CDF Memories, Circa 1992
    By Tommaso Dorigo | May 11th 2013 10:46 AM | 20 comments | Print | E-mail | Track Comments
    About Tommaso

    I am an experimental particle physicist working with the CMS experiment at CERN. In my spare time I play chess, abuse the piano, and aim my dobson...

    View Tommaso's Profile
    In 1992 the top quark had not been discovered yet, and it did not make much sense for the CDF collaboration to have a full meeting devoted solely to it; rather, analyses targeting the search of the top quark were presented at a meeting which dealt with both bottom and top quarks. This was called back then "Heavy Flavour meeting".

    The Heavy Flavour meeting was held every second Thursday afternoon, during busy “on-weeks”, which alternated with “off-weeks” when no physics meetings were scheduled. I was back then a summer student, and together with my buddy Esov I had been instructed by our supervisor Luca Stanco to attend the works of the group and start building an understanding of how top physics data analyses were performed, and to get familiar with the tools and the experimental techniques which we would be using.

    To me, attending a meeting of the Heavy Flavour group was a real event. I remember very well my first time. I got to the Hirise with Esov in his Camaro 25th anniversary, we parked illegally, walked to the front door of the Hirise, went to the elevator on the right of the entrance, and got in with a half dozen people who told jokes to each other. I stood little chance of understanding what they were saying: my English back then was horrible. Similar occasions were a supplement of embarassment and frustration to the other usual sources for a newbye in a particle physics laboratory.

    Esov and I got off at the 12th floor where, on the north corner of the building, stood a square conference room equipped with a very large oval table in the middle, and chairs everywhere else. The wall opposite to the entrance was lined with blackboards and a large square screen; in front of the latter, a plastic transparency projector stood on top of the table. On the left of the screen two large television sets were placed side to side, with a video camera on top: it was the videoconferencing system which allowed remote sites to follow the meeting. The left wall of the room was all glass, with a stunning view of the woods and hills of the Fox River valley.

    As we got in we realized we had arrived a good ten minutes earlier than the start time: the room was still empty. Around the oval table stood nice soft-padded rotating chairs which looked a way better sitting option than the cheap ones lined along the walls. Only half-conscious of the assertive power of our action, we chose two with a nice view of the screen and sat down.

    A few minutes later a tall, lean guy in his mid-thirties, better dressed than the average physicist, arrived and sat down on the opposite side of the table, next to the remote controls of the videoconferencing screens and the slide projector. Turning on the revolving chairs with curious eyes, leaning back and jolly, we certainly looked more like tourists than physics students. The guy stared at us with a perplexed look behind his round glasses. He opened a logbook, wrote something in it; then he addressed us with the best smile he could put together.

    “Good afternoon... I do not think I have seen you at this meeting before. May I ask you to introduce yourself ? Are you students ?”

    My English was trembling and broken, but I managed to put together an answer of sorts.

    “Er... I am Tommaso Dorigo. I am undergraduate student in the University of Padova. I work with Luca Stanco.”

    “And my name is Esov Velazquez. I am also an undergraduate, from the University of Puerto Rico. I am here as a Summer Student, working under the advisorship of Dr. Stanco.”

    “I see. Luca Stanco...Yes, I know him. Nice guy. What are you working on, then, top physics I guess?

    Esov’s English was on another league, and I was happy to let him answer first:
    “Yes, we are putting together a search for all-hadronic decays of the top quark pairs, in the multijet dataset.”

    Liss nodded, and he went back to scribbling in a logbook. That would have been the end of it for a less naive person than me, but I feld obliged to continue the conversation:

    “...And who are you?”

    Tony’s perplexed look at seeing two undergraduates sitting in the places that full professors would usually occupy turned now slightly hopeless, but he kept his cool. His tone was however slightly stiffer as he explained:

    “My name is Tony Liss. I am professor of Physics at the University of Urbana-Campaign, and I chair this meeting with Claudio Campagnari.”

