Vibration fatigue of bond wires in electronic circuits - not all it's cracked up to be

TWI Bulletin, January/February 1999

After graduating in mechanical engineering, Terry Dickerson studied at Nottingham University researching contact problems associated with engineering ceramics. He gained his PhD in 1991 and currently works in the Advanced Materials Structures section of the Structural Integrity Department.

Vibration can cause fatigue failures of wire bonds in electronic components; however, Terry Dickerson shows that it is not as serious a problem as you might think.

Electronic components must often remain reliable for long periods whilst being used in severe environmental conditions. Perhaps the most damaging of these conditions is temperature change that drives differential thermal expansion, which in-turn produces thermal stresses. However, there are other conditions in the environment that can also cause damage. Vibration is probably the next most severe of these and this is covered here.

Long and slender structures are most susceptible to vibration damage. Of all the parts of modern electronic components some of the most delicate are bond wires because they have this slender geometry. In many cases the wires are not encapsulated and hence are free to vibrate. These types of bond wires can respond to external vibrations by resonating, this can cause large numbers of stress cycles in the wire in a relatively short space of time. Therefore only short periods of vibration may be needed to cause damage. The stresses tend to concentrate near to the ball and wedge bonds where there are changes in thickness that act as stress raisers. Such stresses can cause fatigue damage, cracking and separation of the wires and loss of electrical continuity, see Fig.1.

This article describes an investigation into the behaviour of bond wires in response to environmental vibrations. In particular, the aims of this study are to:

• give a visual representation of bond wire vibrational behaviour (using mode shapes) and
• quantify resonant frequencies for realistic bond wire geometries.
  

Designing against vibration

Like all matter, wires have the essential properties of mass and elasticity that are required for them to vibrate. A bond wire of a certain size and shape will have a series of discrete natural frequencies and each of these frequencies will have a mode (direction and shape) of vibration associated with it. The first mode will be at the lowest natural frequency, the second mode at the next highest frequency etc. Environmental vibrations at or near one of these natural frequencies can cause the wire to resonate.

At resonance, the amplitude of movements of environmental vibration can be magnified in the bond wire; however, its amplitude will be limited by damping in or on the wires.

The amount of damping can be estimated from published hysteresis loss data and from the viscosity of the fluid around the wires or it can be derived from direct measurements. However, damping estimates from calculations tend to be simplistic and direct measurements are very difficult on small components such as bond wires. Hence damping factors tend to be unreliable. This may seem to present a problem for designing bond wires to be vibration resistant. However, the solution is to design the bond wires so that their natural frequencies do not correspond to any of the environmental frequencies that the device will be subjected to. If there are many frequencies in the environment this may be difficult. However, there is a simple solution that can often be applied - to make the lowest natural frequency of the bond wire higher than the highest environmental frequency to which the device will be exposed. This simple design solution means that the bond wire will never reach resonance.

A little theory

The n ^th natural frequency, ω_{n, of a straight wire-like structure can be described by:}

Where E, I and m are the Young's modulus, second moment of area and mass per unit length of the wire respectively and l is the length between the ends of the wire. For round bond wires of diameter d, Eq.[1] becomes:

Where ρ is the density of the wire material. H _n is a factor that depends on the boundary conditions at the ends of the wire. H _n has to be determined for each natural frequency ( H ₁ is for the lowest natural frequency); for this relatively simple case this can be done analytically.

Equations [1] and [2] relate to straight wires but bond wires are rarely straight. More accurate solutions are available ^[5] - but these come with the price of greater complexity. The equations above give acceptable results when the loop height is less than about 5% of the bond length. In practical terms, if the developed length of the wire, l _d (see Fig.2), is used instead of the distance between the fixed ends, l, the equations have a much larger range of validity.

Investigation method

Like previous work, ^[3] analytical solutions ^[5] for arched wires could have been used to predict natural frequencies. These solutions would give acceptable accuracy but the visual story and hence a deeper understanding could not be so easily conveyed.

A three-dimensional finite element model (FEM) was used to meet both objectives set out at the end of the introduction. The FEM also had the advantage that more geometric detail could be modelled than in analytical solutions. A PC based parametric FEM package was used to calculate natural frequencies and mode shapes of bond wires. The parametric facilities in the package allowed easy and quick changes to the model so a range of bond wire parameters could be rapidly assessed.

Figure 2 shows a mesh created with the parametric model. The inset 3-D view shows that the bottom of the ball and wedge bonds have fixed displacements applied; this, not unrealistically, models the wire bonded to relatively rigid die and substrate.

All models were for 25µm diameter round section gold wire. If necessary, the theory in the previous section can be used to extrapolate the results to different materials and geometries.

A typical bond wire

An illustration of the vibration of a typical bond wire is shown in Fig.3; this wire had l = 2mm and h = 0.3mm.

Vibration modes

The modes of vibration can be split into two main types; the lateral mode and the in-plane modes. Figure 3 illustrates these two types of mode; it shows end-on and side views of the model of the bond wire for the first four modes of vibration. The lateral modes are as shown in the top diagrams (mode 1) and third from top diagrams (mode 3). There is no in-plane movement in these lateral modes. The in-plane modes are illustrated in the second from top diagrams (mode 2) and the bottom diagrams (mode 4). There is no lateral movement in the in-plane modes. Higher modes follow the pattern of alternate lateral and in-plane modes, the shape of each subsequent mode getting more and more complex. At much higher frequencies it is possible to get a third mode type - a torsional mode.

