The Story Behind the Red Phosphorus Mold Compound Device Failures
A family of mold compounds with red phosphorus flame retardant was introduced as an environmentally-friendly encapsulant for semiconductor devices. However, these mold compounds introduced product failures, including current leakage and resistive shorts between adjacent leads inside the leaded semiconductor device package, and resistance increases and open circuits of the wire bonds. This paper presents the family of mold compounds with the red phosphorus flame retardant, the failure mechanisms and the root cause of the failures.
In the 1990's, a new family of mold compounds for semiconductor devices, known as EME-U series, was developed by Sumitomo Bakelite. The EME-U product family included EME-7351UL, -7351UT, -7351UR, -7351UQ, -9730UC and -6730UC. It was an extension of its precursor brominated flame retardant products, such as EME-7351LS, and was expected to have better performance and reliability [1-3]. Characteristics of EME-7351LS can be found in . Sumitomo Bakelite's characterization concluded that the EME-7351UL (one product in the EME-U family), had similar properties in terms of flowability, curability and electrical properties as its precursor EME-7351LS . Its high temperature storage life (HTSL) performance was believed to be improved due to the absence of bromine and antimony, because bromine and antimony ions could accelerate gold-aluminum intermetallics, leading to premature failure of wire bonds [3,4].
The EME-U family mold compounds were presented as "green" mold compounds due to the composition of the flame retardant used in it. In the EME-U family, particled red phosphorus was used as flame retardant, instead of the commonly used bromides and antimony oxides, which are considered environmentally unfriendly . The particled red phosphorus flame retardant was purchased from Rinkagaku Kogyo (RKK), a major supplier of phosphorus flame retardants.
The red phosphorus flame retardant particles used in Sumitomo Bakelite's mold compound were coated with aluminum hydroxide (Al(OH)3) and phenol resin to prevent spreading of red phosphorus ions through the mold compound . The phenol resin was incorporated as a binder to both the aluminum hydroxide coating and the resin of the mold compound .
The coated red phosphorus flame retardant was produced in several steps (see Figure 1). The process started with the conversion of natural phosphorus stone to purified red phosphorus. A coating of aluminum hydroxide layer and phenol was then applied to the red phosphorus particles
. Subsequently, a sieve was used to attempt to control the dimension of the particles. The sieve size was initially 180 um when Sumitomo Bakelite started production . After sieving, the flame retardant particles were blended with epoxy resin, phenol resin curing agent, curing accelerator and inorganic filler to make the mold compound . Flame retardant particles could agglomerate into a large cluster of particles. This agglomeration has been observed by the authors and forms of agglomeration are shown in Figures 2 and 3.
Figure 1: Coated red phosphorous flame retardent production process (Click to Zoom)
Figure 2: One form of agglomeration of red phosphorous particles (Click to Zoom)
Figure 3: One form of agglomeration of red phosphorous particles (Click to Zoom)
Devices packaged with the EME-U series mold compound started to fail after a few months of operation . In particular, the Cirrus Logic Himalaya 2.0 ICs used in Fujitsu's hard disk drives were reported to have failures . Among the failed units, 92% were reported to have failed at or below 7 months of field operation and 90% failed at or below 3,500 power-on hours . For one of Vitesse's semiconductor devices used in network systems, failures started to occur after four weeks of operation . The causes of the failures did not appear to be dependent on the design of the semiconductor device itself, since failures occurred in various types of devices from various manufactures, for various applications.
The most common failure mode was identified as resistive short and current leakage between adjacent semiconductor package leads. In many cases, the failures were intermittent, and appeared to be dependent on humidity, voltage and temperature. Another failure mode was increased resistance or open circuit due to wire bond failures.
After field failures were reported, Sumitomo Bakelite decreased the sieve size for the red phosphorus flame retardant particles from 180 to 150 um in June 2000 , and then decreased it again to 75 um in June 2001 , although there was no product change notification (PCN) produced . Sumitomo Bakelite did consider reducing the particle size to 20 um, or even 10 um , but this was not implemented in production. However, a reduction in the sieve size was not sufficient to eliminate all oversized particles since most of the particles were not spherical 
and thus particles longer than the seize size could get through the sieve . Particles as long as 237 um have been measured by the authors (see Figure 3), which is larger than the smallest lead-to-lead gap of most IC packages. From a failed unit from IDT, a red phosphorus flame retardant particle as large as 300 um was found .
