A closer look at the elementary fermions
Maurice Goldhaber
Physics Department, Brookhaven National Laboratory, Upton, NY 11973
Contributed by Maurice Goldhaber, October 31, 2001
Proc. Nat. Acad. Sci., USA, 99, 33 (2002) (With Minor Corrections)
BNL - 68960

Abstract
Although there have been many experimental and theoretical efforts to measure and interpret
small deviations from the standard model of particle physics, the gap that the model leaves in
understanding why there are only three generations of elementary fermions, with hierarchical
masses, has not received the attention it deserves. I present here an attempt to fill this gap.
Although our findings are mostly only qualitative, they nevertheless may be of heuristic value.
Rules concerning the elementary fermions, some previously known and some new, lead to a
number of conclusions and questions that seem worth pursuing. Some clarify the standard model,
and others suggest possible modifications, the implications of which are discussed.

What Can We Learn from the Known Facts About Elementary Fermions?
In many developing fields of science, rules deduced from empirical data often can be considered
as qualitative proto-theories, having some predictive power as well as pointing the way to a final
theory. A well known example is what Mendeleev called his law of the periodic system of the
chemical elements, the "elementary particles" of the 19th century. The law, although then far
ahead of a plausible theoretical interpretation, nevertheless allowed predictions of new elements.
About six decades later it proved helpful in formulating quantum mechanics that in turn explained
the periodic system quantitatively, including some of its exceptions. On the way to the final
explanation, the periodic system's hierarchical atomic weights were replaced by atomic numbers,
equal to the number of protons in the atomic nucleus, with no known theoretical limit.
We now face a situation, similar in some respects, for the elementary fermions. A century
of research established a "periodic system" of elementary spin 1/2 fermions, confined, however, to
only three generations. Each generation consists of two kinds of leptons (charged leptons of charge
-1 and their associated neutrinos of charge 0) and two kinds of quarks (charges +2/3 and -1/3),
with hierarchical masses increasing from one generation to the next.
From the partial width for decay of the neutral gauge boson Z0 of the electro-weak theory
into neutrinos, measured ~12 years ago at SLC (SLAC) and LEP (CERN), it was concluded that
there are only three kinds of weakly interacting neutrinos, all of low mass (then believed to be
zero). If further weakly interacting neutrinos should exist, their masses would have to approach or
exceed 1/2 mZo. Thus, there cannot be more than three generations of elementary fermions with
light neutrinos. Although the possibility was considered that leptons and quarks might be built of
"more fundamental" particles, no indications of complexity (e.g., excited states of the elementary
fermions) have been found.
A closer look at the intrinsic properties of the elementary fermions, some measured
directly and some deduced with the help of the standard model (SM) of particle physics (1-3),
reveals interesting rules about them.



2

Some Salient Facts Known for the Three Generations of Elementary Fermions
The SM starts out by postulating chiral symmetries for the interactions of the elementary fermions.
Because such symmetries would lead to elementary fermions of zero mass, their finite masses are
attributed to symmetry breaking by the so-called Higgs mechanism, which can accommodate
masses but not predict them.
Table 1 shows the periodic system of the three generations of elementary fermions with
their symbols and masses arranged, as is common, in the order of their discovery. The generations
are numbered by a generation number i, increasing from 1 to 3 from the lightest to the heaviest
generation.


Table I
The Three Generations of Elementary Fermions

t (1.743  0.05) x 105) (1.777 x 103)
i = 3 ( b (4  4.3) x 103) )
(<2.8 x 10-6)


c (1.15  1.35) x 103)  (105.67)
i = 2 ( s (75  170) )
 (<2.8 x 106)


u (1  5) e (0.51)
i = 1 ( )
d (3  9) e (<2.8 x 10-6)

