-1-

Minimizing the Impact of Solder Spatter on Memory Module Gold

Connector Fingers
By Ross B. Berntson; David W. Sbiroli, Jeffery J. Anweiler
Indium Corporation of America

Solder Spatter: Problem or Cosmetic Issue?

Following the reflow process, memory module gold connector fingers occasionally
exhibit solder spatter contamination. Solder spatter is considered a type of surface
contamination, and the primary concern with it, although not proven, is that it could
interfere with the connector interface on the motherboard when installed. Proven or not,
all surface contamination types are not of equal concern, since they represent not only a
quality issue but a reliability issue for the product, and indicate that there may be a
problem in the manufacturing process.

Surface Contamination Types

Solder Spatter is only one type of surface contamination. Other types include watermark
stains and glossy-flux spatter. To a certain extent, these other types are less of an issue, in
part because they do not represent the depth of issue that solder spatter, in which solder
has actually wetted to the gold surface, does.

Causes of Solder Spatter

Solder spatter has a variety of causes, and is not necessarily the result of hot or molten
solder explosively “spattering” or outgassing during reflow. For example, observing
procedures to ensure optimum cleanliness during solder paste printing can reduce or
eliminate a spatter problem. Any manner in which solder paste spheres can be deposited
on gold fingers – and remain there during the reflow process – can produce spatter. This
includes failure to wipe the underside of a stencil between prints (dirty stencil); an
improperly cleaned misprint; and careless handling during the printing process step.
Other causes of spatter outside of the printing process can include excessive moisture in
the board material, contamination of the board, and a radical ramp rate in excess of 4
degrees C/sec. In these latter causes, violent outgassing of flux can cause essentially
small explosions in molten solder joints, causing solder particles to become airborne in
the reflow chamber and scatter about the PCB, thereby contaminating gold fingers. The
same is true for entrapped moisture in the board material, which upon release has the
same effect as outgassing flux. Similarly, foreign contaminants on the board surface can
contribute to solder splatter.

Impact of Solder Spatter

Although the concern that solder spatter may have a deleterious impact on the connector
interface remains as yet unproven, it is nonetheless a viable concern because the slight
“bumps” of solder spatter create a disruption of the planar surface of the gold finger.
Furthermore, these bumps are non-pliable and the solder metal itself has less conductivity
than gold, and is subject to oxidation. The first and easiest cause of solder spatter to
eliminate is in the solder paste stencil printing process. If this area or process is
established as the cause, then control is established by good “house-keeping” practices
 -2-

through proper printer set-up and operator training. If this is not the cause, then other
areas in the process need to be examined.

Watermark Stains
The root causes of watermark stains are not completely understood. It is considered likely
that multiple sources are involved. Since it has been shown that clean, fresh, non-pasted,
and non-populated boards may exhibit watermark stains after reflow, it is suggested that
these stains may have a variety of causes including PCB fabrication residues,
condensation from ovens, dried flux spatter, residues from cleaning boards, and heat -
induced gold discoloration. There may be others.

Impact of Watermark Stains

Watermark stains are, on the average, difficult to detect, but they are unlikely to have any
impact on the connector interface. In fact, memory module customers are less concerned
with this type of surface contamination (often referred to as gold discoloration).

Glossy-Flux Spatter
The phenomenon of glossy-flux spatter is understood to be caused when flux droplets are
ejected into the air in the reflow oven during the heating/reflow process. These droplets
disperse and are deposited over the entire board, including the gold fingers. There are two
main theories to account for glossy flux spatter: the solvent expulsion theory, and the
coalescence theory (Cleanliness during the printing process, once again, has an impact,
but can be controlled).

The solvent expulsion theory recognizes that solvents used in flux pastes must volatilize
during the reflow process. If excessive temperature is applied to the flux, these solvents
could flash-boil to gas (similar to dripping water on a hot pan), and this violent boiling of
the flux solvents will carry solids into the air. These particles are randomly re-deposited
on the board as glossy-flux spatter.

In an effort to either confirm or discount this theory, tests were conducted during which
observations were made during conductive heating of sample boards using a hot plate.
Temperature setpoints used were 190oC, 200oC, and 220oC, respectively. Paste flux
(containing no solder powder) did not exhibit spattering under any conditions; however,
solder paste (paste flux with powder) consistently exhibited spatter during the melting
and coalescence of the solder. The results are summarized in Table 1 and 2.

