THE glittering reputation of a physics institute responsible for some of the most important discoveries of the 20th century has been severely shaken. A spectacular string of advances in superconductivity and nanotechnology reported by researchers at Lucent Technologies鈥� Bell Labs since 1998 were probably more fiction than fact, according to a panel established in May to look into allegations of scientific misconduct at the company.
The panel, chaired by physicist Malcolm Beasley of Stanford University in California, places the blame squarely on researcher Hendrik Sch枚n. His work promised to revolutionise the way electronic circuits and ultimately computers are made. But the panel concludes that 鈥渢he evidence that misrepresentation and manipulation of data occurred is compelling鈥�.
Everyone else named on Sch枚n鈥檚 papers was cleared of any misconduct, but other scientists working in the field have told New Scientist they think the Beasley report raises serious questions about practices at Bell Labs and other research organisations, and at the journals that published Sch枚n鈥檚 work.
Advertisement
In May, Bell Labs in New Jersey asked Beasley to head a five-member panel to look into mounting allegations of scientific misconduct in 25 of Sch枚n鈥檚 recent papers. Most were published in the highly prestigious journals Science, Nature and Applied Physics Letters. The committee鈥檚 report, released last week, details misconduct in 17 of those papers. That includes cases of data substitution, where the same graph was used to represent results from different experiments (see Graphic), cases where data was generated artificially, and claims of what panel members are calling 鈥渃ontradictory physics鈥�.
Sch枚n, now aged 32, joined Bell Labs as a postdoc in 1998. Between then and 2001 he published more than 70 papers that made him a star researcher in his field. They include the widely reported creation of a single-molecule transistor and the discovery of superconductivity in carbon-60 molecules (see story 鈥淲ith hindsight, it was a hell of a lot of papers鈥�).
In May this year, Sch枚n admitted to mistakenly including the same graph in different papers to describe the electrical behaviour of two different materials. But that admission led to other allegations. Paul McEuen of Cornell University in New York state noticed a third figure in an earlier Science paper by Sch枚n, similar to the two he had admitted were duplicates.
McEuan and Lydia Sohn of Princeton University in New Jersey immediately took a closer look at more of Sch枚n鈥檚 papers and noticed other cases where unrelated graphs seemed to contain the same data. 鈥淥nce you had the idea in your head it was simple to find these examples,鈥� says McEuen. They notified Bell Labs, triggering the investigation.
The case raises far-reaching questions about the effectiveness of peer review. If McEuen and Sohn could cast doubt on Sch枚n鈥檚 work simply by reading his papers, why couldn鈥檛 the journals who published his work do the same? Nghi Lam, editor of Applied Physics Letters, which published four of the discredited papers, says journal referees don鈥檛 have time to cross-check every paper by an author. Even if they did, it鈥檚 not clear they would have noticed the problem. 鈥淚t requires a very big change of viewpoint to say 鈥業鈥檓 looking for fraud鈥�,鈥� says McEuen.
Perhaps that explains why journal referees didn鈥檛 question Sch枚n鈥檚 findings, for example when he reported making molecular-scale transistors that performed far better than anyone thought possible. When the panel asked Sch枚n to substantiate his claims by showing his raw data, he said he had deleted the relevant files after running out of space on his computer.
While the panel couldn鈥檛 prove Sch枚n was lying, such poor record keeping constitutes scientific misconduct, says Beasley. Other research organisations should take note. 鈥淚t鈥檚 time to go back and re-examine procedures and how you keep records in this computerised age,鈥� he says.
Don Kennedy, editor-in-chief of Science, says journals can鈥檛 be held responsible for catching researchers who set out to deceive. 鈥淚 don鈥檛 think any peer-review process can be proof against clever fraud,鈥� he says.
Bell Labs also denies culpability, describing the incident as an 鈥渁nomaly鈥�, but says it will review its guidelines on supervisors鈥� responsibility. So should Sch枚n鈥檚 supervisor, Bertram Batlogg, have questioned the integrity of his researcher鈥檚 data? The panel steers clear of such a suggestion, saying it 鈥渄oes not consider itself qualified to make a specific judgement in the case鈥�. Batlogg, now at the Swiss Federal Institute of Technology in Zurich, did not respond to requests for comment but has issued a statement saying he believes he did ask probing questions of Sch枚n and was satisfied by his replies.
Others say the fact that no co-authors and internal colleagues ever witnessed Sch枚n鈥檚 most stupendous discoveries runs counter to normal scientific practice. 鈥淚t is unthinkable not to participate in the measurement process,鈥� says Phil Anderson at Princeton, who won the 1977 Nobel Prize in Physics for work carried out at Bell Labs on the theory of electrical conductivity.
Anderson suggests scientists should look again at the practice of including co-authors on papers when they don鈥檛 have the expertise or time to critically assess the results.
As New Scientist went to press Sch枚n was unreachable after being fired by Lucent last week. In a public statement he apologises to the scientific community for mistakes made in his work, but stands by the physics he found. 鈥淚 truly believe that the reported scientific results are real, exciting and worth working for,鈥� he says.