Page images
PDF
EPUB

consider two single-level wire bonds for which the same tensile force in the wire is requires to produce rupture. Let the contact angles that the wire makes with each bonding surface be 10 degrees for one wire bond and 30 degrees for the other. The pull strength measured for the one with contact angles of 30 degrees will be about three times as large as for the other (see figure 2).*

If information about the wire tensile strength is not provided the pull strength data of different wire bonds cannot usually be used to compare wire bond quality. This is because failure is usually in the wire and in this case the pull strength is dependent on the tensile strength of the wire. Wire tensile strengths may vary significantly. For example, aluminum ultrasonic wire bonds may be made from wire with an initial tensile strength of from 12 to 20 gf; depending on the thermal history of the wire the final tensile strength of the wire may be as little as 20 percent of its initial value [70P1]. A useful way of normalizing the data from pull tests of different wire bonds with wires having different tensile strengths is to employ a bond efficiency [67R1] defined as the ratio of the tensile force in the wire at rupture to the tensile strength of the wire.

There has been some concern about two aspects of the pull test procedure which possibly can affect the pull strength measured. One is the speed of the pulling stress applied. Therefore, the rate of pull is occasionally specified for the pull test. It is given either in terms of the speed of the pulling element [66R1], [67S3], [69K1], or the rate of increase of the force, as measured at the pulling element [70B8]. In some pull tests, the only specification is that the force must be applied "slowly" [67H1]. Leedy and Main [71B4] reported no dependence of pull strength on pull rates in the range of 1 to 77 gf/s (equivalent to a range of from about 0.4 to 30 mm/min) for single-level, unannealed wire bonds with 1-mil diameter aluminum wire bonded ultrasonically to an aluminum film on silicon. The higher rate may be comparable with the speed of some pull test machines used in the industry where the pull may be likened to a jerk. It should be noted that the wire bonds used by Leedy and Main were constructed so that rupture occurred in the heel of the bond; the independence of the pull strength on pull speed in the range reported may not hold for wire bonds where the failure mode is rupture or peel at the bond interface or where the two bonds are on different levels [71B4].

The other aspect of concern arises if the applied pulling force is directed at some angle, a, out of the plane of the wire loop because of possible tearing of the bond heels. Studies by Leedy, Sher, and Main [72B4] on 1-mil diameter aluminum ultrasonic wire bonds have been shown that the measured pull strength decreases only slightly as a increases; the decrease is more pronounced for wire bonds with greater bond deformation. If a is maintained at less than about 10 deg the pull strength will not be affected significantly by variations in a, except when testing bonds with bond deformationst of about three wire diameters or more.

*For wire bonds made on the same plane, 10 degrees and 30 degrees correspond to values for the ratio of the bond separation to loop height (d/h) of about 11 and 3.5, respectively. + Bond deformation refers to the width, as viewed from above, of the wire deformed at the weld or attachment point.

2.2. Destructive, Single-Bond Test

The destructive, single-bond test consists of pulling on a wire at some angle with respect to the bonding surface with increasing force until rupture occurs. For the test, either the wire bond is cut somewhere along its span or only one bond is made. The purpose of a single-bond pull test is usually to afford better control of the angle of pull (with respect to the bond and contact surface) and also thereby minimize flexing and hence weakening the heel of the bond.

A means of gripping the wire is required in this test. Wasson [65W1] has described the use of special tweezers to grip the wire. Methods to reduce the grip-stress on the wire have been described by Adams and Anderson [68A1] and by Harman [69B5]. The former described the use of a black wax and a separate heating element to liquify the wax at the end of a probe which engulfs the wire by capillary action. The ability to apply a tensile force of 150 gf was reported. The latter described the use of a high-tensile-strength, hot-melt glue with a discrete melting point, at the tip of a nichrome wire-loop probe. The glue is melted by passing current through the wire loop which was electrolytically thinned so that most of the joule heating occurs at the tip. The tensile strength and adherence of such glues are sufficient to test 1-mil diameter aluminum wire if a length of about 0.13 mm is bonded with the glue. The glues do not adhere as well to gold wire so that a longer length of wire must be bonded.

2.3. Nondestructive, Double-Bond Test

The nondestructive test consists of pulling on the wire span until a predetermined force is applied. The test is meant to be nondegrading as well as nondestructive to satisfactory wire bonds but destructive to those bonds that are unsatisfactory. Such a test has been suggested as a 100 percent screen test by Slemmons [69S1] and Ang et al. [69A1]. То substantiate the claim that the test is nondegrading to those that pass, both papers show that after a group of wire bonds has been so tested and then pulled to destruction, the bond strength frequency distribution is merely truncated at the preselected stress level.

Polcari and Bowe [71P3] have reported the results of some preliminary evaluations of a nondestructive pull test. They concluded that while the nondestructive test could be a valuable reliability tool the proper use and adjustment of the tester would be of critical importance. Furthermore, if the nondestructive pull test were to be implemented, they recommended that the manufacturer control loop heights to minimize variations in the stress imposed by the test.

There has been skepticism that the test is actually nondegrading. The idea of using devices whose wire spans have been pulled and, in the process, altered in shape is new and disturbing to some. On the other hand, there does not seem to be a similar reluctance to ase devices stressed in the centrifuge test described in the next section.

3. Centrifuge Test*

In the centrifuge test a constant centrifugal force stresses the wire bonds. Poorly adhering wires are expected to rupture while wires which have a large loop or which have been improperly placed are expected to shift so that they will make electrical contact with adjacent parts of the device. The resultant open or short circuits are detected by subsequent electrical tests.