Table 1: Simulation of Solvent Expulsion:

Test Description Material Result
Flux vehicle (no powder) was Flux Vehicle B No flux spattering was apparent on the
printed on copper coupons and Flux Vehicle D coupons. Test was completed a second
then placed on a hot plate set at time with similar results.
190oC, 200oC, and 220oC.
Solder pastes were printed on Solder Paste B at 90% metal Flux spatter was visually seen with both
copper coupons and then load,Sn63Pb37, -325+500 metal percents. Higher metal % may
reflowed on a hot plate at Solder Paste B at 92% metal have produced less spatter, but it was
 -3-

190oC, 200oC, and 220oC load, Sn63Pb37, -325+500 difficult to tell. Test was completed a
second time with similar results.

It was reasoned that if boiling flux caused spattering, it should have been observed when
flux alone was heated. But since the spattering phenomenon was observed during solder
coalescence, it was reasoned that the mechanism should be found here. The test
illustrated that the solvent expulsion theory does not explain glossy-flux spatter.

Coalescence: A Possible Cause?

As solder melts and coalesces, entrapped flux is squeezed by the denser and heavier
solder surrounding it. Surface tension of the molten solder is a powerful force, and the

melting solder's increased surface tension exerts pressure on entrapped flux. When the
molten-solder surface tension is high enough, the flux is expelled violently. This theory is
supported by work completed by Dr. Ning-Cheng Lee and William Casey on voiding in
BGA assemblies, wherein both describe the link between surface tension and flux
exclusion (flux exclusion rate model)1 . This forcible ejection, then, is the most likely
cause of glossy-flux spatter.

Continued laboratory simulation of flux spatter illustrated the impact of coalescence even
when solder paste was dried prior to reflow. However, complete drying did reduce
spattering significantly. The results are summarized in Table 2.
 -4-

Table 2: Simulation of Flux Spatter due to Metal Coalescence:

Test Description Material Result
 Solder pastes (with powder)  Solder Paste B, 90%,  At all temperature settings, Solder
was printed on copper Sn63Pb37, -325+500 Paste B visibly wets faster and
coupons and then placed on a  Solder Paste D, 90%, coalesces more aggressively than
hot plate set at 190oC, 200oC, Sn63Pb37, -325+500 the Solder Paste D resulting in
and 220oC. more flux spatter.
 Flux spatter was seen with the
Solder Paste D at all temperatures
as well, but to a lesser extent than
the Solder Paste D.
 Spattering was more severe at
higher temperatures.
 Soak Zone (drying)  Solder Paste B, 90%,  Soaking at higher temperatures for
simulation- solder paste was Sn63Pb37, -325+500 more than 2 minutes reduced or
printed on copper coupons eliminated flux spattering.
and preheated (soaked) on a
hot plate set at various  See Table 3.
temperatures for various
lengths of time. Soaks
ranged from 150’C to 170’C
for 1 to 4 minutes. Coupons
were then transferred to a
second hot plate set at 220’C
for reflow and inspected for
flux spatter.
 Solder paste with SN62 was  Solder Paste B, 90%,  Flux spatter was observed with
compared against SN63 to Sn63Pb37, -325+500 both SN62 and SN63 and a
see if SN62’s slower  Solder Paste B, 90%, difference in amount of splatter
coalescence speed would Sn62Pb36Ag2, could not be determined visually.
reduce spattering. -325+500 The slower coalescence speed of
SN62 was observed.

Table 2: Simulation of Flux Spatter due to Metal Coalescence – Drying Study with Solder
Paste B at 90%, Sn63Pb37 Alloy:
1 minute 2 minute 3 minute 4 minute
150’C Flux spatter 1-2 spatters No spatter No spatter
observed
160’C 1-2 spatters No spatter No spatter No spatter
170’C No spatter No spatter No spatter No spatter
 -5-

Because the coalescence model appears promising, the wetting speed of various materials
was investigated. Wetting speed is affected by alloy type, temperature, flux vehicle, and
reflow atmosphere. As illustrated in Figure 1, temperature has a dramatic impact on

Figure 1: Wetting speeds of Material B at different temperatures.

wetting speed; higher temperatures result in faster wetting speeds. Flux vehicle can also
impact wetting speed. Materials A and C are known to have the fastest wetting speed
while Materials D, E, and F have slow wetting speeds. Dr. Ning-Cheng Lee reports in his
paper on Optimizing Reflow Profiles via Defect Mechanism Analysis that inerted
atmospheres (nitrogen) will also increase wetting speed. Reports by Jeannie Hwang and
others indicate that wetting speeds of eutectic alloys tend to be faster than non-eutectic.
Consequently, Sn63Pb37 generally has a faster wetting speed than Sn62Pb36Ag2. Table
3 is taken from Dr. Lee’s paper. A summary of factors that impact wetting speed and
therefore coalescence and potentially spattering is summarized in Table 4.

