Learning from microvia failure in lead-free assembly: a trio of companies studied a set of microvias that failed during lead-free assembly.

Traditionally microvias have been considered to be the most reliable interconnect structure within a PCB. But with the advent of lead-flee assembly, the vulnerability of high-density interconnects to failure has increased. This is due to the elevated temperatures experienced during assembly and rework. Microvias have been found to fail during assembly and in their end-use environment, a trend noticed in North America and Europe in spring 2003.

Curtiss-Wright experienced opens (and intermittent opens) at electrical testing of their PCBs after assembly. Assembly includes double-sided surface-mount components in additional rework procedures when applicable. Assembly and rework procedures were well defined and tightly controlled. The opens demonstrated sensitivity to thermal change and mechanical stress. Specifically, connections became resistive or open under thermal or mechanical stress. Using infrared thermal imaging techniques the defective interconnections were located, and subsequent microscopic evaluation demonstrated random failures associated with microvias. The first microvias reviewed demonstrated a target pad to copper plating separation. A small percentage of microvias exhibited barrel cracks and, on occasion, knee/corner cracks.

These failures were found to be related to specific fabrication date codes and specific vendors. Some of the boards with failing microvia were fabricated by Coretec and were certified compliant to established acceptance criteria. The date codes involved had been tested at PWB Interconnect Solutions and met the established criteria for acceptance for thermal cycle testing of IST coupons.

The reliability test method used throughout this study was interconnect stress testing (IST), as per IPC TM650--2.6.26, DC Current Induced Thermal Cycling Test. The associated coupons were preconditioned by exposing them to five thermal excursions to 230[degrees]C, in exactly three minutes. This method of preconditioning was established as the standard for Curtiss-Wright based on a significant volume of data points used in previous testing. It is meant to emulate, but not replicate, three assembly cycles and two rework cycles. IST thermal cycling to 150[degrees]C was completed on the suspect lot of microvia coupons, the results surpassed the customer's acceptance criteria of a lot mean of 150 cycles, and a minimum of 100 cycles to failure. The suspect lot had achieved a mean of 750 cycles with some coupons achieving a maximum of 1,000 cycles (end of test).

In support of IST evaluations, microsections were used to determine the type and cause of failure. The microsection methods used were in accordance to IPC -650 2.1.1 Microsectioning Manual Method. Microscopic evaluations of microvia were preformed after a mild microetch.

A multidiscipline study was initiated by Curtiss-Wright and undertaken to understand the various types, and mechanisms contributing, to microvia failures focusing on:

1. How to prevent reoccurrence of microvia failures.

2. How to improve test methods to identify suspect microvias and accept reliable microvias.

The joint effort between Curtiss-Wright (OEM and assembler), Coretec (PWB supplier) and PWB Interconnect Solutions was established to resolve these microvia reliability concerns.

This investigation included improvements to metallurgical methods (metrology) necessary to identify and evaluate microvias, analysis of various process parameters required to eliminate the underlying cause of these failures, development of reliability test methodologies to quickly and accurately determine the reliability of microvias and the establishment of criteria for product acceptance or rejection. This study was extended knowing that stress of thermal excursions in a lead-free environment would be significantly higher.

During routine assembly at Curtiss-Wright, a trend was identified where some PCBs had circuits that were open while others exhibited high resistance. This condition was found during testing and rework on PCBs with a specific date code. The first step was a careful review of the assembly and rework processes for any anomalies in practice or procedure. All processes and process parameters were found to be within acceptable limits. Controls were found to be in place and appropriate. There was no evidence of elevated or prolonged thermal excursions. It was demonstrated that the established assembly and rework procedures did not degrade the integrity of the PCBs.

It was concluded that the established IST testing at 150[degrees]C did not adequately identify the level of quality on the failing microvias. The coupon from the failing lots had exceeded the established minimum requirements of a mean of 300 and no coupon undergoing less than 150 IST cycles to failure. The failing lot achieved a mean of 782 cycles with a minimum of 563 cycles to failure; one coupon survived to end of test at 1,000 IST cycles.

