(posted for Gordon Davy)
Dr. Zinn has responded promptly to one of my comments on the abstract of the talk he plans to present. While acknowledging that a small particle's melting point is less than that of the bulk, I said that the nearly 900-degree difference between bulk copper and his copper nanoparticles is "a stretch." He chided me for not checking my facts before making "such 'bold' statements." Had I known of the work that has been going on in his field, I would have alluded to it, but the facts as I now understand them do not alter the substance of what I posted, much of which he ignored.
For interested TechNet subscribers, here is my (readily accessible) allusion: the Wikipedia article on melting point depression. It expresses the ratio of melting temperatures (nanoparticle to bulk) by an equation of the form TMn/TMB = 1 - C/d, where C is a constant and d is the nanoparticle diameter. In fact, this equation predicts a zero K melting temperature – for a non-zero diameter. While that might seem startling at first, it correlates with the common-sense notion that a metal atom bonded to no other metal atoms is in effect melted, regardless of its temperature. The graph in the article also indicates that if metal nanoparticles could be made small enough, they would melt at, or even well below, room temperature.
Wikipedia also reveals that colloidal gold and silver were known in antiquity. Gold doesn't form an oxide, and silver oxide is not strong. Copper of course does oxidize, and coalesce, which makes Dr. Zinn's work significant.
As a digression, as the number of atoms in a nanoparticle diminishes, the distinction between a solid and a liquid (and therefore the concept of melting) gets pretty murky. Above 0 K, a metal's crystal lattice at equilibrium has a certain number of vacancies. From time to time, a vacancy and a metal atom trade places – they jump. That's the mechanism for solid-state bulk diffusion. Grain-boundary and surface atoms jump more readily – the mechanisms for solid-state grain-boundary and surface diffusion. Metal atoms in a liquid do not maintain a fixed arrangement. How frequently do the atoms in a cluster (a teeny-tiny nanoparticle) have to jump for it to be counted as a liquid? For example, bulk copper has a face-centered cubic crystal structure. But one cube, with 14 atoms (8 corner and 6 face), has not even one bulk atom. How likely are the atoms in a 14-atom cluster of copper to be arranged in a face-centered cube, or, above 0 K, to stay that way?
Dr. Zinn, you discovered how to make nanoparticles of copper small enough to melt at 200°C (a 65% MP depression). Based on the Wikipedia graph, their diameter must be ~ 5 nm. Since an atomic diameter is ~ 0.3 nm, I estimate they contain only ~ 5000 atoms. I, and I think many others, would be interested to have you explain briefly on TechNet how you determined the diameter distribution and corresponding melting temperature range.
However, even given that they melt at 200°C, your mechanism seems to require most of these nanoparticles to:
a. Melt simultaneously (a challenge, even assuming they all had identical diameters),
b. Wet (react with) the atoms on adjacent surfaces (i.e., board land and component termination), and
c. Form a bridge (of millimeter dimensions) between them,
before most of the coalescing (and growing) melted copper globules freeze. I propose it be dubbed the attachment-by-copper-nanoparticle-melting-and-wetting-before-freezing (ABCNMAWBF) mechanism, which is how soldering works, with S instead of CN.
I find the ABCNMAWBF mechanism to be a much bigger stretch than non-coalescing copper nanoparticles small enough to melt at 200°C. For this reason, until I have more facts relating specifically to your R&D, I still favor the notion that the connections your paste makes are due to copper nanoparticles reacting with tin (ABCNRWT) in the termination finish.
What facts am I looking for from you? Nothing proprietary. You can resolve this question by addressing at least one of my prior requests for:
1. Differential thermal analysis of your paste (plots), or
2. Connection microstructure (photomicrographs), or
3. Whether you have evaluated the connection to non-tin finishes such as immersion silver, and more importantly, NiPdAu.
Assuming you can get specimens, number 3 seems simplest, and the most convincing. If, as I suspect, your paste does not make connections to non-tin finishes, then this would support my ABCNRWT hypothesis.
Dr. Zinn, as a person trained as a scientist, I can overlook your disdain for my not getting all the facts related to your field that you apparently expected me to get before I publicly commented on your abstract. (I also admit – in advance – to not checking to see if colloidal Cu reacts with Ag or Pd at 200°C, or even if Cu reacts with Ag or Pd at its bulk MP.)
But as a person who had a 34-year career in engineering, I can assure you that the issue you took issue with is not the most important one. Even if your paste works by melting (and does make connections to non-tin finishes), that is still a long way from getting the engineers whom you hope to interest in your process to regard it as a "drop-in replacement" for conventional solder paste.
Most of the subscribers to TechNet are not scientists. They don't get paid to go looking for facts in the kind of publications you cited, and frankly, many don't care whether or not the copper in your paste melts at 200°C. Instead, they expect you to provide the facts – at least rudimentary answers, with evidence, to all of the engineering-related issues I raised:
1. Process The width of the process window – times and temperature range, for surfaces with marginal solderability (or wettability), including Ag and Pd.
2. Process and reliability The benefits of reflowing your paste at 200°C instead of reflowing, at a higher temperature, the paste they are now using.
3. Quality assurance How well the connection's strength can be judged by appearance.
4. Reliability How well the (rigid) connections made with this copper survive thermal cycling (plots of cycles to failure).
Getting satisfactory answers for SAC solder took over a decade – you missed a lot of excitement during your nine years in Alcatraz.
Finally, here's a fact for you, and two questions. Rework is an unavoidable part of manufacturing the kinds of electronic products for which most engineers who subscribe to TechNet are responsible. Regardless of mechanism, the connections your paste makes have a very high melting temperature, so:
1. Can they be reworked?
2. Can they even be repaired?
Gordon Davy
Peoria, AZ
______________________________________________________________________
This email has been scanned by the Symantec Email Security.cloud service.
For more information please contact helpdesk at x2960 or [log in to unmask]
______________________________________________________________________
|