Flux Residue Cleaning for High-Lead Bumping

Mary Pat McCurdie Agilent Technologies, Fort Collins, Colo. -- Semiconductor International, 10/1/2000

Flip-chip on ceramic packages require high-lead solder on the chip side due to the temperature hierarchy of subsequent reflow processes. High-lead solder is deposited sequentially (lead followed by tin) by an evaporative method. An RMA rosin-based flux is spun over the entire wafer, and the wafer is reflowed in an IR oven under a hydrogen-enriched atmosphere. The wafer then must be cleaned to remove the residue left by the flux.

In a rosin-based flux the most prevalent compounds are organic acids that convert during high-temperature reflow to carboxylic acid. It is important to remove carboxylic acid residues because they are hydroscopic and will contribute to corrosion of the solder (Fig. 1). Additionally, during reflow at 340-360°C, the flux is polymerized, and the residue can be very difficult to remove. Since residue cannot be identified as harmful by a visual inspection, it is best to remove all residues (Fig. 2). Post-cleaning processes and properties, including die-attach reflow, underfill adhesion and long-term reliability, may be compromised by the presence of these residues.^1-3 Wafer cleanliness must meet the following criteria:
.Active residues that cause pitting or corrosion during storage may not be present.
.The solder bumps must reflow and wet the substrate contact pads during the subsequent assembly process.
.The surface must be clean enough to ensure good underfill adhesion.

While it is desirable to have shiny silvery bumps, this objective is not necessary. The presence of color does not necessarily indicate corrosion products, nor does the absence of color prove the lack of corrosion products.⁴ Therefore, the ultimate test of wafer bump cleanliness is the performance of the bumps in subsequent packaging processing and use.

Cleaning process selection criteria

Requirements for the cleaning process include adequately cleaned wafers, environmentally agreeable chemistry, minimized hazardous waste, no compromise of operator safety, and, of course, the process (including operation, maintenance and process control) must not be cost-prohibitive. Also, to minimize initial capital costs and simplify the manufacturing process, the cleaning system has to serve multiple uses.

As part of the low-cost criterion, a triple-purpose cleaning system is desirable to clean not only wafers but also flip-chip and BGA assemblies. Because the equipment has to clean between the die and the substrate in flip-chip manufacturing, in-line aqueous spray cleaners were eliminated from consideration. To limit the use of stored Class 1 solvents for safety reasons, vapor-degreasing equipment also was eliminated from consideration.

1. Wafers not cleaned after reflow have residues that collect on the bumps as well as the satellites.

2. Cleaning not only removes the residue, but also some of the satellites.

A batch centrifugal cleaner with different fixtures provides the capability of cleaning several package assembly sizes and both 150 mm and 200 mm wafers. Centrifugal cleaners are an excellent choice to deliver the cleaning chemistry in small gaps and through small channels between solder bumps. Finally, centrifugal cleaning systems are compatible with multiple cleaning chemistry options, including hydrocarbon, semi-aqueous, saponifier and aqueous cleaning agents. The equipment selected for use in this evaluation was the MicroCel from Speedline-Accel. In this equipment the wafers are submersed and rotated in the cleaning solution at 80°C. Following the clean cycle, the wafers are rinsed with DI water at 60°C and dried with 200°C air.

3. Solder bumps cleaned at 70°C still have many satellites.

The primary requirement for the cleaning chemistry is good flux solvency. It also should be free-rinsing and have low surface tension and low viscosity so the fluid will go into small channels and gaps. The chemistries evaluated included semi-aqueous, hydrocarbon, aqueous with saponifier and full aqueous chemistries. In the very early phases of the program, five chemistries were selected for beaker-scale testing on flip-chip assemblies. Wafer testing was not performed because it was felt the flip-chip assembly cleaning would be more difficult due to the geometry of the parts.

Beaker-scale testing eliminated full aqueous and aqueous with saponifier chemistries. Semi-aqueous and hydrocarbon chemistries performed nearly the same in beaker-scale tests. Ionox HC, from Kyzen Corp. (Nashville, Tenn.), a semi-aqueous hydrocarbon solvent, was selected because it had more desirable EHS characteristics than the other chemistries under consideration. It has a high flux solvency yet very low flammability, a mild odor, and is biodegradable.

