Checks and Stripes

Screen Shot 2014-02-10 at 10.08.49 AM

What do you see here? Stripes? Or feedback?

Leaping Robot note: Historians, like most of us, like controversies. They provide a valuable lens to look at a range of issues. This post was inspired by an article that appeared recently in Science. It concerns a simmering feud between two groups of researchers over how to interpret some images of gold nano-particles made with a scanning tunneling microscope (STM). My colleague, Cyrus Mody, a historian of science at Rice University, wrote a prize-winning book called Instrumental Community that tells the story of the invention and spread of scanning probe microscopy.1 So Cyrus is an excellent person to offer a more nuanced reading of this controversy. One might think that the nano-feud is, as one person notes, just a “minor storm in a nano teapot.” Mody’s guest blog post goes deeper than this and shows how controversies like this, besides being rather common, also tell us something important about how scientists (and science journalists) communicate with each other and the public. Here’s Cyrus…

***

The January 24 issue of Science has a news item by Robert Service with the juicy titleNano-Imaging Feud Sets Online Sites Sizzling” describing a multiyear tussle over a decade’s worth of science based on some scanning tunneling microscope (STM) images of gold nanoparticles.  The STM images are from Francesco Stellacci’s group, first at MIT then at the Swiss Federal Institute of Technology in Lausanne.  They purport to show “stripes” of organic molecules attached to particles with diameters of 20 nanometers or less, in an arrangement resembling lines of latitude.  However, a number of critics have insisted – in blogs, Twitter feeds, and elsewhere – that Stellacci’s stripes are actually instrumental artifacts.

I’m not going to weigh in on whether Stellacci’s stripes are real or not.  I know some very smart STMers on either side, so I doubt consensus will be reached soon.  Instead, let me do what historians of science usually do: put controversial research in perspective, followed by a point for the prosecution and a point for the defense.

Controversies like this are pretty common.  The point of doing forefront research is to push our ability to make and measure stuff right to the limits.  Robert Service has made a career out of reporting such disputes, for which historians of science should be grateful.  For instance, I’ve made frequent use of his article on a related dispute from 2003, “Molecular Electronics – Next Generation Technology Hits an Early Mid-Life Crisis.”  If you look at the history of science you’ll find lots of disputes, sometimes extending over decades, over matters of fact that you would think could be easily resolved.  One of my favorite examples is the more than thirty year debate from the early 1920s to the late 1950s in which cytogeneticists nearly uniformly agreed that a normal human somatic cell contains 48 chromosomes (rather than the now accepted 46).2 How hard can it be to count chromosomes that are almost a thousand times larger than Stellacci’s nanoparticles?  And yet, even something as simple as counting can remain interdeterminate (or determined but incorrect) for a very long time.

The instrument Stellacci used to image his nanoparticles, the scanning tunneling microscope, has a particularly rich history of disputes.  STM images are made by bringing a sharp metal probe very close to the surface being imaged while maintaining a voltage difference between the probe and the sample.  This encourages some electrons to “tunnel” from the sample to the probe and vice versa.  Generally, the closer the probe is to the surface the higher the probability of tunneling in one direction rather than the other, so the number of tunneling electrons is a reasonable proxy for the z-height of the sample for any given x-y position of the probe.  As you move the probe around in x and y, you build up a matrix of z values for the strength of the tunnel current, which you can then convert into a three-dimensional image of the sample.

Screen Shot 2014-02-10 at 11.56.26 AM

Schematic of how an STM works

Roughly, the computer to which the STM is hooked up registers a single value for the tunnel current per increment of probe position.  That is, the STM periodically samples the sample, and each sampled value is fed both to the imaging output and to the circuit controlling the height of the probe at the next increment.  Too strong a feedback can make the probe constantly overshoot and then play catch-up, so even a smooth surface will seem to have undulations.  Unfortunately, the phenomena researchers are looking for on a surface are also often periodic in nature – as, for instance, are Stellacci’s stripes.  If you think you’ve made a sample with stripy features, you want to get an STM image with alternating patches of light and dark, up and down.  A nice analogy, suggested to me by my Rice colleague Kevin Kelly, is of a strobe light illuminating water coming out of a tap.  If the strobe samples the water at the right rate, we see a series of drops accelerating under the force of gravity.  If the strobe frequency and duration are varied, though, we might see a smooth flow of water or we might see water drops that appear to move upward into the tap.

So it’s not hard to find people who are skeptical of claims that a particular STM image indicates the existence of some periodic nanostructure, particularly if the distance between periodic features is near the limits of the instrument’s resolution.  Perhaps the most famous such case involved STM images of DNA made in the late 1980s and early 1990s.

Screen Shot 2014-02-10 at 11.53.35 AM

2009 cover of Nature showing false-colour STM image with a DNA molecule running from bottom left to top right.

