Quantum Chromodynamics on Lattice: Direct Minimization of QCD-QED-Action with New Results ()
1. Introduction
The Quantum Chromodynamics (QCD) is based on the SU(3)-color interaction, and is in general considered as a theory, which is both mathematically complicated and numerically hard-to-handle.
We present here a new calculation method (on-lattice minimization of action), which is numerically simpler than the other methods, and which uses a new ansatz for wavefunctions.
We also present calculation results for energy-mass of first-generation hadrons, which agree well with the observed values, and new results for internal component distribution, which give interesting insights into the symmetry and internal structure of these hadrons.
In Chapter 2 we compare the different calculation methods.
In Chapter 3 we describe the ansatz for the component quark and gluon wavefunctions.
In Chapter 4 the numerical algorithm is described.
In Chapter 5 the calculation results for energy-mass and component distribution are presented for the three families of first-generation hadrons: nucleons, pseudo-scalar mesons, and vector mesons.
2. Solutions Methods in Lattice-QCD
[1] [2] [3] [4]
Basically, there are four solution methods in lattice-QCD (LQCD):
Perturbative analytic Feynman solution
[5] [6] [7]
Here one calculates the reaction cross-sections from Feynman diagrams evaluating the corresponding Feynman-integrals in analogy to the QED. As the QCD is renormalizable, all Feynman integrals can be made finite.
However, this works only for convergent Feynman series, i.e. if the interaction constant gc < 1. This is the case for large energies E > EΛ = 220 MeV.
Non-perturbative on-lattice Wilson-loop method
[8] [9] [10]
Here the expectation value of an operator (e.g. energy = Hamilton operator) is calculated using path integrals
(1)
with interaction matrix M[U], under
the local gauge transformation, with coupling constant g, lattice step size a, gluon field
, action
, on closed loops on the lattice.
We get the Wilson action on equidistant lattice
with lattice constant a
the fermionic action becomes (2)
with lattice translations
.
And the gluon action is
(3)
with Wilson plaquette action
The Wilson action
is minimized in
on lattice
.
The solution
yields the corresponding masses
for fermions and energies
for gluons.
Non-perturbative on-lattice eom solution
[11]
The QCD equations-of-motion (eom) are derived from the minimal-action-principle as the Euler-Lagrange-equations corresponding to the QCD Lagrangian. They are
the Yang-Mills-equations for the gluon wavefunction
and the color-field-tensor
(4)
or taking into account self-interaction with the self-current
the Yang-Mills field equations become
(5)
and the Dirac equation
(6)
or
with the usual convention
with the color-covariant-derivative
and the quark-wavefunction
.
These are nq + ng partial differential equations (pdeq) first order in
, for the nq = 2 or nq = 3 quarks and ng = 8 gluons, adding a gauge condition and a boundary condition for
.
They must be solved numerically on a lattice as an eigenvalue problem of the Dirac equation, which is very difficult and time-consuming for a one-dimensional lattice of, say, n1 = 100 points (total number of points n = n14 = 108).
Non-perturbative on-lattice minimization of action (new)
The starting point is the minimum-action-principle for QCD:
(7)
with a gauge condition and a boundary condition for
.
It can be extended to QCD + QED
for the quarks
, QCD-gluons
, QED-photons
.
In order to carry out the minimization numerically, we introduce an equidistant 4-dimensional lattice
,
extract a small random sub-lattice Lsub.
And approximate the integral by a sum over Lsub:
(8)
where
is the elementary integration volume in spherical coordinates, and model the quark wavefunctions as parameterized Gauss functions
.
And the gluon-wavefunctions as Ritz-Galerkin series on a function system
with coefficients
:
,
accordingly the photon-wavefunctions
.
We impose the gauge condition for
:
and a boundary condition:
, the quark-wavefunctions are normalized
.
The minimization is carried out in dependence on
,
where par0 yields information about the energy (= mass), the sizes and the inner structure of the considered hadron.
3. The Ansatz for the Quark and Gluon Wavefunctions
Gluon wavefunction
For the gluon wavefunction we apply here the full Ritz-Galerkin series on the function system
with coefficients
,
where
is a Fermi-step-function which limits the region
of the hadron with “smearing width” dr0.
,
where
is the phase angle between the particle and the anti-particle part of the gluon, and with the Ritz-Galerkin-expansion
with energies
.
Because of color-symmetry, the active (non-zero) gluons are (9)
all gluons for nucleons.
6 non-diagonal gluons for vector-mesons.
3 quark-antiquark gluons for for pseudo-scalar mesons.