The values of the masses (in MeV) are taken from ref. 4 except for neutrinos, for which
only upper limits are known. For e the limit is based on the work reported recently by the Mainz
and Troitsk collaborations (5). Because of the small upper limits for the mass differences within
the neutrino triplet of mass eigenstates, deduced from neutrino-oscillation experiments (see
below), the mass limits for  and can be taken as approximately the same as that for e.
For the lightest quarks (u, d, s and c) so-called "current masses" are quoted; they are not
measurable directly but are derived from the SM. The masses of complex particles containing
these quarks such as, for example, nucleons, are equal to the equivalent of the quarks' potential and
kinetic energies. The masses of the heaviest quarks are essentially measured directly.
For elementary fermions in corresponding positions in each generation we use the
following symbols: within the ith generation, we refer to the quarks of charge +2/3 and -1/3, as ui
and di, respectively, and to the charged leptons and their associated neutrinos as ei and i,
respectively. Our knowledge of the elementary fermions of the first generation stems from
experiments in atomic and nuclear physics, whereas that of the second and third generation stems
from high-energy experiments, first with cosmic rays, which copiously produce particles of
relatively low effective thresholds, and later with high-energy accelerators. Because the
chronological order of discovery of the three generations is correlated with the increase of
effective collision energies with time, it coincides with the hierarchical order of the masses. The
elementary fermions of the first generation, u, d, and e, are widely believed to be the ultimate
building blocks of which the visible universe is built. They were either produced directly or as final
decay products of the quarks and charged leptons of the heavier generations.
Table 2 shows the four known elementary interactions (forces) exhibited by the different
types of elementary fermions.



3

Table II
Elementary Interactions of the Elementary Fermions

Relative Leptons Quarks
Interactions Strength vi ei ui , di

Strong 1 X

Electro-magnetic 10-2 X X

Weak 10-5 X X X

Gravitational 10-39 X X X


The approximate relative strengths of the interactions, hierarchically arranged, are found
to be independent of the generation number (universality). Because the interactions vary with
energy they are characterized here at a scale of ~ 1 GeV.
By using some experimentally determined properties of the three generations of
elementary fermions as input, the SM allowed some important conclusions (see refs. 1-3). For the
known interactions the number of quarks and leptons as well as the "flavor" (generation number)
of charged leptons are very nearly conserved. The near absence of flavor-changing neutral currents
can be explained by the so-called GIM mechanism (named for Glashow, Iliopoulos, and Maiani,
ref. 6) as caused by approximate cancellations.
For the Cabbibo-Kobayashi-Maskawa (C-K-M) matrix (7, 8), for which one might have
assumed mixing of either u
i or di mass eigenstates through their weak interactions, one has chosen
to stay with di mixing following Cabbibo's original suggestion when only the three quarks u, d,
and s were known. The leptonic decays of quarks can then be represented by a unitary matrix:


d Vud Vus Vub d
s V
) = ( ) ( )
cd Vcs Vcb s
( b Vtd Vts Vtb b


The recent review by Gilman, Kleinknecht and Renk (9), gives the following measured ranges for
the matrix elements:


0.9742 to 0.9757 0.219 to 0.226 0.002 to 0.005
0.210 to 0.225 0.9734 to 0.9749 0.037 to 0.043
( )
0.004 to 0.014 0.035 to 0.043 0.9990 to 0.9993


We tabulate in Table 3 for different i-j the mean values of the experimental ranges of
the matrix elements u . Some matrix elements are derived from more accurately measured
i dj

ones, assuming a unitary matrix. For each pair u and the ranges partly coincide.
i dj uj di



4


Table III
The Mean Values of the Absolute C-K-M Matrix Elements u
i dj
i-j 0 1 2
i 1 2 3 1 2 2 3 1 3
j 1 2 3 2 1 3 2 3 1
0.9750 0.9742 0.9992 0.223 0.223 0.040 0.039 0.004 0.009

From Tables 1-3 we can deduce the following rules.

Rule 1. Corresponding elementary fermions of different generations are associated with identical
elementary interactions (universality).
Rule 2. Within each generation, there is a correlation between the mass of an elementary fermion
and the relative strength of its dominant interaction (elaborated further later).
Rule 3. Besides its dominant interaction, each elementary fermion possesses all weaker ones.
Rule 4. As the generation number i increases, the mass differences m(ui) - m(di) and m(ei) - m(vi)
also increase.
Rule 5. The matrix elements of the C-K-M matrix decrease as |i-j| increases from 0 to 2, with
u .
i dj uj di

According to the SM, each elementary fermion may emit or absorb elementary gauge
bosons connected with its elementary interactions, either virtually or really, depending on energy.

