7 was the simulation result[1]: a low flow rate caused laminar behavior while high flow rate caused transitional or turbulent behavior,both of which might resulted into inconsistent or asymmetric FAB.Many experiments recommended the range from 0.4 to 0.6 l/min as an optimized window.

Fig.7 Gas flow rate model of Cu kit

The minor H2played an important role in copper FAB forming: it not only reduced the oxidation on FAB surface,but also helped get a bigger and more spherical FAB.The lighter hydrogen molecule could apply a higher discharging voltage with the same discharging current compared with other gases,shown in figure-8[2],so it was expected to provide more heating energy and resulted into larger FAB.

Fig.8 Effect of gases on arc electric field

At the same time,the lighter hydrogen molecule and lower binding energy of H-H made H2gas tend to expand outward in the discharging arc and cool the arc periphery,so the arc tended to localize the arc path to minimize the heat loss.Compared with the spreading arc,this constricted arc could get a more spherical FAB,shown in figure-9[3].

The discharging current（EFO current）and discharging duration（EFO time）were two major EFO parameters,which determined the final diameter of FAB.An experiment of FAB forming with 25.4 μm copper bonding wire was implemented on KNS maxum-plus wire bonder and found the relationship between EFO parameters and FAB diameter which was described as equation-1.HereImeant EFO current within the range of 74～90 mA,tmeant EFO time within 286～354 μs and Fr meant forming gas flow rate within 0.21～0.89 l/min.The relationship was parabolic due to the conduction,convection and emission of heat.

Fig.9 Arc discharge behavior on the wire tail

Some researchers proposed a more complex but comprehensive equation based on many experiment data.In equation-2[4],FAB meant FAB diameter,Imeant EFO current,tmeant EFO time,Dw meant bonding wire diameter and n was a constant.The constantsCi（i=1,….,6）were gotten with the curve fitting technique and varied with Dw.

EFO parameters also affected the peak temperature of melted drop which could determine the hardness of solidified FAB and final ball bond.Figure-10 showed the Vickers hardness in different zones of copper ball bonds gotten with different EFO currents[5],and it was found that higher the EFO current,softer the ball bond.

Based on the on-line measurement results,a similar hardness changing trend for copper ball bond while an opposite trend for copper FAB were gotten,which were shown in figure-11[6].To get FAB with the same diameter,higher EFO current and shorter EFO time resulted into a higher peak temperature of FAB as well as a more dramatic cooling during the solidification,and this would get finer and harder grains in FAB; while during the subsequent wire bonding,this harder FAB suffered less work hardening and got a softer ball bond.A softer deformed ball bond reduced the stresses under the die pad during wire bonding and decreased the risk of damaging the beneath structure,so a higher EFO current setting was recommended for copper wire bonding.

Fig.10 Effect of EFO current on ball bond hardness

Fig.11 Effects of EFO current on hardness of Cu FAB and ball bond

During the formation of FAB,the heat also travelled along the wire and the size of grains which formed during the solidification varied gradually from the very large grains in the FAB to the finer grains in the wire far from the source of heat.The zone over which the grain size varied was known as heat affected zone（HAZ for short）.And the length of HAZ had an important effect on the looping,which would be discussed in the loop reversing phase.

2.2 EFO-Open Detection

During the electric discharging,the internal electronic system of the wire bonder was detecting the current in the electric circuit.Once the applied discharging voltage was increased up to the preset voltage threshold while the current was lower than another preset current threshold,the wire bonder would consider the circuit as open and show an alarm of EFOOpen.

The EFO-Open issue was common during wire bonding and there were many factors caused this issue.If the setting of voltage threshold was too low or that of current threshold was too high,there would be many false EFO-Open alarms.An insufficient input energy caused by low EFO current or short EFO time couldn’t achieve the electric discharging and caused this alarm.If the wire tail was too short or dirty,the electric discharging also couldn’t be achieved and there would be an alarm; the dirty wire tail may be caused by a dirty wire path,a poor storage condition for bonding wire or a poor assembly room cleanness.Figure-12 showed the sketch of bonding wire path including wire spool,wire tensioner system,wire clamp and capillary.In addition,the improper installation of EFO wand might also introduce this issue.

Fig.12 Sketch of bonding wire path

3 Phase II - 1st Bonding

3.1 Searching-1

After the FAB forming phase,the capillary descended and once descending to a preset height,the wire clamp was opened.Then the capillary descended with a fast velocity（searching speed-1）from the preset height to another preset height above the 1stbonding surface.The two preset heights were called the searching height-1 and the target height-1 respectively,and this phase of descending from the searching height-1 to the target height-1 was called the searching-1 phase（1 here meant 1st bonding）,shown in figure-13.

