Voids in solder joint under high environmental pressure

[Chris Jones]

Hi, Anyone have any clues on judging voids in solder joints for cyclic
high pressure applications(0 -350 bar, 0-5000 PSI)? More specific how
would a 0.1mm3 gas pocket under a D2pack on FR4 laminate with 135u
copper layer affect that solder joint if environmental pressure is
cycled between 0 and 350 bar?

Best regards,

Rune

To prevent voids in our compounds wrapped around electronics, we used to
epoxy them in low atmospheric pressure, 10-3 torr [approx Bell Jar]
where we would watch with surprise at how they foamed like crazy before
settling down.

Perhaps, the soldering should be done in rarified atmosphere also, then
your question would become a moot point.

I think you would only really be sure to have no voids if, after
potting/soldering under vacuum, you then let the air back in whilst the
solder / potting compound / whatever is still molten (liquid).

When you make a vacuum around a liquid and it froths as lots of air
comes out, some of the last few bubbles don't always pop. There is very
little air/gas still inside the bubbles, (depending on surface tension
of the bubble and how good your vacuum is), but the bubble is still
there and it still doesn't have the potting compound / solder in it.
Such bubbles full of vacuum could still cause you problems later, if you
let the potting compound / solder solidify whilst still under vacuum. In
the case of potted high voltage assemblies, bubbles full of rough vacuum
probably break down at least as easily as air does, and in the case of
solder, voids full of vacuum will still allow stresses to occur when the
space around your assembly is highly pressurised.

Therefore, if the vacuum soldering is to help, the vacuum would have to
be released just after the peak temperature is reached and the solder
has all reflowed, but before the solder has solidified.

Chris

 

[Tim Williams]

Hmm, at the elastic limit (EL)? I'm not sure what the EL is.
At a guess when you start making dislocations in the lattice.
I should be able to do a Purcell, back of the envelope, calculation
(Binding energy of one atom to another.. over an atomic distance?)
I'll have to think about it.

It's my understanding that most materials are inverse, so aluminum,
solder, etc. get you 10x more cycles for 10x less strain and so on (it's
not a direct proportion, but probably an inverse power law or something --
I forget), and materials with a "fatigue limit" (steel, titanium, others?)
are inverse exponential (sort of), so for 10x less, you get 10^10 more
cycles, or something like that. Point being, it's not hyperbolic
(infinite fatigue below a threshold), but being exponential, it might as
well be (below a certain point, even rapid ultrasonic stresses won't do
jack over the age of the universe).

I don't know how well that applies to solder, because solder, or some of
the metals in it at least, are prone to creep. I don't have the numbers
on tin, or the actual alloys, but lead at least is notorious for creep.
Alloys need not exhibit creep because of crystal pinning and stuff
(probably characteristic of the hard, high-melt, brittle, lead-free
solders). In essence, because the melting point is so low, the crystal
sort of anneals itself at room temperature, allowing deformation at a
certain rate. An apparent example: take some solder from your spool and
make it stand erect at an angle on your bench. Observe. If the solder
droops down to the bench over time, that's creep. If it stays there
forever (more or less), it must be the sucky kind!

I don't know how well any of this is related to things like lattice energy
and coordination number and differential electronegativity (for alloys)
and so on. Materials science is messy. I know a bit of metallurgy but I
couldn't tell you anything that in depth.

Tim

 

[Rune]

You never said what fluid would be under pressure in contact with the
PCB, I just assumed it might be air?

Sorry my bad The PCB will be encapsulated in a potting material, and
placed subsea at 350 bar water pressure. I was thinking some type of
soft and flexible (low shore value), maybe PUR. But I am not sure how
soft and flexible PUR will be at 350 bar, any ideas?

New entry in "Not to do list": Do not put electronics inside scuba tank!...

 

[Rune]

Go to Harbor Freight and get one of their hand operated hydraulic pumps.
If memory serves, these produce up to 10,000 psi. A bottle of fluid & a
steel block with a suitable hole would produce a test chamber.

Yes, that is a nice one. We actually had a mechanical engineer here now
retired that did exactly that we used water instead of oil in the pump,
not a very god long term solution I know but it worked for the test at
350Bar.

