— Atoms, nuclei, nucleons, and elementary particles, read through the hierarchy of “size” and “fluctuation.” Explanation and derivations for the companion simulator (elementary-particles-simulator-en.html)
In materials science and condensed-matter physics, most problems can be solved by treating an atom as “a sphere of radius about 1 Å,” an electron as “a point in an orbital,” and a nucleus as “a point charge of +Ze.” But once we start thinking about nuclear fusion, we run into questions: why does a nucleus not fly apart when the protons inside it should repel one another, is there any reason for neutrons to stick together, and would it not be more stable, after all, for a proton and an electron to combine into a neutron? These are simple but fundamental questions that the language of condensed-matter physics alone can no longer answer. This explanation answers such questions one by one, starting from the hydrogen atom and the helium atom.
The one backbone of the story is the view that “at every level, what determines the size is the quantum fluctuation of the constituents.” The size of an atom is set by the fluctuation of its electrons (the spread of the 1s orbital), that of a nucleus by the fluctuation of its nucleons, that of a nucleon by the fluctuations of quarks and gluons, while electrons and quarks themselves are “points” down to a resolution of 10−18 m. And the force that governs the fluctuation differs from level to level: the electromagnetic force for atoms, the nuclear force (the leftover of the strong force) for nuclei, the color force (the strong force itself) for nucleons, and the weak force, which converts protons and neutrons into each other. Gravity is only 10−36 times as strong as the electromagnetic force and becomes the leading player only at the scale of astronomical bodies.
First, a map of the whole picture. The numbers are the values of PDG 2024 and CODATA 2022 and of the papers cited in the text (the “sizes” of the electron and the quark are experimental upper limits).
| Level | Size (rms radius, etc.) | Constituents | Size of the constituents | Amplitude of the constituents’ fluctuation | Kinetic energy of confinement | Dominant force |
|---|---|---|---|---|---|---|
| Atom (H, He) | 0.5–1 Å | Electrons + nucleus | Electron < 2.8×10−19 m | 0.5–0.9 Å | 10–40 eV | Electromagnetic force |
| Nucleus (4He) | 1.68 fm | 2 protons + 2 neutrons | 0.84 fm (actual size) | 1.45 fm | ≈29 MeV per nucleon | Nuclear force (residual strong force) |
| Nucleon (p, n) | 0.84 fm | 3 quarks + gluons + sea quarks | Quark < 4.3×10−19 m | ≈0.6 fm | Hundreds of MeV | Color force (strong force) |
| Quark, electron | < 10−18 m | (point) | — | — | — | (The range of the weak force, 2.5×10−18 m, lies around here) |
If a particle is confined to a region of width Δx, its momentum acquires a spread of at least Δp ≈ ħ/Δx. The kinetic energy associated with this momentum is, in the nonrelativistic case,
Using ħc = 197.3 MeV·fm and ħ²/2m = 3.81 eV·Å² for the electron, the orders of magnitude at each level become visible at once.
| Confinement | Δx | Δp ≈ ħ/Δx | Kinetic energy | Binding energy for comparison |
|---|---|---|---|---|
| Electron in an atom | 1 Å | 1.97 keV/c | ≈ 4 eV (13.6 eV for an exact 1s state) | 13.6 eV |
| Electron in a nucleus (impossible) | 1 fm | 197 MeV/c | ≈ 200 MeV (relativistic) | The Coulomb attraction inside a nucleus is a few MeV |
| Nucleon in a nucleus | 1.5 fm | 130 MeV/c | ≈ 9 MeV (actually ≈29 MeV per nucleon) | Nuclear force −30 to −70 MeV |
| Quark in a nucleon | 0.5 fm | 400 MeV/c | ≈ 400 MeV (relativistic) | Confinement: ≈1 GeV at 1 fm |
In short, “the tighter the confinement, the more violently the constituents move.” The ladder of energies, atom (eV) → nucleus (MeV) → nucleon (hundreds of MeV), is the mirror image of the ladder of lengths, Å → fm → 0.5 fm. Tab ② of the simulator (“Ladder of Energies”) puts this correspondence into a single figure.
Experimentally, the size of the electron is at most 2.8×10−19 m (electron–positron scattering at LEP [Bourilkov 2001]) and that of a quark at most 4.3×10−19 m (electron–proton scattering at HERA [ZEUS 2016]); the Standard Model treats both as point particles of zero size. On the other hand, the range over which that electron is found inside an atom (the amplitude of its fluctuation) is 10−10 m, and the range over which a quark is found inside a proton is 10−16 m. The “size of a particle” and “the range in which the particle is found” are entirely different quantities, differing by a factor of 103–108.
Nucleons and atoms have internal structure, so their size itself can be measured. The proton’s 0.84 fm is the rms radius of its charge distribution (from muonic-hydrogen spectroscopy [Antognini 2013]; 0.84075(64) fm in CODATA 2022), and the radius of an atom is the spread of its electron distribution. The size of a nucleon is the fluctuation of the quarks inside it, and the size of an atom is the fluctuation of the electrons inside it — that is, the “size” of a composite particle is the “fluctuation” of its constituents.
In a crystal of metallic copper, against an atomic radius of 1.28 Å, the rms displacement of thermal vibration at room temperature is only about 0.08 Å (from the Debye–Waller factor of X-ray diffraction, B ≈ 0.55 Ų, ⟨u²⟩ = B/8π²), and the zero-point motion makes up about 0.04 Å of it. The fluctuation is only 1/30 of the size — so atoms can be placed as “spheres,” and classical lattice models can be used. The reason is that atoms are 105 times heavier than electrons and the lattice confinement is loose on the Å scale, so the kinetic energy of Eq. (1) is only of the order of meV. The deeper we descend into the world of elementary particles, the more this ratio is reversed, until the fluctuation exceeds the size by a factor of 103–108 (the ledger in §8).