Fig.13 Sketch of searching-1phase

During the searching-1 phase,the wire clamp kept open and with the backward dragging force applied by the wire tensioner system,the FAB was entered into and held by the inner chamfer of capillary.

Shown in figure-14[7]was the schematic diagram of capillary.The critical dimensions included wire diameter（WD）,hole diameter（H）,inner chamfer（IC）,inner chamfer angle（ICA）,chamfer diameter（CD）,tip diameter（T）,face angle（FA）,and outer radius（OR）.

Fig.14 Critical dimensions of capillary

If CD and ICA was too small,there would be a risk of FAB off-centering.In this case,IC volume wasn’t big enough to hold the FAB fully and the FAB might be placed into CD asymmetrically.Increasing CD and ICA may bring benefit to solve this issue,just as shown in figure-15.

3.2 Impacting-1

The target height-1 in the searching-1 phase was called the impacting height-1 here.When the capillary descended down to this height,it slowed down to a low constant velocity until contacting the 1st bonding surface.This phase from impacting height-1 to contacting the bonding surface was called the impacting-1 phase,which was shown in figure-16.

Fig.15 Effects of CD and ICA on FAB centering

Fig.16 Sketch of impacting-1 phase

The low constant velocity was called the impacting speed-1 which made the bond head contact the bonding surface smoothly.Too much work found these two parameters,impacting speed-1 and impacting height-1,affected the quality of 1stbond greatly.A high setting of impacting speed-1 or a low setting of impacting height-1 increased the UPH（unit per hour）and reduced the duration of impacting-1 phase which may cause bad effects:（1）poor FAB centering and final 1stbond symmetry;（2）smashed 1stbond;（3）poor IMC coverage.

The impacting-1 phase ended when the capillary with FAB contacting the 1stbonding surface and there were two methods to detect this contacting signal commonly.One method was ‘position control’: the bonder would learn and record the height of bond pad before wire bonding; once the capillary descended down to the learnt height,the bonder would consider the FAB contacting the 1stbonding surface.However,the actual leant heights varied with different heights of bond pads when the actual pad heights were different or the bonding surface was unstable due to the vibration or floating of die; at the same time,the actual FAB diameters also had a variation.All these would make the actual result of‘position control’ not be as good as the designed goal and there might be a big variation of 1st bonding formation.So the other method,‘force/velocity control’,was more popular.In this method,a piezo-electric sensor detected the counterforce applied on the FAB by the 1st bonding surface during the impacting-1 phase; once the force became higher than the preset force threshold,the bonder would consider the FAB contacting with the bonding surface.Because the velocity could be transformed to the force physically,‘velocity control’ was another type of ‘force control’ and there was a control threshold of velocity change accordingly.During the impacting-1 phase,the impacting speed would decrease and if the velocity decreasing was bigger than the preset threshold,the bonder would consider the FAB contacting with the bonding surface.

Before continuing the introduction,a basic understanding of the procedure and mechanism of bonding was necessary.Although the application of bonding started about 50 years ago and the final conclusion about the mechanism was still not made,some points were accepted by the worldwide researchers.The FAB was plastically deformed by the applied force during bonding,and its slip lines resulted into the stepladder-like fresh surface（grooves）which broke the oxidation of FAB,just as figure-17 showed; then this breaking provided an opportunity to form a close contact between fresh metal surfaces and a condition of diffusion for joining.At the same time,the ultrasonic also played important roles:

（1）ultrasonic vibration cleared the contamination and broke the oxidation of both bonding wire and bonding surface;

（2）ultrasonic energy absorbed by the metals produced many dislocations which softened and plastically deformed the metals,in fact ultrasonic had a high efficiency of softening than heat[8].Just as described above,all these provided an opportunity for joining.A previous view considered that the ultrasonic friction might produce heat which resulted into micromelted zones and macro-joints,while the heat was regarded as just a by-product and not the key factor based on the later experiment of achieving Al joints in the liquid nitrogen[9].

Fig.17 Micro-deformation and oxidation breaking

According to the bonding mechanism,the bonding procedure could be divided into 3 phases in figure-18.