 

[Rune]

This reference:
http://www.fujitsu-ten.com/business/technicaljournal/pdf/33note2.pdf

suggests that a fractional strain of 1% gives a fatigue life of 1000
cycles to failure with PbSn eutectic, and that the lifetime goes as
1/(epsilon**2). Lead-free solder has about a quarter of the fatigue
life of PbSn.

This reference: http://www.jim.or.jp/journal/e/pdf3/46/06/1271.pdf
says that the Young's modulus of Pb-free solder is around 50 GPa.

So assuming the void is large enough that its area is half that of the
solder joint, a pressure of 350 bar (35 MPa) will cause a strain of
35 MPa/50 GPa = 0.07%.

Thus a rough estimate of the fatigue life due to pressure cycling is

N ~ 0.25*1000*(1%/0.07%)**2 = 50k cycles, even with a pretty huge void.

Thermal cycling may well be a bigger effect.

These data come from chip resistors, so the stress concentration in the
D2PAK may be a bit different.

Cheers

Phil Hobbs

Wow! Thanks, Nice numbers and references. I am far lesser concerned
about those voids now.

Cheers
Rune

 

[Rune]

That was sorta my 'gut feeling' but I really don't know! (Dang, I
sometimes hate it when people think my advise is useful... I guess
it's mostly when I don't really know what I'm talking about.) I
assume there are others here who have perhaps done some underwater
circuits and will advise.

Assuming that the length expansion relates to bulk modulus as

dl = l * -dp / E

dl is shrinkage in any direction dp is pressure difference in Mpa E is
the bulk modulus Gpa. l is the original length in that direction.
assuming two materials, potting (dlP), and PCB FR4 (dlF).

then the differences in length compression would be

dlP-dlF=l*dp(1/EF - 1/EP)

Assuming PCB length 5cm and pressure difference 35MPa EF = 9Gpa EP =
2.3GPa this gives a a length compression difrence of
50*35e9*(1/9e9-1/2.3e9) = -0.57mm. I am not entierly sure about that EF
an EP value, just some suggestions found on the Internet.

So the potting material must be elastic or viscous enough to take up
that expansion difference of 0.57mm without transferring any significant
forces on the SMD components. Any suggestions of such a material? can
polyurethane rubber be used? or maybe some kind of gel like material?
How to calculate this?

An other concern would be the bulk modulus difference between copper on
the board and the FR4 PCB. copper has bulk modulus of 140 Gpa according
to wikipedia. So the compression difference would be approx 0.2mm over
the pcb length of 5cm. this does not seems so much but I would need to
consider the elasticity in the bond between PCB and the copper and also
the elasitcity of copper it self.

Is the potting to keep the water out?
Yes.

Well this reminds me of differential thermal expansion. (I wonder how
the numbers compare?) I've got this project that's been "on hold" for
a while now. But I'd like to use little surface mount components at
low temperatures. I've been told that they aren't reliable.. but of
course I want to try some testing of my own. (I haven't done it yet
so I can't say much more.) In your case, if it's possible, I might
try through hole components. The leads can flex a little and take up
any strain.

Yes I am considering hole components but I guess the potting material
will still need to be some degree of elastic or viscous.

So have you done underwater circuit before? Maybe there's an engineer
in your group, that has made all these mistakes already? (So you
don't have to :^)

Yes but only with encapsulation i bulky 1-atmospheric containers. And I
wish I had a large group of experienced subsea enginners on my back,
even one would be fine. Unfortunately its only me and my boss here
"Small company" the large upside is, I get to do everything .

Regards,
Rune

 

[Tom Miller]

Hi, Anyone have any clues on judging voids in solder joints for cyclic
high pressure applications(0 -350 bar, 0-5000 PSI)? More specific how
would a 0.1mm3 gas pocket under a D2pack on FR4 laminate with 135u copper
layer affect that solder joint if environmental pressure is cycled between
0 and 350 bar?

Best regards,

Rune

You should dig around for information on the structure of the black boxes
used in aviation. I believe they have active pingers that can go to 20k feet
or more.