Failure Mechanism Analysis
Failure analysis was conducted on field failed devices. On the devices that had resistive shorts and leakage current failures, we observed copper and silver filaments spanning across the gap between the inner-leads of the leadframe in the plane parallel to the package (see Figure 4). This finding was also observed by others and can be found in internal reports [11,12,20,21,22]. In many cases, a pitting was observed on one of the innerleads, indicating that the copper was chemically attacked. The copper/silver filaments can provide a conductive (low resistance) path between the leads and cause the failure. The failure mechanism is identified to be electrochemical migration of copper and silver during field operation. EDS elemental analysis on the field failed samples showed a strong presence of phosphorus ions in the same location as the migrated copper and silver filaments.
Figure 4: Red phosphorous induced copper migration between adjacent leads of the packaged semiconductor device (Click to Zoom)
To assess the devices encapsulated with the EME-U series mold compound, we polished multiple packages from the top and observed the cross-sections under optical microscope. In all cases, red phosphorus particles were observed distributed throughout the mold compound. Many of them lay between the inner leads of the leadframe, as shown in Figure 5. The distribution of the red phosphorus flame retardent particles was studies in . EDS elemental mapping confirmed that the contents of the particles was phosphorus. The red phosphorus flame retardant particles varied in size. Most particles were smaller than the inner lead spacing, but large particles bridging or nearly bridging the adjacent inner leads were observed.The root cause of the failures was identified to be insufficient coating of the red phosphorus flame retardant particles. Insufficient coating may have occurred due to coating quality problems, quality problems when mixing the phosphorus flame retardant into the resin composition, package assembly quality problem and/or from stress generated during device to circuit card assembly and operation. In our analysis, devices were already encapsulated, so individual particles and coating dimensions could not be directly measured. However, our assessment of the mold compound uncovered phosphorus ions throughout the encapsulant, indicating an insufficient coating. Figure 6 shows the diffusion of phosphorus through the mold compound from the agglomerated particles shown in Figure 2. When the red phosphorus is exposed to moisture and oxygen absorbed in the package encapsulant, phosphoric ions (POx) and phosphoric acids can be created. It is well known that phosphoric acids are corrosive and can attack any aluminum, copper and silver in the electronic package. The insufficient coating was due to some combination of a lack of initial coating integrity and thermal degradation.Regardless of the integrity of the applied aluminum hydroxide coating, the coating could breakdown due to thermal decomposition. In particular, aluminum hydroxide can decompose to boehmite (AlO(OH)) and then to alpha-Al2O3, at a temperature over 190°C and the transformation will be completely at 250°C . This thermal decomposition can be identified by monitoring the weight of the Al(OH)3 sample. RKK's measurements indicated a weight loss of Al(OH)3 coating when the temperature was 60°C and the weight loss increased as the temperature increased . The decomposition of aluminum hydroxide coating results in alumina and water,
Figure 5: Flame retardant particles between inner leads of leadframe (Click to Zoom)
Figure 6: EDS phosphorus mapping of the red phosphorus flame retardant agglomeration in Figure 2 (Click to Zoom)
