The Emperor’s New Stripes

Philip Moriarty

Philip Moriarty

This is a guest post by  Philip Moriarty, Professor of Physics at the University of Nottingham

Since the publication of the ACS Nano and Langmuir papers to which Mathias Brust refers in the previous post, I have tried not to get drawn into posting comments on the extent to which the data reported in those papers ‘vindicates’ previous work on nanoparticle stripes by Francesco Stellacci’s group. (I did, however, post some criticism at ChemBar, which I note was subsequently uploaded, along with comments from Julian Stirling, at PubPeer).  This is because we are working on a series of experimental measurements and re-analyses of the evidence for stripes to date (including the results published in the ACS Nano and Langmuir papers) and would very much like to submit this work before the end of the year.

Mathias’ post, however, has prompted me to add a few comments in the blogosphere, courtesy of Rapha-z-Lab.

It is quite remarkable that the ACS Nano and Langmuir papers are seen by some to provide a vindication of previous work by the Stellacci group on stripes. I increasingly feel as if we’re participating in some strange new nanoscale ‘reimagining’ of The Emperor’s New Clothes! Mathias clearly and correctly points out that the ACS Nano and Langmuir papers published earlier this year provide no justification for the earlier work on stripes. Let’s compare and contrast an image from the seminal 2004 Nature Materials paper with Fig. S7 from the paper published in ACS Nano earlier this year…

moriarty comparison

Note that the image on the right above is described in the ACS Nano paper as “reproducing” high resolution imaging of stripes acquired in other labs. What is particularly important about the image on the right is that it was acquired under ultrahigh vacuum conditions and at a temperature of 77K by Christoph Renner’s group at Geneva. UHV and 77 K operation should give rise to extremely good instrumental stability and provide exceptionally clear images of stripes. Moreover, Renner is a talented and highly experienced probe microscopist. And yet, nothing even vaguely resembling the types of stripes seen in the image on the left is observed in the STM data. It’s also worth noting that the image from Renner’s group features in the Supplementary Information and not the main paper.

Equally remarkable is that the control sample discussed in the ACS Nano paper (NP3) shows features which are, if anything, much more like stripes than the so-called stripy particles. But the authors don’t mention this. I’ve included a comparison below of Fig. 5(c) from the ACS Nano paper with a contrast-enhanced version. I’ll leave it to the reader to make up their own mind as to whether or not there is greater evidence for stripe formation in the image shown on the right above, or in the image shown on the right below…

moriarty 2

Finally, the authors neglect any consideration at all of convolution between the tip structure and the sample structure. One can’t just assume that the tip structure plays no role in the image formation mechanism – scanning probe microscopy is called scanning probe microscopy for a reason. This is particularly the case when the features being imaged are likely to have a comparable radius of curvature to the tip.

I could spend quite a considerable amount of time discussing other deficiencies in the analyses in the Langmuir and ACS Nano papers but we’ll cover this at length in the paper we’re writing.

About these ads

18 thoughts on “The Emperor’s New Stripes

  1. Phillip,
    This is a great post, and very convincing. I’m still struck as to why the STM experts on the Stellacci papers can’t provide a simple response that either confirmed that the 2004 paper images are crap or at least hint at why the simple side by side comparison here is not correct. Someone will be very wrong at the end of all this.

    Just some things that you imply, but I thought would be useful to confirm to your readers:

    -The 2004 Nat. Mat and 2013 ACS Nano papers are at comparable scales (I presume nanoparticles can vary in size by at least an order of magnitude).

    -The interpretation of the these images is straightforward, that the absence of stripes in the 2013 ACS Nano image is not a result of being an STM muggle.

    Looking forward to your paper.

    • The nature 2004 image is about 15 nm wide and the new image is about 7nm, so the stripes should appear larger in the new image. Also the new image, is as far as I can tell (I don’t have the raw data), is a full image. It probably was recorded with about 500 pixels left to right. The image on the left is a zoom see (http://raphazlab.files.wordpress.com/2013/09/2004-20131.jpg). Interestingly the “full” image here is a zoom too. These zooms have been done in a program which interpolates the data so the pixels have nice round edges (almost all drawing programs and many SPM analysis programs do this by bi-cubic interpolation automatically while expanding images, I am not trying to accuse anyone of foul play, I am just explaining how the zoom works!).