Quark wavefunction
The first-generation (u, d)-hadrons consist of three quarks (nucleons) or three color-symmetric quark-antiquark-combinations (vector-mesons) or two quark-antiquark-combinations (pseudo-scalar mesons) (10).
For nucleons
For vector-mesons
or
For pseudo-scalar mesons (pi+, pi0)
or
A Ritz-Galerkin series for quarks would blow up the complexity of calculation, therefore we use here a simpler model, based on the asymptotic-freedom property of quarks: gaussian “blobs”
,
where
is the energy,
and
is the position (r, θ) and its width,
is the quark-antiquark phase and the antiquark is
The ansatz and the color symmetry
The form of the quark color-wavefunction and the corresponding set of active gluons are enforced by the color-symmetry and the number of particles equal to the number of combinations.
The 8 gluons of the SU(5) form 3 families:
• The diagonal
, which map color indices into itself,
• The non-diagonal
, which exchange color-index with a different color index, and
• The non-diagonal
, which exchange color-index with a different anti-color index.
The nucleons consist of three quarks with color (r, g, b), and the color wavefunction q is mapped into itself under color-permutations, therefore the full set of 8 gluons Agi is required, and there are only two possibilities for first-generation hadrons: p = uud and n = ddu.
The vector mesons consist of quark-antiquark pairs, where the color wavefunction q has three identical components.
q is mapped into itself under the corresponding set of 6 non-diagonal gluons
(each flips two color indices).
It is seen immediately that the three combinations listed above are the only possible ones, which is confirmed by the existence of the three v-mesons omega0, rho0, rho+.
The pseudo-scalar mesons consist of quark-antiquark pairs, where the color wavefunction q has two non-zero components. The corresponding gluon set are the 3 non-diagonal color-anticolor gluons
, which exchange a color-quark with a different anti-color-quark. For example,
flips color-indices (3, 1) and transforms
into
. So in reality, the wavefunction is a superposition of the three
and is mapped by the gluon set into itself.
Again, one can see immediately that there are only two possible combinations, which correspond to the two known ps-mesons pi+ and pi0.
The corrected coupling constant
In the original Callan-Symanzik relation the QCD coupling constant has the (asymptotic) energy dependence
where
Λ ≈ 220 MeV critical QCD energy;
nf = 3 generations, N = 3 QCD charges.
For energies μ ≈ Λ it must be modified to avoid the singularity
(11)
for the numerical calculation we set
,
which is consistent with the Callan-Symanzik relation for
, as shown below (Figure 1).
4. The Numerical Algorithm
The energy, length, and time are made dimensionsless by using the units: E (
), r(fm), t(fm/c) fm = 10−15 m. The hadrons have axial symmetry, so we can set φ = 0 and use the spherical coordinates (t, r, θ).
We choose the equidistant lattice for the intervals
with 21 × 21 × 11 points and, for the minimization 8× in parallel, 8 random sublattices:
.
For the Ritz-Galerkin expansion we use the 12 functions
Figure 1. QCD coupling constant gc(μ), μ in Λ-units [12] .
The action
(12)
becomes a mean-value on the sublattice
(12a)
where
the
-volume and
is number of points of the sublattice, we set
.
We impose the gauge condition and the boundary condition for
via penalty-function (imposing exact conditions is possible, but slows down the minimization process enormously).
is minimized 8× in parallel with the Mathematica-minimization method “simulated annealing”, the execution time on a 2.7 GHz Xeon E5 is 9100s for the proton p = uud, the complexity
million terms.
The minimization is performed in the parameters
, for the proton is the number of parameters
,
,
.
The proper parameters of the quarks and the gluons are:
,
Criteria for correctness of the ansatz
1) Convergence of minimization
As we found out during the computation, a wrong ansatz, e.g. lacking color symmetry, leads to a non-convergent minimization. We chose a high goal precision of prec = 10−4, so there was a high probability that a convergent minimizations hits a real (global) minimum.
2) High relative deviation between solutions
Strongly differing solutions indicate a non-correct ansatz, as we found out e.g. for the nucleons with too many degrees-of-freedom for the gluons: the relative deviation for crucial variables, like energy, should be no more than 2% for the nucleons and 6% for the ps-mesons.
3) Vanishing parameter-derivatives
A true minimum must satisfy the derivative-condition
, where
is one of the minimization parameters, Normally, the parameter-derivatives are close to zero, otherwise the minimum is not genuine, or the ansatz is wrong.
4) Boundary condition and gauge condition
The boundary and gauge condition must have values close to zero, otherwise the weight for the penalty function is too low.