The phosphorus release will be accelerated when the temperature is over 150°C , and an increase in PO4 ion concentration will occur when the temperature is increased. From measurements on a molded sample using water extraction method to determine PO4 ions , the PO4 ion concentration was 170 ppm at 175°C and 2,500 at 250°C . Since the typical temperature for the transfer molding process is around 175°C, and temperatures for reflow soldering are around 220°C, it is possible that thermal degradation of the coating occurred during normal package manufacture and board assembly. In fact, experiments have shown the positive correlation of the temperature and the concentration of POx released from the flame retardant . With the water generated by the aluminum hydroxide decomposition, and the moisture and oxygen gradually absorbed in the mold compound, POx will be converted to phosphorus-containing acids,
Electrochemical migration of the copper lead-frame and silver leadframe plating occurs when red phosphorus flame retardant exists between two adjacent inner leads, resulting phosphoric acids, which can act as an electrolyte . If the adjacent leads are electrically biased, copper and silver can migrate from the anode (the lead with higher voltage) to the cathode (the lead with lower voltage) when both an electrolyte and an electrical bias are present. The migrated copper and silver then form a conductive path between inner leads within the packaged semiconductor device, as shown in Figure 7. In addition to the failures caused by electrochemical migration, wire bond failures (increased resistance or open circuit) in packages with the EME-U series mold compound also occurred [29,31,32]. Failure analysis showed the ball bond corroded with intermetallics remaining on the aluminum pad. To compound this problem, copper and silver atoms have also been reported to migrate from the wedge bond through the wire surface and arrive at the ball bound . This failure displays similar characteristics to the ball bond failures caused by bromine ions from brominated flame retardant , however, the concentration of bromine ions was negligible in the EME-U series mold compound. Amkor concluded that "the mechanism is dependent upon an abnormally high mold compound ionic concentration which accelerates bond pad intermetallic formation and results in a time and temperature dependent increase in parts' ball bond resistance" [29,31]. We concluded that the failure of red phosphorus flame retardant which causes the resistive short failures is the cause of the wire bond failures.
Figure 7: Schematic of the electrical migration of copper and silver between adjacent inner leads, with the absence of leaked red phosphorus flame retardant (Click to Zoom)
Summary and Conclusions
The EME-U series mold compound incorporated a red phosphorus flame retardant. Red phosphorus is very reactive, and when exposed to moisture and oxygen, it tends to oxidize and form phosphorus-containing ions and acids which are corrosive to electronics interconnects. To stabilize and insulate the red phosphorus, a coating was applied, which consisted of aluminum hydroxide and phenol resin. Before the launch of these mold compounds, some evaluation tests had been conducted and showed positive results. However, the testing may not have been adequate because failures still occurred in field use.
We identified the root cause of this problem as insufficient coating of the red phosphorus flame retardant particles. It was shown that the red phosphorus content was not completely contained in the particle coating. In addition, the coating can degrade due to the thermal decomposition of the aluminum hydroxide content during manufacture processes. The insufficient coating can lead to the formation of phosphoric ions and acids, which can serve as electrolyte and cause electrochemical migration of the lead material. The migrated lead material forms an unwanted conductive path and can cause a resistive short and leakage current failure.
Sumitomo Bakelite terminated the production of the EME-U series mold compound with red phosphorus flame retardant. The final shipment to Amkor Technology was scheduled in February 2002 . Sumitomo Bakelite released a product change notice  in August 2001, which stated Sumitomo Bakelite would change away from the red phosphorus flame retardant.