Evaluation methods and results

4. Solder bumps cleaned at 80°C have few remaining satellites.

One of the difficulties in evaluating wafer cleanliness is finding an analytical technique. Surface analysis techniques such as X-ray spectroscopy for chemical analysis (ESCA) or time-of-flight secondary-ion-mass-spectroscopy (TOF-SIMS) are precise, but are practical only to measure a few solder bumps out of the thousands on the wafer. ESCA cannot assign the elemental results to molecules, so it is unknown if the residue is flux or cleaning agent or from another source. Visual techniques, microscopic inspection and scanning electron microscope (SEM) are not quantitative and are difficult to evaluate over the entire wafer.

Gas chromatograph-mass spectroscopy (GC/MS) has the potential to be a valuable tool. The wafer is rinsed with a known quantity of solvent (in this case, acetone), which is collected and analyzed via gas chromatography. The different fractions coming out of the gas chromatograph are analyzed with a mass spectrometer. Theoretically, in this manner one can obtain molecular fragment identification and quantitative comparisons of residue. This method was tried with several product wafers. Unfortunately, the technique is so sensitive it identifies molecular fragments from photoresist and the polyimide passivation layer, in addition to flux and cleaning residues.

ESCA results (Table 1) show cleaned samples have much less organic contamination than uncleaned samples. All carbon residues are eliminated after a slight sputter on the surface. The uncleaned samples still have a significant amount of carbon present even after a surface sputter.

Table 1. ESCA Results
Sample	Carbon	Nitrogen	Oxygen	Tin	Lead
Cleaned	50.8	1.5	30.4	4.6	12.7
Cleaned	45.9	3.0	32.0	2.2	16.8
Cleaned (sputtered 6 nm)	0.0	5.7	18.8	8.3	67.2
Cleaned (sputtered 6 nm)	0.0	6.0	20.5	7.5	66.0
Uncleaned	89.7	0.2	8.8	1.1	0.2
Uncleaned	87.4	0.0	9.2	1.6	1.8
Uncleaned (sputtered 6 nm)	81.8	3.0	1.9	2.3	11.0
Uncleaned (sputtered 6 nm)	46.3	6.5	0.0	5.1	42.1

A similar experiment using GC/MS confirms these results (Table 2). The wafers used in both the ESCA and GC/MS experiment did not have polyimide passivation on the surface. The polyimide tends to complicate the analysis since it provides a source for organic contamination. The molecular fragment eluded at 12-14 min from the cleaned wafer rinses has been matched with the cleaning chemistry. The amounts shown in Table 2 for the cleaned wafers represent ion abundance measured by the gas chromatagraph and translate to sub-microgram levels of contamination.

Table 2. GC/MS Results Without Polyimide Passivation
Sample/ elution time	12.8 (flux)	18.5 (flux)	20.7 (flux)	12-14 (cleaning chemistry)
Cleaned	0	0	0	30,000
Cleaned	0	0	0	14,000
Uncleaned	30,000	290,000	260,000	0
Uncleaned	17,000	0	70,000	0

Many fragments eluded from uncleaned wafer rinses match with RMA flux. Types of flux fragments include aliphatic amides, aliphatic carboxylic acids and nonylphenol. For clarity, only two of the fragments' abundance levels are shown in Table 2. The elution time of 18.5 min corresponds to an unsaturated aliphatic acid and 20.7 min to an aliphatic amide. The amount of residue is quite variable on the uncleaned wafers.

Finally, SEMs provide a visual illustration of how well the cleaning chemistry removes RMA flux residues (Fig. 1 and Fig. 2). It is clear that the dark residue present on the uncleaned bumps is removed during cleaning. Additionally, some of the small, extraneous solder balls (satellites) surrounding the main solder bump are removed during centrifugal cleaning.