5) Minimum value
The minimum value should be −30, …, 30 for the considered parameter range. Very large positive values result in the case of too high penalty weights. Very large negative values may come out, if the Ritz-Galerkin parameters
are not bounded appropriately.
6) Correct energy scale and number of particles
The three types of first-generation hadrons have energy scales: E (nucleon) ≈ 0.98 GeV, E (v-meson) ≈ 0.78 GeV, E (ps-meson) ≈ 0.14 GeV, and these values emerge automatically with 8, 6 and 3 gluons respectively.
Furthermore, with the above ansatz, the number of possible particles is 2, 3, 2 respectively.
5. The Results for First-Generation Hadrons
[13] [14]
nucleons n, p quarks (3), gluons (8), spin = 1/2.
Masses (Table 1).
Energy for quark-number (n = 1, 2, 3), gluon-number (n = 4, …, 11), both sorted with increasing energy (Figure 2).
Distribution quarks (r[fm], θ) sorted with increasing energy (Figure 3).
Figure 2. Energy for quark-number (n = 1, 2, 3), gluon-number (n = 4, …, 11) in nucleons [12] .
Figure 3. Distribution of quarks (r, θ) in nucleons [12] .
The quark distribution differs largely between the nucleons: the proton is ring-symmetric (no θ-component), the neutron has two orbitals with an angle of α = π/2. The small mass difference is due to the electromagnetic contribution, which is about 1% of the total mass.
The mass of the nucleons, as is the case for all first-generation-hadrons is generated almost exclusively by the energy of the gluons and the quarks, the rest masses of u and d (mu = 2.3 MeV, md = 4.8 MeV) contribute very little to the total mass.
The gluon distribution is practically the same for both nucleons, which is to be expected, since the two particles are identical for the color interaction.
The radius of the nucleons can be assessed from the above diagram: r(p) ≈ 0.8 fm, r(n) ≈ 1 fm.
Proton p = uud.
m = 0.938 GeV, r0 = 0.84 fm.
Etot = 0.945 GeV, ΔEtot = 0.032, dEem = −0.013 (Table 2).
Mean calculation error ΔEui ΔEAi Δai ΔaAi Δdrui Δrui Δθui [12] .
Gluons Agi (Figure 4).
The proton p has one rotation plane (orbital), the two quarks (u, d) are close at r = 0.15 low energy E ≤ 0.03, the second u-quark further outside r = 0.4, and high energy E = 0.2. The “smearing” width is comparable, δr ≈ 0.3.
The electromagnetic correction is negative and much larger than with the neutron, dEem = −0.013 GeV, which is apparently the reason for the proton’s smaller mass.
Neutron n = ddu
m = 0.939 GeV, r0 = 0.84 fm.
Etot = 0.945 GeV, ΔEtot = 0.018, dEem = +0.0017 (Table 3).
Mean calculation error ΔEui ΔEAi Δai ΔaAi Δdrui Δrui Δθui [12] .
Gluons Agi (Figure 5)
The neutron n has two orbitals with an angle of α = π/4, the u-quark is at the center with low energy E = 0.05, the two d-quarks sit in the orbitals with higher energies E = 0.09, 0.013. The “smearing” width is comparable, δr ≈ 0.4 and higher than with the proton.
The gluon distribution is practically the same as for the proton, which is to be expected, since the two particles are identical for the color interaction.
The electromagnetic correction is positive and much smaller than with the proton, dEem = +0.0017 GeV, which is apparently the reason for the proton’s smaller mass.
Table 2. Energy of quarks in the proton, Eui and gluons EAi sorted by energy, quark amplitude ai and gluon amplitude aAi, quark position rui θui, quark radial smear-out drui.
Table 3. Energy of quarks in the neutron, Eui and gluons EAi sorted by energy, quark amplitude ai and gluon amplitude aAi, quark position rui θui, quark radial smear-out drui.
Magnetic moment of nucleons
The magnetic moment is
, for a rotating charge distribution:
,
where
is the momentum of charge, in analogy to the momentum of inertia
.
For a rotating solid sphere with radius r0 with constant charge density
.
The magnetic moment of the nucleons is measured in nuclear magnetons
, which is the magnetic moment of a rotating solid sphere with constant charge density
.
The actual momentum of charge is therefore:
We have to take into account the “smearing” Δri of radius ri
,
so it becomes
We get for the neutron.
Iqn = −0.1766e, IqNn = 0.106e, so
= −1.766, measured
= −1.91.
And for the proton
Iqp = + 0.2226e, IqNp = 0.0909e, so
= + 2.448, measured
= +2.793.