    So that was the CDF debut of my utter incapability of keeping silent when advisable; the first in a long row of verbal hemorrhages during my long career in the experiment. On that occasion, I decided that was enough self-embarrassment for the afternoon, and I made my best effort to shut up for the rest of the day. So I acted nonchalantly as dozens of other physicists started to swarm in, taking chairs along the walls and leaving the seats next to ours free for the most authoritative members of the collaboration. The feeling of their eyes pointing at my back slowly faded away, as all the chairs were eventually occupied, and the meeting started.
    I was a total newcomer, but luck was on my side that time. Physicists are by and large not really attached to formal issues; on that occasion, two summer students sitting in the front row did not raise more than a few ounces of eyebrows. At the meetings of the experiment, but also during the review process of analyses, the drafting of publications, the exchange of public e-mail messages, and all the other situations when CDF members had to interact and bounce opinions off each other, there was a total disregard of academic titles. It was not abnormal for the undergraduate to sit next to the full professor; it was just as normal for the former to interrupt the latter during a debate, showing no reverence if a mistake of the full professor needed to be exposed. This was a real point of strength for the experiment. One which, I believe, was totally American in spirit. There just existed two categories of persons: collaborators, and the rest of the World.

    The physicist directing the Heavy Flavour meeting with Tony in 1992 was a tall, handsome young man with dark hair, a big nose, and a clean-shaven face. He spoke good English but he could not hide a certain Italian accent. Everybody called him Claudio, and I stared at him in admiration: at less than 35 years of age being at the centre of the attention, in the heart of one of the most important particle physics experiments in the world, looked to me like the crowning of a dazzling career, the one of a surely brilliant mind. My opinion on the deductions that could be made on the scientific merits of a physicist by observing his or her chairing an analysis meeting would change in the forthcoming years, but my esteem of Claudio Campagnari, who appeared in full command of the scientific output of CDF on heavy flavour physics, would stay high for the years to come. I would later discover he was the co-author of the algorithm which identified b-quark jets using soft leptons: "heavy machinery unsuitable to be used by monkeys", as described by the other author of the code, Avi Yagil.

    After an introduction by the chairmen, there were short update reports from the subgroup leaders. In rapid succession Esov and I could hear in several "sub-group reports" the status of searches for top quarks in the their various characteristic signatures, as well as other analyses of high interest and relevance for the top hunt, which was back then in full swing. (In June 1992 the CDF detector had just started to collect data after a shutdown during which it had been endowed with a new silicon microvertex detector, which would be instrumental in finding evidence of top quark pairs.) After those summaries, a graduate student went to the screen to present the slides of his work. It was the first real talk of the meeting, where he was to show the results of a new kinematical fitter.

    A kinematical fitter is an algorithm which attempts to tailor the measured energy and direction of all particles observed in an event to a given hypothesis. If the hypothesis matches the event characteristics, or in other words if this "tailoring" succeeds, the fit yields an estimate of a free parameter of interest. The algorithm that the student was presenting used a sample of Monte Carlo events, which simulated the production of top-quark pairs followed by their decay into the single-lepton topology: two b-quark jets, two additional jets from a W boson, and a high-momentum lepton and a neutrino from the other W. The program was designed to determine the most probable mass of the top quark; this was of course known from the start in the used simulated sample, which allowed one to test and tune the fit.

    The first introductory slides explained the algorithm and received no comment from the audience, but once the speaker landed on the slide projector a transparency displaying a histogram of the top mass reconstructed by the algorithm, the guy sitting to my left started complaining. It was Avi Yagil, a tall, lean, neatly dressed guy with dark hair, olive skin, and quickly flying hands which accompanied his speech everywhere it went –very far, sometimes. Handsome, in his late thirties, he spoke with a strange tune, which I could describe as the one of Yogi the bear, if you are familiar with Hanna and Barbera’s cartoons.

    “Wait a minute, what’s that chickenshit scattered around at random masses ? Your algorithm is failing big time on those events!”

    With his funny expression Avi was pointing at several bins of the histogram displayed on the screen, each containing very few entries. Those bins were separated from the core of the distribution, where bins contained tens or even hundreds of entries each. Those non-empty outliers were indeed exposing some pathology of the reconstruction algorithm, and Avi was right in his assessment, although one could have argued that his question could have been formulated in a more polite way. The speaker hesitated, as the questioner continued his diatribe, using the well-known tactical trick of cutting off the person he was questioning, addressing someone else.