Both ESCA and GC/MS analyses indicate some of the cleaning chemistry remains on the solder bumps. This observation is confirmed by TOF-SIMS analysis. Several wafers were evaluated using TOF-SIMS in an attempt to identify different types of coloration observed on the surface of the solder balls. While no conclusions were made regarding color, all of the wafers have residue identified as the cleaning chemistry. Table 3 displays these results. By far the most prevalent contaminants are sulfur, polydimethyl siloxane (PDMS) and dioctylphthalate (DOP).

Table 3. TOF-SIMS Analysis of Cleaned Wafers
Prevalence of Contamination	Wafer 1 (control)	Wafer 2	Wafer 3	Wafer 4
1	Sulfur	PDMS	PDMS	Sulfur
2	PDMS	Sulfur	Sulfur	PDMS
3	DOP	DOP	DOP	Copper
4	Cleaning chemistry	Cleaning chemistry	Cleaning chemistry	Cleaning chemistry
5	Copper	Copper	Copper	DOP

Comparisons with other contaminates found by TOF-SIMS show the cleaning chemistry residue is insignificant. TOF-SIMS is an extremely sensitive tool for surface analysis, but it identifies only the top monolayer of material. Therefore, because tin and lead are by far the most prevalent materials identified in this analysis, we concluded the contamination would have little effect on the performance of the solder bumps. Additionally, no adverse adhesion or voiding occurred in the underfill reliability testing of the assembled packages.

Cleaning temperature effects

As expected, the cleaning chemistry properties change with temperature. The recommended operating temperatures are 50-70°C, where the optimum solvency, free-rinse properties and surface tension occur. Data supplied by the vendor (Table 4) illustrate temperature effects on the solvent. While surface tension actually increases with temperature, the viscosity and rinse properties are enhanced. Viscosity was measured by a capillary tube method. Rinse time is the amount of time it takes a dye-doped sample of chemistry to be removed by DI water from beneath a 2-in. die at a 1.5 mil standoff. Time-to-wet is a similar measurement: the time it takes for a dye-doped droplet to travel under the length of a 2-in. die at a 1.5 mil standoff. Most of the enhancements to performance occur between room temperature and 60°C.

Table 4. Cleaning Property Changes with Temperature
Temperature, °C	Viscosity, centipoises	Time to rinse, mm:ss	Time to wet, sec	Surface tension, dynes/cm
22	9.06	12:30	90	21.2
49	4.60	No data	No data	25.4
60	3.31	3:15	15	26.8
77	2.33	1:14	13	27.5

5. GC/MS of solvent after 272 and 816 process cycles both have an undetectable amount(<0.01%) of flux.

Due to the nature of the evaporative solder deposition process, satellites remain around the main solder bump after reflow. They are not known to be a reliability problem, except for the following hypothesized effects. First, flux residue tends to collect around the base of the satellites, and since this is an active flux, it must be removed entirely to avoid corrosion. Second, satellites form on the passivation layer and do not adhere as well as the solder bumps to the under bump metalization. Therefore, satellites are a potential underfill failure-mode during long-term use and during thermal excursions.

During initial process development, experiments showed flux removal was better at 80°C than at 70°C. However, at 80°C the cleaning agent is more aggressive and causes some degradation in fixtures and seals in the cleaning equipment. An evaluation to compare cleaning at these two temperatures was only partially successful. Because the wafer samples had polyimide passivation on the surface, GC/MS results were not conclusive. Organic residues other than flux residue dominated the results. SEM photographs show visually that the flux removal is approximately the same for the two temperatures. Satellite removal was better at a temperature of 80°C (Fig. 3 and Fig. 4), so that temperature was chosen. Remaining satellites are removed during subsequent process steps.

Minimizing solvent use and handling

6. FTIR comparison of fresh and used (~200 cycles) solvent shows no detectable change in composition.

During development of the manufacturing process, the cleaning solvent was changed every 200 cycles. This was calculated to maintain a flux loading level below 1-2% by volume, as recommended by the manufacturer to maintain the solvency of the cleaning fluid. These calculations were "worst-case" and based on the amount of flux dispensed on the parts, not on the amount remaining after reflow. As manufacturing demands increased, the replacement schedule proved to be a bottleneck. The solvent replacement procedure and subsequent heat-up time was impacting throughput. Additionally, the amount of solvent handled, used and disposed was excessive. New calculations based on estimated flux residue after reflow indicated the cycles prior to solvent replacement could be increased to at least 800.

Determining contamination at such low levels in the cleaning solvent is difficult. GC/MS proves to be an excellent tool for this type of information and has been used in a similar investigation.⁵ GC/MS data on the cleaning chemistry after 272 process cycles and 816 process cycles are presented in Figure 5. There are no indications that flux is present in the solution at the detection limit of ~0.01%. Figure 6 displays absorbance spectra from fourier transform infrared spectroscopy (FTIR) tests that also show the amount of flux residue in the cleaning solvent does not change the chemistry and is below detection limits.

Process control

Some of the water from the cleaning solution evaporates during each cleaning cycle, due to the equipment design and the elevated process temperature. When the water level drops, the cleaning capability decreases, so the water content must be carefully controlled. The refractive index is an indication of the amount of water in solution. As the water level drops, the refractive index increases. Therefore, the refractive index is used to indicate when water must be added to the reservoir tank.

To maintain process control, a hand-held refractometer is used once a week to measure the refractive index of the cleaning agent. A process control action limit has been set to indicate when water must be added to the solution. A process control chart for the past year shows that responding to the action limit keeps the refractive index inside the control limits (Fig. 7).

Conclusion

A batch centrifugal cleaning system combined with a semi-aqueous alcohol-based cleaning chemistry has been very effective in cleaning RMA flux from the surface of wafers in a wafer bump/flip-chip assembly line. In two years of production, wafer cleanliness has never been identified as the cause of suspect parts or part failure. Regularly performed pull strength tests from assembled packages show excellent wetting of the wafer bump to the substrate pad. Analytical tests and visual inspections indicate flux residues are removed adequately from the solder bumps. Costs have been minimized by using the centrifugal cleaner for multiple cleaning processes and by increasing the time between cleaning solvent replacement.

Mary Pat McCurdie, Ph.D., joined Agilent Technologies as a manufacturing development engineer in 1998. She works in wafer bump and assembly, specializing in wet etch and cleaning. She received her Ph.D. and M.S. in chemical engineering from Colorado State University in 1997 and 1994, respectively; and her B.S. in chemical engineering in 1984 from the University of Utah. She also has worked for Ball Aerospace and Hercules Aerospace as a materials engineer.

---

REFERENCES

1. K. Dhaneshwar, N.C. Lee, "Post Reflow Solder Paste Residue: Sources, Properties, Chemistry and Concerns," Indium Corp., 34 Robinson Rd., Clinton, NY 13323, USA.
   
2. S.M. Scheifers, C.J. Raleigh, "Effects of Flux Contamination on Flip Chip Reliability, Sensors in Electronic Packaging," MED-Vol.3/EEP-Vol. 14, ASME, pp. 101-109, 1995.
   
3. A. Sinni, M.A. Palmer, "Kinetics of Flux Residue Formation in a Humid Environment," IEEE/CPMT International Electronics Manufacturing Technology Symposium, pp. 152-156, 1997.
   
4. H.H. Manko, "Solders and Soldering," McGraw-Hill, New York, 1979.
   
5. C.P. Wong, W.O. Gillum, R.A. Walters, P.J. Sakach, "Reactions of High-lead Solders with BIOACT EC-7R Semi-Aqueous Cleaning Reagent," IEEE 45th Electronic Components and Technology Conference, pp. 1016-1027, 1995.
   

Acknowledgments

The author would like to thank Paul Mazurkiewicz of Hewlett-Packard for developing techniques for FTIR and GC/MS analysis; David Niles of Agilent Technologies for ESCA analysis; Charles Evans and Associates for TOF-SIMS and FTIR analysis; and John Sanders of Speedline-Accel and Mike Bixenmen of Kyzen Corp. for sharing technical data on their products. .

Talkback

We would love your feedback!

Post a comment

» VIEW ALL TALKBACK THREADS

Related Content

• Topics
• Author

Related Content

 

By This Author

There are no other articles written by this author.

SPONSORED LINKS