The calculation does not take into account the orbitals, and there is also the statistical uncertainty of the order 7%, so the results are satisfactory.
Pseudo-scalar mesons pi+, pi0 quarks (2), gluons (3), spin = 0-
[6]
Masses (Table 4)
Energy for quark-number (n = 1, 2), gluon-number (n = 3, 4, 5), both sorted with increasing energy (Figure 6)
Distribution quarks (r[fm], θ): independent (θ) = spherical (Figure 7)
Figure 6. Energy for quark-number (n = 1, 2, 3), gluon-number (n = 4, …, 11) in ps-mesons [12] .
Figure 7. Distribution of quarks (r, θ) in ps-mesons [12] .
The pseudo-scalar mesons are spherically-symmetric, there is no θ-dependence: θ ≈ 0 in the quark-distribution, the gluon-wavefunctions show little θ-dependence, and the gluon amplitudes are much smaller (factor 30) for pi0 than for pi+.
For the pi0,
and
sit at r = 0.4 E ≈ 0, and at r = 0.75 E ≈ 0.1.
For the pi+, the u and
have practically equal radii, but different energies: r = 0.6 E ≈ 0.001, and r = 0.6 E ≈ 0.01.
The measured masses of the ps-mesons (0.135, 0.139) are reproduced by the calculation (0.155 ± 0.025, 0.129 ± 0.026), but only roughly within the error bounds.
Ps-meson pi0 =
.
m = 0.135 GeV, r0 = 0.66 fm.
Etot = 0.155 GeV, ΔEtot = 0.025, dEem = +0.007 (Table 5).
Mean calculation error ΔEui ΔEAi Δai ΔaAi Δdrui Δrui Δθui [12] .
Gluons Agi (Figure 8).
Ps-meson pi+ =
.
m = 0.139 GeV, r0 = 0.66 fm.
Etot = 0.129 GeV, ΔEtot = 0.026, dEem = +0.0014 (Table 6).
Mean calculation error ΔEui ΔEAi Δai ΔaAi Δdrui Δrui Δθui [12] .
Gluons Agi (Figure 9).
Vector mesons rho0, rho+, omega0 quarks (2), gluons (6 non-diagonal), spin = 1.
[15]
Masses (Table 7).
Energy for quark-number (n = 1, 2), gluon-number (n = 3, …, 8), both sorted with increasing energy (Figure 10).
Distribution quarks (r[fm], θ) (Figure 11).
Figure 8. Selected gluon wave functions in the pi0-meson [12] .
Table 5. Energy of quarks in the pi0-meson, Eui and gluons EAi sorted by energy, quark amplitude ai and gluon amplitude aAi, quark position rui θui, quark radial smear-out drui.
Figure 9. Selected gluon wave functions in the pi+ -meson [12] .
Table 6. Energy of quarks in the pi+ -meson, Eui and gluons EAi sorted by energy, quark amplitude ai and gluon amplitude aAi, quark position rui θui, quark radial smear-out drui.
Figure 10. Energy for quark-number (n = 1, 2, 3), gluon-number (n = 4, …, 11) in vector mesons [12] .
Figure 11. Distribution of quarks (r, θ) in vector mesons [12] .
Table 7. Vector meson masses [12] .
The vector mesons are spin-1 bosons but only rho+ shows an explicit θ-dependence of quark-distribution: it is ellipsoidal. The gluons show explicit θ-dependence and are, as for the nucleons, practically equal for all three particles.
For rho0: the quarks
and
have identical parameters r = 0.5, δr = 0.3, E = 0.1.
For omega0: the quarks
and
again have identical parameters, are at center, δr = 0.25, E = 0.1.
For rho+: the heavier quark
has r = 0.5, δr = 0.05, E = 0.05, the light quark u has r = 0.9, δr = 0.5, E = 0.07, rho+ has two orthogonal orbitals. Its two quarks have completely different width; the
quark closer to the center has a small bandwidth, the light u quark is strongly “smeared” like all the other quarks in the 3 particles.
The measured masses of the v-mesons (0.775, 0.775, 0.782) are reproduced correctly by the calculation (0.771 ± 0.0052, 0.779 ± 0.012, 0.782 ± 0.007).
V-meson rho0 =
.
m = 0.775 GeV, r0 = 0.75 fm.
Etot = 0.771 GeV, ΔEtot = 0.0052, dEem = +0.002 (Table 8).
Mean calculation error ΔEui ΔEAi Δai ΔaAi Δdrui Δrui Δθui [12] .
Gluons Agi (Figure 12).
V-meson rho+ =
.
m = 0.775 GeV, r0 = 0.75 fm.
