When Science Goes Wrong (41 page)
So far, it looked as if the problem was a technical one involving the data-analysis software. Either Ninov’s software was finding events that didn’t exist, or Loveland’s was failing to find events that did exist. A second investigatory committee was formed in June of 2001, headed by Darleane Hoffman. The committee delegated a post-doc and Victor Ninov to independently search for the reported decay chains in the output files from both the 1999 and the 2001 runs: neither of them could find any of the chains. Meanwhile, Ken Gregorich got into the act. He wrote an entire data-analysis program from scratch in an attempt to resolve the conflict. When he applied this program to the data, he again failed to see any sign of the four decay chains that Ninov had reported.
Because the problem now seemed to be in the data-analysis software rather than the equipment, a third committee was assembled, this time consisting of Berkeley computer experts, including Gerald Lynch, Augusto Macchiavelli and two others. At some point in this process, some of the magnetic tapes that carried the original output from the detectors went missing, and they have remained missing ever since. The Lynch-Macchiavelli group also found that some of the crucial files from Ninov’s own computer were missing. Loveland had copied all of Ninov’s directories onto his own computer, however, and the investigators relied in part on those copies in their further studies.
One thing that greatly aided the computer sleuths in their investigations was the existence of a ‘log file’ in the data-analysis software. This file was a record of all the activities that had taken place during the process of analysis, as well as who had performed them. The log file showed that Ninov had indeed run an analysis on the data from the 2001 run and that it had come up with the element 118 decay chain, just as he had told his colleagues. However, another similar analysis run later with the same software had not shown any such decay chain. The sleuths went through these two files, checking every tiny detail, and they found evidence that the first file had been manipulated – some of the details, such as the timing of the supposed events, just didn’t make sense, and the lengths of the pages were not what would have been generated during normal operation. They then showed that the apparent record of a decay chain in this file could easily have been inserted by cutting and pasting code from elsewhere in the software. When the investigators turned their attention to the records from the 1999 runs, they found similar evidence of manipulation. In all cases, the suspect manipulations had been made from Victor Ninov’s computer account. Furthermore, in 1999 at least, Ninov was the only person in the group who knew how to run the data-analysis software.
Once they realised that the alpha decay chains were not in the data, Gregorich and Loveland hastened to let the scientific community know – Loveland mentioned it at a research conference in New Hampshire in late June of 2001, for example, and the Berkeley group also put out a news release on the topic at the end of July. They knew that it was also necessary to send an official retraction to the journal in which the original report had appeared, and Gregorich sent off the retraction statement at the beginning of October. But Victor Ninov sent his own letter to the journal, in which he asked for his name to be removed from the retraction statement. This put the editors of
Physics Review Letters
in a quandary. They sat on the letter for a year, while correspondence went to and fro between the editors, Ninov and the other authors. Finally, in July of 2002, the journal announced that ‘all but one’ of the authors had requested a retraction – they didn’t name Ninov as the standout. That put the paper into a kind of permanent vegetative state in which it still lingers today, although everyone knows that the Berkeley Lab’s claim to have discovered element 118 is dead.
As soon as Ninov’s onetime collaborators in Germany heard of what had transpired, they were concerned that similar invalid data might have been used to document the discovery of some of the elements he had worked on there. So the GSI scientists went back and examined all 34 decay chains that had been used to document the discovery of elements 110, 111, and 112. They found that two of the chains – the second of the chains reported in the element 110 study of 1994, and the first of those reported in the element 112 study of 1996 – were based on ‘spuriously created’ data. Sigurd Hofmann, the team’s leader, told the Berkeley group about this, and in due course the GSI group published a retraction of those specific chains. This did not put their claims for discovering elements 110 and 112 into doubt, however, because additional, genuine chains were observed in the same and subsequent experiments. Although the published retraction did not go into details, Sigurd Hofmann told colleagues and reporters that the spurious chains had been created intentionally – by Victor Ninov.
‘That was the killer,’ said Walter Loveland. The world was now closing in on Ninov. The university convened a fourth and final investigative committee, headed by Rochus (‘Robbie’) Vogt, a retired Caltech physicist. The committee surveyed the results of all the prior investigations and interviewed the participants – except for Ninov, who declined to attend committee meetings after the first introductory session. In March 2002 the committee issued its report, which named Victor Ninov as a perpetrator of scientific fraud.
The other investigators took some heat too – not for faking data, but for failing to independently analyse the data before publication. ‘The Committee finds it incredible,’ Vogt wrote, ‘that not a single collaborator checked the validity of Ninov’s conclusions… The claim of an important discovery demanded no less.’
Neither Gregorich nor Loveland was very happy with this part of the report. Gregorich told me that he first set eyes on the report when it was sent to him by a journalist, and he felt aggrieved that he hadn’t been given the chance to ‘check it for accuracy’. Loveland took issue with the general notion that scientists are obligated to validate the authenticity of their colleagues’ work. ‘In theory the answer is, yes, we should do this,’ he said, ‘but in practice the way science is carried out in large collaborative experiments is that people have assigned responsibilities, and if they’re reliable and there’s no question about their work, that kind of double- and triple-checking doesn’t occur. There’s just not enough time for anyone to do that sort of thing, so you trust people to do various portions of the experiment.’
Of course, the main focus was on Ninov. In November of 2001, even before the Vogt committee met, he was put on paid leave and banned from the Berkeley Lab. In May of the following year he was fired.
After Victor Ninov left Berkeley, he took a teaching position at the University of the Pacific in Stockton, California. When I tried to reach him there in the fall of 2006, I was told that he had left the college and that no one knew how to contact him – something that turned out not to be quite true, as (unknown to me) Ninov’s wife was still a professor at the college. Eventually I was able to track him down, and he agreed to a telephone interview, which took place late one evening in December of 2006.