    “You see, Claudio? I am telling you, that is exactly what we should avoid: methods that lose efficiency by attempting a full-fledged reconstruction, and cannot be kept under control. Only a small part of those top events are clustering at the right mass! The inefficiency is huge! We need to avoid unstable fitters like this one...” And on, and on. He was like a flood.

    Claudio was a good friend of Avi –they had developed the soft-lepton tagging algorithm together-, and he knew he sometimes needed to restrain him.

    “Wait a minute, Avi, let him speak” he interrupted. The graduate student took courage:

    “Yeah, I guess the algorithm still requires some tuning... I will have a look at those events. But I’d like to stress the fact that here I have ran it over the whole sample containing a lepton and at least three jets, without applying the tight cuts we usually make to select the cleanest topologies...”

    Avi did not appear placated, and the discussion went on for a while. A similar fate awaited the other talks in the line, so that by the time the meeting ended, three hours later, Avi had questioned with similar verve and vis polemica all of the speakers in at least a couple of occasions each. After observing this performance, my first impression of the guy was thus that of a really smart person, who seemed to know everything way better than those who were doing the work; but also of some kind of a nagger, one of those sorry souls with a career who like to use their position and their experience to please themselves by shooting at sitting ducks, every time the chance arises –always, that is. Despite that, I could not hide to myself that I found Avi Yagil a quite fascinating person, in a peculiar way.


    John Duffield
    I just love "history of physics" books Tommaso, and the way they bring physics to life. I hope this piece is part of one you're writing.
    Thanks John! Indeed this is taken from a book I've written... But not published.
    A very sentimental scientific post. I really liked it.

    Tommaso, you said about Fermilab in 1992
    "... There just existed two categories of persons:
    collaborators, and the rest of the World. ...".

    Is that the underlying reason for the controversy
    between Fermilab and Dalitz, Goldstein, and Sliwa ?

    Was Sliwa considered a traitor for doing physics with "rest of the world" people ?


    Hello Tony,

    well, you know the story better than I do, but there are several players in it. Fermilab, as you say, as a lab could not allow that physicists not participating in an experiment could try to steal an important discovery, or not less bad, create confusion about a false signal (note that it is quite likely that the top candidate event from the 1988 run was indeed a genuine top pair, but that is beside the point). CDF, as a collection of physicists, did feel betrayed by Chris Sliwa. I do not know how clear was the violation of internal rules of the experiment, but for sure that was the sentiment circulating those days in the corridors of the CDF trailers.

    Also note that I am not justifying that sentiment or judging the matter here. I have personally been the subject of attacks by some CDF colleagues when I went so far as to post a blog article commenting a public CDF graph, speculating on whether there could be some SUSY Higgs decays in there (as a result of making a back-of-the-envelope calculation based on a tentative Higgs excess in some analogous DZERO dataset)... So I know how quick and reckless can be the judgement of some of the less balanced colleagues.

    Excellent post, I'd like to read more of the same kind.

    the first in a long row of verbal hemorrhages during my long career in the experiment.

    "heavy machinery unsuitable to be used by monkeys"

    verve and vis polemica

    Just three little gems that tell me you really should publish your book.  Quite apart from your wit, you explain the methodology so clearly that even I - who understand slightly more than nothing of particle physics - can get smarter without having my head spin faster than a fermion on steroids.
    nice post!

    To add a stroke to the picture, I'll take the liberty to paste here a comment I received privately from what will remain an anonymous source:

    If anything Kuni [the late Kuni Kondo, TD] was the first to propose what later sliwa/dalitz tried to do. He never got proper credit for it (partly because he was discovering a new particle every week). They all claimed to have discovered the top at 130 GeV (and some of them at other mass values as well).

    Here is a bit more context for the added "stroke to the picture".
    Quotes are from the book "The Evidence for the Top Quark" by Kent W. Staley (Cambridge 2004).

    Kondo, at Tsukuba Unhiversity, came from the Nagoya group that included Kobayashi and Maskawa who did the 3-generation mixing matrix.