      I have the raw data from the image on the left which is 512px wide, the “full” image in the link above is only 122px wide section of that data, or in terms of area only 6% of the full image. Trying to recreate the image above I get an image of 48px wide, or in terms of area less than 1% of the original scan. Also the colour scale is over saturated in the raw data (see http://raphazlab.files.wordpress.com/2013/05/comparison-natmat-fig1a-vs-raw-data.gif) the area between the particles is not blacked out, in fact it is possible to see the stripes continue between the particles.

      On the topic of interpretation being straight forward I would say “yes and no”. No because the image is a convolution of the tip and the sample so if the tip is big and flat you wouldn’t see stripes even if they are really there. However, the fact that we don’t see stripes in the image show that this can’t be used as evidence for the stripes.

      Here we end up on the point about proving a negative. Without clear images of where the ligands are we can’t conclude that the stripes don’t exist. The point I want to make is I have never seen any convincing evidence that stripes do exist, as the early work appears to be image artefacts. The new image does not present any evidence for stripes which I can see, so we are still without any convincing evidence for stripes.

    • Thanks nanoymous. Julian’s comment (Nov. 15 at 7:53 am) covers the key points in his usual careful and comprehensive style.

      It’s also worth adding that if the stripes are not there in the raw real space STM images, moving to an analysis in reciprocal space [as Biscarini, Stellacci et al. do in their 2013 Langmuir paper, Langmuir 29 13723 (2013)] is not going to make them magically appear!

      We had a lengthy e-mail exchange with Fabio Biscarini yesterday on the matter of the power spectral density analysis in that Langmuir paper. Here is not the place to discuss whether the reciprocal space analysis in the Langmuir paper stands up to scrutiny – this is part of the paper we’re writing at the moment. To his immense credit, Fabio has very kindly promised to send us the PSD data for the paper (despite some rather intemperate and tetchy e-mails from me at times. This is another reason why I don’t have a Twitter account…).

      One thing that is worth mentioning, however – and which Julian pointed out to me – is that the following sentence appears in the Langmuir paper: “PSD exploits the whole information content present in the image, as opposed to the analysis of manually drawn line profiles in real space.”

      A power spectral density approach clearly does not exploit the “whole information content” in an image. The PSD is effectively the modulus squared of the Fourier transform. It thus cannot represent “the whole information content” of the image because all phase information is lost. If the modulus squared of a complex-valued function contained the “whole information content” not only would Fourier analysis be a very different beast indeed, but we’d have to rework the entire framework of quantum mechanics!

      This might seem like ‘nit-picking’ but it’s important to be clear as to the limitations of the analysis techniques for those who may not be familiar with them.

  2. Update added 17 Nov 2013.

    There has been considerable e-mail traffic about this post over the weekend. In those e-mail exchanges one of the authors of the papers in ACS Nano and Langmuir has claimed that I have been “unethical” (their term) by not pointing out that the image from the Nature Materials 2004 paper and that from the ACS Nano 2013 paper shown side by side above are of particles with different ligand types.

    My credulity in this stripy nanoparticle saga has been strained to breaking point many times but on this particular occasion it’s snapped entirely. To claim that the comparison of images above is unethical beggars belief.

    The author in question willfully misses the entire point of the post. My argument is clearly laid out above but I’ll reiterate the key point one more time. The STM instrument which should be most capable of resolving stripes (due to its enhanced environmental and instrumental stability compared to the other instruments used for the results in the ACS Nano and Langmuir papers), i.e. the LT UHV STM in Geneva, has produced only one image which was deemed worthy of inclusion. (And even then, it’s included only in the supplementary information for the ACS Nano paper). It’s shown above. This particular image shows no evidence at all for the type of stripes reported previously by Stellacci and co-workers, as exemplified by the image from the Nature Materials 2004 paper.

    To argue that the UHV LT STM couldn’t resolve the stripes because the wrong type of particle had been sent to Renner’s group is an exceptionally weak argument. As I said to the author in question in an e-mail yesterday:

    “The onus is on *you* to use the same particles and to do the direct comparison. There’s a very simple way to prove me wrong – send the same type of nanoparticles used for the Nature Materials 2004 paper to Christoph and have his group show that it’s possible to get images where the contrast is the same as in the Nature Materials paper while the tunnel current is being tracked reliably . *Then, and only then, will I modify the post at Raphael’s blog.* Do this and not only will I retract that blog post but I’ll donate £1000 to a charity of your choice.“

    Predictably, this suggestion was rejected by the author in question.

    • Very confusing for observers like me who are not subject specialists.

      Surely preparing the same type of nanoparticles and running good STM and demonstrating images that are comparable to the ones in the 2004 Nat Mat. paper would essentially settle the issue? With the successfull acquisition of those images, we could be reasonably certain that those image features were not a result of poor STM practice. One could probably just post the images and associated data to this blog and send some nanoparticles over as well….there would be no need to spend anymore time on this by anyone, I imagine?

      • That’s exactly it, nanonymous. The obvious experiment to do, which, if successful, would silence all of Stellacci’s critics, would be to reproduce the images seen in the Nature Materials 2004 paper (and the subsequent JACS, J SPM etc… papers) using UHV STM at 77 K (or 5 K).

        This hasn’t happened and when I suggested to one of the authors of those papers that this is the experiment they should do, they outright refused to do it. One has got to wonder why…

      • Clearly someone on the other side is reading this blog (given Phillip’s emails). I’d ask them to please write in (why not do it anonymously?) and elaborate on these two points:

        1) Are the images in the 2004 Nat Mat paper valid data and not merely the result of scanning artifacts?

        2) If the answer to 1) is “yes they are valid data”, then wouldn’t it make sense to put an end to the debate and scan some particles put some pictures here (you can use imgur) and send them off to another group so that they can verify? Why not do this instead of things like PSD analysis? Could you suggest why this would not be a good idea?

      • @nanonymous.

        This blog is indeed being read by Francesco Stellacci and, most likely, his co-workers. But Francesco has repeatedly said that he will not communicate via blogs.

        I can answer your questions on the basis of recent e-mail exchanges (which were cc-ed to a number of people and were not labelled as confidential):

        1. Francesco claims that the images in the Nature Materials 2004 paper are valid data. We critique those images in detail in the paper we are writing. Francesco was good enough to (eventually) send us the tunnel current image acquired in parallel with the topographic images included in the Nature Materials 2004 paper. In short, from the tunnel current map it is clear that the feedback loop was not tuned correctly. This problem with feedback loop tuning was raised by Pedja Djuranovic many years ago.

        I stress again that feedback loop ringing is just one possible source of artefacts in STM images.

        2. It would make complete sense. But Francesco has point-blank refused to do this.

        Philip

      • What a shame. Professor Stellacci should understand that many prestigious academics (outside of chemistry/materials science) use informal electronic mediums (like blogs) for the dissemination of information, it is a very efficient way of doing things. There is a very simple, clear question here (the validity of the 2004 images) being put forth by credible scientists. There are many more reading silently. I am guessing that Pep is looking at this blog at least occasionally, surely he must agree that we can leave all the detailed arguments behind as long as the 2004 images can be reproduced? Reproducing images would be far faster than writing entire papers as a response.

        Why Professor Stellacci and other members of the response papers wouldn’t be eager to quickly sort this out leaves me puzzled and disappointed. The silence that we see here and on by Molly Stevens on her papers is deafening.

    • I’m not a super Heisenberg physicist. But just basic design of experiments would say that FS should have sent the same particles. How can he simultaneously say that ‘the 2013 papers validate the 2004 work’, and then ‘but you can’t compare them because the ligands are different’.

      Really, the whole corpus of work is a mess with how he never did a proper study in the beginning (just even like a six sigma person in a factory writing down a list of variables and examining response in the parameter space…this is basic logic). He has all these papers that are supposed to support each other but are in different systems and do different things. It’s a mess.

  3. Apologies. One sentence in the comment above needs clarification. When I say that “the LT UHV STM in Geneva, has produced only one image which was deemed worthy of inclusion” this should read, “the LT UHV STM in Geneva, has produced only one image which was deemed worthy of inclusion in the ACS Nano paper”.

    There is one other nanoparticle image produced by the Geneva group’s STM system. This is in the Langmuir 2013 paper, specifically Fig. 1(c) of that paper. We discuss this image, and the other images in Fig. 1 of Biscarini et al., Langmuir 2013 (and the associated PSDs) in the paper we are preparing.

    This clarification doesn’t affect any of the arguments in, or the general ‘thrust’ of, the preceding comment

  4. I certainly did not post that at Fakerapha, in very bad taste for the author to attribute words on the fakerapha to my my anonymous handle….

    I might not care if was written in a funny tone…but it’s just sort of whiny…

    It is cues like the fakerapha blog that suggest that there is really something odd going on with the whole stripy affair…

  5. Pingback: The stripy controversy as a window into the scientific process* | Rapha-z-lab

  6. Pingback: Open science to settle stripy controversy? | Rapha-z-lab

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s