IEEE TRANSACTIONSON COMPONENTS,HYBRIDS, AND MANUFACTURING TECHNOLOGY, VOL. CHMT-9. NO. 4, DECEMBER I986 379 Effect of Bromine in Molding Compounds on Gold-Aluminum Bonds SYED SAJID AHMAD, RICHARD C. BLISH, 11,. MEMBER, IEEE, TIMOTHY J. CORBETT, JERROLD L. KING, AND C. GLENN SHIRLEY, MEMBER, IEEE Abstract-Degradation rates of gold wire ball bonds on aluminum bonding pads were studied in two molding compounds as a function of bromine concentration, temperature, and time at temperature. The measure of degradation was the resistance increase of the aluminum-gold contact. The rate of degradation was observed to increase with increasing bromine content and temperature for both molding compounds. How- ever, the bromine content dependence of the activation free energy of the degradation reaction indicates a different degradation mechanism for each molding compound. INTRODUCTION I N PREVIOUS STUDIES [I]-[41 it has been shown that the presence of bromine in flame retardant epoxies accelerates the degradation of gold-aluminum wire bonds. In this study the presence of various proportions of flame retardant materi- als in molding compounds has been correlated to gold- aluminum bond degradation rates. Earlier work [l] employed a nondestructive technique of measuring changes in the resistances of wire bonds by a four- point (Kelvin) method. This technique is improved in the present work by using a more efficient bonding scheme. EXPERIMENTAL DETAILS Bond degradation was measured nondestructively using a special test chip. Variations in Kelvin resistance of bonds as a function of temperature and time provided degradation rates of the bonds. Special dice were prepared with unpassivated Al/l percent Si metallization over oxidized silicon. These dice were attached to 40-lead leadframes (Fig. 1). These leadframes had gold-plated lead fingers for gold wire bonding. The dice were wire bonded with 1.3 mil gold wire. The bonding scheme (Figs. 1, 2) produced four wire Kelvin measurement modules capable of measuring the resistance of one wire in series with the resistances of ,the Au-Al interface at the chip end of the wire and the Au-Au interface at the leadfinger end. As only the Au- Al interface degrades appreciably due to thermal aging [l], therefore, changes in the measured0 resistance are taken as a measure of degradation of Au-Al interface. This arrangement saves the trouble of elaborate calculations as was necessary in the bonding arrangements previously used [ 11. In this arrange- Manus 682 ript received March 22, 1986; revised July 10, 1986. This paper was pr nted at the 36th Electronic Components Conference, Seattle, WA, May 5-7, 1986. The authors are with Intel Corporation, 145 South 79th Street, Chandler, AZ 85226. IEEE Log Number 8611132. \ 35 34 . . n 2 33 32 Fig. 1. A is specially prepared die. Its surface consists of unpassivated Al/l percent Si metallization over oxidized silicon. Pins 1 and 21 make Kelvin modules in conjunction with pins (2, 3), (4, 5), . . . , (14, lS), (16, 17), and (22, 23), (24, 25), . . . , (34, 35), (36, 37) to measure resistance of gold- aluminum bonds a, b . . . g, h and i, j, . +., o, p, respectively. Squares represent gold-gold bonds while circles represent gold-aluminum bonds. Typical module is described in Fig. 2. Pins 16-20 and 36-40 provide four gold-gold reference modules as described in Fig. 3. , ment, resistance of a bond is simply obtained by dividing electrical potential by the forced current. A minimum value of current was chosen in such a way that the resistance repeatability fell in the range of f 0.1 mQ or less. Measure- ments were taken at room temperature and were not done at the storage temperature. Enough time, a minimum of 8 h, was allowed for the modules to reach a steady temperature so that all readouts are taken at the same temperature because the resistance depends on and varies with the temperature at the time of measurement. Each die had 16 gold-aluminum Kelvin modules and 4 gold-gold Kelvin modules for reference (Figs. l-3). A minimal resistance increase has been seen previously in gold-gold reference bonds [l]. Reference bonds were used to test equipment malfunction, misconnections, or similar anomalies. A Keithley 350i parametric test system was used for resistance measurements. The bonding diagram is symmet- ric. This helped reduce any adverse effect on measurement results if a device were reverse-socketed during a measure- ment. Current forcing and voltage measuring terminals were double bonded for redundancy. Two flame retardant molding compounds, henceforth called compounds X and Y, were used for encapsulation of these Kelvin test chip devices. Vendor provided each compound 0148-641 l/86/1200-0379$01 .OO @ 1987 IEEE