Table 3: Wetting speeds of Sn63Pb37 and Sn62Pb36Ag2 with different flux vehicles
from Dr. Lee’s Reflow Paper (Note that the flux vehicles are not the same as included in
this study)
Wetting Time (seconds)
Flux Sn63Pb37 Sn62Pb36Ag2
200C 240C 200C 240C
1 2.87 1.48 4.68 2.52
2 1.03 0.50 1.44 0.63
3 1.65 1.00 2.60 1.03
4 2.50 1.44 3.48 2.10
 -6-

Table 4: Factors that may contribute to solder spatter:

Factor Mechanism Impact on Spatter
Fast coalescence will increase the
Different activators promote different probability of flux entrapment and will
Flux vehicle
wetting and coalescence speeds during likely increase the force on the
activator
reflow entrapped flux thereby causing
explosive expulsion of the flux
Increase solvent content during
Flux vehicle solvent Solvent type and quantity will affect the
coalescence may cause a more violent
and solvent content extent of drying during preheat
expulsion of entrapped flux
Fast coalescence will increase the
probability of flux entrapment and will
Alloys affect wetting and coalescence
Alloy type likely increase the force on the
speeds during reflow
entrapped flux thereby causing
explosive expulsion of the flux
Fast coalescence will increase the
Inert (nitrogen) atmosphere increase probability of flux entrapment and will
Reflow atmosphere wetting and coalescence speeds during likely increase the force on the
reflow entrapped flux thereby causing
explosive expulsion of the flux
Fast coalescence will increase the
Higher temperatures will increase probability of flux entrapment and will
Temperature of
wetting and coalescence speeds during likely increase the force on the
molten solder
reflow entrapped flux thereby causing
explosive expulsion of the flux

Impact of Glossy-Flux Spatter

Although glossy-flux spatter is visually detectable, it is not always easy to see. Because it
is insulative, it is possible that glossy-flux spatter may cause contact problems at the
connector interface, but this as not been proven, and is unlikely when the volume of
spatter is very low and thus virtually undetectable. Nonetheless, because it is a real type
of contamination, memory module customers are concerned about this type of flux
spatter, and would prefer to prevent or eliminate it. Solutions to the glossy flux spatter
problem involve a three-fold approach: prevention, minimization, and detection and
cleaning.

Solutions to Solder Spatter and Glossy-flux Spatter: Prevention,

Minimization, Detection and Cleaning

Prevention
One method of preventing deposition of glossy-flux spatter is to coat the gold fingers
with a peelable solder mask. This solder mask would be applied after the solder paste had
been printed on the PCB, and removed after reflow. This process is as yet unproven, but
would be expensive (time-consuming, probably manual application/removal) and
difficult to process since it would have to be applied to a selective area of the board, and
would interrupt production flow. The application of temporary tape to the fingers is
another option, however it suffers from the same drawbacks as peelable solder mask,
with the additional disadvantage of causing poor gasketing of stencil to board if applied
 -7-

prior to solder paste printing. Consequently, it makes sense to look at minimization as a

solution to the problem.

Minimization
Optimization of the flux vehicle chemistry and reflow profile will produce minimized
solder spatter. Standard solder paste systems were evaluated with the support of a
memory module manufacturer to assess the impact of material and reflow profile
optimization. The experimental matrix was based on studies conducted at Indium
Corporation of America and outlined above. These studies gave clear indication that
activator, solvent, alloy, and reflow profile impact the extent of spattering.
Consequently, the problem of solder spatters was approached with confidence that proper
adjustment of these parameters could produce a solution that minimized spattering to the
point of virtual elimination.

Non-standard materials such as polymerizing flux systems were not included in the study
because of their higher cost, shorter shelf life and stencil life, a tight process window, and
difficulty of rework. However, in the future polymerizing fluxes offer a possible
minimization solution because potential spattering materials are entrapped during the
thermally activated polymerization process. In other words, no liquid flux remains that
could spatter.

The test vehicle was a 6-up array memory module, upon which no components were
placed. Components have been found to decrease the impact of spatters presumably by
blocking expelled flux from the gold fingers. Current production materials and profiles
were used as the baseline performance. Spatter levels on production boards was roughly
one spatter per 100 panels. Two engineers inspected all boards were with a 20X power
microscope to evaluate the extent of spattering.

Standard solder paste materials with varying characteristics were used for the line study.
These materials were selected because of their different wetting speed and solvent
characteristics. In order to reduce the variable in the study, the same alloy was used for
all solder pastes, Sn63PB37 at –325/+500 mesh.

