Us vs Them

All experiments have a personality. Especially these large collaborations -- sometimes the members have a collective sociology. I think there was even a study once by a sociologist!

The two Tevatron experiments -- DZERO and CDF -- are no different. CDF started first -- it had almost 3 or 4 years head start on DZERO. As a result of that DZERO always feels like it is playing catchup -- weather or not it actually is! It is just part of this experiments collective sociology.

I do my best not to get caught up in this -- I've been on both experiments in my career. But every-now-and-then something goes by that makes me cringe. I was scanning the papers on XXX today and I saw nice looking overview paper. Heavy Quark physics is quite exciting right now, but I don't work on it -- so this sort of thing is perfect for me. But I got caught in the forward by the following:

What is being refered to is the discovery of the top quark -- a huge achievement. And certainly worthy of mention in this article's forward. But there is one problem... it was discovered by both CDF and DZERO -- they submitted their papers minutes apart from each other. I scanned down to see who authored this paper. An fellow I've known for years -- and a member of CDF. :-)

Comments

[Tommaso]

Hi Gordon,

I agree with you, the fellow from CDF should have payed the due homage to D0 as well.

However, if you stop the story there, you misrepresent the picture. As you well know, CDF did observe in 1994 a small sample of ttbar events, and from it measured a mass of the top quark which is dead-on the most precise measurements to date (174 GeV was 1994's CDF measurement, 173.5+-3.5 is what is now the best world average). This was published in April 1994, one year before the official "discovery" papers by CDF and D0, back to back on PRL.
What happened was that in 1995, with additional statistics (which, however, included the same core events used in the 1994 paper) CDF decided to claim for observation of the top quark, and gave D0 a week before going public, as previously agreed. D0 did put together what they had, to avoid being left out of the discovery, but in the pressure of time they could not provide a precise mass measurement (they measured 199 GeV, a value now excluded by more precise measurements by both D0 and CDF), and their signal's significance was somehow arguable. No blame on that, of course their top candidates were just as good as CDF's. But I think I am not in my political correct mood today, and I believe that once the story is told fully, few would disagree that CDF did observe top quark production one year before D0 did.

[Dmitry]

I think everyone knows that CDF discovered top quark. D0 is there only due to this rule of one week warning.

[Bob]

Seems like the November Revolution of 1974 has never ended.

[Dmitry]

Once a speaker who weren't paying attention was actually heckled at a conference for a phrase "D0 and CDF discovered Top Quark" (the phrase puts the rightful authors in incorrect order even alphabetically).

I was at CERN in 1994 and I clearly remember the news that CDF discovered the top quark.

Some skeptics were saying though that the top would have been discovered event if it weren't there given the pitiful state of HEP after SSC shutdown.

[Dmitry]

yeah, for completeness, I refer to this paper:
Phys.Rev.D50:2966-3026,1994

[Gordon T Watts]

Yeah. As you can see, some people feel rather strongly about this. In fact, that paper you refer to was my thesis (I was on CDF at the time). That 1994 thing wasn't an officail discovery -- it was a 2.9 sigma observation (that is why the paper was called evidence, and not discovery).

And the 1 week rule has nothing to do with this paper -- DZERO claimed, when 1994 rolled around they didn't see anything and, in fact, argued that CDF may have made a mistake. The next year, 1995, both of them submitted the offiical discovery papers at the same time.

Science here is tricky, right? Since it wasn't a 5 sigma bump, do you call it discovery? Why didn't we, on CDF, decide to call it discovery? It is hard to have it one way back then and the other way now. But we can be very happy we were right with that first paper!

[Gordon T Watts]

I wonder how many comments this post will end up getting? ;-)

[Dmitry]

So D0 did not see anything in 1994 and got 5sigma signal next year. Did D0 get all their 5 sigma effect in 1995 run only? Or did they recover signal in old data?

CDF sees 3sigma and grows it into 5 sigma.

In case of CDF I see consistent road to discovery in case of D0 I see an inconsistency which can be explained by fluctuation (either down in earlier data or up in the later data) or by a problem in the 1994 analysis.

Week rule I mentioned refers to 1995 paper not 1994. The events of 1994 were (correct me if I am wrong) - D0 pubishes upper limit on top crossection and within a couple of months CDF publishes evidence (OK call it evidence) creating controversy.