In classical electromagnetism, an electron circling a nucleus loses energy by radiation and, in hydrogen, would fall into the nucleus in about 10−11 s. It does not fall because confining the electron to a region of radius r requires the kinetic energy of Eq. (1). The energy obtained by adding the Coulomb attraction,
is minimized at r = a0/Z (a0 = 0.529 Å). This rough estimate agrees with the exact 1s energy because the equality ΔxΔp ≈ ħ holds almost exactly for the 1s state, which is close to a Gaussian. If one tries to push the electron into the nucleus (r → 1 fm), the first term jumps to 200 MeV while the second is only 1.4 MeV — the reason the electron does not fall into the nucleus is the uncertainty relation itself. The simulator’s “Electron in an atom” view has a figure that shows this curve E(r) for different values of Z. ▶ Simulator ①: atom → electron
Because the virial theorem ⟨T⟩ = −E holds for the Coulomb system, the kinetic energy of the 1s state is 13.6 Z² eV. Setting this equal to ½m⟨v²⟩ gives
For hydrogen, v = 0.73% c = 2,190 km/s. The time scale of the fluctuation is a0/(αc) = ħ/Eh = 2.42×10−17 s (the atomic unit of time). In helium the two electrons screen each other’s nuclear charge; using the effective nuclear charge of the variational method, Z′ = 27/16 = 1.6875, one gets v = 1.23% c (the per-electron value from the exact ⟨T⟩ = 2.9037 hartree is 1.24% c). This Z′ is exactly the variational calculation of Lecture 3 of the Computational Materials Science course (the page on the variational method (in Japanese)).
Eq. (3) gives v/c = 0.58 for Z = 79 (gold) and 0.67 for Z = 92 (uranium). The point-nucleus 1s level of the Dirac equation is
and for gold the nonrelativistic −84.9 keV deepens to −93.6 keV, while the radius of the 1s orbital shrinks to √(1−(Zα)²) = 0.82 of its nonrelativistic value. This contraction propagates through the orthogonality condition up to the 6s orbital, lowering the d→s transition energy from 3.7 eV (ultraviolet) in silver (4d→5s) to 2.4 eV (blue) in gold (5d→6s) — which is why gold absorbs blue light and looks yellow. In mercury, the contracted 6s² becomes an inert closed shell, the metallic bonding weakens, and mercury is a liquid at room temperature (Pyykkö & Desclaux 1979). The common wisdom of condensed-matter physics that “heavy elements are relativistic” starts from an elementary-particle fact: the 1s electron is fluctuating at half the speed of light.
Classical electromagnetic intuition suggests that a positive proton and a negative electron combining into a neutral particle would be the most stable arrangement. But look at the particle masses:
so the neutron is heavier than a proton plus an electron. Hence p + e− → n + νe (electron capture) cannot occur unless 0.782 MeV is supplied from outside, and conversely a free neutron decays as n → p + e− + ν̄e (lifetime 878.4 s, §5). There are two reasons for the stability of the hydrogen atom: (i) quantum mechanics (§2.1) stops the electromagnetic fall, and (ii) energy (Eq. (5)) forbids conversion into a neutron.
This 0.782 MeV is by no means a number to be taken for granted. The neutron–proton mass difference of 1.293 MeV (measured) is reproduced within the errors by the lattice calculation, which finds the net balance of +2.5 MeV from the d quark being heavier than the u quark and −1.0 MeV from the electromagnetic self-energy of the proton’s charge to be 1.51(16)(23) MeV (a first-principles lattice QCD+QED calculation by the BMW Collaboration [Borsanyi 2015], §4.6). If the difference were smaller than 0.511 MeV, hydrogen atoms would turn into neutrons by inverse β decay and neither chemistry nor life would exist — the same paper says as much.
So where is the condition for “combining” met? In dense matter where the electrons are degenerate, the Fermi energy of the electrons can pay this cost. For a free gas of protons and electrons, the condition is a kinetic energy EF ≥ 0.782 MeV, and
From the upper end of the central densities of white dwarfs (106–108 g/cm³) onward, protons and electrons really begin to “combine” (neutronization). For protons inside nuclei, higher densities (109–1011 g/cm³) are needed, and in the outer crust of a neutron star this electron capture turns nuclei one after another into neutron-rich ones. In other words, the intuition that “proton + electron → neutron is stable” is correct only in a world where gravity has squeezed the electrons up to 0.8 MeV.
Before the discovery of the neutron in 1932, the nucleus was thought to be “made of protons and electrons.” It is Eq. (1) that buried this idea. Confining an electron to the size of a nucleus (1–5 fm) gives a kinetic energy ħc/Δx = 40–200 MeV. On the other hand, the Coulomb attraction that the nucleus exerts on the electron is Ze²/(4πε0R), which is only 14 MeV even for Z = 50 and R = 5 fm, and the energy of the electrons that emerge in β decay is only about 1 MeV — a nucleus cannot confine an electron. So the neutral part inside the nucleus must be not an electron but a new particle about as heavy as the proton. That is the neutron (Chadwick 1932), and in the same year Yukawa began to work on the problem of “the force that binds protons and neutrons” (§3).
The “motion” of electrons, nucleons, and quarks in the simulator is not a trajectory of quantum mechanics. It uses the diffusion process of Nelson’s stochastic mechanics [Nelson 1966], a random motion that preserves the probability density |ψ|²:
The diffusion coefficient is ħ/2m, and the drift is the sum of the “current velocity” v and the “osmotic velocity” u; the stationary solution of the Fokker–Planck equation of this process is exactly ρ = |ψ|². For the 1s state (ψ real, v = 0), u = −(Zαc) r̂: it is no accident that the magnitude of the inward drift velocity equals Zαc, the same as vrms of Eq. (3). For a (real) wave function with v = 0, ⟨p²⟩ = ħ²∫|∇ψ|² d³r = m²⟨u²⟩ holds identically, so the osmotic velocity carries all of the kinetic energy. For 1s, in addition, the decay constant of |ψ| ∝ e−Zr/a0 is constant, so the magnitude of u is ħZ/ma0 = Zαc everywhere. For nucleons and quarks, the ground state of a harmonic oscillator (Gaussian ψ) is used, and the drift u = −(ħ/mb²) r reproduces a Gaussian distribution with rms radius b√(3/2). The distribution is exact, but a “single path” has no physical meaning — that is the convention of this visualization (§12).
The carrier of the electromagnetic force (the photon) is massless, and the Coulomb potential reaches to infinity as 1/r. Hideki Yukawa (1935), noting that the nuclear force reaches only about 2 fm, supposed that its carrier has a mass. The equation for a field φ of mass m produced by a static point source gδ³(r) (the static limit of the Klein–Gordon equation) and its solution are
For m → 0 it reverts to 1/r. From a range of 2 fm, Yukawa predicted m ≈ 100–200 MeV/c², and in 1947 the π meson (139.6 MeV) was found: ħ/mπc = 197.3/139.6 = 1.41 fm. Applying the same formula to the weak force, the mass of the carrier W, 80.4 GeV, gives a range of 2.5×10−18 m (§5). Range of a force = ħ/(carrier mass × c) — this single line explains half of the difference in the “apparent strengths” of the four forces.