Sumitomo Bakelite Co., Ltd, "Most valuable low cost and high performance compound `super EME' for super high reliability/HTSL, anti-popcorn and non-PMC," Sumitomo presentation to Philips Semiconductors, August 1998
Sumitomo Bakelite Co., Ltd, "High temp. storage life & high anti-solder crack," presented to Philips Semiconductors, January 22, 1997
Sumitomo Bakelite Co., Ltd, "New molding compound, super HTSL performance, EME-7351UT, upper compatible grade to EME-7351LS", presentation to Philips, September, 1998
P. McCluskey, D. Das, J. Jordan, L. Condra, T. Torri, J. Fink and R. Grztbowski, "Packaging of electronics for high temperature environments," International Journal of Microcircuits and Electronics Packaging, Vol. 20, No. 3, 1997
C. A. de Wit, "An overview of brominated flame retardants in the environment," Chemosphere, vol. 46, no. 5, pp. 583-624, 2002
Shigehisa Ueda (to Sumitomo Bakelite Co., Ltd.), "Epoxy resin composition comprising red phosphorus,"
U.S. Patent 5 869 553, February 9, 1999
Sumitomo Bakelite Co., Ltd, "Al coating issue", communication letter to Amkor Technology Inc., October 30, 2001
Leon Lantz II and M. Pecht, "The measurement of ion diffusion in epoxy molding compounds by dynamic secondary ion mass spectroscopy", submitted to IEEE Transactions on Components and Packaging Technologies, 2003
Thomas Goida (Analog Devices), "Meeting With Sumitomo", communication email to Asif Chowdhury (Amkor Technology), October 9, 2001
Don Foster (Amkor), "Cirrus H2.0 lead short failure issue final findings", presentation to Cirrus Logic, September 24, 2001
Vitesse, "VSC7123 RMA Analysis", presentation to Cisco, Vitesse, Amkor and Sumitomo, July 24, 2001
Sumitomo Bakelite Co., Ltd, "Short problem between leads on EME-7351UL," communication letter to Amkor Technology Inc., June 21, 2001
Sumitomo Bakelite Co., Ltd, "Re: Sumitomo's reply regarding inquiries from Seagate", communication letter to Amkor Technology Inc., November 16, 2001
Sumitomo Bakelite Co., Ltd, "Re: Technical Information of PFR", communication letter to IDT, January 30, 2001
Sumitomo Bakelite Co., Ltd, "New phosphorus flame retardant for short circuit issue," presentation to Amkor, June 2001
Sumitomo Bakelite Co., Ltd, "EME-7351UT shorting failure", August 2, 2001
Sumitomo Bakelite Co., Ltd, "Re: Technical information on EME-7351UL PFR related (2)", communication letter to Amkor Technology Inc., June 15, 2001
Sumitomo Bakelite Co., Ltd, "RE: Phosphorus Issue of EME-7351UT," communication letter to Amkor Technology Philippines, August 23, 2000
Amkor Technology Inc., "Amkor 7351UL history in ePad TQFP", 2002
Amkor Technology Inc., "Failure analysis report for LSA 0266 LQ 128 LD electrical fail", January 7, 2001
Cirrus Logic, "Failure analysis report 8D (SH8671-450E)", to Fujitsu, October 1, 2001
Y. Huang, S. Zhan and D. Bigio et al.,
"Distribution of a Minor Constituent in a Transfer Molded e-Pad Leadframe Package", IEEE Transactions on Components and Packaging Technologies, 2004, (to be published)
S.K. Mehta and A. Kalsotra, "Kinestics and Hydrothermal transformation of gibbsite", Journal of Thermal Analysis, Vol. 367, pp. 267-275, 1991
Rinkagaku Kogyo, "Chemical composition analysis results of NVE140 classified products", report to Sumitomo Bakelite, September 16, 2001
Sumitomo Bakelite Co., Ltd, "Re: Sumitomo's Reply regarding inquiries from Motorola", communication letter to Amkor, October 19, 2001
Rinkagaku Kogyo, "Measurement of specification items after NVE140 (93012, 93022, 93032) heat treatment)", report to Sumitomo Bakelite, December 10, 1999
R.Souto, M.Laz, and S.Gonzalez, "Copper electrofaceting: corrosion of copper in phosphoric acid aqueous solutions," Materials Science Forum, Vol. 192-194, pp. 525-34, 1995
Amkor Technology Inc., "Potential reliability issues with 7351UL EMC", January 17, 2000
Gary (Amkor), "Executive summary: Sumitomo P issue," (date unknown)
Amkor Technology Inc., "WW quality direction, strategy and systems, PRF EMC position," Quality Council Meeting, Seoul, Korea, August 1-3, 2001
Don Foster (Amkor), "Philips HTS ePad Failures, the Customer in Mind for `99", letter to Mick Cowan, November 11, 1999
Ker-Chang Hsien and Theo Martens, "Ag and cu migration phenomena on wire-bonding", Journal of electronic material, October 2000
Don Foster (Amkor), "HTS failures with 7351UL EMC", November 9, 1999
Sumitomo Bakelite Co., Ltd, "EME-7351UT & EME-7351UL final shipment notification," to Amkor Technology Inc, December 28, 2001