jump to navigation

First observation of single top production from CDF!!! March 5, 2009

Posted by dorigo in news, physics, science.
Tags: , , ,

The paper, submitted to PRL yesterday evening, is here.
I will discuss the details later today…

UPDATE: a reader points out that the above link was broken. Now fixed.


1. Ptrslv72 - March 5, 2009

Hi Tommaso, as of 9:52 on Thursday morning the link is password-protected. Cheers Ptrslv72

2. Muzio Gambit - March 5, 2009

It’s on the arxiv: http://arxiv.org/abs/0903.0850

3. dorigo - March 5, 2009

No Muzio, that is the D0 paper, who also was submitted to PRL yesterday.
The two collaborations have decided to publish together the observation, since they both have a similarly significant result.

4. Marco - March 5, 2009

The CDF paper is apparently password-protected on the arXiv too (http://arxiv.org/abs/0903.0885)😦

5. dorigo - March 5, 2009

See update above.

6. Daniel de França MTd2 - March 5, 2009

This name single top is funny. It sounds like that the Top is not confined…

7. Andrea Giammanco - March 5, 2009

Too bad that the CDF authors insist in calling the sigma(observed)/sigma(SM) ratio a direct measurement of |Vtb|^2…

8. dorigo - March 5, 2009

Andrea, please explain… Why is it not true ?

9. Andrea Giammanco - March 5, 2009

Sure, I can explain.
CDF assumes sigma(observed)/sigma(SM)=|Vtb|^2, with the idea that the production vertex is Wtq (with q = any quark) in the numerator and Wtb in the denominator.
But if you look at the Feynman diagrams carefully, you may notice that there are two Wtq (or Wtb) vertices. Unfortunately, CDF declares in page 2 that BR(t->b) is assumed =1 through the entire paper. (By the way, this is true for D0 too.)

This is just the most evident inconsistency, but if it were only for this, it would be easy to correct it a posteriori.
Another one, almost as evident, is that both analyses (although this is explicitely declared only in the D0 paper, as far as I remember) assume sigma(t channel)/sigma(s channel) equal to the SM value.
The problem is that if you relax the unitarity constraint on the 3×3 CKM matrix, you should relax it for the entire third row and not only for Vtb: so |Vtd| and |Vts| can be larger than what you get from the indirect measurements which assume unitarity constraints.
How much larger? Not so much (large values would affect several precision measurements from B factories), but remember that the PDF for d and s quarks (especially d quarks) would then give huge enhancement factors to the t-channel production (and not to s channel, whose initial state is different).

But if the list of problems amounted only to these two, still we could extract |Vtb| from the numbers quoted in the CDF and D0 papers.
This was done, using Tevatron data available in 2006 (when only D0 had yielded a 3-sigma evidence for single top), in one of the sections of http://www.arxiv.org/abs/hep-ph/0607115 (I apologize because this qualifies as spam, since I am a coauthor of this paper).
A few months ago I started the task of updating the results:
I contacted the CDF authors in order to get feedback, and while they agreed that their original statement about “direct Vtb measurement” (now repeated, alas) was wrong, they also pointed out that my approach, although it improves the situation, is still unsatisfactory: there is no way that one can get Vti from the final single-top selection numbers (even after separating them by Njets and Ntags, as I did), because too much information has been lost in the process: the NNs and likelihoods and BDT and whatever multivariate analysis you use, you always need to “train” it to recognize your signal, and unfortunately they only used the SM single top as signal. I.e., |Vtb|=1 was deeply embedded in the analysis in each step.
Just to give you one example: a powerful discriminating variable, thus one of the inputs to almost any multivariate analysis, turns out to be the invariant mass formed with the lepton, the missing energy (with W-mass constraint) and the b jet. But if |Vtb| is significantly less than 1, let’s say 0.9 (which is the central value quoted by CDF), this means that 10% of the s-channel events would not have t(->Wb)b final state, but t(->Wq)b, so it would be classified in the Ntags=1,Njets=2 block (supposedly populated preferentially by t-channel) and, what is worse, the invariant mass would be formed by taking a b jet which is NOT coming from the decay of the top quark.
Another problem, which would affect other variables, is related to what I already mentioned about initial states with d and s quarks contributing significantly as soon as |Vtb| is a bit less than 1. This has a subtle effect on the kinematics, coming from the fact that d quarks (which can be valence quarks, in the proton) would contribute a lot more than b quarks in the high-x region, therefore giving less central events.

Bottom line: the CDF and D0 analyses, in order to achieve the incredibly challenging task of separating this elusive signal from the huge background, were obliged to make some reasonable assumptions on the signal. The most reasonable assumption was that its mechanism is SM, with all that this implies: BR(t->b)=1, only b contributing in the initial state for t channel, etc.
This strategy worked, as we may celebrate now given the publication of these two 5-sigmas evidence, but there was a price to pay.
The price to pay is that if there are non-SM contributions to single-top production, these contributions are naturally treated as background, and mostly lost.

I want to be clear on one point: I’m not saying that looking at the sigma(observed)/sigma(SM) is pointless, on the contrary I say that it is the first thing to do: you need a hypothesis testing, to check whether your assumption (i.e., “single-top production is completely described by SM”) holds at the end of the day. In the scenario where |Vtb|<1, i.e., |Vtd|<<|Vtb| and |Vts|<<|Vtb|.
So, it’s interesting to constrain new physics models where the CKM matrix has additional rows but only the t quark can mix with the additional quarks (as in the main sections of the paper that I spammed before, sections where my contribution was exactly 0): in that case we could have an effective Vtb whose modulus is significantly less than 1, and Vtd and Vts have the SM values.