Fig.18 3 phases to achieve bonding

In the 1stphase,initially the FAB was deformed and the surface was cleared.Then in the 2ndphase,the deformation and clearing continued; the FAB slipped on the bonding surface and some micro-weld zones formed.In the last phase,the micro-weld zones became into macro-joints and the FAB/bond was stuck to the bonding surface; and the ultrasonic as well as the bonding force accelerated the atom diffusion,softened and deformed the FAB into the final bond.In fact,the beginnings and ends of these 3 phases were difficultly to be distinguished accurately,especially for the 2ndone and the 3rdone,the transition from slipping to sticking.In this paper,the impacting phase was considered as the 1stphase,contacting as 2ndand bonding as 3rd.

3.3 Contacting-1

The contacting -1 phase lasted from detecting the contacting signal to the end of contacting time-1,shown in figure-19.In the duration of contacting-1 phase,USG power（contacting USG-1）,force（contacting force-1）and heat（bonding temperature）were applied to achieve a good pre-treatment for the 1st bonding.After this contacting-1 phase,the FAB/bond was deformed partly.

Fig.19 Sketch of contacting-1 phase

Although the FAB/bond wasn’t formed finally,the contacting parameters had important effects on the quality of 1stbond.Figure-20 showed the IMC coverage of two contacting parameter settings（in table-1）.The samples were built with 33.02 μm gold wire on an ASM Eagle-60 wire bonder with same bonding parameters;while parameter #1 and #2 had different contacting-1 parameters and got different bond qualities.Compared with parameter #2,#1 got a smaller but stronger ball bonds and this could be explained by the better IMC coverage of parameter #1.In figure-21,the experiment with 25.4 μm copper wire on a KNS maxum-plus wire bonder got a same result.A normal parameter setting with pre-bleed（an equivalent parameter to contacting USG-1）got a 6.6 gf-higher ball shear strength compared with that of without pre-bleed.

Table 1 Different contacting parameter settings

However,what should be noted was that the socalled optimized setting got stronger bonds but it might damage the structure beneath the bond pad at the same time.Figure-22 showed the ball shear failure modes of samples built with 25.4 μm copper wire on a KNS maxum-plus wire bonder and thermal aged 168 h at 175℃,the setting with pre-bleed got metal lifts which indicated that there were some damages under the pad,and this might be introduced by the heavy contacting parameters.So the contacting parameters need a further optimization to find a balance between the bond strength and the integrity of die structure.

Fig.20 IMC coverage of different contacting parameters

Fig.21 Comparison of ball shear strength between without pre-bleed and with pre-bleed

Fig.22 Ball shear failure modes after thermal aging

The reason why different contacting parameter settings got different IMC coverage and bond qualities would be explained in the next phase.

One issue occurred easily here was known as called golf bond or off-center bond,just as figure-23 showed.In fact,the occurrence of golf bond was related to many phases,including contacting-1,impacting-1,searching-1,FAB forming and even 2ndbonding.

The abnormal wire tail might be melted into an asymmetrical FAB which would result into a golf bond.Excessive 2ndbonding parameters,a dirty 2ndbonding surface,a damaged capillary,a poor storage condition for bonding wire and a poor assembly room cleanness all might cause this curly or dirty wire tail.For copper wire bonding,the forming gas was used for FAB forming; as discussed above,the flow rate and H2content were key factors affecting the symmetry of FAB.Too high flow rate and less H2content might get an asymmetrical FAB and a golf bond[3,10].And the installation of EFO wand had effects on EFO discharging and bonding processes.If the installation wasn’t improper,the FAB might be asymmetrical formed and cause this issue.

Fig.23 SEM picture of golf bond

For a symmetrical FAB,if it wasn’t placed into the CD（chamfer diameter）of capillary well,it would also be made into a golf bond.During the searching-1 phase or the impacting-1 phase,if the motion speed was too high or the distance was too short,there would be not enough time for the motion of drawing back and placing the FAB; the hurried motion might make the capillary to hold the FAB asymmetrically and got a golf bond.In the other case,if the capillary geometry was improper,small ICA and CD,most part of FAB would be hung out of CD and this partly holding might cause this issue too.

The last kind of root cause occurred during the contacting-1 phase.If the contacting force was too low,the capillary would not press the FAB well and during the application of too high contacting USG,the FAB would jump on the bonding surface and this unstable bonding process might result into a golf bond.At the same time,if the capillary was not installed vertically,bonding force and bonding USG would not be applied on the FAB symmetrically during whether the contacting-1 phase or the bonding-1 phase,also resulting into a golf bond.（未完待續(xù)）