At the time, many STMers harbored hopes that their instrument could resolve the base pairs in a strand of DNA and might therefore be used in genetic sequencing.  Lots of STMers tried to image DNA; several groups published such images; one Caltech group even managed to get an atomic resolution image of DNA on the cover of Nature that appeared to show a helix with the right pitch distance (the linear distance between two turns of the helix).  And yet, there were lots of good reasons to think that an organic molecule like DNA should be difficult to image via electron tunneling.  Whispers began to circulate that some of the best images of “DNA” were actually images of complex defects in a graphite substrate that might or might not have any DNA deposited on it.  In the end, images such as the 1990 Nature cover were never definitively disproved, but the uncertainties surrounding their validity became so insurmountable that almost everyone moved away from trying to image DNA with an STM.  In the late ‘90s and early 2000s, the three or four remaining groups showed that DNA could be imaged, but only under very specific and difficult conditions completely unlike those used in the early ‘90s, conditions that, so far, have barred using an STM for genetic sequencing.

The STM of DNA case, then, leads to my two points about Stellacci’s stripes.  My point for the defense is that even a result that turns out, after vigorous debate, to be wrong can be extraordinarily productive.  Lots of people got into STM on the basis of those images of “DNA”, even if it’s still uncertain whether there was actually any DNA there.  The debate about those images led to a general tightening of standards for STM image production and interpretation, and a better understanding of which applications STM was and was not good for.  The DNA boom provided an early market for commercial STMs that gave manufacturers the revenue to build better versions that could be used in more appropriate ways.  Since the vast majority of scientific findings are ignored, a finding that turns out to be questionable but which is actually taken up for productive debate is doing pretty well.  I don’t have the expertise to say whether Stellacci is correct or not, but even if the consensus emerges that he is wrong, it looks to me like he and his skeptics will still have managed to move the field forward, not back.  (Conversely, if his skeptics turn out to be wrong, they also will still have done the field – and Stellacci himself – a great service).

My point for the prosecution is that some of the criticism of Stellacci’s skeptics’ methods is misplaced.  Service’s article offers a number of quotes from Stellacci and his allies complaining that the skeptics have used extrascientific means to carry out the debate – that they have resorted to blog posts and non-peer-reviewed articles instead of remaining within the arena of peer-reviewed journals.  That’s a rather ahistorical view of how scientific controversies proceed.  Peer reviewed journals are, of course, an important mechanism for fostering the validity of scientific contributions, but we all know that peer review is slow and hardly error-free.  Often, peer-reviewed articles only make sense within some ecology of other forms of communication.3  In the “STM of DNA” controversy, the thread of argument left quite a light footprint in peer-reviewed articles – it’s difficult to piece together, just from published texts, who said what when, much less when and why various actors changed their minds about STM of DNA.  Most of the influential voices used conference presentations and post-presentation conversations to persuade themselves, each other, and the rest of the community that there were serious problems with STM of DNA.  Some of the forms of communication used by Stellacci’s skeptics today (blogs and Twitter) didn’t exist back then, but if they had you can bet they would’ve been used too.  If history is anything to go by (and I hope it is!) there’s nothing inherently unscientific about using any mode of communication you can to get your point across.

Screen Shot 2014-02-10 at 12.02.38 PM

If you liked analysis and narrative here, you’ll probably like Cyrus’s book (which won the 2013 Cushing Memorial Prize)

  1. By forming a community, he argues, these researchers were able to innovate rapidly, share the microscopes with a wide range of users, and generate prestige (including the 1986 Nobel Prize in Physics) and profit (as the technology found applications in industry). Mody shows that both the technology of probe microscopy and the community model offered by the probe microscopists contributed to the development of political and scientific support for nanotechnology and the global funding initiatives that followed. []
  2. described in Aryn Martin, “Can’t Any Body Count? Counting as an Epistemic Theme in the History of Human Chromosomes,” Social Studies of Science 2004 []
  3. For a particularly engaging article on this topic with a self-explanatory title, see Bruce Lewenstein, “From Fax to Facts: Communication in the Cold Fusion Saga,” Social Studies of Science (1995). []

10 thoughts on “Checks and Stripes

  1. Really nice post. I will go check out more of you and of your posts in general. I think you do a good job of describing the meta aspects.

    I admit to not believing in the stripes. I’m not an STM expert but have a wide background in chemistry and analytical techniques and seeing human nature of people not wanting to admit mistakes.

    For me, the biggest thing with that image is the unlikeliness of all the stripes lining up in the same direction (perpendicular to the scan). You should see some different angles on different balls and even some “archery targets”. There’s a bunch more stuff when you dig into it (feedback current 1000 times the norm, raw data showing streaks and noise in many areas, offline zooming (essentially creating pixels that were never recorded), and many others.

    There are a couple times that Raph and Phil have possibly over-reached on a blog science argument (Radon plots perhaps are not really additive). But in general, they are really good and there’s just many arguments against the stripey images.

    On the meta side, I think they’ve also been fairly easy going on the condemnation aspect. I would have been way harsher, not for FS being wrong, but for not wanting to see if he was wrong (instead trying to prove himself right, or argue himself right). It’s not a big deal to be wrong, but digging your heels in, not sharing data and samples, crying “cyberbully”…they’re not how truth seeking scientists should welcome inspection of their work (either it shows a small or BIG problem, or it validates the work even more.) FS even had a grad student call out the problem in 2005 but did not take the opportunity to go back and check his work and retract the 2004 paper. Instead, he just kept generating high profile follow-on papers.