Etot = 0.779 GeV, ΔEtot = 0.012, dEem = +0.002 (Table 9).
Mean calculation error ΔEui ΔEAi Δai ΔaAi Δdrui Δrui Δθui [12] .
Gluons Agi (Figure 13).
V-meson omega0 =
.
m = 0.782 GeV, r0 = 0.75 fm.
Etot = 0.782 GeV, ΔEtot = 0.007, dEem = +0.002 (Table 10).
Mean calculation error ΔEui ΔEAi Δai ΔaAi Δdrui Δrui Δθui [12] .
Gluons Agi (Figure 14).
Table 8. Energy of quarks in the rho0-meson, Eui and gluons EAi sorted by energy, quark amplitude ai and gluon amplitude aAi, quark position rui θui, quark radial smear-out drui.
Table 9. Energy of quarks in the rho+ -meson, Eui and gluons EAi sorted by energy, quark amplitude ai and gluon amplitude aAi, quark position rui θui, quark radial smear-out drui.
Table 10. Energy of quarks in the omega0-meson, Eui and gluons EAi sorted by energy, quark amplitude ai and gluon amplitude aAi, quark position rui θui, quark radial smear-out drui.
6. Conclusions
We present here (Chapter 2) a new calculation method for on-lattice QCD, namely on-lattice minimization of action.
It works by direct minimization on parameters of action
,
where the action integral is approximated as a summation on a random sublattice of an equidistant lattice.
This has the following advantages:
• The minimal-action principle is a fundamental principle, from which the equation-of-motion (eom), i.e. the Dirac equation for QCD is derived. The parameters
of quarks
and gluons
in the Lagrangian
(e.g. energy-mass of a quark) can be calculated from it in principle exactly without solving the eom. In comparison, the Wilson loop method uses an approximation in order to make the path integral tractable numerically.
• The calculation is a simple summation, which is very fast numerically, as opposed to the analytical integral calculation of the perturbative analytic Feynman solution.
• The calculation is scalable, i.e. the precision can be increased arbitrarily, simply by making the step size of the lattice smaller, or the size of the sublattice larger.
• The calculation can be carried-out in parallel by Np processes on Np different sublattices with the same number of points (in this implementation we have Np = 8). The mean of the Np resulting values is then the calculation result for a parameter, whereas the standard-deviation assesses the calculation error.
• On-lattice minimization of action uses parameter minimization algorithms instead of solving partial differential equations (as in on-lattice eom solution) or instead of calculation of parameterized integrals (as in Wilson-loop method or in analytic Feynman method). Nowadays, there exists a large selection of powerful algorithms for parameter minimization which can be used for this purpose.
• On-lattice minimization of action, as opposed to the other solution methods, yields information about radial and axial distribution within hadrons.
The calculation ansatz in Chapter 3 for the quark wavefunctions:
For nucleons
For vector-mesons
or
For pseudo-scalar mesons (pi+, pi0)
or
And gluon wavefunctions
all 8 gluons for nucleons.
6 non-diagonal gluons for vector-mesons.
3 quark-antiquark gluons for pseudo-scalar mesons.
Explains effectively the mass scale for the three types of first-generation hadrons.
Nucleons
.
Vector-mesons
.
Pseudo-scalar mesons
.
Simply by using gluon configurations compatible with the SU(3) symmetry, i.e. corresponding to subgroups of the SU(3).
In Chapter 5 we present the results of calculations for first-generation hadrons.
The calculated masses agree well with the observed values, and the calculation shows that individual gluons on average contribute as much as individual quarks to the total energy-mass of hadrons.
On-lattice minimization of action, as opposed to the other solution methods, yields information about radial and axial distribution within hadrons, which gives interesting insights into their symmetry, “smearing-out” of components, and internal energy-mass distribution.
• Structure of nucleons
The proton is spherically symmetric, the neutron is axial with two orbitals.
The gluon distribution is practically the same for both.
The small mass difference is due to the electromagnetic contribution, which is about 1% of the total mass.
• Structure of pseudo-scalar mesons
The pseudo-scalar mesons are spherically-symmetric.
The gluon amplitudes are much smaller (factor 30) for pi0 than for pi+.
• Structure of vector mesons
The neutral v-mesons are spherically-symmetric.
For neutral rho0: the quarks
and
have identical parameters with distribution peak at r = 0.5.
For neutral omega0: the quarks
and
again have identical parameters, with distribution peak at the center.
The charged rho+ is axial, has two orthogonal orbitals, its two quarks have completely different width; the
quark closer to the center has a small bandwidth, the light u quark is strongly “smeared-out”.