Quantitative Nanomechanics with AFM: Are we there yet?
A new year. A new start. It’s always an opportunity to assess where things stand and wipe the slate clean if necessary. I have been thinking a lot about the state of quantitative nanomechanical measurements with AFM. It is certainly where I have focused my technical career and is my personal technical area of expertise. And in all fairness, the title of this blog is the title of the talk I delivered at MRS in December (see my MRS December 2017 debrief blog here). It is the area where I do most of my own measurements for customers, as well as the area where I continue to contribute research in development of new methods.
AFM is inherently a mechanical probe for mechanical interactions between tip and substrate. Sure you can manipulate the interaction in a lot of fancy ways (oscillate it at resonance, put it in contact with the sample and oscillate the tip-sample contact on or off resonance, feedback off of amplitude/phase/frequency). You can also probe a lot of non-mechanical properties – e.g. magnetic, electrical, or optical properties. But at its core, AFM is a mechanical measurement and the first measurements that came out of AFM were mechanical (think lateral force measurements and force curves – among the earliest measurements performed with the instrument.)
So how is AFM doing? How well can it quantitatively measure mechanical properties such as stiffness (modulus) or adhesion?
So let’s do the easy part first: AFM does pretty well with adhesion measurements. Using conventional force curves where the AFM “pokes” a surface, upon the retraction the AFM tip typically probes an adhesive dip. As long as you have matched your AFM cantilever pretty well to the contact stiffness of your material (more on that in a later blog), the AFM should do a pretty good job measuring the adhesion.
A force curve image
How about modulus (what is modulus? I tried to address it recently in my blog on modulus confusion?) The answer is a less happy one. To cut to the chase, in my opinion, the answer to the title of my blog is: no, we are not there yet. At best (and by best I mean with a lot of patience and hard work and possibly involving a Ph.D. thesis), we can get 20% accuracy on elastic modulus. Working with viscoelastic samples? The picture is even worse because of the complicated frequency dependence in viscoelastic materials, and we don’t fully understand the frequency response in AFM measurements.
In my talk I had outlined 5 hurdles that remain in the pursuit of quantitative nanomechanical measurements with AFM which are:
- Tip shape and size
Is the AFM tip a sphere or a cone? It matters!
- Cantilever calibrations uncertainty
- Frequency mismatch between AFM measurements and typical mechanical measurements [for viscoelastic materials]
- Maintaining a linear viscoelastic regime [again, for viscoelastic materials]
- Limitations of current contact mechanics models for real-world samples.
Poor (JKR) fit to force curve on elastomer
At MRS, one audience member asked why 20% accuracy was not sufficient and why I had such a pessimistic view of AFM nanomechanical measurements?
So I would like to think that our field should aim better than 20% accuracy! Remember, 20% accuracy is best case scenario. 50% + accuracy is more realistic. And I am guided by the instrumented nanoindentation world that can achieve 5% accuracy routinely (granted, it has some inherent advantages). Furthermore, I don’t find the methods we have in place to be particularly easy or user-friendly – they require significant calibrations and tweaking during operation. And no, 20% accuracy is simply not good enough for many real-world, industrial applications that is the area I work in.
And second, I don’t consider my perspective pessimistic. On the contrary, I think it’s optimistic. It’s job security! It means we still have a lot of work to do to improve the AFM capabilities, and there is nothing pessimistic or negative about that.
Dalia Yablon, Ph.D.