kosower@harvard.UUCP (David A. Kosower) (01/17/86)
There might possibly be a "fifth force", or sixth force, or
all sorts of other strange things we can't yet measure, but in my
opinion (and the opinion of others around here), none are revealed
by the article of Fishbach et al. which appeared in the January 6
issue of Physical Review Letters. As I will detail below, the
article is so hopelessly riddled with errors that the only sensible
thing at the moment is to pay no attention to it at all. Don't
be misled by the eagerness with which the popular media have latched
onto the story; NOTHING has been discovered yet.
The article refers to three different experiments: a set of
measurements by an Australian geophysicist of the gravitational
constant in an mine Down Under; the old Eotvos (spelling with
accents: E\"otv\"os) torsion-balance experiment; and kaon
experiments at Fermilab. The authors claim that (a) all three
experiments show unexplained anomalies (b) these anomalies can
be explained in terms of a new force, similar in nature to
electromagnetism, but much weaker (indeed, weaker than
gravity), and coupling to a different kind of charge, which
they dub ``hypercharge''. The first of these claims is dubious,
and the second is incorrect.
The first thing that struck me, upon reading the article, is
that the geophysical data they quote show an effect only at one
standard deviation; with two-standard deviation error bars, the
effect disappears. Normally, three or four standard deviations'
significance are necessary before anyone will take the data seriously.
Upon closer examination, the paper reveals itself to be internally
inconsistent; the magnitude of the effect supposedly indicated by geophysical
experiments is 20 times smaller than that indicated by their re-analysis of
the old Eotvos experiment. This discrepancy is NOT going to be explained by
local variations in the Earth's composition or any of the approximations they
have made in modelling the shape of the earth; one might believe a factor
of 2, but not 20. If the geophysical data give the right order of magnitude,
then the error bars in Eotvos's data must be larger than he indicated (which
there is reason to believe anyway), and NO effect at all can be extracted
from the his data.
Fishbach et al. also choose to disregard a later experiment by
J. Renner (performed in the early 30's; the last Eotvos experiment
was done in 1909). Renner did his statistical analysis incorrectly;
but once one corrects for this (as R. Dicke has done), there is no
reason to believe the Eotvos experiment more than the Renner one.
The reason they disregard the Renner experiment is obvious, however,
once one plots his data: the later experiment is consistent with
absence of any effect, and not consistent with the existence of
a "fifth" force as supposedly seen in Eotvos's data.
The really damning thing about their re-analysis of the Eotvos
experiment, however, is the fact that the data indicate an effect of the
OPPOSITE sign, i.e. attractive, if they indicate any effect at all!
As a plot in the article shows, the total acceleration of an object
towards the Earth INCREASES as the baryon number (which is the same as
``hypercharge'' for ordinary matter) per unit mass increases.
But a repulsive force would have the opposite effect, since the Earth
contains lots of baryons. So the factor of 20 aluded to earlier isn't
merely 20, it's -20!
None of us have bothered to look in detail at the supposed
"anomalous" effects in the kaon system; but I for one doubt the
effect they claim to be seeing is a genuine anomaly.
The idea of this particular kind of "fifth" force is an old one,
first proposed in 1964 by Bernstein, Cabibbo, and Lee to explain
a then-new observation of certain kinds of rare kaon decays (they
are actually due to the violation of time-reversal invariance).
It didn't last long; Weinberg quickly showed that as originally
proposed, the idea was completely inconsistent with particle-physics
data. It was then deservedly discarded, until ``resurrected'' on
extremely shaky grounds by Fishbach et al.
All in all, it's unfortunate that the popular media (NY Times,
Time magazine) have lent so much exposure to proposals of such
dubious merit. To say that these proposals have "jolted physics",
as Bob Alpert (alpert@chovax.DEC) read in the Jan. 12 issue of
the Philadelphia Enquirer is just plain silly.
To those who are interested in more details on Eotvos-type experiments,
I recommend two articles by R. Dicke. One, in the December 1964 issue
of Scientific American, is a gentle, non-technical introduction, while
the other, in an issue of Annals of Physics from the same year, is
more technical, and goes into great depths about the sources of noise
and error one has to worry about in experiments of this sort. The
experiments are quite difficult, which is undoubtedly one reason they
have not been repeated more frequently in the last 50 years.
I should note that although many of the above observations were
made by others here (Andrew Cohen, Sidney Coleman, Sheldon Glashow,
Aneesh Manohar, Greg Moore), any inaccuracies are due solely to me.