The real nuclear force is complicated, depending on spin, isospin, tensor, and spin–orbit terms (Argonne v18 has 18 operator terms; chiral effective theory has π exchange plus contact terms), but the skeleton of the central force is well represented by “a sum of two Yukawa-type terms.” The simulator uses the Malfliet–Tjon np (³S₁) potential [Malfliet & Tjon 1969]:
| r [fm] | 0.3 | 0.5 | 0.7 | 0.8 | 1.0 | 1.5 | 2.0 | 3.0 |
|---|---|---|---|---|---|---|---|---|
| Nuclear force VMT [MeV] | +574 | +30 | −70 | −77 | −69 | −32 | −13 | −1.9 |
| Proton–proton Coulomb e²/4πε0r [MeV] | +4.8 | +2.9 | +2.1 | +1.8 | +1.4 | +1.0 | +0.7 | +0.5 |
Inside 0.5 fm there is a strong repulsive core (nucleons do not overlap — which is why the nuclear density is almost constant at 0.16 nucleons/fm³ for every nucleus), around 0.8 fm there is an attraction of −77 MeV, and beyond 2 fm a weak tail of π exchange remains. The Coulomb repulsion between protons is a mere 1.4 MeV at 1 fm. The answer to “protons repel one another, yet the nucleus holds together” is that the attraction at 1 fm is 50 times the repulsion. But the Coulomb force reaches far and accumulates as Z², so heavier nuclei need more neutrons, and beyond Z ≈ 100 they become unstable against fission (§3.6). ▶ Simulator ②: the four forces
The “residual force” most familiar to materials scientists is the van der Waals force. Neutral atoms, although they carry no net charge, attract each other weakly through fluctuations of their internal charge distributions (induced dipoles), and the strength (10 meV) is far weaker than the bonds inside an atom (eV). The nuclear force has the same structure. Nucleons are color-neutral (the three colors mix to give “white”), yet they attract each other through fluctuations of the quarks and gluons inside — specifically, through the exchange of color-neutral mesons (π, σ, ρ, ω …) — and the strength (tens of MeV) is orders of magnitude weaker than the color force that confines quarks (GeV/fm). That is why the nuclear force is called “the leftover of the strong force” (the residual strong force). The difference is that, whereas the van der Waals force falls off as 1/r6, the nuclear force is cut off as e−r/1.4 fm because its carrier is massive.
Confining four nucleons to a radius of 1.5 fm should give, by Eq. (1), a kinetic energy of several tens of MeV per nucleon. The result of a quantum Monte Carlo calculation (GFMC) using the Argonne v18 two-body force plus the Illinois three-body force [Pieper & Wiringa 2001] puts numbers on this intuition:
| ⁴He (GFMC, AV18/IL2) | Value [MeV] | Remarks |
|---|---|---|
| Nucleon kinetic energy ⟨K⟩ | +115 | 29 MeV per nucleon, prms = √(2mT) = 233 MeV/c, v/c ≈ 0.25 |
| Two-body nuclear force ⟨vij⟩ | −136 | Of which 1π exchange is −105 (77%): Yukawa’s picture is still the leading player, quantitatively |
| Three-body force ⟨Vijk⟩ | −8.4 | 6% of the two-body force, but responsible for 30% of the binding |
| Electromagnetic (Coulomb, etc.) ⟨vγ⟩ | +0.9 | Repulsion between the two protons; 1/150 of the nuclear force |
| Total | −28.3 | Measured −28.296 MeV (7.07 MeV per nucleon) |
The binding energy is the net balance of an enormous kinetic energy and an attraction that exceeds it only slightly. For the deuteron it is even more delicate: in AV18, ⟨T⟩ ≈ 19.8 MeV, ⟨V⟩ ≈ −22.0 MeV, and the net is −2.22 MeV. The nucleons spend nearly half of their time outside the reach of the nuclear force, the charge radius swells to 2.5 times that of the proton (2.13 fm), and there is not a single excited state. The time scale of the nucleon fluctuation is ħ/⟨T⟩ ≈ 2×10−23 s. The nucleus view of the simulator slows this time down by a factor of about 1022.
The intuition that “there is no reason for neutrons to stick together” is correct: the dineutron (nn) does not exist. The reason, however, is not the Coulomb force. For two identical fermions to be in a relative s wave (the configuration that makes the best use of the attraction), the Pauli principle requires the spins to be antiparallel (¹S₀, the spin singlet). A proton and a neutron, on the other hand, are distinguishable particles, so the spin-parallel ³S₁ (triplet) is also allowed, and there the tensor force of π exchange mixes in a ³D₁ component (probability ≈5%) and produces extra attraction. This is what decides the binding of the deuteron, and ¹S₀ does not have it. Comparing scattering lengths a (negative means unbound):
| System (channel) | Scattering length a [fm] | State | Remarks |
|---|---|---|---|
| np (³S₁) | +5.42 | Bound: the deuteron, B = 2.22 MeV | The tensor force mixes in a D wave |
| np (¹S₀) | −23.7 | Unbound (virtual state, ≈70 keV short) | The spin-singlet np does not bind either |
| nn (¹S₀) | −18.6 | Unbound (virtual state, ≈100 keV short) | Charge independence of the nuclear force: nearly the same as the ¹S₀ of np |
| pp (¹S₀) | −7.8 (including Coulomb) | Unbound | Same nuclear force as nn, plus Coulomb repulsion |
The nuclear force is strong enough that “100 keV more would bind nn,” yet not strong enough to “bind ¹S₀” — this delicacy determines cosmic nucleosynthesis (deuterons form but diprotons do not) and why the Sun’s pp reaction needs the weak force (§5.3). With the “Bring two neutrons together” and “Bring two protons together” buttons in the nucleus view of the simulator, you can watch them drift apart.
⁴He (the α particle) is bound by 7.07 MeV per nucleon, far more strongly than any other light nucleus (deuteron 1.11 MeV, ³He 2.57 MeV). The 1s orbital can be “fully occupied” by the four combinations of spin up/down × proton/neutron, so all four nucleons enter the configuration that makes the best use of the attraction. The fifth nucleon is pushed into the 1p orbital by the Pauli principle, so neither ⁵He nor ⁵Li is bound. For heavy nuclei, the semi-empirical mass formula
summarizes the story. The nuclear force has a short range, so it acts only as the volume term aVA, whereas the Coulomb term accumulates as Z². In this competition the binding per nucleon is greatest at iron and nickel (8.8 MeV) (the reason fusion releases energy; see the B/A curve in the nuclear fusion simulator (in Japanese)), and heavy nuclei require an excess of neutrons (N > Z) and, beyond Z ≈ 100, fall apart by spontaneous fission. The periodic table has an end because the electromagnetic force is a long-range force.