    ".. in light of the 1978 "Fukuda initiative" .. Physicists from Tsukuba University .. joined CDF in 1979
    Three main "kinematic" .. top .. searches .. began as attempts to understand the 1988-9 data ..
    One group .. was led by physicists from the University of Pisa .. Hans Grassmann and Marina Cobal
    another was .. Kumi Kondo, from Tsukuba University
    the third was.. Krzysztof Sliwa, of Tufts University .. Gary Goldstein .. at Tufts .. and Richaed Dalitz ..[of].. Oxford
    The Pisa event structure analysis .. relied heavily on the use of Monte Carlo computer models .. at the time many CDF physicists were skeptical of the reliability of Monte Carlo
    Kumi Kondo's Dynamical Likelihood Method .. would give a kinematical reconstruction of events and then calculate the likelihood of that reconstruction using the dynamics of the hypothesized decay process .. Kondo found that .. the lone dilepton candidate found during the 1988-9 run .. could be reconstructed with his method as the decay of a top-antitop pair, with a top mass of around 130 GeV/c2, but "it has a very broad error"..
    Kondo .. was concerned about siilarities between his .. method and the method developed by Krys Sliwa with Gary Goldstein and Richard Dalitz .. Kondo .. wanted to publish .. a .. paper in PRL to secure credit ..[but].. This paper met resistance from the collaboration, who felt it too closely resembled a top discovery claim .. as with the .. Pisa .. event structure analysisKondo's method elicited skepticism because of its heavy reliance on Monte Carlo calculations
    The Dalitz-Goldstein Sliwa method .. encountered .. difficulties .. as a joint venture of phsycists from inside and outside the collaboration .. they were trying to apply their method to the first CDF dilepton event, the same published e-mu event from the 1988-9 run that Kondo had analyzed .. One distinctive aspect of the Dalitz-Goldstein-Sliwa method .. was [that it].. did not treat all possible solutions equally .. Instead, it .. assign[ed].. each reconstruction a "weight" based on how well it satisfied certain constraints - a measure of how "top-like' that reconstruction was ..
    In February 1992 .. Goldstein and Sliwa wer invited to present their method .. at a meeting of the heavy flavors group (the precursor to the top group) .. Sliwas showed .. a bump .. at a top-quark mass of about 120 GeV/c2 .. After that .. Goldstein was told not to come to Top Group meetings ..".

    What folowed was a very unpleasant confrontation precipitated by a New Scientist article, the details of which I think may have been discussed earlier on this blog.

    The bottom line as of now is that the 1988-9 event may indeed be a Tquark event
    the 130 GeV/c2 Kondo and 120 GeV/c2 Dalitz-Goldstein-Sliwa mass values are not accepted by anyone nowadays
    (except me, who thinks that the Higgs-Tquark system based on Higgs as T-Tbar condensate has three mass states (for Tquark around 130, 174, 220 GeV/c2 and Higgs around 126, 200, 250 GeV/c2).

    With respect to the 200 and 250 GeV/c2 Higgs states, the LHC results for the Golden ZZ 4l channel are now inconclusive so answers must await results of a new run (maybe around 2016 or so).

    With respect to the 120 and 220 GeV/c2 Tquark states, I think that proper analysis of event data would support them, but nobody else does, so I am a lonely outcast on that point.


    PS - It is interesting that Tsukuba remains at the cutting edge of physics. For example in arXiv 1206.0516 they investigate Terahertz radiation based on BSCCO high-temperature superconductors that have intrinsic Josephson Junction structure.

    Maybe the history of Fermilab 1992 teaches a useful lesson:

    The differences between Fermilab 1992 and HHC 20 years later seem to be;

    1992 - Monte Carlo type stuff was suspect
    Now - Statistical stuff is the dominant way to go
    (and, for Higgs digamma where you never see a single event as clearly Higgs,
    just a bump over background, statistical stuff is the only way to go)

    1992 - Individual candidate events were closely studied and considered of primary importance
    Now - Individual events studies are discouraged as not being in line with sophisticated statistical analysis

    Constructive Proposal for Now:

    The number of candidate events in the Higgs to ZZ to 4l channel
    is small enough to do 1992-style detailed analysis of individual events
    during the long shutdown of the LHC
    it would be nice if ATLAS and CMS analysts did 1992-style detailed analysis
    of individual events of the Higgs to ZZ to 4l channel.


