Austin American Technology Corporation

"A New Process for Cleaning Flip Chip Devices:
Reliability and Process Control"

Jose A. Berbel, Austin American Technology
Philip Chen, Sulzer Intermedics
Robert Moran, Austin American Technology

Abstract

A new cost effective process for cleaning low gap flip chip packaging applications is proposed. This cleaning process has proven to be reliable for effectively removing RMA fluxes located under a low gap (less than 0.003 in.) of flip chip devices (less than 0.004 in. bump size and less than 0.008 in. pitch).

The development of a cleaning process for flip chip applications has been deliberately approached under the following parameters and constrains: effective and reliable cleaning, real time process control capabilities, minimal chemical usage (consumption), and practical safety considerations for manufacturing environments.

Effective cleaning action is accomplished by the design of a cleaning agent (solvent blend) with certain physical and chemical characteristics related to thermodynamic solubility parameters as well as the interfacial tension between the fluid (solvent) and the solid substrate. Although these properties can be predicted to some extent by studying the Gibbs free energy change as a function of the enthalpy, entropy, and interfacial tension, a practical/empirical approach has been chosen. Changes in solubility and surface tension properties as a function of the concentration of the individual components in the solvent blend will be discussed based on experimental data.

The real time process control of the cleaning process is based on quantitative conductivity detection as a function of the ionic contamination concentration in solution. The Onsager theory predicts a relationship between the ionic concentration and the resistivity of a solution. By incorporating a resistivity cell connected to a resistivity controller, the amount of ionic contamination in solution is measured. The solvent usage is minimized by a solvent recovery technique based on ion exchange mechanisms.

Safety and environmental considerations have been a focal point in the chemical design of the cleaning agent in terms of ecological toxicity, ignitability, and environmental persistence.

Introduction

Microelectronics technology has been widely used in the medical device industry, from hearing aid devices, to implantable neurological devices for neuromuscular stimulation or pain control, to implantable pacemaker and defibrillator for cardiac arrhythmia management. For implantable pacemaker and defibrillator applications in particular, these devices are typically designed and tested to have a survival rate of 99% with a 90% confidence level over the projected device life. Continual efforts to miniaturize the overall device size, which is essential to patients comfort, has resulted in a never ending challenge for high density electronic device packaging and design. Flip Chip Technology (FCT) offers a unique alternative for high reliability and high-density device packaging.

Effective cleaning of a Flip Chip substrate after reflow of solder is essential to prevent device failure due to corrosion resulting from ionic contamination. In addition, flux residues, even no-clean fluxes, will likely restrict the underfill material flow, resulting in weak adhesion at interface creating voids that will eventually degrade solder bumps joint reliability.

Process engineering and design

1. Cleaning agent

Considering the high reliability of cleaning applications, cost parameters, and safety factors, a specific organic solvent blend was designed and optimized according to solubility and surface tension characteristics. Cleaning tests were performed on a standard chip test die (0.250 in. X0.230 in.) with 96 eutectic solder bumps mounted on a ceramic test substrate. A standard RMA type flux was used. The RMA or R type flux is used not just to provide wettability to the solder joint, but also to hold the chip down to the substrate, taking advantage of the tackiness nature of the flux. The substrate is then reflowed on a conveyor belt with conductive heat. After reflow, the typical bump height (standoff) is about 0.003 in.

In principle, five different types of cleaning agents were considered): alkaline aqueous agents, low molecular weigh alcohols, high molecular weight alcohols, HFCs/HFEs (Hydrofluorocarbons/Hydrofluoroethes), and Propylene Glycol Ethers.

1. Alkaline aqueous agents.

These are clearly effective, particularly when used with pressurized sprays. The surface tension is reduced to the order of 28 to 33 dyne/cm by the action of ionic surfactants. This method presents possible reliability problems due to the high surface tension of the rinse water (72 dyne/cm). An incomplete rinse of the alkaline agents (i.e. Monoethanolamine) will inevitably result in corrosion and/or dendrite formation.

2. Low molecular weight alcohols (Ethanol, 2-Propanol).

These have the advantage of very low surface tensions. However, the solubility of reflowed Rosin flux is poor in such agents. The addition of cyclic hydrocarbons (i.e., cycloheptane, cyclohexane) improves the solubility. However, the ignitability (flash point) of such blends is reduced to less than 32° F in some cases. This clearly presents a safety problem.

3. High molecular weigh alcohols.

These are effective cleaning agents. The Rosin flux is readily soluble in these and their surface tensions can be reduced down to 26 to 29 dyne/cm by the addition of non-ionic ethoxylated surfactants. However, due to their low vapor pressure, they will require a rinse with different agent such as a low molecular weigh alcohol. Possible problems can arise as a result of cross-contamination of the rinse agent. Also, low flash point considerations of the rinsing agents may be a safety concern.

4. HFCs/HFEs azetropes.

These are very effective cleaning agents when used under controlled conditions in the vapor phase. The advantages of these agents can be summarized as low surface tensions (17 to 23 dyne/cm), excellent solubility for rosin fluxes, high vapor pressure (high evaporation rates), and no measured ignitability. Cost considerations ($200.00/gal to $300.00/gal) however make these agents a less attractive alternative.

5. Propylene Glycol Ethers.

Some of these agents have the following desirable characteristics: high solubility for Rosin fluxes, relatively high vapor pressure (>3 mm Hg at STP), relatively high flash points (>117° F), and relatively inexpensive ($20/gal to $50.00/gal). However, the surface tension may be slightly high for flip chip cleaning applications.

Based on the previous observations, it became apparent that by modifying the surface tension of certain Propylene Glycol Ethers, a reliable, efficient, and cost effective cleaning process can be achieved. Moreover, by selecting these agents, the cleaning process can be incorporated into the Megacleanerä equipment, which operates as a closed-loop process with real time process control.