A free neutron undergoes β decay in 15 minutes, but a neutron in ⁴He is stable forever. Turning n → p would convert ⁴He into ⁴Li, but ⁴Li is unstable against proton emission, and the energy required falls short by as much as 22.9 MeV. The difference in binding energies created by the nuclear force and the Pauli principle shuts out the weak force. Conversely, if the daughter nucleus is slightly more deeply bound, as in tritium (³H → ³He, Q = 18.6 keV, half-life 12.3 years), the decay does occur. Whether “a neutron inside a nucleus is stable” is decided not by the weak force but by the energy balance of the nucleus.
Protons and neutrons are not elementary particles but composite particles with three valence quarks (Gell-Mann and Zweig 1964; confirmed by deep inelastic scattering at SLAC in 1968). There are six kinds (flavors) of quarks, and their charges are the fractions +2/3 or −1/3 of e:
| Generation | Quark (charge +2/3) | Mass | Quark (charge −1/3) | Mass | Charged lepton | Mass |
|---|---|---|---|---|---|---|
| 1st | u (up) | 2.16 MeV | d (down) | 4.70 MeV | e− | 0.511 MeV |
| 2nd | c (charm) | 1.273 GeV | s (strange) | 93.5 MeV | μ− | 105.7 MeV |
| 3rd | t (top) | 172.6 GeV | b (bottom) | 4.18 GeV | τ− | 1777 MeV |
(The masses of the light quarks are values in the MS-bar scheme at 2 GeV, PDG 2024. Each generation also includes a neutrino: νe, νμ, ντ.) Ordinary matter is made only of the first generation: proton = uud (+2/3 +2/3 −1/3 = +1), neutron = udd (+2/3 −1/3 −1/3 = 0). The charges of nucleons come from quarks, and the difference between a proton and a neutron is only that one u has been swapped for one d — so when the weak force turns d ⇄ u, protons and neutrons convert into each other (§5). One convincing piece of evidence for the quark model is the magnetic moment: the model predicts μn/μp = −2/3, and the measured value is −1.913/2.793 = −0.685.
In addition to electric charge, quarks carry a “color charge” that takes three values (red, green, blue). When all three colors are present they make “white” (color-neutral), and a nucleon is a color singlet with one quark of each color (a meson is a quark plus an antiquark: color plus anticolor makes white). Gluons carry the color force, and there are 3×3 − 1 = 8 kinds of them, from the combinations of color and anticolor (in group-theoretic terms, the adjoint representation 8 of SU(3); see §7 of the Lie group page). Unlike photons, gluons carry color charge themselves, so gluons interact with each other. This is what separated the fates of the electromagnetic and strong forces: the field lines of the photon spread out and thin as 1/r², whereas the field lines of gluons attract one another and bundle into a thin tube (a string), transmitting a constant force independent of distance (§4.4).
The coupling constant depends on the energy (distance) at which it is measured. In one-loop perturbation theory,
| Scale Q (distance ħc/Q) | 91 GeV (0.002 fm) | 10 GeV (0.02 fm) | 2 GeV (0.1 fm) | 1 GeV (0.2 fm) | ≈0.3 GeV (0.7 fm) |
|---|---|---|---|---|---|
| αs | 0.118 | ≈0.18 | ≈0.30 | ≈0.5 | ≳1 (perturbation theory breaks down) |
The electromagnetic α slowly grows at high energies (1/137 → 1/128 at 91 GeV), whereas αs shrinks (asymptotic freedom, Gross–Wilczek and Politzer 1973, Nobel Prize 2004). So in high-energy experiments (HERA, LHC) quarks behave as nearly free point particles, whereas at the scale of the nucleon size, 1 fm, the coupling exceeds 1 and quarks can no longer be pulled apart (confinement). The world of quarks inside a nucleon is a strongly coupled world where perturbation theory cannot be used, and lattice QCD is the only first-principles tool.
The quark–antiquark potential is well described by the Cornell form, from lattice QCD and the spectroscopy of heavy-quark systems (charmonium):
The first term is the Coulomb-type contribution of single-gluon exchange (4/3 is the color factor), and the second term is the confining string. The force pulling quarks apart is constant, independent of distance, and its magnitude is the weight of 10 cars — pulling them apart by just 1 fm costs 0.9 GeV, the energy of six π mesons, so the string breaks on the way and a quark–antiquark pair (a meson) is created, and a lone quark never emerges. For comparison, the electromagnetic force between two protons at 1 fm is 231 N (23.5 kg-force). Tab ② of the simulator lines up these two curves (the color force between quarks and the nuclear force between nucleons) on the same yardstick as the electromagnetic force, the weak force, and gravity.
Against the proton mass of 938.27 MeV, the sum of the masses of the three valence quarks is 2.16 + 2.16 + 4.70 = 9.0 MeV, and even for the neutron only 11.6 MeV. Nearly all of the mass is the kinetic and field energy of the confined quarks and gluons (the reverse of E = mc²). Using the energy–momentum tensor of QCD, the rest-frame mass decomposes as
(the quark mass term, the quark kinetic energy, the gluon field energy, and the trace anomaly), and the breakdown from lattice QCD is [Yang et al. 2018]:
| Breakdown of the proton mass (χQCD, MS-bar 2 GeV) | Fraction | Meaning |
|---|---|---|
| Quark kinetic energy ⟨HE⟩ | 32(4)(4)% | Motion due to confinement (Eq. (1)) |
| Gluon field energy ⟨Hg⟩ | 36(5)(4)% | Energy of the color electric and color magnetic fields |
| Trace anomaly ¼⟨Ha⟩ | 23(1)(1)% | Quantum breaking of scale invariance (the mass generated by Λ) |
| Quark mass term ⟨Hm⟩ (from the Higgs) | 9(2)(1)% | Contribution of the u, d, and s condensates. The bare masses of the three valence quarks alone would give 1% |
In the language of condensed-matter physics, a quark inside a nucleon is not a “particle with a bare mass of 2–5 MeV” but a quasiparticle dressed in a cloud of gluons, and its effective mass (the constituent quark mass) is ≈ 330 MeV. The mechanism that produces this effective mass is the spontaneous breaking of chiral symmetry, which Yoichiro Nambu formulated taking the generation of the gap in BCS superconductivity as a hint [Nambu & Jona-Lasinio 1961, Nobel Prize 2008]. The vacuum of QCD is filled with a condensate of quark pairs ⟨q̄q⟩, like a superconductor, and a quark moving through it acquires the gap (≈300 MeV) as its mass — for those who know superconductivity, I think this is the explanation that makes the most sense.
