W Bosons Weighted To One Part In 5359
    By Tommaso Dorigo | April 19th 2012 09:12 AM | 5 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
    A couple of weeks ago the CDF and DZERO experiments have produced a combination of their measurements of the W boson mass. Besides two older determinations of this fundamental parameter of the Standard Model, the new 2.2/fb measurement by CDF and the 4.3/fb measurement by DZERO have been averaged together, accounting for correlated systematics. [x/fb is a shorthand for the amount of collisions from which the W boson datasets have been extracted by the experiments: 1/fb is about 80 trillion proton-antiproton collisions.]

    The W boson is one fundamental ingredient in our recipe of elementary interactions. It is the particle sought by alchemists who tried to turn lead into gold: the only particle we know that has the power to change the flavour of quarks or leptons. The W was discovered together with its neutral partner, the Z, by the UA1 experiment in 1983 at the CERN SppS collider, and since then it has been one of the main characters in all high-energy studies: nowadays W boson decays are used to identify top quarks, to search for Higgs bosons, and to determine detailed properties of the electroweak interaction.

    The result of the averaging exercise of the Tevatron experiments will stand as one of the major legacies of the Tevatron: thanks to it the electroweak boson is now known with a relative uncertainty of less of a five-thousandth; such a precision is relatively rare even for macroscopic quantities (how many of us know our body weight to a dozen grams ?), and for subatomic ones it is really exceptional.

    The combination can be seen in the picture below, which lists the four Tevatron ingredients on the top, then their combination, and then the result of combining the four determinations of the W boson mass produced ten years ago by the LEP II experiments ALEPH, OPAL, DELPHI and L3. You should remember that LEP II collided electrons and positrons at twice the W boson mass and above, and was in fact dubbed a "W factory" since a large fraction of the collisions yielded pairs of W bosons. The two Tevatron experiments together have halved the W mass uncertainty of LEP II, bringing it from 35 to 15 MeV.

    I have already written in some detail about the W mass measurement of CDF. As for the DZERO result, more succintly described here, it is based on W->eν decays only (CDF uses also W->μν decays), so the actual number of W bosons used is in the same ballpark in the two datasets. DZERO calibrates the electron energy scale using the Z->ee decay process as a "standard candle", and its measurement is thus effectively that of the ratio between W and Z boson masses. This allows the reduction of many sources of systematic uncertainties; the catch is that the accuracy is then limited by the statistics of the Z->ee sample. At a hadron collider the ratio of production of W and Z bosons is roughly three to one; in addition, the Z boson decays only 3.3% to electron-positron pairs, as opposed to the 10% fraction of W decays to electron-neutrino pairs.

    It would be nice to see these measurements updated using the full data samples collected by CDF and DZERO in the course of Run II - each experiments has collected over 10/fb of data - but this is most likely not going to happen. As you add data, your statistical uncertainty decreases with the square root of the data size, but systematic uncertainties often do not behave as graciously. Furthermore, data collection during Run II proceeded in varying conditions: instantaneous luminosity kept increasing, the detector aged, triggers were modified. Accounting for these variations brings in additional systematic uncertainties, such that there is little to gain by insisting that the measurement use as much data as possible. Further, the LHC experiments are already producing their own results on the W boson mass, and these will eventually surpass the Tevatron precision anyway.

    A summary of the combination of CDF and DZERO W boson mass determinations is given at this link.


    One implication of the result for theorists is that the more accurately you know the fundamental constants, like the W mass, the more credibility combinations of constants that show relationships between them have - a one part of five thousand accuracy constant is harder to make appear with mere numerology than a one part in ten accuracy constant.

    Also, since we know that the Z boson mass is related in the Standard Model to the W boson mass via the Weinberg angle (with the mass of the W divided by the mass of the Z equal to the cosine of the Weinberg angle), any improvement in the accuracy of the W boson mass also tightens the boundaries on the value of the Z boson mass and Weinberg angle arising from pre-existing experimental data.

    For example, the more accurately we know the W and Z boson masses, the more stringently we can test the hypothesis that the Higgs boson mass is related to the W and Z boson masses by the relationship 2H=Z+2W, which could simply be crude numerology. If we know both the W and Z boson masses quite accurately on can put very narrow error bars around the value predicted by that relationship and then know quite soon as the Higgs boson mass (which is already predicted to a fairly narrow mass range) is know with just a bit more specificity - basically the instant that the Higgs boson mass error bars get small enough. And, if there was such a simple mass relationship, this would suggest that there is some really profound but subtle missing piece to electroweak unification out there waiting to be explained.

    Of course, it could be that this relationship is ruled out and that some other one is supported. But, the narrower the error bars, the more fruitful it should be to juggle numbers looking for deeper relationships.

    Mr. T- Why was the D0 run 1 so far off the mark?

    What do you mean ? It is one sigma away. You get this kind of behaviour 32% of the time.
    Tommaso, that's true of individual events in a normal distribution, but in a sampling distribution of the mean, variance decreases with sample size. Are you saying these experiments sample intrinsic variance? Then why the average to get "the mass"?

    I must say it doesn't help that our statistical theory was developed by Bernoulli and Laplace for classical objects and errors of observation, when nothing like quantum theory was envisaged.

    Here, from control theory/electronic engineering, are treatments of:

    Quantum non-locality:

    Non-linear oscillator phase and amplitude estimation:

    Not easy to find, in a competitive arena!

    Sorry, I do not understand what you are referring to. Does it refer to my comment that the early D0 measurement is completely compatible with the average ? That average is driven by the last two results, so effectively the error bar of the first D0 measurement can be compared to that without correlations.