 

[[email protected]]

The first paper I cited shows PbSn eutectic being just about exactly
quadratic, i.e. 1/10 the strain gets you 100x the cycles to failure.

The threshold for fatigue failure exists in most materials, especially
steel. (See e.g. the last figure in
http://www.lesjoforsab.com/technical-information/durability.asp)

All solids have a range of stresses over which their creep rate is
zero--that's the definition of a solid. Even paint. Some paint is a
liquid, but some is a very soft solid, which helps amazingly in reducing
running and drips. (Folks of a certain age will remember the Lucite
Wall Paint ads from the '60s--"Like having an army of painters on the
job". Lucite was the first soft-solid paint formulation AFAIK.)

Good latex paints are non-Newtonian fluids. Their viscosity is
inversely proportional to the pressure applied, often only in one
direction. Is that what you mean by a "soft-solid". I guess I'm not
seeing it because solids don't tend to "wet" things.

 

[George Herold]

I don't know how well any of this is related to things like lattice energy
and coordination number and differential electronegativity (for alloys)
and so on. Materials science is messy. I know a bit of metallurgy but I
couldn't tell you anything that in depth.

Grin I agree with that! My materials science knowledge is next to zilch.

George H.

 

[Phil Hobbs]

Good latex paints are non-Newtonian fluids. Their viscosity is
inversely proportional to the pressure applied, often only in one
direction. Is that what you mean by a "soft-solid". I guess I'm not
seeing it because solids don't tend to "wet" things.

A solid is certainly not a Newtonian fluid.

The criterion for being a solid is that the creep rate goes to zero at
nonzero stress.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510

hobbs at electrooptical dot net
http://electrooptical.net

 

[[email protected]]

A solid is certainly not a Newtonian fluid.

It's certainly not a fluid at all.

The criterion for being a solid is that the creep rate goes to zero at
nonzero stress.

So solids can wet other solids? Gotta process that.

 

[[email protected]]

Solder on clean copper. If it doesn't wet, it doesn't stick.

Solder doesn't wet copper when it's in the solid phase. Latex paint
will wet (whatever is being painted) just sitting there. It takes
work to move it around though (by design).

 

[[email protected]]

Sure it does, if the contacting surfaces are clean enough. But it's
really slow at low temperatures, and we run out of patience.

Sounds like the "glass flows" and "windows are thinner at the top" old
wives' tale.

 

[[email protected]]

Hmm, that seems like cheating. Force is being applied.

Are you sure the friction is not causing it to heat and melt?

I wouldn't think so.

How about surfaces that are machined and polished so flat that they bond
when in contact?

Metal migration? Still seems like cheating the terminology.

 

[Tom Miller]

Solder, for instance.

You can make indium wet metal or glass by rubbing it on.

Cheers

Phil Hobbs

Are you sure the friction is not causing it to heat and melt?

How about surfaces that are machined and polished so flat that they bond
when in contact?

 

[Glen Walpert]

Sounds like the "glass flows" and "windows are thinner at the top" old
wives' tale.

Not an old wives tale, although the one about glass being a supercooled
liquid certainly is. A good example is clamping a thin foil of gold
between two clean and flat blocks of silver. The gold will completely
diffuse into the silver in a few years at room temperature, apparently
vanishing into what becomes a solid block of silver, although all of the
gold will still be there not too far from the original position of the
foil. It happens faster at higher temperatures of course. Silver plate
is used on the bonnet seal rings of many high pressure main steam valves,
diffusing into the steel valve body and bonnet, creating a hermetic
seal.

Most metals will undergo some creep with zero externally applied stress
at room temperature if you look close enough, due to internal stresses.
For this reason it is necessary to let the hardened tool steel for micro-
inch accurate gage blocks age for a few years before producing the gage
block. The atoms in solids move around a lot more than most people
realize. Tin and zinc whiskers are good examples. Stress free materials
are virtually nonexistent, with the exception of dislocation free single
crystals, which are not too common.

The structure of latex paint, which contains no latex, is produced by
microscopic beads of acrylic diffusing into each other and the surface
they are applied to, a process which starts when the carrier evaporates,
allowing the beads to make contact. Latex paint which is not labeled
100% acrylic contains clay fillers which significantly weaken it and
reduce the life of the paint dramatically in many cases, definitely not
worth saving a few bucks for paint that won't hold up.

 

[[email protected]]

Not an old wives tale, although the one about glass being a supercooled
liquid certainly is. A good example is clamping a thin foil of gold
between two clean and flat blocks of silver. The gold will completely
diffuse into the silver in a few years at room temperature, apparently
vanishing into what becomes a solid block of silver, although all of the
gold will still be there not too far from the original position of the
foil. It happens faster at higher temperatures of course. Silver plate
is used on the bonnet seal rings of many high pressure main steam valves,
diffusing into the steel valve body and bonnet, creating a hermetic
seal.

Sounds like metal migration which is a completely different kettle
than wetting.

Most metals will undergo some creep with zero externally applied stress
at room temperature if you look close enough, due to internal stresses.
For this reason it is necessary to let the hardened tool steel for micro-
inch accurate gage blocks age for a few years before producing the gage
block. The atoms in solids move around a lot more than most people
realize. Tin and zinc whiskers are good examples. Stress free materials
are virtually nonexistent, with the exception of dislocation free single
crystals, which are not too common.

The structure of latex paint, which contains no latex, is produced by
microscopic beads of acrylic diffusing into each other and the surface
they are applied to, a process which starts when the carrier evaporates,
allowing the beads to make contact. Latex paint which is not labeled
100% acrylic contains clay fillers which significantly weaken it and
reduce the life of the paint dramatically in many cases, definitely not
worth saving a few bucks for paint that won't hold up.

Yes, quite like corn starch and water. ...and the shape of the
"beads" changes the characteristics of the paint (directionality of
the shear).

I'm still not buying the "solid" part.

 

[Tim Williams]

Are you sure the friction is not causing it to heat and melt?

Yes. See also: weighted wire drawn through ice. In this case, the
pressure causes the ice to liquify under the wire, and freeze above it,
leaving the block otherwise intact. This can be done at any arbitrary
speed, so friction is not causing it to melt.

Solid state metallurgical reactions have been known since antiquity: one
can "smelt" mercury from cinnabar (its sulfide) by pounding it in a mortar
and pestle with another metal more reactive towards sulfur (I think the
story uses a silver mortar, though I'd think the mercury metal would stick
to that rather well given its propensity for amalgamation).

These days, reactive ball milling is a common laboratory practice (though
I don't know if it's seen much industrial use). Especially "high energy"
ball milling probably does proceed through contact frictional heating, but
solid state diffusion is nonetheless a necessary mechanism.

Tim

 

[Tim Williams]

Silver plate
is used on the bonnet seal rings of many high pressure main steam
valves,
diffusing into the steel valve body and bonnet, creating a hermetic
seal.

Hmm, seems like a potentially bad idea -- diffusion can leave porosity:
http://en.wikipedia.org/wiki/Kirkendall_effect
Maybe not an issue for hermetic seals.

Similar stuff goes on as normal industry practice in silicon, especially
MEMS stuff: take two slabs of commercial silicon -- high purity and single
crystal, a materials theoretician's dream -- and press them together in
the middle and heat. Usually a charge is applied, especially for bonding
glass. The charge causes ions to diffuse to the surface, effectively
reducing the melting point locally. The perfectly planar surfaces begin
to bond by diffusion, and the bonding wave can be observed by infrared
interferometry.

Tim

 

[Spehro Pefhany]

Well, that makes two of us. Somehow I was reading 0.1 as .01. That
makes all the numbers go up by two orders of magnitude, to some millions
of cycles if you believe they scale that far.

Generally there is some threshold strain below which fatigue failure
doesn't occur no matter how many cycles you give it.

Cheers

Phil Hobbs

Not likely important for the OP, but what we were taught in first
year...

http://en.wikipedia.org/wiki/File:S-N_curves.PNG

... does not seem to be 100% true..

http://www.diva-portal.org/smash/get/diva2:210661/FULLTEXT02.pdf

--sp


Best regards,
Spehro Pefhany