As we saw in §2.4, this small difference (1.293 MeV) underpins the existence of the hydrogen atom. A first-principles calculation that incorporates QED into lattice QCD [Borsanyi et al. 2015] obtained
— showing that the difference is the net balance of two effects. Because d is heavier than u, the udd neutron becomes heavier, while the charged proton becomes heavier by the amount of its electromagnetic self-energy. If α were twice its present value, the QED term would be −2 MeV and the difference would shrink to nearly the electron mass, 0.511 MeV, leaving the hydrogen atom on the verge of stability; conversely, if the difference were too large, too many neutrons would be lost in Big Bang nucleosynthesis and heavy elements would hardly form — a textbook example of “cosmic fine-tuning.”
Inside the proton’s charge radius of 0.84 fm, the position of a valence quark fluctuates by about 0.6 fm rms (the rest comes from the π-meson cloud). On the other hand, the size of the quark itself is at most 4.3×10−19 m from electron–proton scattering at HERA [ZEUS 2016]: the fluctuation is more than 1,400 times the size. From Eq. (1) the momentum is p ≈ ħ/(0.5 fm) ≈ 400 MeV/c — exceeding the bare mass of 2–5 MeV/c² by orders of magnitude, so the quark moves at nearly the speed of light. Moreover, the content of the proton is not just “three quarks.” Gluons carry about half of the proton’s momentum (⟨x⟩g ≈ 0.4–0.5), and quark–antiquark pairs (sea quarks) are constantly created and annihilated. These are what the simulator’s nucleon view draws as small purple pairs. The time scale of the fluctuation is ħ/(pc) ≈ 1.6×10−24 s.
The charge of the neutron is 0, but the mean square radius of its charge distribution is negative: ⟨r²⟩n = −0.1155(17) fm² (PDG 2024). This means that there is positive charge on the inside and negative charge on the outside, and it is explained by both (i) the “π− cloud,” in which the neutron temporarily fluctuates into p + π−, and (ii) the spin-dependent color magnetic force pushing the d quarks outward. The magnetic radius is 0.86 fm, similar to the proton’s — even though the neutron is “neutral,” its inside is a whirl of charge and current. ▶ Simulator ①: nucleus → neutron
The β decay of the neutron, n → p + e− + ν̄e, is, at the quark level, d → u + W−, W− → e− + ν̄e. The weak force is not a force that “sticks things together” but a force that changes the type (flavor) of particles, and of the four forces it is the only one that can change the type of a quark. Fermi (1934) formulated it as an interaction in which four fermions couple at a single point, and its strength is the Fermi constant GF = 1.1663787(6)×10−5 GeV−2. In 1983 the carriers W± (80.37 GeV) and Z0 (91.19 GeV) were discovered at CERN, and it became clear that the Fermi constant can be written in terms of the W mass and the coupling constant g:
αW ≈ 1/32 is larger than the electromagnetic 1/137 — the coupling of the weak force itself is not weak. Substituting into Eq. (15) gives GF ≈ παW/(√2 MW²) ≈ 1.1×10−5 GeV−2, reproducing the measured value in both order of magnitude and coefficient.
Applying Yukawa’s Eq. (8) to the W, the range is ħ/MWc = 197.3 MeV·fm / 80,369 MeV = 2.5×10−3 fm = 2.5×10−18 m. This is 1/340 of the size of the proton (0.84 fm) and 1/500 of the distance between nucleons inside a nucleus (1–2 fm). The Yukawa factor e−r/rW at r = 1 fm is e−400 ≈ 10−174 — between one nucleon and another, the weak force is literally zero, and it “shows its face” only at the instant d → u occurs inside a single nucleon. The effective strength at low energies is GFE² ∝ (E/MW)², which at the scale of the proton is GFmp² ≈ 10−5: this is the number for the weak force in the often-quoted “strong force 1, electromagnetic force 10−2, weak force 10−5–10−6, gravity 10−39.”
The slowness of the weak force shows up in lifetimes. A decay by the strong force (Δ → Nπ) takes 10−23 s, a decay by the electromagnetic force (π0 → γγ) 10−16 s, and decays by the weak force take 2.6×10−8 s for π+ and 878.4 s for the neutron (PDG 2024; half-life 10.2 minutes). The neutron is especially long-lived because Q = 0.782 MeV is small and the phase space scales as Q5 (Sargent’s rule). Indeed,
(Vud = 0.974 is an element of the Kobayashi–Maskawa matrix, gA = 1.275 is the axial coupling of the nucleon, f = 1.689 is the phase-space integral, and ΔR ≈ 3.9% is the radiative correction). In this way, the Fermi constant and the electron mass alone give almost the correct lifetime. The weak force is “weak” because GF is small, GF is small because the W is heavy — and the W is heavy because of the Higgs field (§5.4).
The first step of the pp chain, the energy source of the Sun, is
As we saw in §3.5, the diproton (pp) is not bound, so a deuteron cannot form unless, within the instant (10−22 s) in which the two protons are within reach of the nuclear force, a u → d occurs inside one of the protons and turns it into a neutron. The Coulomb barrier (500 keV against 1.3 keV at the center of the Sun, overcome by the tunnel effect) and the slowness of the weak force combine to make the reaction rate extremely small: the mean time before a given proton reacts at the center of the Sun is about 1010 years, and the heat generation density at the center is a mere 276 W/m³ — comparable to human metabolism or a compost heap. The Sun is bright only because its volume is enormous.
This is a blessing, not a defect. If the pp reaction proceeded by the strong force alone, the Sun would burn out in a few thousand years and there would have been no time for life to evolve. By contrast, fusion reactors choose D + T → ⁴He + n because this reaction does not need the weak force (the neutron is already inside the tritium) and its rate is as much as 1025 times that of pp (see the “Four Forces” and “Walls and Tunneling” tabs of the nuclear fusion simulator (in Japanese)).