What Can the Rules Teach Us?
Rule 2 suggests that self-interactions may be responsible for a major part or all of the mass of an
elementary fermion.
Extrapolations from rules may have predictive value. There are many examples in the
history of science where an extrapolation or generalization from existing knowledge was taken as
a prediction, often only qualitative but worth pursuing, and in many cases resulting in important
progress. This was especially so in the early steps leading to the SM. If one wants to be more
cautious, however, one might consider extrapolations as questions worth pursuing until they are
either confirmed experimentally or integrated into theory, not necessarily in that order.
Rules 1-3 invite the intriguing question: Are there additional members of each generation,
elementary spin 1/2 fermions, subject solely to the gravitational interaction, and thus expected to
have extremely small masses? Can such gravity fermions be integrated into the theory of general
relativity without running into inconsistencies?
Rules 1 and 2 imply that within each generation the order of the masses of the elementary
fermions is correlated with the hierarchy of their dominant self-interactions.
Rule 4 indicates that the ui and di (except for the special case of the first generation
discussed later) have large mass differences, contrary to expectations from quantum
chromodynamics (QCD). Should we therefore treat them, in analogy with the ei and i, as
elementary fermions with different interactions, implying that a new dominant hyperstrong
elementary interaction may have to be added to the QCD interaction for the ui?
Rule 2 is consistent with the existence of finite neutrino masses as deduced from recent
neutrino oscillation experiments. This rule also allows us to sharpen and extend the conclusion
drawn from the Z0 experiments; because the dominant self-interaction of neutrinos is the weak



5

one, we can expect only low-mass neutrinos. If a new neutral elementary particle of large mass and
spin 1/2 should be discovered, e.g., as a component of cold dark matter, it would have to be
considered sui generis.
Rule 5 suggests that the amount of weak mixing of quarks depends on the "closeness" of
the generations involved. In his different parameterization of the C-K-M matrix, Wolfenstein
already emphasized the hierarchical reduction in the extent of mixing (see refs. 9 and 10).
From a particular version of the superstring theory, Candelas et al. (11) predicted the
existence of four generations of elementary fermions. However, Erler and Langacker (see ref. 4)
concluded from a review of several precision measurements of quantities sensitive to virtual
effects of elementary fermions of a hypothetical fourth generation that there is no evidence for it
even if its members were too heavy to have been detected directly at presently available energies.
Many theoretical approaches attempt to understand some particular empirical input to the
SM or to reproduce some of its results in a new way. These approaches usually are based on
special assumptions going beyond the SM and not on a comprehensive, generally accepted theory.
For two interesting examples, see refs. 12 and 13.

Attempts to Understand the Empirical Regularities
Several questions remain: What do the rules imply, what causes the hierarchy of the masses of
corresponding members of different generations, and why are there only three generations of
elementary fermions?
The SM does not have satisfactory answers to these questions; however, the empirical data
suggest modifications of the SM that may qualitatively explain the observations. Although the SM
assumes no a priori difference between the different generations, we conclude from rules 4 and 5
that the elementary fermions "know" to which generation they belong and that the generation
number is not just a label, as is usually assumed, but stands for a new physical property equal for
all members of a particular generation but changing systematically from one generation to the
next. It is apparently a property that enhances self-interaction, more so the larger i becomes. What
can such a physical property be? While retaining the universality of the interactions, for which
there is good empirical evidence, a possible tentative interpretation is to forgo the assumption of
equal point sources for the three generations and replace them with source shapes of finite size,
identical for each type of elementary interaction within a generation but decreasing systematically
in volume as i increases; as the source volume becomes more "singular," the self-interactions will
increase. The masses of the elementary fermions then have to be considered as secondary
quantities that take on hierarchical values.
When we find nature repeating itself with variations, it is worth asking: Why should the
repetition stop at three? I conjectured some time ago that the volume of the source shapes might
decrease "naturally" if the existence of just three generations were connected with the three-
dimensionality of space (14). The source shapes of the three generations might resemble, for
example, a sphere, a disk, and a rod (with no zero thickness and no sharp edges); as i increases
from 1 to 3, their dimensionality decreases from 3 to 1, and their shapes become more singular.
Although the value of the masses of the elementary fermions is correlated with the strength
of their dominant self-interaction, nondominant self-interactions also must play a role, as they do
when they are the dominant self-interactions in lighter elementary fermions of the same
generation. The effect of a nondominant self-interaction on the mass may be positive or negative
depending on whether it has the same or opposite sign from that of the dominant interaction. This
effect may be especially important for the strong and electromagnetic self-interactions, which are
near in relative strength and presumably have opposite signs (the dominant interaction is attractive