Table 5: Materials Tested

Flux Vehicle Description Relative Solvent Reflow Solvent
(Material) Wetting Content Atm. Volatility
Designation Speed
Current production material at
memory module manufacturer. Prefer
Flux A Unknown Moderate High
It is a moderate residue RMA Inert
based Material
Advanced, high performance,
Air or
Flux B long stencil life, moderate Fast Moderate Low
Inert
residue material
Flux C Advanced, high performance, Fast Moderate Air or Low
long stencil life, moderate Inert
 -8-

residue material
High performance, RMA type,
Air or
Flux D long stencil life, moderate Slow Moderate Low
Inert
residue material
Low residue, high solvent
content air or nitrogen reflow Prefer
Flux E Slow High Moderate
material (inert atmosphere Inert
preferred)
Extremely low residue, inert
Flux F Slow High Inert Moderate
reflow material

Results of Minimization Testing

Reflow Profile Selection

During the laboratory testing, it became evident that both the reflow profile and material
type would need to be adjusted to minimize spattering. The memory module testing
added credibility to this theory. The two main profiles tested differed in their soak zone
characteristics. A linear ramp profile as depicted in Figure 2 with no plateau soak zone
Figure 1: Example of linear ramp profile.

250

200
Temperature (C)

150

100

50

0
Time (S econds)

resulted in some spattering for all materials and increased spattering for the base-line
production material. Consequently, this basic profile shape was not investigated further.
Based on our hypothesis for a spattering mechanism, the conclusion was that this linear
profile did not dry the flux sufficiently.

A more promising basic profile shape included a high temperature soak (dry-out) at
160oC to evaporate all solvents. An example of this type of profile is depicted in Figure 3.
This loss of solvent increases the viscosity of the flux remnant and reduces the volatile
content thereby minimizing spatter. Potential problems with such a dry-out are, however,
poor wetting and voiding. The use of an inert (nitrogen) processing atmosphere can help
improve wetting and reduce voiding but appears to have no effect of spattering. This
profile was also a ‘long’ profile, eliminating the need for excessively fast ramp rates.
With this profile, the maximum ramp rate is 1.75oC/sec.
 -9-

The results of all the profile studies are summarized in Figure 4 and Table 6.

Figure 3: Example of a high temperature soak profile

250

200
Temperature (C)
150

100

50

0
Time (S econds)

Figure 4: Summary of spatter results for each material on a six-up array of memory
modules.
 - 10 -

Table 6: Results from material studies.

# of Flat dull, flux droplets #of defined, shiny flux
Paste Type belt speed Atm.
(watermarks) droplets
Flux A 0 34 26"/min N2
Flux A 0 42 26"/min N2
Flux B 12 5 26"/min N2
Flux B 4 20 26"/min Air
Flux B 0 21 26"/min Air
Flux B 0 21 22"/min Air
Flux B 0 21 26"/min Air
Flux B 0 29 26"/min Air
Flux C 2 7 26"/min Air
Flux C 0 35 26"/min Air
Flux D 0 0 26"/min Air
Flux D 0 2 26"/min Air
Flux D 0 2 26"/min Air
Flux D 0 4 26"/min Air
Flux E 0 3 26"/min Air
Flux E 0 3 26"/min N2
Flux F 2 0 26"/min N2
Flux F 1 0 26"/min N2

As described above, the level of spatter measured on these bare boards would be
diminished significantly on populated production boards. Estimates from the engineering
staff indicate that less than 10-20 spatters on an unpopulated board will result in no
spatters on a populated board. Consequently, Flux types D, E, and F all provide viable
solution to spatter. The D type flux vehicle has the added advantage of a wide process
window and air reflow capability. All three of these materials have slow wetting speeds
but different solvent types. This indicates that all solvents can be dried effectively and
that wetting speed is a key factor in flux spatter.

Detection and Cleaning

Detection and subsequent cleaning will result in clean connectors but is an expensive and
time-consuming correction to the defect of spatter. Through formulation changes in the
solder paste residues, detection could be simplified through dies and fluorescing
chemicals. Cleaning could also be improved with proper residue design. Unfortunately,
as with the prevention methods, cost and time make detection and cleaning undesireable.

Conclusions
Solder pastes combined with the correct profile can result in virtual elimination of solder
and flux spatters. Pastes that have high content of relatively volatile solvents and slow
wetting speeds give the best results. Many of these high solvent materials require an inert
atmosphere. Masking connector fingers and detection and cleaning may provide interim
fixes but do not address the root cause of spattering.

References
Dr. Ning-Cheng Lee, Voiding in BGA, Indium Corporation of America
William Casey, Voiding in Micro-BGA, SMI 1998 Proceedings, MCMS