Natural frequencies

The natural frequencies associated with the first three modes are also shown in Fig.3. The lowest frequency is about 8kHz - this is a relatively high frequency when compared to the frequencies of potential environmental sources of vibration. In this case the second mode has a frequency only 2.5kHz higher than the first mode. The third mode frequency is about 22kHz. and the fourth mode 29kHz.

Influence of material, size and geometry

The example described above is for one typical wire size and geometry. There are some other parameters that can also be changed but their influence was expected to be minimal, for example the radius ( r) and the drop distance from the die surface ( x) - see Fig.2. Changing these parameters within sensible limits did not have a significant effect on the predicted natural frequencies. And so for the subsequent analyses x was fixed at a value of 0.3mm and r was taken as 0.2 or 0.3mm ( r = 0.3mm for all but the analyses with h = 0.3mm).

The influences of bond wire length ( l), loop height ( h) and bond wire material (gold or aluminium) were the main parameters examined. This is because these are the features that affect the mass and stiffness of bond wires.

Bond wire material

The influence of the material that the bond wires are made from can be predicted using eq.[1] and [2]. H _n is a function of wire geometry and boundary conditions only, the material specific terms are E and ρ.

The vibration modes of gold and aluminium wires will be the same, only the natural frequencies will be different. Given the material properties, the results for on wire type can be scaled to another wire type. This was done for Fig.4 which shows the lowest natural frequency for both gold and aluminium wires.

The natural frequencies of an aluminium wire bond are 2.7 times higher than a similar gold wire bond.

Vibration modes

For all parametric variations to the model, the first mode was always predicted to be a lateral mode as illustrated by mode 1 in Fig.3. Higher modes were alternately in-plane and lateral.

Natural frequencies

Figure 4 shows the predicted relationship between bond wire length, height, wire material and the lowest natural frequency. The frequencies for both Au and Al bond wires are shown.

As expected from Eq.1, as the bond gets longer and higher the frequency of the first mode falls. Even so, for very long tall bond wires ( l = 5mm, h = 1.5mm) the lowest natural frequency for a 25µm gold wire remains above 1kHz; for the same size aluminium bond wires the mode 1 frequency is about 2.8kHz.

Discussion

The potential for vibration induced fatigue failure of wire bonds has been demonstrated by Riddle ^[1] and considered a potential problem by others. ^[2,3] Failures that have possibly been caused by vibration have been observed at TWI. However, such failures seem to be quite rare, or at least rarely reported.

There are many sources of environmental vibration. As the lowest natural frequencies of typical bond wires are in the region of 4-10kHz it is worth considering what in the environment might produce such high frequencies.

Possible sources of vibration

During production, the assemblies containing the bond wires may undergo ultrasonic cleaning which has been shown to cause damage. ^[1] Vibration based tests, will typically go up to 2-3kHz. Screening of production components in this way therefore may or may not induce damage depending on the bond wire size.

In service, there may be mechanical sources of vibration. Gas turbines or turbochargers will typically operate at up to 100,000 rpm, this equates to 1700 rev/sec. Hence vibrations due to out of balance forces are unlikely to create the required vibration frequencies. Piston engines, pumps etc. will operate at much lower revs and cause lower frequency excitation. Hence mechanical vibrations are not likely to be the source of any failure.

Other potential vibration sources are acoustic (a soprano shattering a glass is the common example). For instance, cavitation of fast flowing fluids, piezoelectric effects and even loudspeakers can cause high frequency vibration.However, airborne acoustic vibrations usually have low energy levels and so the effects are relatively easily damped out.

It is clear from the above discussion that there will be few sources of environmental vibration that will be of high enough frequency to cause bond wire resonance and hence fatigue damage and failure. However, the longer and thinner bond wires are the more susceptible they will be to environmental vibrations. Furthermore, the much higher density of gold than aluminium gives gold bond wires lower natural frequencies. Hence gold wires are much more likely to resonate and become damaged.

'Tuning' the bond wires

In the unlikely event that resonance of the bond wire does take place, this article shows that the wire geometry can be adjusted and its natural frequencies changed. For instance, bond wire height can be used to tune the vibration response of the bond wires to avoid resonance problems.

Life is not that simple

Even though the lowest natural frequencies of typical bond wires are much higher than likely excitation frequencies, there are a number of ways that vibrations can still cause problems:

• The relatively high mass and low modulus of some encapsulants will mean that embedded wires will be forced to move (vibrate) with the encapsulant. For instance, the very low modulus of silicones will mean the natural frequency of the encapsulation made from this material will be much lower than the bond wires that it contains.
• Vibrations could induce relative movement of the bond touch down points, eg circuit board bending or movement of wire feed-throughs, both of which can stress the wires and cause fatigue damage.

Conclusions

Whilst it is rather too much of a simplification to say that bond wires will not suffer from vibration induced fatigue, this article has shown that such failures are unlikely for most bonds, even in relatively harsh vibration environments.

• Because of their higher density, gold bond wires have lower natural frequencies than aluminium bond wires of the same size. Therefore gold wires are more likely to be susceptible to vibration fatigue.
• Generally the lowest natural frequency of wire bonds is higher than any normal source of environmental vibration and hence in most applications there is no real risk of the wires resonating.
• There are still dangers for resonance if the bond wires are subject to severe acoustic vibrations, set in a low modulus encapsulant, or bonded onto flexible components.
  

Acknowledgements

The author wishes to thank Bob Clements for his practical advice on bond wires and wire bonding.

References