6

and the nondominant one is repulsive, independent of the sign of the electric charge), leading to a
net reduction in the self-interaction. Compared with QCD expectations alone, ui and di both will be
depressed but ui more than di. In the first generation, this might be the cause of the lower u than d
mass, which would explain the long-standing puzzle of why the neutron (containing one u and two
d) is heavier than the proton (containing two u and one d). But then one may ask: why is ui not
lower than di for all i (see later discussion)?
With the discovery of oscillations of atmospheric neutrinos , near-maximal mixing
of the neutrino mass eigenstates m2 and m3 was established (for the latest results, see refs. 15 and
16). Of several potential oscillation solutions still under consideration for solar neutrinos, a large
mixing angle solution is preferred (see ref. 17), compatible with measurements of the solar 8B
neutrino spectrum by SuperKamiokande (18, 19) and the Sudbury Neutrino Observatory (20).
From the oscillation experiments which yield a value for |m 2 2
3 -m2 | of (2 - 5) x 10-3 (eV)2
(16, 17) and for |m 2 2
2 -m1 | < 7 x 10-4 (eV)2 (21), assuming maximal mixing, a spread of <0.1 eV
for the mass eigenstates of the neutrino triplet was deduced (22). The near degeneracy of the mass
eigenstates makes large mixing plausible. The often-conjectured hierarchical order of the neutrino
masses might be expected from the source shapes assigned to the three generations.

Epilogue
We have seen that the known properties of the elementary fermions lead to rules through which
nature speaks to us, suggesting clarifications and modifications of some assumptions of the SM as
well as hinting at possible new avenues worth exploring.
Some comments may be useful. We learned that elementary interactions are connected
with appropriate elementary fermions from each generation and raised the question of whether this
should be extended to the gravitational interaction.
Although the assignments of corresponding elementary fermions to different generations in
hierarchical order of their masses, including leptons and quarks that have no other known
associations, can be ascribed to the happenstance of the chronological order of their discovery, we
have made it plausible that this is to be expected for the assumed source shapes.
Although many successes of the SM were not discussed here and usually will not be
affected by our considerations, small deviations from its predictions may result from some of the
suggested modifications.
For finite source shapes, the dependence of the C-K-M matrix elements on the generation
distance (Rule 5) can be expected to depend on the amount of overlap between the source shapes
of the initial and final quarks. This allows an alternative interpretation to di mixing; if the
"intrinsic" charge current weak interaction was assumed to be independent of generation change,
the matrix elements, as long as the off-diagonal elements are small, would be a measure of the
overlap of the source shapes, and the unitarity limit would be expected to be approached.
The large increases in the mass difference between ui and di with i (rule 4) indicate that the
postulated hyperstrong interaction of the ui would have more impact on the self-interaction as the
shapes become more singular. This would be so if it is mainly a short-range high-energy
interaction, whereas the effective QCD self-interaction is a comparatively low-energy interaction at
short distance. The assumed source shapes may cause the ui masses to be lifted well above the di
masses for i > 1. For the spherical source shape assumed for i = 1, the u mass would be less
affected by a high-energy interaction. The relative closeness of the di to the ei despite the large
difference in their dominant interaction strengths may be caused in part by differences in the



7

energy dependence of their interactions and in part by the depression of the d iby their Coulomb
interaction.
If a hyperstrong interaction exists, one would expect that the cross sections for producing
t + t at energies that are well above their threshold would deviate from QCD predictions.
If a gauge boson is connected with the hyperstrong interaction, it might be detectable by its
characteristic mass, decay modes, and lifetime.
Bardeen et al. (23 and references therein) discuss dynamical symmetry breaking of the SM
by a tt condensate where a new interaction (less general than the hyper-strong interaction proposed
here), called topcolor (24 and references therein), is ascribed to the t. The Higgs boson is then
considered to be a tt condensate, with a mass <500 GeV, estimated by Chivukula (25).
The existence of a hyperstrong interaction also would affect the estimate of the grand
unification mass and thus the predictions of Grand Unified Theories for the proton lifetime.
If elementary fermions with only gravitational interactions exist, the "gravity fermions"
might be a link between general relativity and quantum mechanics. They also would contribute to
hot dark matter (as would spin 2 gravitons that often are ignored in this connection). The
gravitational wave detectors now being built are only sensitive to coherent waves of gravitons and
would not detect individual gravity fermions that might only be detectable through interactions at
extremely high energies, of the order of the Planck mass (1019 GeV), but such high energies have
not been found among the cosmic rays.
Fermions of spin 3/2, with gravitational interactions only, are predicted by the theory of
supergravity (26).
Ongoing improved experimental and theoretical studies of the values of the C-K-M matrix
elements, corrected, in the case of bound quarks for the different wave functions of the initial and
final quarks, may make the a priori assumption of unitarity unnecessary.
Although our considerations seem to yield a coherent picture, with important parts
independent of the assumption of finite source shapes, clarifying some aspects of the SM and
raising the possibility of modifying and extending it, a crucial question remains: Can our
qualitative approach be changed into a quantitative one without running into contradictions? Can
one find source shapes (compatible in effective size with the present experimental limits of ~10-17
cm) that, unlike point sources, would yield finite masses and lead, for the well established self-
interactions, to the masses found for the elementary fermions (including the mass difference d-u)
e.g., by inversion and iteration, and would such source shapes also explain the C-K-M matrix
elements?