    Hi Tony,

    interesting spin. I would say that regardless of the gamma-gamma signal, in most cases a new signal will always be constituted by a scarce number of events. This quite generally applies to "clean" resonances, for instance (Z' signals, e.g.). It is true that we also go after bumps in jet-jet masses and such, where the implied strong-production processes make these large cross-section kind of things.

    Individual events are still studied, but rarely so. This was not too different 20 years ago. We had the top candidate of the 1988 run, and we later had the infamous "ee γγ met" event, but mostly we extracted our results by statistical analyses even back then. To go to single-event studies one has to go, IMO, 40 years back. That was the time of the omega discovery (one event did it), or neutral currents (3 events of neutrino-electron scattering did it).

    While I can understand why statistical analysis is used (the volumes of data produced are truly staggering), to really understand what your analysis is doing you really need to look at the actual data.

    When looking at huge volumes of data with routines, usually the routines only look at stuff the designer expects to see, in other words if you don't know something is there, you frequently don't look for it.
    Never is a long time.
    That's a simplistic thing to say to any physicist with decades of experience. Physics made p-value and the look elsewhere effect famous precisely because they know what pitfalls to avoid in a way that other sciences do not. There is no sole designer in a physics experiment that is skewing the results.

    In my previous career, we made fun of any result that used Monte Carlo, because we did full-wave analysis, but when searching for signals a statistical analysis is the way to go.
    That's a fair comment, but isn't the Monte Carlo simulation based on what is expected?
    Even if there's separate MC results for different versions of String theory for example or whatever is fashionable, I would think it would have to be something someone thought might happen? Does p-value and look elsewhere really detect the unknown?

    As I'm definitely not a physicist, and I am not versed in Statistics if you say they will, I'll accept that as a valid answer.
    Never is a long time.
    I'll let Tommaso (or Tony) address the sociological flaws of humans in physics but, generally, a lot of cranky people who want to trip each other up don't let significance inflation happen.  If this were a small group looking for an effect it would be a worry - in these huge datasets not only will they 'detect the unknown' they are certain to get it.  :)

    Statistical analysis helps weed out the real signals from the 'if a psychic makes enough predictions one will be accurate' kind.
    significance inflation
    That's a new one for me.  I thought I grasped it's meaning, but checked* anyway.  It's good to know that my linguistics skills aren't entirely wasted.

    * Wikidpedia, natch. ;-)
    Wikipedia is right on something? Even a blind squirrel finds a nut once in a while. 
    Wikipedia is more often right than wrong, but tends to repeat "historical facts" without checking.  A bit like most of us, I guess.  Yes, me too, but rarely, I hope.
    Wrong !
    Lead–acid batteries, invented in 1859 by French physicist Gaston Planté,
    Wrong !
    Gaston Planté (1834–1889) was the French physicist who invented the lead-acid battery in 1859.
    Right !  German Wikipedia.
    1854 entwickelte der deutsche Mediziner und Physiker Wilhelm Josef Sinsteden den ersten Bleiakkumulator. Sinsteden stellte zwei große Bleiplatten, die sich gegenseitig nicht berührten, in ein Gefäß gefüllt mit verdünnter Schwefelsäure.  Durch Anschließen einer Spannungsquelle und häufiges Ent- sowie Aufladen (Formieren) erreichte er nach einer gewissen Zeit eine messbare Kapazität. An einer der Platten bildete sich Bleidioxid (Blei(IV)-oxid), und an der anderen reines Blei. 

    In 1854 the first lead-acid battery was invented by the German physician and physicist William Joseph Sinsteden. Sinsteden put two large lead plates that did not touch each other into a vessel filled with dilute sulfuric acid.  By connecting a voltage source and by frequent charging (forming) and discharge, he achieved a measurable capacity after a certain time. On one of the plates lead dioxide (lead (IV) oxide) was formed, and on the other pure lead.

    This is in my backlog of articles.  I still have to OCR and transcribe the source paper.


    My favorite examples of 1990s-style individual event analyses are:

    From D0:

    From CDF:

    They are examples of what I would like to see for Higgs to ZZ to 4l events.