In the Glashow–Weinberg–Salam theory, the electromagnetic and weak forces are four forces (W+, W−, Z0, photon) arising from a single SU(2)×U(1) gauge symmetry, and the coupling constants g, g′ are tied to the electric charge e = g sinθW through the mixing angle θW (sin²θW = 0.2313). The vacuum expectation value of the Higgs field, v = 246 GeV, gives the masses MW = gv/2 = 80 GeV and MZ = 91 GeV, and only the photon remains massless — and this is the whole reason why the electromagnetic force reaches to infinity while the weak force reaches only 10−18 m (the Higgs boson, 125.2 GeV, was discovered in 2012). When the collision energy exceeds 100 GeV (when the distance becomes shorter than 10−18 m), the Yukawa factor returns to 1 and the weak force acts with a strength comparable to the electromagnetic force. If you lower the distance to 10−18 m in tab ② of the simulator, you can see the curve of the weak force catch up with that of the electromagnetic force.
The weak force has one more feature that the other forces lack: it breaks mirror symmetry (parity) maximally and acts only on left-handed particles (the Lee–Yang prediction and Wu’s experiment, 1957). Neutrinos feel only the weak force, so they pass straight through the Earth and the Sun (solar neutrinos: Koshiba 2002; neutrino oscillation: Kajita 2015).
The size of a particle is measured by scattering. If the charge distribution is extended, the scattering amplitude at momentum transfer q is multiplied by a form factor F(q²) = 1 − q²⟨r²⟩/6 + …, and when ħ/q becomes shorter than the size, deviations from a point charge become visible. The proton’s 0.84 fm is the actual size measured in this way (by electron scattering and muonic-hydrogen spectroscopy). For the electron, e+e− → e+e− scattering at LEP (√s ≈ 200 GeV) probed below 10−18 m and found no deviation, yielding an upper limit on the radius of 2.8×10−19 m (95% confidence level) [Bourilkov 2001]. For quarks, ep scattering at HERA gives 4.3×10−19 m [ZEUS 2016]. The anomalous magnetic moment of the electron, g−2, agrees with QED to 13 digits (g/2 = 1.001 159 652 180 59(13) [Fan et al. 2023]), and using a relation of composite models (Brodsky–Drell), the size of any structure is estimated to be below 10−22 m [Dehmelt 1988] — but this is a model-dependent estimate, not a measurement.
The classical electron radius often quoted as the “radius of the electron,” re = e²/(4πε0mc²) = 2.82 fm, is “the radius of a classical sphere whose electrostatic energy equals mc²,” and is not a size (it is 104 times larger than the actual upper limit). It only has meaning as the length scale that appears in the Thomson scattering cross section (8π/3)re².
So how finely can a “point” electron be localized? From the relativistic version of Eq. (1),
If an electron is confined to less than the reduced Compton wavelength, the energy exceeds mc², electron–positron pairs are created from the vacuum, and one can no longer say “which one is the original electron.” This is the lower limit at which the position of the electron can be given meaning as a single particle.
The same thing appears in the Dirac equation. The eigenvalues of the Dirac velocity operator cα are only ±c, and the time evolution of the position operator of a free electron is
so even an electron at rest has its position trembling with an amplitude of ƛC/2 = 1.9×10−13 m, 1021 times per second (Zitterbewegung, Schrödinger 1930). This oscillation comes from the interference between positive-energy and negative-energy (positron) solutions; for a purely positive-energy wave packet it disappears, but instead the position operator itself is blurred over a range of ƛC (Newton–Wigner). Either way, the conclusion is the same: “an electron as a stationary point charge” does not exist on scales finer than 10−13 m. An observable trace is the Darwin term of the hydrogen atom. The correction, equivalent to averaging the Coulomb potential over a range of ƛC,
raises only the s states and has been measured as part of the fine structure of hydrogen.
In QED the vacuum also has zero-point fluctuations of the electromagnetic field, and the electron is constantly shaken by them. In Welton’s (1948) intuitive calculation, one sums the equation of motion m d²δr/dt² = −eE(t) over all modes of the zero-point electric field to obtain
(EB ≈ 17 Ry is Bethe’s mean excitation energy, and the logarithmic factor is ≈ 8). When the electron is blurred over a range of 10−13 m, the potential it feels near the nucleus is averaged and the s levels become shallower — this is the intuitive explanation of the Lamb shift (measured 1057.8 MHz), which lifts the degeneracy of 2s1/2 and 2p1/2, and even this rough estimate gets the order of magnitude and the coefficient nearly right. If Zitterbewegung is “a property of the Dirac theory of a single electron,” Welton’s trembling is “a shaking from the medium called the vacuum,” and both are on the scale of 10−13 m.
The Born–Oppenheimer approximation is an approximation that stops the nucleus, not an approximation that stops the electron. If an electron is “placed” as a Gaussian wave packet of width σ0, the free solution of the Schrödinger equation is
| Initial width σ0 | τ = 2mσ0²/ħ | After 1 fs | After 1 ps | After 1 ns |
|---|---|---|---|---|
| 1 pm (2.6 times the reduced Compton wavelength ƛC) | 1.7×10−20 s | 58 nm | 58 µm | 5.8 cm |
| 1 Å (the size of an atom) | 1.7×10−16 s | 0.59 nm | 0.58 µm | 0.58 mm |
| 1 nm | 1.7×10−14 s | 1.0 nm | 58 nm | 58 µm |
The more tightly it is confined, the faster it spreads — because the momentum uncertainty ħ/2σ0 becomes the spread of velocity. The reason the electron does not spread inside an atom is that the Coulomb attraction of the nucleus pulls the spreading back and forms a stationary state (|ψ1s|²); that 0.5 Å cloud is not “standing still” but “a fluctuation in which the tendency to spread and the attraction balance.”