I started the interview, as I usually do, by asking whether I could tape the conversation. Ninov said that I could not, because previous journalists had twisted his words; the only journalists who had treated him fairly, he said, were Germans. Casting about for something that might change his mind, I offered to conduct the interview in German if I could use the recorder. To my surprise he agreed, and our conversation proceeded in that language for a few minutes. Unfortunately, I soon found that my German – which has been languishing in a dark corner of my cerebellum since my student days in that country – wasn’t up to the technicalities we were discussing, so I lapsed into English. Ninov quickly pounced. ‘Are you still taping the conversation?’ he demanded. ‘Yes,’ I said. ‘So you have already broken your word,’ he replied, and he demanded that I stop the tape and erase what I had already recorded.
At that point I thought the interview was about to come to a premature end, but in fact Ninov kept talking for an hour longer. He talked fast, sometimes about highly technical matters, and he didn’t always pause to deal with my requests for clarification. Thus I don’t claim to have understood everything he said, and I certainly wasn’t able to write much of it down. However, I did take home a number of key points.
Most importantly, Ninov emphatically rejected all the accusations that have been made against him. ‘I did not do anything wrong,’ he said. ‘I stand by the integrity of everything I’ve done in my scientific life.’ And he denied he was ever less than frank in his dealings with his colleagues. ‘I presented everything openly to everyone; I never tried to hide anything.’
Ninov told me that he had never liked the Berkeley Lab. The intellectual atmosphere was stuck in the 1950s, he said, and he didn’t trust anyone there except Gregorich. The equipment was outdated and the resources meagre. In building the gas-filled separator, for example, he had been forced to recycle decades-old magnets that he found in a storeroom. The main thing that kept him at Berkeley was its coastal location: he kept an ocean-going sailing boat in the Berkeley marina, and since leaving Berkeley he has sailed it across the Pacific. (The GSI, he said disdainfully, is surrounded by forest.) In addition, he married an American soon after his arrival.
He also told me that he had never wanted to participate in the search for element 118. ‘I resisted to the bitter end to do the experiment,’ he said. His reasons had partly to do with the fact that the gas-filled separator wasn’t ready – it wasn’t yet possible to accurately measure the number of krypton atoms impinging on the target, for example. This criticism is substantiated to some extent by what I read in Hoffman and Ghiorso’s book: they spoke of modifications being made to the setup even after the runs got under way. They also wrote of having to put pressure on Ninov to do the experiment, as mentioned earlier.
He was also unenthusiastic about the experiment because of his scepticism about whether Robert Smolanczuk’s reaction would work. He told me that he asked Smolanczuk how far off base his ‘670 picobarn’ estimate might be, and Smolanczuk replied, ‘Three orders of magnitude’ (a thousandfold); this could put the actual figure around 1 picobarn. Again, Ninov’s statement is substantiated by others. ‘Everybody expected this [670 picobarn] estimate to be wrong,’ Loveland told me. ‘Robert himself thought it would be wrong.’
What about the three alpha decay chains that he reported finding in the data? Ninov denied that he had ever referred to them as decay chains. He called them ‘events’, and he said that they could equally well have been artefacts caused by ‘damped oscillations’. A damped oscillation would be a series of waves in the electronic equipment that rapidly diminished in amplitude. He suggested that each wave might cause a signal resembling an alpha emission, and the decreasing heights of the waves would cause the appearance of decreasing alpha energies, as expected for a decay chain. Even if they were actual decay chains, he said, there was no way to tell whether they arose from the decay of element 118 nuclei, because it was impossible to answer the key question – how many protons they possessed.
As far as I could tell from the Vogt report, Ninov never brought up his damped-oscillation explanation during the investigations. I asked him how such phenomena could give rise to precisely the half-lives and energies that were predicted by Smolanczuk’s theory. His answer was that the electronic filters were set to only accept the predicted events out of a huge sea of candidate events, so it wasn’t surprising that a few would match the requirements. Ninov emphatically denied that he had ever said, ‘Does Robert talk to God or what?’ as claimed by Hoffman and Ghiorso.
Of course, I asked Ninov why he had gone ahead and published the claim for element 118 if, in his own mind, the data was so suspect. This is where he really surprised me. He told me that Gregorich and Loveland had jumped to the conclusion that the ‘events’ were alpha decay chains, and that they had submitted the paper to
Physics Review Letters
without his knowledge or consent. He himself wrote an entirely different paper, he said, but the other two men rejected it, citing his poor English. (In speaking with me, Ninov’s English was flawless – which is remarkable, considering that it is probably his fourth language.)
These claims don’t square very well with the quotation cited earlier from the Berkeley Lab’s magazine, in which Ninov reportedly talked in enthusiastic terms about the findings at around the time the paper appeared. Nor do they square with Ninov’s refusal to sign the retraction letter. If he had been as sceptical about the authenticity of the data as he now claims, he should have been the first to call for a retraction. I didn’t ask Ninov about these apparent contradictions. When I asked Gregorich to comment on Ninov’s version of events, he sent me three excerpts from Ninov’s draft of the paper, all of which interpreted the findings as detections of element 118. With regard to the investigations that led to his identification as the person who had inserted the evidence for decay chains into the computer files, Ninov denounced the committees as biased against him and politically motivated. His colleagues also betrayed him, he said, especially Gregorich. With regard to the Lynch-Macchiavelli committee, he expressed contempt. ‘You know already from their names what’s up,’ he said, implying that they were a lynch mob and were following the cynical recommendations of Macchiavelli’s 15th-century namesake. ‘The investigation was absolutely political,’ he said.