First Bond Failures in Leaded Packages
12/01/2003
BY DICHEN JAMES HUA, PATRICK KO AND RAFAEL HUERTA
A fabless semiconductor company using an outsourcing model relies on a foundry for wafer fabrication, a testing house for wafer sort and an assembly house for chip assembly. The multiple manufacturing processes involved make it difficult for the fabless company to isolate the cause of an assembly failure. This article addresses bond lift — the most common wire bonding failure for leaded packages. Two major causes of this type of failure are examined — bond pad contamination and bond pad damage at wafer sort. Cooperative solutions to each of these issues are also addressed.
In leaded packages, a gold wire is used to connect a bond pad on the die to a lead finger. A first bond refers to the joint between one end of the gold wire and a bond pad, while a second bond is the joint between the other end of the gold wire and a lead finger (Figure 1). During the bonding process, a gold ball at the tip of the gold bond wire is attached to a bond pad on a die (silicon chip) by forming a joint of gold and aluminum compounds on the pad. Then, the other end of the wire is attached to a lead finger. Two important conditions for reliable gold ball bonding are cleanliness of the pad and minimal damage to the bond pad during wafer sort. A first bond failure occurs when the gold ball bond lifts from the bond pad.
Bond Pad Contamination
Generally, a suitable bond forms if a gold ball is attached by thermosonic bonding to a clean aluminum pad. An Au-Al bond forms fully over the ball attach area. The quality of the bond is subjected to a ball shear test. If the pad is contaminated with foreign substances, the area may not form a good Au-Al joint and a weak bond may result. A weak bond will be stressed by the pull of the gold bond wire during the second bonding process, which may cause the first bond to lift. This may also occur during the wire pull and ball shear test. Bond pad contamination is usually caused by silicon dust, aluminum oxidation or etchant residual such as fluorine from a wafer process. Contamination can also be a combination of these substances.
Silicon dust contamination. The process of sawing a silicon wafer into individual die (chips) generates silicon dust. It is difficult to remove all of the residual dust completely from the bond pad using deionized water, especially on corners. One solution is to adjust the deionized water pressure to force the dust away from pad corners.
In a case study of silicon dust contamination, bondability was compromised by not only silicon dust, but also fluorine contamination and wafer probe damage. Pad abnormalities involving probe damage and certain contamination were first detected at the assembly house during wafer incoming quality check. Due to production urgency, the wafer lot was released to wafer saw. Problems arose during the wire bonding process when the sample bonding test could not clear the wire pull and ball shear criteria, and the ball lifted. The wafer foundry performed a failure analysis to find the problem's cause. Silicon dust contamination was found to be causing the bonding failure. Given this information, the assembly house examined their wafer sawing process and proposed a corrective plan to reduce silicon dust contamination. The assembly house conducted experiments and provided a corrective option plan (Table 1). Option 3 of the corrective plan, which is shown in Table 1, achieved the best results in cleaning silicon dust from bond pads.
 Figure 1. Leaded package cross section. |
null
Oxidation contamination. Oxidation on the pad is caused by a chemical reaction between the aluminum of the pad and the oxygen in ambient air. Oxidation impairs the formation of an Al-Au joint in the bonding process. Although a wafer is normally stored in a nitrogen-inerted cabin to prevent oxidation, improper handling may result in oxidation. Improper handling includes failure to maintain the nitrogen purity level of the wafer storage cabin, in which case the cabin may contain enough oxygen from the air to gradually oxidize the wafer's bond pads. Another type of mishandling is frequent exposure of wafers to air. This occasionally happens when a process engineer needs to study the wafer between multiple wafer probings. Most foundries contend that oxidation contamination is a result of improper storage after the wafers leave the foundry; they are often correct.
A solution to oxidation contamination is to check wafer storage conditions for nitrogen purity levels. An important, practical method of isolating a wafer from oxygen and moisture is to vacuum dry pack each case of wafers, and then put the case in a nitrogen cabin. Using this process, a wafer can be safely kept for 1-2 years — even if the purity of the cabin changes from opening the door. In addition, test engineers must strictly follow wafer storage requirements. Once a wafer becomes oxidized, a plasma cleaning is necessary to remove the oxide layer.
In a case study of oxidation contamination, a semiconductor company experienced a downturn and inventory was moving relatively slowly. During these conditions, wafers tend to be "old" before they are assembled. The older a wafer, the more likely it is to become oxidized. The company discovered an oxidized wafer at the assembly house, stopped before bonding and sent it off to for plasma cleaning. The clean wafer was bonded, and the pull and shear data indicate that bonding quality improved.