It was determined that two products were affected: one, already partially delivered and another was a "development model." Delivered product was recalled; both produces were scrapped as per procedure.

In response to these developments two parallel actions were launched. Curtiss-Wright initiated failure location, microsectioning of suspect interconnection with failure analysis on a defective PCBs that had been identified as having several sites of opens or higher resistance interconnections. Microsections indicated some microvias with separation of the target pad and the base of the microvia. The second activity was initiated by PCB Interconnect Solutions to determine how this particular defect could be better detected. A proposal was accepted to explore the possibility to use the remaining IST coupons in an experiment to determine what temperatures may be used to improve the detection of defective microvias.

In conjunction with these actions a more robust material was implemented for PCB fabrication that was better suited to withstand the rigors of assembly and rework, and had a lower coefficient of thermal expansion (CTE) that would reduce the strain generated during the thermal excursions associated with assembly and rework.

The Test Vehicle

IST coupons are specifically designed to have interconnect structures that are sensitive to increases in resistance. A proprietary method allows IST coupons to be engineered to focus on the robustness of the interconnection of interest. The coupon is also a design large enough to contain hundreds of interconnect structures in order to test a statistically significant sample of interconnects. IST coupons have at least two circuits: a power circuit, P1, used to heat the coupon (and test internal interconnections), and a sense circuit, $1, for testing the interconnect structure of interest. The coupon used in this study, TV16002 (FIGURE 1), has one power circuit, P1, with a PTH interconnect (layers 1/16) and 3 senses circuits where S1 is a PTH (1/16), S2 is a core via (2/15) and S3 is a microvia (1/2 15/16). Each TV16002 coupon contains 340 microvias.

[FIGURE 1 OMITTED]

In an IST evaluation, testing automatically stops when a circuit reaches a 10% increase in resistance. The failing coupons have circuits that are still electrically conductive but have a modest increase in resistance. Because the test stopped well before catastrophic failure (open), it is possible to identify the exact microvia that is contributing the highest resistance to the circuit using thermo-graphic techniques.

The one worst-case microvia in a group of 340 may be identified using thermal imaging cameras in a technique called "failure location." Since the failed circuit has experienced only a 10% increase in resistance, a DC current can be applied causing the failing interconnection to heat. The compromised microvia has a higher resistance, compared to other connecting traces and other (robust) microvias, and it will become the hottest structure in the coupon, and easily found using a thermographic camera. The thermographic camera allows direct visualization of the exact location of the hottest microvia. The worst-case failing microvia is seen as a high-temperature hot spot on the surface of the coupon (FIGURE 2). This is a powerful tool for finding failure locations for subsequent microsection analysis.

[FIGURE 2 OMITTED]

Failure Analysis of Microvias

The established method for evaluating microvias by microscopic technique requires that the microvia to target pad interface to be examined without the use of a microetch. Examining an unetched cross-section of a microvia can be used to good effect in fabrication; however, with thermally stressed microvias in the process of failure, it is more effective to used a mild microetch. Typically a mild microetch is used on microsections to enunciate internal structures within the interconnection. With a microetch, the crystalline structure of the electrolytic copper, the layers and thickness of the electroless copper and microinclusions within or between copper layers are easily seen. It is necessary, for the effective evaluation of a microvia that is beginning to fail and has only a 10% increase in resistance, to use a mild microetch for failure analysis. An aggressive microetch can hide subtle internal structure and create artifacts that could be confused with known defects.

FIGURES 3a and 3b show a microvia that was photographed before and after etching to demonstrate the degree of detail that may be achieved. In Figure 3a the layer 1 foil interface, electroless copper, interface between the bottom of the microvia and the target pad, and the plating on the top of the target pad are easily visualized after a microetch. In Figure 3b all internal detail is not visible. Microvias evaluated in this study had a mild microetch. It was demonstrated that a well-controlled microetch, using hydrogen peroxide and ammonium hydroxide, greatly enhanced the physical details of the failing interconnect and improved the ability to objectively evaluate failures and causative mechanisms.