The stress applied to the wire bond is dependent on the maximum acceleration (expressed in gravity units, g's†), the shape of the wire loop, and the direction of the centrifugal force relative to the wire loop. The typical range of centrifugal forces used is from 30,000 to 50,000 g's. The typical duration at the maximum stress level is one minute in any one direction of applied centrifugal force. The directions are chosen so that the force is directed either away from, toward, or parallel to the bonding surfaces. The test is often preceded by other tests intended to weaken unreliable wire bonds and thereby promote their failure in the centrifuge test.

For the case where the centrifugal force is directed away from the bonding surfaces, the tensile stress in the wire at the heel of the bond on the terminal, Fwt, and on the die, Fwd, are given in grams-force by the following relations where it is assumed that the centrifugal force is sufficiently large so that the wire loop takes the shape of a catenary:

[merged small][merged small][merged small][merged small][merged small][ocr errors][merged small][merged small]

h = vertical distance between the terminal contact surface and the peak of the wire loop after the centrifugal forces have deformed the loop to describe the catenar curve (cm),

H = vertical distance between the terminal and die contact surface (cm),

[blocks in formation]

G = centrifugal acceleration (in units of gravity),

and where a is given by the relation h + H + a = a cosh (D/2a) in which D/2 is the lateral distance between the bond at the die and the apex of the wire loop. For d 2(H + h), a good approximation for a is given by

[blocks in formation]

A graphical representation of eq (8) is shown in figure 4. Here, the tensile force the wire at the heel of the bond of a single-level, l-mil diameter wire bond subjected to centrifugal force of 10,000 g's is shown as a function of d for different values of d/h.

*[64U1], [65C5], [65R1], [66G1], [66L4], [66P1], [67G1], [68B1], [68D2], [6811], [68R2], [69B2], [6901], [69S4], [70D3].

†See footnote on page 1.

[blocks in formation]

d (mm)

2.0

0.05

10

8

0.04

6

0.03

0.02

0.01

Fw(gf)- aluminum wire

Figure 4. Tensile force, Fw, in the wire at the heel of the bond of a single-level, 1-mil diameter wire bond subjected to a centrifugal force of 10,000 g's for different values of d/h. The left and right-hand vertical scales are for gold and aluminum wire bonds, respectively.

The tensile force for gold wire is shown on the left vertical axis and for aluminum wire on the right vertical axis. As can be seen, Fw increases as d is increased but Fw decreases as h is increased. Fw begins to increase with h for hd/3.

To use the centrifuge test for accelerations much greater than about 50,000 g's requires that special fixturing be used to hold each device and incurs the risk of damaging other satisfactory components of the device. Therefore, there are practical limitations on the maximum acceleration that can be used. Because of this and the low density of aluminum, only gold wire bonds can be screened satisfactorily with centrifuge tests. While the centrifuge test is widely used for this purpose there are some who believe that even for gold wire the centrifuge test is only acceptable for culling out grossly defective wire bonds [67A1]. For example, using figure 4 and considering a gold wire bond with a separation of 0.15 cm and a loop height of 0.015 cm the tensile force in the gold wire for an acceleration of 30,000 g's will only be about 0.55 gf. As a basis of comparison, such a tensile

stress would be produced in a pull test if the hook, placed at mid-span, were pulled with a force of 0.25 gf.

Judging whether a centrifuge test may be useful, as with many of the other tests, is dependent on the values of d and h of the wire bonds to be tested. When d is sufficiently large and h sufficiently small the test may stress the wire bonds sufficiently in those cases when the centrifugal force is directed away from the bonding surfaces. For tests where the centrifugal force is directed either into or parallel with the bonding surfaces no calculations on the stresses imposed have been found in the literature. Again, however, it would be expected that the larger is d the greater will be the stress applied to the wire bond simply because of the greater mass of the wire involved.

4. Mechanical Shock Test*

In a typical test the device is first accelerated either by free-fall or by pneumatic means and then brought to a sudden halt on striking an impact pad. The test conditions that may be specified are the maximum deceleration (usually a few thousand g's), the duration of the impact or shock pulse width (between 0.1 and 1.0 ms), the direction of the stress (usually along one or more of the principal axes of the device package), and the number of shocks per direction (usually less than ten).

In essence, the basic stresses imposed by a mechanical test to a wire bond result from the amplitude and the number of vibrations induced in the wire loop by the mechanical pulse. The number of vibrations depends on the damping of the wire loop. Thus, if the wire has been softened, through earlier exposure to high temperatures, fewer oscillations with smaller amplitudes occur and the resulting flexure stress may be less. However, irreversible changes occur in annealed wire at smaller bending and torsional stress than in unannealed wire.

To discuss the test in terms of the magnitude of the induced deflection and stress, it is useful to define a dynamic-to-static deflection or stress ratio, K:

K

=

maximum wire deflection or stress by a mechanical pulse
wire deflection or stress induced if peak acceleration
of mechanical pulse were applied statically

As a rule of thumb, the upper bound for K is 2 [48F1], [65R3, p. 368], [6711]. How much less than 2 K is depends on the shape and duration of the pulse and on the lowest resonant frequency of the wire loop. However unless the product of the shock duration and the reso nant frequency of the wire bond is less than about 0.4, K is not less than 1 [6711]. The results of estimates made of the lowest frequency of wire bonds [72S2] indicate that this might occur for 1-mil diameter gold wire bonds with a bond separation, d, greater than about 2 mm and for 1-mil diameter aluminum wire bonds with d greater than about 3 mm.

Changes in the shape of the excitation pulse will produce differences in K. Consequently the International Electrotechnical Commission (IEC) in its recommended test method

*[6711], [68D2], [69B2], [69S4], [70D3].

« PreviousContinue »