Thursday, September 10, 2020

The power of a timely collaboration

Sometimes it takes a while to answer a scientific question, and sometimes that answer ends up being a bit unexpected.  Three years ago, I wrote about a paper from our group, where we had found, much to our surprise, that the thermoelectric response of polycrystalline gold wires varied a lot as a function of position within the wire, even though the metal was, by every reasonable definition, a good, electrically homogeneous material.  (We were able to observe this by using a focused laser as a scannable heat source, and measuring the open-circuit photovoltage of the device as a function of the laser position.)  At the time, I wrote "Annealing the wires does change the voltage pattern as well as smoothing it out.  This is a pretty good indicator that the grain boundaries really are important here."

What would be the best way to test the idea that somehow the grain boundaries within the wire were responsible for this effect?  Well, the natural thought experiment would be to do the same measurement in a single crystal gold wire, and then ideally do a measurement in a wire with, say, a single grain boundary in a known location.  

Fig. 4 from this paper
Shortly thereafter, I had the good fortune to be talking with Prof. Jonathan Fan at Stanford.  His group had, in fact, come up with a clever way to create single-crystal gold wires, as shown at right.  Basically they create a wire via lithography, encapsulate it in silicon oxide so that the wire is sitting in its own personal crucible, and then melt/recrystallize the wire.  Moreover, they could build upon that technique as in this paper, and create bicrystals with a single grain boundary.  Focused ion beam could then be used to trim these to the desired width (though in principle that can disturb the surface).

We embarked on a rewarding collaboration that turned out to be a long, complicated path of measuring many many device structures of various shapes, sizes, and dimensions.  My student Charlotte Evans, measuring the photothermoelectric (PTE) response of these, worked closely with members of Prof. Fan's group - Rui Yang grew and prepared devices, and Lucia Gan did many hours of back-scatter electron diffraction measurements and analysis, for comparison with the photovoltage maps.  My student Mahdiyeh Abbasi learned the intricacies of finite element modeling to see what kind of spatial variation of Seebeck coefficient \(S\) would be needed to reproduce the photovoltage maps.  

From Fig. 1 of our new paper.  Panel g upper shows the local crystal
misorientation as found from electron back-scatter diffraction, while 
panel g lower shows a spatial map of the PTE response.  The two 
patterns definitely resemble each other (panel h), and this is seen
consistently across many devices.

A big result of this was published this week in PNAS.  The surprising result:  Individual high-angle grain boundaries produce a PTE signal so small as to be unresolvable in our measurement system.  In contrast, though, the PTE measurement could readily detect tiny changes in Seebeck response that correlate with small local misorientations of the local single crystal structure.  The wire is still a single crystal, but it contains dislocations and disclinations and stacking faults and good old-fashioned strain due to interactions with the surroundings when it crystallized.  Some of these seem to produce detectable changes in thermoelectric response.  When annealed, the PTE features smooth out and reduce in magnitude, as some (but not all) of the structural defects and strain can anneal away.  

So, it turns out it's likely not the grain boundaries that cause Seebeck variations in these nanostructures - instead it's likely residual strain and structural defects from the thin film deposition process, something to watch out for in general for devices made by lithography and thin film processing.  Also, opto-electronic measurements of thermoelectric response are sensitive enough to detect very subtle structural inhomogeneities, an effect that can in principle be leveraged for things like defect detection in manufactured structures.  It took a while to unravel, but it is satisfying to get answers and see the power of the measurement technique.



dislocations and disclinations and stacking faults and good old-fashioned strain.......
still homogenous?

Anonymous said...

Is there any indication that this is an effect specific to gold? Or would you expect to also see such effects in other metals/semiconductors?

It may also be useful/interesting to measure amorphous metals where you can have strain but no clear grain boundaries or crystalline defects.

Douglas Natelson said...

Anon, this should not be specific to gold. Indeed, we had previously seen the spatial variation in Seebeck response in other metal structures (e.g. nickel), and strain can definitely influence Seebeck response in semiconductors and other materials. Here is a recent paper showing related effects in graphene, for example:

jonah said...

I'm surprised one can anneal the gold on silicon oxide at such high temperatures and not have it short to an underlying doped silicon. Our devices have issues with this at much lower temperatures, around 500-600C or so. Do you observe that or is this somehow avoided?

Douglas Natelson said...

Hi Jonah, the fabrication procedure for the gold single-crystals (and bicrystals) is described in the Methods section of this:
The substrates start off with growth of 50-200 nm of SiO2 on top of them before anything else happens. That seems to be thick enough to avoid the Au penetrating that top layer upon annealing.

Anonymous said...

This guy formulates Schrodinger equation from classical gravity.