[FIGURE 3 OMITTED]

IST equipment has the inherent ability to stop testing an individual coupon automatically when it achieves a 10% increase in resistance, in any test circuit. This assures that the failing circuit is stopped before the coupon has achieved a catastrophic failure. FIGURE 4 is an obviously failing microvia, while FIGURE 5 has a more subtle failure mode. Both coupons were stopped at a 10% increase in resistance and then identified, with thermal imaging techniques, as the microvias contributing the highest resistance to the failing circuit.

[FIGURES 4-5 OMITTED]

Armed with these tools, a comprehensive failure analysis can be achieved. Failure analysis suggested that there were two distinct failure mechanisms at work in these coupons. As seen in FIGURE 6, barrel cracks were found at the base of the hole wall of a few microvias. More commonly debris was under or above the electroless copper deposit, at the interface with the target pad leading to pad separations. The analysis suggested that the dominant underlying contributor to the microvia failures in this study was an inherent weakness at the interconnection between the base of the microvia and the target pad. The electroless and electrolytic copper thinned from the knee to the bottom of the microvia and was believed to he a contributing factor to the barrel cracks.

[FIGURE 6 OMITTED]

Characteristics of Failure

After analysis of a large number of failed microvias, some common characteristics became apparent. One of the characteristics of failure due to a target pad separation is that they usually exhibited inclusions between the electroless copper and the target pad, or between the electroless and the electrolytic copper (FIGURE 7). Because we are viewing failing microvias that have not failed completely (fully separated), we have the opportunity to review the failure in progress. The microsection captures the failure's moment in time. We were able to review the most electrically degraded microvia and the neighboring microvias that are in various stages of failing.

[FIGURE 7 OMITTED]

Frequently failing microvias exhibited precursor "black dots" at the target pad/via-plating boundary. The black dots enlarge into small cracks that then coalesce into large cracks. This process resembles a tearing along perforations in a piece of paper. Often the cracks would appear in the center and at the perimeter at the base of the microvia. Microvias that were advanced in the failing process had the appearance of a crack that started on one side of the base of the microvia and progressed completely across the interface.

FIGURES 8 and 9 exhibit the precursor black dot above and below the electroless copper. FIGURE 10 shows coalescing micro-cracks both above and below the electroless copper.

[FIGURES 8-10 OMITTED]

Another observation is that the electroless and electrolytic copper quickly thins below the knee of the microvia. Electroless that measured 120 millionths of an inch on the surface of the pad can be unmeasurable at the base of the microvia. The thickness of the electroless and electrolytic coppers was reduced, in some cases dramatically, as the deposit descends into the microvia. The effect of thinning electroless can produce a condition commonly called step plating.

FIGURE 11 demonstrates the reduction in electroless, which is plainly visible on top of the surface foil and only barely discernable at the base of the microvia. FIGURE 12 is an extreme example of wedge plating on a microvia processed with direct metalization. Usually microvia barrel cracks are associated with process conditions that are rejectable to IPC-6012 or other applicable documents. If these conditions were to be discovered using random microsections, the lot would have been rejected. The problem with thinning copper plating is that this condition is random and may occur anywhere on the board. Not all microvias would have the exact same condition. With the failure location method, it is very successful in finding this condition when it is random in nature.

[FIGURES 11-12 OMITTED]

It was found that the shape of the microvia was also contributing to the robustness of a microvias. Bowl-shaped microvias appeared to have a more uniform copper distribution to the base, as compared to cylindrical microvias (FIGURES 13 and 14). Bowl-shaped microvia were found to achieve higher cycles to failure than cylindrical microvias.

[FIGURES 13-14 OMITTED]

Armed with improved microsection techniques and a better understanding of the failure modes Coretec investigated methods of improving microvia fabrication and improved reliability.

Part 2 will be featured in an upcoming issue of PCD&M.

PAUL ANDREWS is quality specialist for PWBs and final quality control manager at Curtiss-Wright Controls Embedded Computing. He can be reached at [e-mail]. GARETH PARRY is CTO of Coretec in Ontario, Canada. He can be reached at [e-mail]. PAUL REID is program coordinator at PWB Interconnect Solutions in Ontario, Canada. He can be reached at [e-mail].