If an electron cannot be stopped in free space, the only way is to confine it in a harmonic potential. Even then the zero-point amplitude of the ground state, √(ħ/2mω), remains. In Harvard’s one-electron g−2 experiment (Penning trap, B ≈ 5.4 T [Hanneke 2008; Fan 2023]), the cyclotron frequency is eB/2πm ≈ 150 GHz, the magnetic length of the ground state is √(ħ/eB) ≈ 11 nm, and at an axial frequency of ≈ 200 MHz the zero-point amplitude is ≈ 0.2 µm. A particle of size below 10−18 m fluctuates over a range of 10−7 m even when it is held “at rest” in the laboratory: a factor of 1011. And this one fluctuating electron fixes g−2 to 13 digits and tests QED more precisely than anything else in human history. ▶ Simulator ①: electron in vacuum
The “fluctuation” of an electron inside an atom (the spread of |ψ|²) becomes a chemical bond when it overlaps with that of the neighboring atom (page on the overlap integral (in Japanese)), and in a solid it spreads over the whole crystal and becomes a band. The electron density n(r) that density functional theory computes is precisely the many-electron version of the stationary distribution |ψ|² of §2.6. In other words, the “distribution of electrons” that condensed-matter physics handles every day is nothing other than a map of the quantum fluctuation of electrons, which are point particles. The electron is a point of 10−18 m, yet its fluctuation creates the chemistry and the materials of the Å scale — the size of this gap (a factor of 108) is the number this explanation most wants to convey.
To compare the “strength” of forces without units, we use the dimensionless quantity (the coupling constant) obtained by dividing the potential energy of two particles at distance r by ħc/r:
| Force | Definition of the coupling constant | Value | Carrier (mass) | Range ħ/mc |
|---|---|---|---|---|
| Gravity | αG = Gmp²/ħc | 5.9×10−39 | Graviton (0, undiscovered) | ∞ |
| Electromagnetic force | α = e²/4πε0ħc | 7.30×10−3 = 1/137.04 | Photon (0) | ∞ |
| Weak force | αW = g²/4π = α/sin²θW | 3.2×10−2 = 1/31.7 | W± (80.4 GeV), Z0 (91.2 GeV) | 2.5×10−18 m |
| Strong force (color) | αs = gs²/4π | 0.118 (91 GeV) – ≈1 (1 GeV) | Gluons, 8 kinds (0) | Confinement (the string breaks at ≈1.5 fm) |
| Strong force (nuclear force) | gπNN²/4π | ≈14 | π (140 MeV), σ, ρ, ω … | 1.4 fm |
Looking at the coupling constants alone, the weak force is stronger than the electromagnetic force, and the nuclear force is stronger than the color force by orders of magnitude. The names “weak” and “residual” are given according to the effective strength, including the range. The often-quoted “1 : 10−2 : 10−6 : 10−39” lines up the strong force ≈ 1, the electromagnetic force α, the weak force GFmp² ≈ 10−5 (written as 10−6 to 10−13 in some normalizations), and gravity αG, and it implicitly compares them at “the scale of the proton (1 fm).” Change the distance and the ranking changes — that is the figure in tab ② of the simulator.
The five potentials used in tab ② of the simulator (for two protons, or for a quark–antiquark pair) are
and the force is F = −dV/dr (1 MeV/fm = 1.602×10−13 J / 10−15 m = 160.2 N). Values at representative distances:
| Distance | Gravity | Electromagnetic force | Weak force | Nuclear force (np) | Color force (q q̄) |
|---|---|---|---|---|---|
| r = 2.5×10−18 m (= rW) | −4.7×10−28 eV | +587 MeV | +934 MeV | — (the nucleons overlap; meaningless) | ≈ −10 GeV (αs ≈ 0.1) |
| r = 1 fm | −1.2×10−30 eV (1.9×10−34 N) | +1.44 MeV (231 N = 23.5 kg-force) | ≈ +10−170 eV | −69 MeV (1.2×104 N = 1.2 tonnes-force) | +0.82 GeV (1.6×105 N = 16 tonnes-force) |
| r = 1 Å | −1.2×10−35 eV | +14.4 eV (2.3×10−8 N) | 0 | 0 (e−1.5×105) | — (the string breaks at 1.5 fm) |
The weak-force potential is an order-of-magnitude estimate that treats Z/W exchange as a Yukawa-type potential; the actual coupling depends on spin and chirality, and its sign is not constant either. The color force uses the fixed value αs = 0.3, so at distances shorter than 0.1 fm it is an overestimate by the effect of the running coupling constant (§4.3).
In particle physics, both lengths and temperatures are expressed through energy. The conversions are L = ħc/E (197.3 MeV·fm) and T = E/kB (1 eV = 11,604 K):
| Energy | Length ħc/E | Temperature E/kB | Main players |
|---|---|---|---|
| 25 meV | 7.9 µm | 300 K | Lattice vibrations, diffusion, room-temperature chemistry |
| 1–10 eV | 200–20 nm | 104–105 K | Visible light, chemical bonds, ionization of atoms |
| 1 keV | 0.2 nm | 1.2×107 K | X-rays, inner-shell electrons, the center of the Sun (1.3 keV) |
| 1 MeV | 197 fm | 1.2×1010 K | Nuclear binding, β decay, pair creation (0.511 MeV) |
| 140 MeV | 1.4 fm | 1.6×1012 K | π mesons, the range of the nuclear force, the QCD phase transition |
| 1 GeV | 0.2 fm | 1.2×1013 K | The mass of the proton, the scale of confinement |
| 100 GeV | 2×10−18 m | 1015 K | W, Z, Higgs, electroweak unification |
| 10 TeV | 2×10−20 m | 1017 K | The smallest length the LHC can resolve |
| 1019 GeV | 1.6×10−35 m | 1032 K | The Planck scale (quantum gravity) |
To sum up the cast of characters that have appeared so far: the matter particles are 6 kinds of quarks and 6 kinds of leptons (three generations), the carriers of the forces are the photon, W±, Z0, and 8 kinds of gluons, and the Higgs particle gives mass — this is the Standard Model. Only gravity lies outside the Standard Model, and its unification with quantum theory remains unsolved. Japanese physicists stand at the milestones of this story: Hideki Yukawa (the nuclear force = meson exchange, 1949), Shinichiro Tomonaga (renormalization of QED, 1965), Masatoshi Koshiba (neutrino astronomy, 2002), Yoichiro Nambu (spontaneous symmetry breaking = the origin of mass, 2008), Makoto Kobayashi and Toshihide Maskawa (the prediction of three generations, 2008), and Takaaki Kajita (neutrino oscillations, 2015).
