|
Schön Shown the Door
By
Alexander Hellemans, Contributing Editor
Investigation of star Bell Labs researcher stops
just short of declaring his work bogus
15 October
2002—When charges surfaced earlier this year that work
published by Lucent Bell Labs researcher Jan Hendrik Schön
might be seriously flawed, the general hope and perhaps even
expectation was that defects would prove minor or incidental
and could be attributed to mere carelessness or
overzealousness. With the release at the end of September of a
blue-ribbon investigative report, all such hopes have been
completely dashed.
"There is
a massive tendency of misrepresentation and altering of data.
It is almost unimaginable what happened," observes physicist
Arthur Hebard of the University of Florida in Gainesville,
accurately capturing the prevailing mood. (Hebard was one of
the many who had tried to replicate some of Schön’s
results.)
Over the
past few years, the 32-year-old German-born physicist had
produced a stunning series of papers at a rate of close to one
per week, reporting what seemed to be seminal work in several
fields: superconductivity in fullerenes and polymers, new
organic field-effect transistors, and even a one-molecule
transistor. Schön claimed to have done the work both at Lucent
Bell Laboratories in Murray Hill, N.J., and at the University
of Konstanz, in Germany, and all the papers had
coauthors.
Then, last
May, several researchers drew attention to oddities in some of
his papers—duplication of data in graphs, for example, and
identical sections in different graphs supposedly representing
the electrical behavior of entirely different samples.
Meanwhile, concern had mounted among scientists having
difficulty reproducing many of Schön’s more spectacular
results [see "Identical
Graphs Chart a Dubious Picture," IEEE Spectrum,
July, pp. 20–21].
Lucent
responded quickly and set up a five-member independent
investigative panel, which was led by Stanford’s Malcolm
Beasley and included IEEE Medal of Honor recipients Herwig
Kogelnik and Herbert Kroemer [see interview
with Kroemer]. Its report, released by the lab on 25
September, is a devastating indictment of Schön, who was
immediately dismissed. At least by implication, the report
also raises serious questions about Schön’s supervision and
the laboratory’s management.
Conduct unbecoming
The panel scrutinized 24 cases of
alleged misconduct and found clear signs of it in 16. "The
evidence that manipulation and misrepresentation of data
occurred is compelling," concluded the panel. "At a minimum,
Schön showed reckless disregard for the sanctity of data in
the value system of science."
The panel
grouped the 24 allegations in three categories: "data
substitution," "unrealistic precision," and "contradictory
physics"; each of the allegations is discussed in detail in
Appendix E of the report. A striking case of data
substitution, for example, was the similarity in the graphs
obtained from inverters created with self-assembled monolayer
organic field-effect transistors (SAMFETs), single-molecule
transistors, and pentacene transistors.
A case of
unrealistic precision, highlighted in the body of the report,
was Schön’s treatment of the breakdown strengths in the
insulating layers of aluminum oxide that he used to insulate
the SAMFET transistor channel from the gate electrode. The
panel found that the breakdown strengths fit a Gaussian
distribution to a highly improbable degree, suggesting that
Schön had generated the reported strengths out of whole cloth,
using equations.
As an
example of results inconsistent with stated device parameters
and prevailing physical understanding, the report cites
measurements done on a unipolar inverter: though both
transistors were n-types, the inverter’s reported behavior
corresponded uncannily to what one would expect from a
complementary inverter with one n-type and one p-type
transistor.
An
especially shocking finding was Schön’s inability to produce
laboratory notebooks or computer records of his experimental
work. Meticulous maintenance of lab notebooks is, of course,
standard procedure in research and a sacred tradition at Bell
Labs. "If you cannot document your results, they should not
count," observes Paul McEuen, a physics professor at Cornell
University (Ithaca, N.Y.), one of the first to blow the
whistle on Schön.
Collaborators cleared
The panel’s report completely
cleared Schön’s collaborators—notably Zhenan Bao and Christian
Kloc, as well as his superior Bertram Batlogg—of any
scientific misconduct. Bao and Kloc were found to have
supplied materials to Schön but not to have directly
participated in his experiments [See "Who’s
Responsible?"].
Schön’s
inability to document the experiments leaves them, and the
journals that published the articles they cosigned, in an
awkward position. Should the articles be retracted, and what
is the procedure for doing that? [See "Taking
It All Back"]
In the
absence of documentation, and in light of Schön’s inability to
reproduce some of his own reported results when challenged by
the panel, speculation is rife about what happened. Did he, in
a panic, destroy computer data and lab notebooks wholesale? Or
did he, in fact, never perform some of the experiments in the
first place?
In a
response to the panel’s report published as an appendix, Schön
maintains that "all the scientific publications that I
prepared were based on experimentation." And he has his
defenders. Norbert Karl at the University of Stuttgart in
Germany says he has "no doubt" that the experiments are real.
"Those measurements are fitting into what we know," says
Karl.
The more
common view, though, is that Schön’s results fit what was
thought was known only too well, and that they too well
confirmed ideas of Schön’s mentor, Batlogg, of the
Eidgenössische Technische Hochschule (ETH) in Zurich. Batlogg,
himself a star, well known for his work in superconductivity,
and a former head of the solid-state physics division at Bell
Labs, signed 17 of the 24 disputed Schön papers.
To judge
from the cautious language of the report, Batlogg’s role as a
manager and cosignatory had the panel in rather a quandary.
"The committee concluded that the coauthors...have, in the
main, met their responsibilities, but that in one case
questions remain that the committee felt unqualified to
resolve, given the absence of a broader consensus on the
nature of the responsibilities of participants in
collaborative research endeavors."
Many in
the community feel that Batlogg got away too lightly in the
report, says Hebard—though, he adds, "if you read between its
lines, then maybe he did not get off so easily."
In a
statement that he circulated to concerned fellow scientists
and also provided to Spectrum, Batlogg expressed contrition:
"As coauthor I acknowledge a responsibility to ensure the
validity of data in publications. I have learned, with the
deepest of regrets, that the control measures I took in this
extraordinary case were not adequate…I had placed too much
trust in my collaborator. What had worked well in the past
failed in this case of intentional wrong-doing."
To
download the full report go to http://www.lucent.com/news_events/researchreview.html.
|