The initial solvent blend selected (Megasolvä JB) has been already tested extensively for Rosin flux removal. Preliminary cleaning tests for the flip Chip devices showed poor results. Although the flux residues located on the ceramic surface outside area of the chip were easily removed, large amount of residues remained under the chip, particularly between solder bumps. These results suggested a high interfacial tension between the solvent and the device surface. This problem was overcome by the addition of 2-Propanol to the solvent blend. The surface tension appears to decrease linearly as a function of the 2-Propanol concentration. The surface tension was experimentally measured by the capillary rise method¹:

g » ½ (h + r/3) r r g, where g : surface tension, h: capillary rise, r: capillary radius, r : density, and g: gravity constant. A least-squares fit of the relationship is shown in the following graph:

Plot of Surface Tension vs. Percentage of IPA in Megasolv™

Surface Tension

There are two restraints that limit the maximum concentration of 2-propanol in the solvent blend: solubility and ignitability. The solubility of the reflowed Rosin flux in the blend decreases substantially whenever the concentration of 2-propanol in the blend exceeds 40% by volume. In addition, the flash point of the blend containing 25% by volume of 2-Propanol was found to be 80° F. It would be expected that a decrease in the flash point would occur as the 2-Propanol concentration increases. Therefore, it can be deducted that the optimal concentration of 2-Propanol is 25% by volume. The surface tension is low enough and a high degree of solubility is maintained while keeping the flash point slightly above room temperature.

Once the optimal blend has been established, a simple and practical analytical tool must be developed in order to measure the concentration of 2-Propanol at any given point in time. Due to its high vapor pressure, it is expected that the concentration of 2-Propanol will decrease over time. Specific gravity measurements are a very simple and accurate method for the purpose of controlling the 2-Propanol concentration. A least-square fit of the specific gravity versus the 2-Propanol concentration is shown:

Plot of Specific Gravity vs. Percentage of IPA
in Megasolv™

2. Process control

The ionic contamination of the solvent blend is continuously measured during the cleaning and solvent recovery cycles by means of conductivity detection. It can be shown that

c = (1/R - a )²/l ² , where a and l are empirical constants characteristic of a specific solvent-solute combination ^2,3. The concentration units are arbitrary. If a standard of Sodium Chloride is used, the recommended concentration units are mgr/L.

The cleaning process can be summarized as follows: the devices to be cleaned are placed in the process chamber. The solvent fluid is agitated internally by means of a pumping mechanism, which sprays the fluid under immersion. The resistivity of the solution is monitored as a function of time. A sharp resistivity drop is normally noticeable within the first five minutes, as the ionic contamination is dissolved into solution (dR/dt<0). At the point where a zero slope is attained (dR/dt = 0), the process controller signals the end of the cleaning cycle, and the solvent recovery cycle is activated. During this cycle, the solution is passed through a set of ion exchange macroreticular resins where the ion exchange reactions take place and the ionic contamination is removed from the solution (dR/dt>0). This cycle will continue until a pre-programmed target resistivity value is reached (see the following graphic illustration):

where the time is in minutes and R (resistivity) in Mohm-cm.

The theoretical solvent recovery cycle as a function of time can be described mathematically according to the following equation²:

(1/R - a )² = (1/R0 - a )² e^-kt, where R: resistivity, R₀: resitivity at infinite dilution, and

k: flow rate/total volume.

Since the previous equation assumes a 100% ion exchange efficiency, by comparing the theoretical and actual recovery times, the % efficiency of the resins can be calculated. The level of efficiency, when lower than 30% indicates that the resins are near exhaustion and need to be replaced.

3. Toxicity/Environmental aspects⁴

Toxicity studies have shown most Propylene Glycol Ethers do not cause critical toxicities of testicular, thymic, or blood injury, and do not produce birth defects. It is believed that the lack of critical toxicity is due in part to the inability of the predominant isomer, which has a secondary hydroxyl group, to be metabolized to carboxylic acids, in contrast with Ethylene Glycol Ethers.

The NOAEL (No Observed Adverse Effect Level) values for the components present in Megasolvä JB are reported as:

Birth defects: 1,524 ppm inhalation

Fetal toxicity: 755 ppm inhalation

Testicular toxicity: 600 ppm inhalation

Blood damage: 600 ppm inhalation

The aquatic toxicity for a representative species group ranges in LC-50s (mg/L) from 161 to >5000.

Propylene Glycol Ethers are biodegradable. A 28-Day BOD (Biological Oxygen Demand) for the components of Megasolvä confirms a 60% biodegradability.

Conclusion

The described cleaning process can be incorporated for Flip Chip cleaning applications with a high degree of reliability and safety. The process control capabilities and the closed-loop nature of the process provide clear advantages over other alternative methods.

The cleaning agent has been optimized and an ideal three-point parameter conjunction has been achieved for solubility, surface tension and ignitability.

References

D. P. Shoemaker, "Experiments in Physical Chemistry," 5^th ed., chap. 10, McGraw-Hill, New York (1989).
J. A. Berbel et al. "Closing The Loop On Solvent Cleaning Systems," Nepcon West 97 Proceedings, Reed Exhibition Companies (1997).
L. Onsager, Phys. Z. 28, 277 (1927); P. W. Atkins, "Physical Chemistry," 3d ed., chap. 27, Freeman, New York (1986); D. P. Shoemaker, "Experiments in Physical Chemistry," 5^th ed., chap. 8, McGraw-Hill, New York (1989).
"Environmental Aspects Report," Arco Chemical Company (1995)

Acknowledgements

Special thanks to the Chemistry Department at The University of Texas at Austin and in particular to Dr. Ruth McKay for the use of her facilities and equipment.