Acknowledgements
I thank M. J. Creutz, M. V. Diwan, A. S. Goldhaber, T. Goldman, R. L. Jaffe, W. J. Marciano, D.
J. Millener, and R. E. Shrock for valuable discussions.

Note
My attention was drawn to three attempts to understand various aspects of the three generations of
elementary fermions. They differ from each other and from the views presented here (27-29).



8


References
1. Collins, B. D. B. , Martin, A. D. & Squires, E. J. (1989) Particle Physics and Cosmology
(Wiley, New York).
2. Perkins, D. H. (2000) Introduction to High Energy Physics (Cambridge Univ. Press,
Cambridge, U.K.)
3. Hooft, G. `t. (1997) In Search of the Ultimate Building Blocks (Cambridge Univ. Press,
Cambridge, U.K.).
4. Groom, D. E. , et al. (2000) Eur. Phys. J. 15, 23-26.
5. Bonn, J. , et al. (2000) Nucl. Phys. B 85, 273-286.
6. Glashow, S. L. , Iliopolulos, J. & Maiani, L. (1970) Phys. Rev. D 2, 1285-1292.
7. Cabbibo, N. (1963) Phys. Rev. Lett. 10, 531-533.
8. Kobayashi, M. & Maskawa, K. (1973) Prog. Theor. Phys. 49, 652-657.
9. Gilman, F. J. , Kleinknecht, K. & Renk, B. (2000) Eur. Phys. J. 5, 110.
10. Chau, L.-L. (1983) Phys. Rep. 95, 1-93.
11. Candelas, P. , Horowitz, G. T. , Strominger, A. & Witten, E. (1985) Nucl. Phys. 258, 46-
74.
12. Gell-Mann, M. , Ramond, P. & Slansky, R. (1979) in Supergravity, ed. Van
Nieuwenhuizen, P. (North Holland, Amsterdam), pp. 315-321.
13. Hall, L. & Murayama, H. (1995) Phys. Rev. Lett. 75, 3985-3988.
14. Goldhaber, M. (1994) in Gift of Prophecy: Essays in Celebration of the Life of Robert
Eugene Marshak, ed. Sudershan, E. C. G. (World Scientific, Singapore), p. 204
15. Sobel, H. , et al. (2001) Nucl. Phys. B 91, 127-133.
16. Kajita, T. & Totsuka, Y. (2001) Rev. Mod. Phys. 73, 83-118.
17. Bahcall, J. N. , Krastev, P. I. & Smirnov, A. Y. (2001) Nucl. Phys. B 91, 29-35.
18. Suzuki, Y. , et al. (2000) Nucl. Phys. B 91, 27-35.
19. Fukuda, S. , et al. (2001) Phys. Rev. Lett. 86, 5656.
20. Ahmad, Q. R. , et al. (2001) Phys. Rev. Lett. 87, 071301-071308.
21. Apollonio, M. , et al. (1999) Phys. Lett. B 466, 415-430.
22. Goldhaber, M. (2000) in Next Generation Nuclear Decay and Neutrino Detector, eds.
Diwan, M. V. & Jung, C. K. (Am. Inst. Phys., Melville, NY), Vol. 533, pp. 23-25.
23. Bardeen, W. A. , Hill, C. T. & Lindner, M. (1990) Phys. Rev. D 41, 1647.
24. Hill, C. T. (1991) Phys. Lett. B 266, 419.
25. Chivukula, R. S., 35th Rencontres de Moriond: Electroweak Interactions and Unified
Theories, March 2000, Les Arcs, France, pp. 11-18.
26. Ferrara, S. , Freedman, D. Z. , van Nieuwenhuizen, P. , Breitenlohner, P. , Gliozzi, F. &
Scherk, J. (1977) Phys. Rev. D 15, 1013-1018.
27. Wilkinson, D. H. (1991) Can. J. Phys. 69, 1459.
28. Dobrescu, B. A. & Poppitz, E. (2001) Phys. Rev. Lett. 87, 030801-030811.
29. Bjorken, J. D. (2001) Phys. Rev. D 64, 085008-085012.



