The International Journal of Impact Engineering has a highly technical retraction of an article whose authors discovered that crucial findings relied on a model that was different from the one they reported using in a particular figure.
The article, “Dynamic behavior of polymers at high strain-rates based on split Hopkinson pressure bar tests,” was published online last November by a pair of researchers from the U.K. and China. As the notice states:
This article has been retracted at the request of the editors and authors. The authors realised recently that the FE model used in the paper is different from that shown in Figure 4 of the paper. The calculations were mistakenly undertaken using an axisymmetric FE model of a tube for the solid SHPB specimen. The authors examined the possible consequences of this mistake. The recalculation of Figure 9 indicates that the transition strain-rate predicted from the correct model has not been observed up to 104 1/s, which is two orders higher than the transition strain-rate predicted by the wrong model (in the order of 102 1/s). Since the strain-rate of 104 1/s is already outside the testing range of conventional SHPB set-ups, the authors can no longer draw the conclusion that there is considerable lateral confinement contribution to the testing data from conventional SHPB tests on polymers when the strain-rate is greater than 102 1/s, which completely undermines the conclusion of the study. The authors apologise for misleading any readers due to this unfortunate mistake.
Unfortunate, indeed.
What’s not clear from the notice, however, is what that wayward figure — the one upon which the whole paper evidently rested — was doing in the paper.
Apology: Infrared Photodetectors Based on Reduced Graphene Oxide and Graphene Nanoribbons
http://onlinelibrary.wiley.com/doi/10.1002/adma.201190182/full
Sounds like the Finite Element (FE) model they should have used was the one correctly given in Figure 4 of the paper. But the calculations were actually made using a different model.
A careless mistake to make – but explainable – when selecting from a multitude of FE models to run with a given set of data.
But what may be more disturbing is that there would have been no conclusions of interest – and therefore no paper – except when using the wrong model.
In advertising it is said that “All publicity – even bad publicity is good”.
It makes me wonder if a retracted publication too, like bad publicity, has its uses in the academic world??