  2. A most interesting perspective. Not being a historian, but having some curiosity for the subject, what is the “position” when the challenges of a technique such as SPM were figured out in the 20-30 years ago, yet so-called “challenging” measurements are published decades later? Is this still real science and part of the process or instead what has been termed variously “zombie science” or “zombie ideas”, so something that has been demonstrated to to work or be true, but which raises its ugly head several decades later?

    • Yeah, great point that connects to the stripes controversy since Stellacci’s critics have claimed that his images contains artifacts of a type that were well understood 20 years ago. In general, I’m sympathetic to veteran STMers who get frustrated when they see well-known artifacts pass into the peer-reviewed literature. Certainly, newcomers to the STM have a duty to learn what kinds of artifacts it’s prone to. On the other hand, STM (and AFM) are sort of victims of their own success – these techniques have become so common that it’s impossible to ensure that everyone uses it in ways that veterans would approve of. So I guess you could call that zombie science, but it’s also an indicator of how successful the STM and AFM have become.

      • Cyrus,

        Your comment nails the problem on the head. I coordinate an 11-partner EU network called ACRITAS (www.acritas.eu) which is focussed on exactly this issue. AFM/STM have indeed become victims of their own success and are now treated as black-box techniques. Not only can this lead to the type of major issues we see with the striped nanoparticle data, but it also very much restricts innovation with regard to the evolution of the instrumentation.

        Philip

        P.S. “Acritas” is the Latin for force or sharpness…

  3. This is a great, well-balanced post – thanks for writing it and being fair to all sides involved. I enjoyed your discussion of STM of DNA – this is something that we also discuss in the introduction to our paper critiquing the evidence for stripes (https://pubpeer.com/publications/B02C5ED24DB280ABD0FCC59B872D04 ).

    I have just one comment, and it relates to Kevin Kelly’s analogy for imaging the stripes on nanoparticles. I’ve met Kevin a couple of times, and he and his group have done some neat STM work, but I’m afraid that the analogy in this case doesn’t hold water (if you’ll forgive the ‘pun’). Kevin’s analogy works well to describe an effect called aliasing or undersampling but the reason that it doesn’t apply to the striped nanoparticle data is that there was absolutely no reason for Stellacci and co-workers to restrict the sampling frequency to such a low pixel density. Nobody does STM like that – we zoom in to get higher resolution images in real time (just like an electron microscopist would do), rather than taking a low resolution scan and relying on imaging processing/manipulation to zoom in ‘offline’ after the image was acquired.

    The reason Francesco et al didn’t reduce the scan area (i.e. increase the strobe/sampling rate, to use Kevin’s analogy) in real time was because, as a postdoc who worked with Francesco put it to me, when they zoomed in “the delicate contrast was lost”. That alone should have sounded major alarm bells because if the features disappear when you zoom in then you really cannot treat those features as real – this is SPM 101. It’s one of the first things probe microscopists learn…

    • Thanks, glad you liked the post, and it’s an absolutely fair point about the analogy. Just to be clear, Kevin bears no responsibility here – he simply gave me a grab bag of analogies for thinking *in general* about artifacts that can happen in STM and other imaging technologies. He wasn’t specifically using the water analogy to explain the criticisms of the stripy nanoparticles images. My bad for muddying things in the way I introduced the analogy. The point I was hoping to get across is that STM is a technique where you can very easily get images with periodic features, even when there’s no periodic feature to be imaged. That can happen through a number of routes, of which aliasing (as in the strobe analogy) is just one [as Philip says, though, aliasing is not the possible artifact being debated in the stripes case]. At the same time, very often in nanoscience people hope to synthesize a structure that has periodic features, so they’re looking for STM images with the hoped-for periodicity. That dynamic encourages STMers to seize on and publish images with periodic features, but it also gives skeptics plenty of ammunition to question such images.

      • Thanks for this post (I’ll have a few other questions for historians of science soon). I think your last point in the comment above is very true “At the same time, very often in nanoscience people hope to synthesize a structure that has periodic features, so they’re looking for STM images with the hoped-for periodicity.”

        A couple of years ago, I showed these STM images to an SPM colleague who looked at them in dismay. When I asked her how she thought such images had ever been published, she made precisely that point. The 2004 paper was accepted – according to her – because at that point in time the community was looking for this kind of evidence of organization of ligands at the surface of nanoparticles. There were already a couple of articles with indirect measurements but this was the first *”{[direct visualization]}”*. I have to admit that even this urge is a very weak excuse given the multiple obvious and very serious problems with the 2004 article.

  4. What about the bad work on DNA from 1990. Isn’t there something wrong that people muttered at conferences that it was wrong, but it was never really repudiated or peer reviewed critiques or corrections made?

Leave a Reply

Your email address will not be published. Required fields are marked *