Fluorine contamination. Wafer pad contamination can occur at a foundry during wafer fabrication. Recently, fluorine residual was left on the bond pad after the pad passivation etching process. Fluorine is an active chemical, which within the time period of three months to a year can cause a thin film of fluorine compounds (Al, F, C and O) to form over bond pads.
 |
null
 Figure 2. a) Fluorine-contaminated pad. b) Fluorine-contaminated pad and EDX results. |
null
In a case study of fluorine residual on the bond pad, a foundry was using SF6 in its etching process to remove passivation on the aluminum pads. The assembly house reported contamination on three separate wafer lots. Given the high possibility of first bond lift failures, the wafers were returned to the foundry for failure analysis. The foundry detected fluorine content on the pad surface. Unconvinced that the fluorine was from its process, the foundry asked that the wafer storage history be checked from the time of fabout to wafer check at the assembly house. The assembly house confirmed that wafer storage conditions were acceptable, and discovered through other foundries that fluorine is routinely used in wafer fabrication. The foundry was asked to accept responsibility and investigate the cause for fluorine inclusion after the etching process (Figure 2a and b).
 |
null
Probe Damage
Wafer probe damage may result from excessive overdrive of a probe during the wafer sort process. These probes are made of tungsten, and can harm the Al pad of a wafer. When the probe drives too much and exposes the material beneath the pad, which is normally TiN, bondability is affected. Three key things need to be done to prevent this type of damage. First, prior to wafer sort, the wafer testing house must examine the bond pad's condition to determine whether contamination, normally in the form of oxidation, is present. If contamination is present, the pad must be cleaned before wafer probe. If the contaminated wafer is probed before cleaning, a normal probe overdrive may result in an "open" circuit. To solve this problem, an engineer or operator may wrongly set overdrive to an excessive degree and cause probe damage. Second, engineers must set an appropriate overdrive for probing. If it is set too low, the probing can result in an "open" circuit that will impact wafer sort yield. Third, operators must be careful when probing wafers, and adjust the overdrive based on actual probe results.
The probe mark area is another factor affecting bondability. If the area is too large, pad roughness will worsen and a gold ball will fail to form a sound joint with the Al surface. The mark area is normally set at one fourth of the whole area.
 Figure 3. More than 25 percent of this wafer pad shows severe damage. |
null
A solution to probe damage is to decide whether the affected wafer pad is bondable. If an assembly house experiences difficulty in handling bonding, it is better to try another assembly house to ensure high assembly yield. Once an assembly house reports damage, instruct them to test bond some die, measure the wire pull and ball shear results and determine whether they meet specifications.
The first case study involved normal damage, since the damaged pad area was less than 25 percent of the entire pad. However, the originally assigned assembly house was not confident they could achieve bonding. To meet their production schedule, the company moved the task to a different assembly house. A test bonding proved that the new assembly house was capable of handling the situation, and assembly was completed with a satisfactory yield.
A second case study involves severe wafer pad damage, where more than 25 percent of the whole pad was damaged (Figure 3). Wafer test was done at a foundry, so the production engineering team was asked to investigate the cause of the damage. The investigation revealed that this lot had special handling by engineering, which included excessive overdrive of the probe on the bond pad, as the wafer test line was newly established. These pad-damaged wafers were not assembled and were returned to the foundry.
Conclusion
Diligence and a cooperative effort among all participants are required to prevent bonding failures. When bonding failures occur, these steps should be taken:
1) Maintain constructive and cooperative relationships with the foundry, assembly house and test house.
2) Appropriate deionized water cleaning is necessary for silicon dust contamination. A test is needed to set up water pressure, time and volume for this cleaning.
3) Plasma cleaning is helpful to remove oxidation on the pad to ensure that the bonding ball passes ball shear and wire pull tests.
4) Investigate the foundry's etching process when fluorine contamination occurs. Often, fluorine residual forms chemical compounds on the pad with oxygen (in the air), carbon (also from a foundry's process) and bond pad aluminum. If the degree of fluorine content affects the bonding, the water should be returned to the source.
5) Appropriate action is determined by the severity of damage in wafer probe damage.
6) First bond failures may be caused by a combination of all of the above factors.
DICHEN JAMES HUA, senior assembly engineer, PATRICK KO, assembly engineering manager, and RAFAEL HUERTA, failure analysis engineer, may be contacted at Integrated Silicon Solutions Inc., 2231 Lawson Lane, Santa Clara, CA 95054; (408) 969-6600; e-mail: [email protected], [email protected] and rafael_ [email protected].