[Dmitry]

yes, last but not the least.
I did not know that you made 1994 analysis. This makes you the author of top discovery (in my book). My sincere congratulations!
(you have just become my personal hero)

[ursula]

Haven't there been many 3 sigma effects seen which came out not to be a discovery in the end? Quark substructure for example, Leptoquarks, 4 jet events, the Higgs we don't know yet, etc? Isn't that the reason why we do call a 3 sigma effect an evidence and not a discovery?

[Dmitry]

Hold your horses Ursula, none of the effects you listed had 3 sigma significance.

My point - today you have evidence , tomorrow you confirm that your evidence is not a fluke by adding statistics.

Unconfirmed 3 sigma is not yet a discovery indeed.

But suppose you did nothing or you run out of statistics and you are left w/ 3 sigma effect. Suppose someone else has exceeding statistics and looks at the same channel next year and sees 5 sigma effect. Who discovered the effect? In my book the one who did 3 sigma thing first.

This is similar to CDF/D0 & Top quark with the exception that CDF did not stop and turned 3 sigmna into 5 sigma (in fact 2.9 sigma into 4.8 sigmna to be precise, so still no 5sigma cigar).

[ursula]

In the 3/5 sigma case you describe, I think what I would say is:
experiment x saw first evidence at the 3 sigma level, and dicovery at the 5 sigma level was announced by experiment y.
One without the other would be incomplete. Is there anything else to it?
Well in case of top maybe only, that electroweak precision measurements were making very good predictions.

PS: I have not at all been involved in top searches. Therefore: is your book available? I could certainly learn something from it.

[Kevin]

Dmitri,

I think the point is that the discovery of a new particle is a statement of statistical signficance. You have to decide on a level of signficance which is enough for discovery - which the comunity has decided is 5 sigma. I assume we will agree on this.

Although it is most likley true that the events CDF saw in 1994 were in fact a top signal, it was not statistically signficant enough to claim discovery - it could have gone away like many other 2-3 sigma effects that you are well aware of.

Apparently, there was one anomalous events at either UA1 or UA2 which had 4 jets, a charged lepton, and met, which had a reconstructed mass at around 170 GeV (I don't think this was ever published but was shown in a conference as possible new physics). Could this have been the top quark? Should we give the discovery to them? According to your argument- perhaps. It was not a statistically signficant signal but it was confirmed by experiments afterwards. Since the probability of producing such a high mass particle at the Sps was so small, this was likely nothing (mismeasurement, mutliple interactions...)

The point I am trying to make is that you have to choose a level of statistical signficance to claim discovery - otherwise you open yourself up to endless speculation of one or more event anomolies. We have and it is 5 simga and CDF did not have it in 1994.

[Gordon T Watts]

The DZERO analysis. The 1994 version was before my time on DZERO, so I'm not sure I remember all the details. I'm sure there wasn't a problem with it, however (nothing to that effect has come up since I've been a member of DZERO). Rather keep in mind the way we work here is we usually remain in "limit setting mode" until we are sure we see something. I don't remember what DZERO had in 1994, but it wasn't as large a significance as CDF saw, so DZERO set a limit. The improved analysis had, I believe, a much larger dataset and also was reoptimized for the higher mass searches (becuase the lower masses had been ruled out). As I mentioned before, the DZERO cross section was stable, and the CDF mass was stable. ;-)

I think Kevin's comment is correct -- there is a definition of discovery. I don't think he or anyone else would argue that some of the events in CDF's 1994 were top quarks. Just not enough of them to be sure. And I'm also sure that some of the events in the 1994 DZERO analysis sample were also ttbar.

[Herbert Greenlee]

I'm a member of D0, and I participated in the top quark discovery during the 1994/1995 discovery period.

I'd like to make a couple of points about the 1994 papers. D0's 1994 top quark paper was the last one in which D0 set an upper limit on the top quark cross section. Something that no one remembers (no one has mentioned it in their comments) is that in this paper D0 also observed an excess of events: 9 events observed with an expected background of 3.8+-0.9. That's a pretty significant excess (fluctuation probability 2.7%), although not as significant as CDF's excess in 1994. So, it isn't the case that D0 had nothing in 1994, as some comments here imply.

The second point I'd like to make is that the CDF cross section measuement in the 1994 evidence paper was 13.9 pb +6.3pb -4.8pb, far above the predicted value of about 5 pb. So, CDF was the beneficiary of some good luck (nothing wrong with that).