METHODS FOR TESTING WIRE-BOND ELECTRICAL CONNECTIONS*

Harry A. Schafft

Abstract

A significant fraction of the failures that occur in integrated circuits are due to failures of the wire-bond electrical connections that are used. Therefore, a critical area for reliability improvement is in the methods for testing and evaluating wire bonds. Several of these methods are surveyed. In particular, analyses with regard to the stress that the test imposes on the wire bond in the pull, centrifuge, mechanical shock, vibration, and temperature cycling tests are presented and used in discussing the capabilities and limitations of these methods.

Key Words: Bonding; electrical connection; failure (wire bond); integrated circuits; microelectronics; reliability; semiconductor devices; testing (wire bond); wire bond.

1. Introduction

Many microelectronic devices use wire bonds to electrically connect the semiconductor die and the package terminals. The diameter of the wire is typically 1 mil.† Unless special precautions are taken, a significant fraction of the failures that occur in these devices is due to wire bond failure. Because of the increasingly large number of devices used in many present-day electronic systems, the reliability of the individual devices must be increased, even from present-day levels. Hence there is considerable interest in methods for testing and evaluating wire bonds.

The term wire bond, for the purposes of this paper, includes all the components of the die-to-terminal electrical connection: the wire, the metal bonding surfaces, and the adjacent underlying supportive material. It is customary to speak of the bond as that part of the wire bond that is associated with the volume of the wire deformed at the weld or attachment point and to speak of the heel of a stitch or wedge bond as that part of the wire at either end of the wire span which is deformed by the edge of the bonding tool. The purpose of this paper is to review aspects of the following tests: pull, centrifuge, temperature cycle, thermal shock, mechanical shock, variable frequency vibration, and #ibration fatigue. In particular, these tests are examined with regard to the stress that

This paper was prepared for presentation at the Third Symposium on Reliability in Electronics, November 13-16, 1973, in Budapest, Hungary.

The data referenced in this paper that are not given in the International System (SI) of nits are followed in parentheses by the values in the appropriate SI unit. General usage dictates that three exceptions be made: (1) acceleration is given in units of gravity (1 g = 9.8 m/s2), (2) the wire diameter is given in mils (1 mil 25.4 μm), and (3) the force exerted on the wire or wire bond is given in grams-force (1 gf

=

=

9.8 mN).

the test applies to the wire bond and what that implies about the capabilities and limitations of the test. These and other test procedures, including visual inspection, have been discussed in detail elsewhere [72S2]*.

2.1. Destructive, Double-Bond Test+

2. Pull Test

The destructive, double-bond test consists of pulling on the wire span by some means (usually with a hook) with increasing force until a rupture in the wire bond occurs. The pulling force required to produce rupture is called the pull strength, and it is used as a measure of quality for the wire bond.

In the test, pull strengths of a sample are taken to be representative of the group. Usually, the location of the rupture is recorded as well as the pull strength which is often expressed in grams, although grams-forces are implied. To facilitate analysis of the data, the distribution of pull strengths can be displayed in a histogram. The magnitude of the pulling force at the peak in the distribution displayed in the histogram gives an indication of the general ruggedness of the wire bond while the spread of the distribution indicates the uniformity of a group.

The actual stress in the wire bond is the tensile force in the wire. If the geometrical variables are defined as in figure 1, wire tensile forces on the terminal side, Fwd, and on the die side, Fwt, are related to the applied pulling force by

It is assumed that the pulling probe is in the plane of the wire loop but inclined at an angle with respect to a normal to the bonding surface. The angles et and ed are the contact angles that the wire makes with the bonding surfaces of the terminal and die, respectively. The ratio Fwt/F as a function of t is given in figure 2 for = 0 and various fixed ratios of alet. The ratio of Fwd/F as a function of d may be obtained by interchanging the subscripts d and t in figure 2.

Although the geometrical dependence is easily visualized in terms of the contact angles, these angles are difficult to measure. The contact angles depend on the height, h, of the wire span above the terminal contact surface, the height difference, H, between the

*Alphanumerics in brackets indicate the literature references at the end of this paper. +[66R1], [67H1], [67S3], [68D2], [69B5], [69K1], [70A1], [7184].

See footnote on page 1.

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

PACKAGE (HEADER)

Figure 1. Geometric variables for the double-bond pull test.

t

ве ф

=

0.

gure 2. Dependence of Fwt/F on t for various ratios ed Horizontal bars above and below the curves for 20d et and ed 20t show the Eect of changing to plus and minus 5 deg, respectively. To obtain the dependence of /F, interchange everywhere the subscripts d and t and change the sign of .

to ot•

The curves are for the

70°

die and terminal contact surfaces, and the horizontal distances, ad and (1 - a)d, from the bonds to the point at which the wire span is contacted by the pulling probe. eqs (1) and (2) in terms of these quantities we have

Expressing

For a normal pulling force applied at mid-span to a single-level wire bond (= 0, H = 0, α = 1/2) these equations simplify further to:

The dependence of F/F and F/F on d/h in eq (7) is graphed in figure 3.

The experimentally determined dependence of pull strength of ultrasonic aluminum wire bonds on the geometrical variables has been compared with the predictions of eqs (3) and (4), where it is assumed that the tensile force to rupture the wire bond remains constant as the variables are changed.* In general the experimental results on unannealed singlelevel wire bonds agreed with the theory. However, differences were observed in the depen

dence of pull strength on loop height, h, for wire bonds with loop height greater than 1/3 the interbond spacing, d. It is felt that this is caused by a weakening of the wire bond because of bond peeling as a result of the relatively high loop heights. Because of the

*[71B1], [71B4], [72B1], [72B2], [72B3], [72B4], [73B1], [73B2] by A. Sher will be published as an NBS Technical Note.

Figure 3. F/F and F/F as functions of d/h for a single-level, double-bond pull test.

elongation of the wire during the test, all annealed wire bonds studied had loop heights greater than 1/3 the interbond spacing at the time of failure, and hence the experimentally obtained values for pull strength differed from the predicted values.

While the pull test is the most widely used test for wire bonds it is also perhaps the most under-specified. To maximize the usefulness of pull test data, especially when used to evaluate and compare wire bonds, it is important to specify fully the wire bonds tested and the test conditions as well as the failure mode.

It is necessary to have information about the average shape of the wire span* because significantly affect the pull strength measured. To indicate the magnitude of the dependence of the pull strength on the shape of the wire span in a double-bond pull test,

it can

*In cases where the elongation of the wire during the test is large, such as in annealed wire, it is necessary to take this into account when providing information about the shape of the wire span.