All this was debated ad nauseam during TOP2008, in the presence of CDF and D0 representatives and of several phenomenologists; this discussion was summarized by Scott Willenbrock in his final summary talk:
slides 67-73.

10. Andrea Giammanco - March 5, 2009

Please read this sentence:
“In the scenario where |Vtb|<1, i.e., |Vtd|<<|Vtb| and |Vts|<<|Vtb|.
So, it’s interesting […]”
“But there are scenarios where |Vtb|<1, i.e., |Vtd|<<|Vtb| and |Vts|<<|Vtb|. For example, it’s interesting […]”

11. dorigo - March 5, 2009

Hi Andrea,

that clarifies it excellently, thank you! And of course yours does not qualify as spam (although my filter agreed with you and I had to fish your comment out of the spam box!). So try harder if you want to really spam… Today you taught us a very interesting lesson.


12. Tom Junk - March 19, 2009

Hi Andrea,

Just some comments on comment 9, from one of the guys doing the CDF work (and I’m the guy you contacted before!). I mostly agree with what’s said in comment 9, but would like to clear up a few details.

As you say, it is very important to establish the presence of single top quark production at the five-sigma level. It’s been 14 years since ttbar was observed, and we’ve been working quite hard to separate the signal from the background. The signal has lots of features we can use to help us (and the background too), and we used everything we could. This means that our selections and discriminants are highly focused on what we’re looking for, and we have not yet explored many interpretations beyond the SM. But we have gotten to a few of them!

Both CDF and D0 assume production in the SM s- and t-channel ratio and say so in their PRL submissions. Both CDF and D0 have relaxed this assumption and measured s- and t-channel cross sections separately. D0 has published these in Phys.Rev.D78:012005,2008. Rainer Wallny showed CDF’s latest s and t separate cross section fits with 3.2 fb-1 in his talk at Fermilab:
http://theory.fnal.gov/jetp/talks/Wallny.pdf. These 2D fits still assume Br(t->Wb)=1 and the signal models also include only tWb couplings in the production.

CDF writes that the assumed Br(t->Wb)=1, and D0 writes that |Vtb| >> |Vtd|,|Vts|, which are meant to express the same assumption. This is not to say that |Vtb| is close to 1 however, just that Vtd and Vts are much smaller. A value of 0.9 for Vtb would not necessarily, as you say, make Br(t->Wb)=0.9. 3×3 Unitarity-constrained values of Vtd and Vts are about 1% and 4% and they get squared in the b.r.’s.

Of course 3×3 unitarity may not be true, which is one reason why we put so much effort into measuring the single top production cross section. A common way to express a result like this under less restrictive assumptions is that we are measuring the cross section times the branching ratio (t->Wb) and let readers and theorists make of it what they will. But life isn’t quite as simple as this even, since a large value of Vts, for instance, enhances the t-channel production for reasons Andrea spells out quite nicely (and are further documented in Tait and Yuan: Phys.Rev.D63:014018,2001).
The problem here is also what Andrea says after consulting with us — we did not search explicitly for ts and tqs production, and thus cannot say what our sensitivity is to those processes. s-channel ts production might end up getting selected as looking like SM t-channel production, since we don’t b-tag the s very much, and the 2-jet 2-tag category is what looks most like s-channel signal to us (which would be depleted if the production has fewer b’s and more s’s). On the other hand, t-channel tqs and tqb are largely indistinguishable to us since the extra b usually goes outside of our b-tagging acceptance (which covers well only to |eta|<1). So an exotic model with large Vts and suppressed Vtb would show up as more t-channel signal and less s-channel signal than expected, but we haven’t done the work to say exactly how much. Naively, I’d say that the s-channel ts production would look more like background than t-channel signal due to the fact that the s won’t have high Q*eta which is one of our golden variables for t-channel signal. The t-channel tqs should look almost exactly like the SM tqb signal, however. If we really wanted to do a thorough job of it, we’d run some signal Monte Carlo for these cases and try these models one by one.

But first things first. We very much wanted to establish the fact that we can observe this process. Our hypothesis test was between two less subtle hypotheses — is single top present or absent? and not the harder ones to try to distinguish one kind of single top signal from another — that will come later as we get able to do such things.

D0 by the way has a nice PRL on testing right-handed and other exotic Wtb couplings: Phys.Rev.Lett.101:221801,2008, another useful avenue studying single top beyond the SM.

13. Andrea Giammanco - March 23, 2009

Hi Tom,
luckily I checked again this thread, and I was able to read you clarification.
Thanks a lot for the further details, and especially for the link to Wallny’s presentation.
In the meantime, I re-read your paper and I have to retreat my original statement: indeed your phrasing was very carefully chosen, and I was fooled by the use of the word “direct” in “direct measurement of Vtb”🙂

14. Wolfgang Wagner - March 25, 2009

Dear Tommaso, Andrea, Tom and colleagues,

in the context of the discussion above I would like to point you to the recent phenomenological analysis by Alexander Lenz et al in arXiv 0902.4883.

Alexander asks the question “How much space is left for a 4th familiy of quarks? using constraints from the mixing of neutral mesons (K, D, Bd and Bs) as well as b->s gamma and assuming 4×4 unitarity.

He finds that values of Vtb as low as 0.8 are still possible, while V_ts and especially V_td are indeed quite small, maximum at a few percent level.
See Fig. 11, 12 and 13. According to this analysis the assumption |Vtb| >> |Vtd|, |Vts |, made by the Tevatron analyses, appears to be well justified.
Enhancements of V_ts and V_td above their 3×3 SM are possible, but even a factor of 3 doesn’t make them comparable to V_tb


15. Andrea Giammanco - March 25, 2009

Hi Wolfgang, thanks a lot!

Sorry comments are closed for this entry

%d bloggers like this: