It is most likely the filler in the epoxy.
As the filler moves it pushes the epoxy out of the way.
Your non-bonds likely didn't have enough resin left in the joint.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed

Ed,
By "filler" do you mean "resin", which is the component affected by magnetic flux?
Why would polarity affect this movement when the resin appeared to be attracted to both poles?

No, I mean the fillers added to the resin.
Often these are oxides, some of which can be weakly magnetic.
I recall testing epoxies for this.
We ended up using a 3M 'B' stage epoxy for assemblies.
No mixing, non-magnetic, just heat to cure (2214 High Temp).
This was used on 2:17, some 1:5, and Alnico assemblies.
We did use a lower temp compound for NEO work.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed

Many of the J-B Weld products contain an epoxy resin, some amount of calcium carbonate, and steel powder. The more active chemicals are listed on the MSDS (Material Safety Data Sheed) that is available for their products on their web site.

"J-B Kwik" is the sales name for a large number of products.

Is there any possibility that the 'correct' magnetic flux can improve the epoxy bonding process (since the opposite seems to be happening).

Did you degrease and use a silicone oil remover on the magnets first?

You really want an adhesive that is unaffected by the field.
Otherwise, some assemblies will be impossible to make correctly.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed

3DDave,

No de-grease was applied to ANY of the magnets.

Think about it. One side has some silicone oil and the other side does not. A thin film is all that is required. Where might this come from? The magnetizing process may have them on a silicone rubber sheet, so all are magnetically aligned and all have the same side contacting the contaminated surface.

One should always degrease and use silicone oil remover for any surface getting epoxied that wasn't exposed to the air by a clean saw.

It is an intriguing question, the manufacturer of the resin won't disclose the details of the resin content, but could be interested in giving some advice.
The only idea that comes to mind is that the steel powder in the mix could be from a hard magnetic steel that acquires some magnetic orientation during the grinding. This orientation could influence how the powder migrates toward the magnet or away from it during the curing.
Anyway, not a good idea to use a steel reinforced resin with magnets

prex
http://www.xcalcs.com : Online engineering calculations
https://www.megamag.it : Magnetic brakes and launchers for fun rides
https://www.levitans.com : Air bearing pads

3DDave,
The silicon contaminant would have been spread over both sides of the 16 magnets. They were handled excessively with bare fingers befoe being assigned to the current project. More 'controlled' testing is in order.
prex and Ed,
Would'nt steel reinforced resin be a good idea if it improved bonding to magnets?

Steel will interfere with bonding to the magnets as it will tend to displace the adhesive and isn't adhesive on its own.

Did you rub the magnets completely and thoroughly to make sure that surface contamination was completely spread and why do you fight the notion that the number one cause of adhesive failure isn't working in this case?

Even if the particles or resin were magnetized, there aren't any magnetic monopoles so they will simply align with the magnetic field.

Any magnetic fillers will tend to migrate to where the field is strongest, usually the edges.
Why this seems to effect one polarity more than the other is likely just an anomaly.
My favorite way to clean magnetized magnets is duct tape and acetone.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed

3D and Ed,

You've convinced me to experiment. Please suggest any steps I have missed below.

1) purchase 10 neodymium magnets
2) clean both surfaces thoroughly with duct tape then acetone
3) line up magnets in a row on clean sheets of paper, facing 5 north and 5 south up,
4) mix J-B KWIKWELD product 8276CAN and spread puddle 1/16" thick
5) dip each plastic knob in puddle and sit on a magnet in alternating order (N,S,N,S,...)
6) Let cure and test bonding strength for tension and shear
7) photograph failed bond surfaces
8) report findings to skeptics

Wear clean gloves for all handling of the magnets.

Thoughts from a Chemist;
I'm a bit confused by 3DDave's comment about silicon oil contamination...why would neodymium magnets have silicon oil only on one side? Why would silicon oil be there at all? I thought these magnets were just small, flat cylinders of a neodymium alloy. If these are the same magnets from the fridge, they've been excessively handled, like you said, and any contamination would be spread out evenly, ruling it out as the cause of unsuccessful bonding. Looking into the chemistry of epoxides, I learned how the polymer is actually formed, but I don't know if I can explain why polarity affects the quality of the epoxy bond. As others have commented, there's a metal mixed in with the resin component, iron powder: https://www.whatsinproducts.com/types/type_detail/.... Although this isn't "kwik" weld so I'm not sure if it's the same.

Iron has 4 unpaired electrons, each with its own valence orbital, so the net spin is +2 giving it a magnetic moment. I know there is an energy cost to initiate a "spin-flip" of their orientation, but how that affects the curing process is beyond my understanding.

• https://files.engineering.com/getfile.aspx?folder=d515c5ba-9e89-4380-af49-7

"The magnetizing process may have them on a silicone rubber sheet, so all are magnetically aligned and all have the same side contacting the contaminated surface."
Or they use silicone rubber padding on their work surfaces so they don't chip or fracture or maybe someone used silicone spray to lubricate a non-magnetic plate to slide them across.

What doesn't matter is understanding where contamination can come from. What matters is understanding all the typical lubricating contaminants that prevent good adhesion and having cleaners suitable to remove them.

For the reasons that Dave states we prohibited any synthetic lubes in our magnet plant.
We built a lot of assemblies that were epoxied together so this was a concern for us.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed

3D & Ed,
Testing was completed as described resulting in no appreciable difference in neodymium/plastic epoxy bond strength at north or south poles.

All that is coincidental may not be causal.