David A. Kosower
kosower@harvard.Harvard.Educlt@newton.ARPA (Carrick Talmadge) (02/04/86)
David Kosower has raised a number of objections to the the paper involving
the reanalysis of the Eotvos experiment, which I feel compelled to respond
to. I am going to apologize ahead of time for not going into too much
technical detail, mainly due to time constraints on my part (I simply don't
have the time to "work up" a longer reponse).
I should comment that I was one of the coauthors of the article, so that
my remarks may be biased.
His substantial criticisms were:
(1) Geophysical data is only a 2 sigma effect.
In the original article we quoted a value for the coupling strength (relative
to gravity) due to the geophysical data of alpha = -0.0076 +- 0.36. Kosower
is correct if the errors were really symmetric. In reality, experimental
errors are rarely symmetric, and here we have no exception. The bounds
on the coupling strength at the 3 sigma levels are -0.0071 <= alpha <= -0.0128.
The tolerance level for alpha = 0 is approximately at the 100 sigma level
(that is, the geo results are consistent with alpha=0 only at the 100 sigma
level).
(2) Magnitudes inconsistent (factor of 20 between Eotvos and geo results),
and sign of interaction is wrong.
If you examine our paper, you'll notice that we assumed a uniform spherical
Earth to model the size of the effect in the Eotvos experiment. This, as
it turns out, is a highly dubious model in light of the short range (200 meters)
nature of the proposed interaction. The results of a more detailed analysis
indicates that it probably isn't possible to extract either the
magnitude or the sign of the effect, without detailed knowledge of the
local mass distributions where Eotvos performed his experiment (this may have
unfortunately been lost to history). What we do know about the local
mass distributions, however, gives the proper sign and magnitude (to within
a factor of 2).
Secondly, we quoted in the paper a limit on the range of the interaction
to be lambda = 200 +- 50 meters. It turns out that the quoted error was
seriously underestimated (Stacey admitted this to us latter), but we
had no choice but to quote that number he wanted us to quote. In reality,
the 3 sigma limit on the range is 50 meters <= lambda <= 6000 meters.
Since the Eotvos result can only measure the quantity alpha*lambda,
you can see that even a factor of 20 can be accommodated by the geophysical
results.
Actually there is a profound consequence of the relative insensitivity
of Stacey's experiment to the range -- there is no direct evidence that
the interaction must really be a Yukawa-type coupling...
(3) Renner's work was disregarded
As Kosower mentioned, Dicke showed in 1964 that Renner's experiment
contained an internal conflict -- that is his quoted errors were
1/3 the theoretical error due to the diffraction limit of his telescope.
If one were to inflate Renner's error by this factor of 3, then the
probability of Renner having correctly performed the experiment and
achieved his quoted outcome was less than 1 chance in 10 million.
As Dicke stated in this same article, he was unable to find any similar
problem with Eotvos' work.
The real question here is why would one be more inclined to believe
Eotvos than Renner. It turns out that Eotvos was doing this sort
of thing for a living for almost 20 years before actually performing the
Eotvos experiment. The apparatus he used was actually a device designed
to measure local gravity anomalies, from which one could deduce the
presense of heavy mineral concentrations. In 1935, Renner actually
used Eotvos' original apparatus (which had been sitting on the shelf
for 20 years) and claimed to achieve sensitivities with the same apparatus
4 times better than Eotvos. This is just a bit hard to swallow, as are
Renner's results -- even before inflating his error bars ala Dicke,
his seven quoted results were all less than 1 standard deviation from zero,
again an improbable outcome.
Dicke also examined Eotvos' work in 1964 and was unable to find
any problem whatsoever with this work (he also was one of the referee's
of our work, by the way, so that if he were aware of such criticisms
of the Eotvos experiment, he would almost certainly have pointed it out
to us!).
(4) Weinberg "showed" that a hypercharge interaction leads to contradictions.
In Weinberg's famous paper on a possible hypercharge interaction, it was
demonstrated that, if a hypercharge interaction were of sufficiently
long range, it would lead to catastrophic decay rates in certain
types of elementary particles, such as the K-meson (hence this is
called "Weinberg's catastrophe").
In the same article, Weinberg also noted that, if the range of the interaction
were less than 10**6 meters or so, that this catastrophe would be avoided.
Since the proposed range of the interaction is on the order of 200 meters,
there is no such difficulty for this theory.
*****
In summary, as far as I am aware, no substantial criticisms of this
paper have yet been forthcoming. The good thing about this work is
that it is easily testable -- we expect that the first result will
be announced within six months.
For people interested in looking into this question on their own,
the reference to our paper is Physical Review Letters, Vol. 56, 1986, page 3.
This paper references most of the other works I've talked about here.
Regards,
Carrick Talmadge