Thursday, November 22, 2012

17 citations in 12 years in not impressive

I can fully imagine a grant reviewer, tenure or hiring committee saying that when reviewing the "impact" of a particular publication of an individual.

Furthermore, one can also say
"the paper was in PRL but it was a full 12 months between submission and publication. Clearly, he was lucky to get in PRL at all..."

But, there is a problem with all this.
These observations apply to Duncan Haldane's 1988 paper Model for a Quantum Hall Effect without Landau Levels: Condensed-Matter Realization of the "Parity Anomaly"

Haldane's paper has been receiving about 100 citations per year for the past few years.
It now has a total of 530 citations in Physical Review journals.
However, from 1988 to 1999 it received only 17 citations.
Hardly impressive.


  1. "Of course one of the things that happened was, as you may or may not be aware, is that they gave me tenure after only two years and with no publication record."

    -Nobelist Kenneth Wilson

    It's interesting how many great thinkers don't fit the cookie cutter mold of a successful scientist. I wish scientists worried more about the significance of their publications than the quantity.

  2. On the other hand, did Haldane miss something significant? Or did he not sell his work adequately (as though he's obliged to!)? In thinking about how to make his system realistic recently, I was forced to write down a spin orbit coupling term. He wrote a spinless model, which is unphysical. You need to break the spin degeneracy in a real model, so you need to write a spin-orbit coupling term. It turns out that the simplest SOC term you can write, is the one that Kane and Mele wrote down in 2005, and Kane may get the Nobel prize for doing so. Not because he wrote it down, but because he and S.C. Zhang realised how significant that term, and the corresponding edge states, was.

    I feel like they weren't really doing anything more than Haldane did, they just followed it up more, and people caught on. And then Haldane's citations flowed.

    So was Haldane 22 years ahead of his time, or was he one of the many people who almost discovered something paradigm shifting, and then didn't quite get as far as he ought?

    I should emphasise that I love Haldane, have depended heavily on his paper for my own understanding, and know that I would never have thought to write what he wrote, nor what Kane-Mele wrote. My science is far too derivative and unimpressive.

    I'm just wondering is all... You can be way ahead of your time, or you can almost discover something and then not really get there. Both of these can lead to substantial delayed citations.

  3. @Tony

    I am not sure I understand how you ended with a Kane-Mele like model with spin-orbit coupling just by making Haldanes model more realistic.

    Haldanes point was to demonstrate, through a toy model, that a finite magnetic flux and Landau levels are not necessary for Quantum Hall Effect. He demonstrated you can get it just by breaking time-reversal symmetry. Preserving time-reversal would not lead to the parity anomaly, which is what he wants to realize. You can also see he writes that he can achieve chiral fermionic modes at the edge of the system, without any modes of opposite-chirality which can cancel the anomaly. The fact that you can have helical modes, like in the Kane-Mele model, without backscattering due to Kramer's theorem is also something that needs more that just "following up" on his work.

    Also note that just by writing down a model with spin-orbit coupling you haven't achieved anything. The non-trivial part (solved by Kane and Mele), is to realize there is a completely new type topological invariant at play and to calculate it. Kane and Mele even mentions words like twisted real K-theory, nothing trivial for most physicists.

    So I understand why Haldane didn't include spin and time-reversal symmetry, that would go against the point of his paper and what people where thinking about at the time. The fact that he demonstrated that QHE is possible without Landau levels, was by itself a big surprise back then, nobody could imagine it could be generalized as much as it can. The (much) more general concept of "Topological Order" by Wen wasn't defined either back then.

    I think both the works of Haldane, and Kane-Mele are highly original and deep and nothing following trivially from Haldanes work.

    Personally I think (partly) the reason why topological insulators were not discovered earlier, is that most condensed matter physicist (especially back then) are not too fond of working with ideas which isn't directly realizable by experiments. Most people like to follow clues from experiments, rather than pursue purely theoretical ideas. You can argue that most of what we call topological insulators and superconductors were discovered by high energy-physicists long time ago based on purely theoretical reasoning. The whole idea of Majorana fermions (and Ising anyons in general) where also discovered through purely theoretical work, for example based on conformal field theory (Moore-Read).

    When I discuss new theoretical toy models realizing certain subtle and deep effects, I always get dismissive attitudes from many condensed matter physicists asking "how can you realize that experimentally"?

  4. Ross, are you suggesting that Haldane would likely not have survived in modern science? From what I understand, this is also a plausible proposition for Feynman.

    Decision-making committees must proceed on the information available to them. I understand you are probably not proposing that every tenure-track scientist should be allowed one or more decades of "benefit of the doubt", in the event that what they are doing becomes well-cited or relevant for previously unforseen reasons?

    Wilson's example, cited above, is interesting: it must have been the case that people in the Department could see that he brought value despite these points. This highlights the continuing need for people with good sense, judgement and human intuition (in addition to scientific expertise and intuition) on bodies that make human resource decisions in science.

    What causes the sort of silly decisions to which the title of this blog entry alludes?

    The first is that there are many more applicants than positions (a "blame-the-baby-boomers" argument, since, in part, the old must make way for the young). This is also what makes grant-rounds effectively random: if there are more good grants than funds, decisions must be made based on arbitrary minutae, or coin-flipping. Citations are an element of minutae that, while partially arbitrary, present a thin veil of relevance. The only answers to this problem would seem to be 1) produce fewer potential applicants or 2) increase funding for positions. It seems difficult in either case, because sometimes larger PhD cohorts can bring funding into an organization, while creating more positions is a considerable budget outlay, and intra-University funding seems to be a zero-sum game.

    The second problem is that the people making human resources decisions are not sufficiently informed, lacking in judgement, or prohibited from exercising good judgement. There is a risk of this at any institution, but it is probably more severe the more "civilian" administration gets in the way of hiring.

    There is no substitute for having people with good human sense, intuition and judgement making human resources decisions. This is particularly true in Human Resources, because machines can't "suss out" people effectively, and it is unlikely they will do so in the near future.

  5. Thanks for the reply HM. You do realise i was basically just being cheeky, right?

    I just wrote a lengthy response and then deleted it! I will say this though:

    I certainly never claimed that topological insulators follow 'trivially' from a Haldane model. In fact, I said "[Kane] and S.C. Zhang realised how significant that [SOC] term, and the corresponding edge states, was." Is this not just a corollary of "Also note that just by writing down a model with spin-orbit coupling you haven't achieved anything. The non-trivial part (solved by Kane and Mele), is to realize there is a completely new type topological invariant at play and to calculate it."

    What Haldane did or didn't achieve was not my point. He achieved a lot, it's a great paper. I've read it many times, and I love it.

    However, I claimed that topological insulators began a revolution, and I claimed that the Haldane model did not. I speculated that Haldane's model was not far off topological insulators.

    You're right of course, Haldane "wanted" chiral edge states and so he needed to break TRS. Topological insulators rely on unbroken TRS to prevent most backscattering. And the real significance comes when you realise the topological source of these results.

    In this sense they are completely different results.

    But I don't think they are so different after all. Perhaps this is a matter of taste and perspective. Maybe I'm too simple minded, but I see very similar models, with very similar physics.