Finally, we put the numbers that formed the backbone of this explanation into a single table (the same as tab ③ of the simulator). “Size” is the actual size (rms radius) for a composite particle and the experimental upper limit for a point particle. “Fluctuation” is the rms of the position of the particle in the situation in which it is placed.
| Particle (situation) | Size | Fluctuation | Fluctuation / size | Nature of the fluctuation |
|---|---|---|---|---|
| Electron (1s of hydrogen) | < 2.8×10−19 m | 0.92 Å | > 3×108 | √3 a0 (§2) |
| Electron (1s of gold, Z = 79) | < 2.8×10−19 m | 1.2 pm | > 4×106 | √3 a0/Z, v = 0.58c (§2.3) |
| Electron (in vacuum, Dirac trembling) | < 2.8×10−19 m | 1.9×10−13 m | > 7×105 | Zitterbewegung ƛC/2 (§6.2) |
| Electron (vacuum zero-point shaking) | < 2.8×10−19 m | 8×10−14 m | > 3×105 | Welton, Lamb shift (§6.3) |
| Electron (Penning trap) | < 2.8×10−19 m | 0.2 µm | > 7×1011 | Axial zero-point amplitude (§6.5) |
| Quark (inside the proton) | < 4.3×10−19 m | 0.6 fm | > 1.4×103 | Confinement (§4.7) |
| Nucleon (inside ⁴He) | 0.84 fm | 1.45 fm | 1.7 | Point-nucleon radius from GFMC (§3.4) |
| Nucleon (inside the deuteron) | 0.84 fm | 1.97 fm | 2.3 | Matter radius (§3.4) |
| Atom (Cu crystal, 300 K) | 1.28 Å | 0.08 Å | 0.06 | Debye–Waller (§1.3) |
From the bottom row to the top, the fluctuation/size ratio grows as 0.06 → 2 → 103 → 108. Atoms can be treated as spheres in the world of materials because the ratio is smaller than 1, inside a nucleus the nucleons fluctuate so much that they overlap one another (ratio ≈ 2), and inside a nucleon and inside an atom a “point” fluctuates over a vast range. And the force that governs the fluctuation at each level changes in turn — electromagnetic force → nuclear force → color force — with the weak force rewriting the types of particles in between: the division of labor among the four forces corresponds, one by one, to the rows of this ledger.
In classical electromagnetism, a proton creates a Coulomb field E = e/4πε0r² in space, and an electron feels a force from that field. In quantum electrodynamics (QED), the electromagnetic field itself is quantized, and its quantum is the photon. The electrostatic field around a proton is a cloud of virtual photons that the proton constantly emits and absorbs, and when an electron is in that cloud, it absorbs a virtual photon emitted by the proton (or the other way around) — this is what is meant by “the force is carried by exchanging photons.” A real photon satisfies E = pc (mass 0), but a virtual photon need not satisfy this relation; instead, the uncertainty relation
allows it to “borrow” energy and fly for a short time (the argument of Wick, 1938). Since the photon has mass 0, there is no lower bound on the energy that must be borrowed, and it reaches arbitrarily far (range ∞; the Coulomb force goes as 1/r²). The π meson must borrow 140 MeV, so it hardly reaches farther than 1.4 fm (the same conclusion as Yukawa’s argument in §3.1). The W boson borrows 80 GeV, so its range is 2.5×10−18 m. The formula of §3.1, “the mass of the carrier sets the range of the force,” is a mechanism common to all forces that exchange a carrier, including the electromagnetic force (mass 0 → infinite range). The color force alone is an exception: although the gluon has mass 0, the range does not become infinite, because the gluons interact with one another and bundle the lines of force into a thin string (§4.2), and that string breaks at ≈1.5 fm (§4.4). ▶ Simulator ④: photon exchange
When “photon exchange” is turned into a calculation, the amplitude for the exchange of a single carrier with momentum ħq (the propagator) has the form 1/(q² + μ²), and the interaction energy between two particles is its Fourier transform:
For μ → 0 (the photon), V = −g²/4πr: reading g² → e²/ε0, charges of opposite sign feel the Coulomb attraction −e²/4πε0r, and charges of the same sign feel repulsion. For μ ≠ 0 it is the Yukawa potential of §3.1. In other words, “photon exchange” and “π exchange” are nothing but the same formula with μ = 0 and μ = mπc/ħ. The propagator 1/q² can actually be seen: in the Born approximation of scattering the amplitude is proportional to Ṽ(q), and the cross section is |Ṽ|² ∝ 1/q⁴ ∝ 1/sin⁴(θ/2) — the angular distribution of Rutherford scattering is precisely the square of the photon propagator.
Classically it is sometimes explained that “if two people throw balls to each other, they recoil away from each other (repulsion),” but attraction does not come out of this picture. Correctly, it is determined by the spin of the carrier and the sign of the charges: exchange of spin 1 (photon, gluon) gives repulsion between like charges and attraction between unlike charges, whereas exchange of spin 0 (σ meson) or spin 2 (graviton) gives attraction even between like charges (the overall sign of Eq. (25) changes with the spin of the carrier). That is why gravity, whose charges cannot cancel, is always attractive and adds up, while the electromagnetic force disappears once matter becomes neutral (§7.3). The π meson is a pseudoscalar (spin 0, negative parity), and the one-pion-exchange potential carries spin and isospin operators:
(S12 is the tensor operator, and T(x) = (1 + 3/x + 3/x²)Y. This is the form of the one-pion-exchange term of Argonne v18 [Wiringa et al. 1995].) The sign of the central part is set by (τ1·τ2)(σ1·σ2)/3: for the deuteron (T = 0, S = 1) it is (−3)(+1)/3 = −1, attractive; for np, nn, and pp in ¹S₀ (T = 1, S = 0) it is (+1)(−3)/3 = −1, also attractive; for T = 1, S = 1 it is a repulsion of +1/3, and for T = 0, S = 0 a strong repulsion of +3. That “the deuteron is bound only in the triplet,” as stated in §3.5, was the result of the tensor force S12T (which acts only in the triplet) being added to this central force.
The “exchange” so far has meant (A) throwing carrier particles back and forth. The “exchange” in the “exchange energy” of many-electron theory is a different thing and refers to the operation of (B) swapping the coordinates of indistinguishable identical particles. The two-electron wave function must satisfy the Pauli antisymmetry Ψ(1,2) = −Ψ(2,1); if the spins are parallel, the spatial part is antisymmetric, that is, it vanishes at r1 = r2. Electrons of the same spin cannot approach one another — the expectation value of the Coulomb repulsion decreases by the amount of this “Fermi hole,” and that decrease is the exchange energy. In terms of Eqs. (2.51)–(2.54) of the author’s doctoral thesis,
where the direct term (Hartree) is the Coulomb integral of the densities with one another, and the exchange term (Fock) is the Coulomb integral of the density matrix ρ(r,r′) = 3ρ(sin t − t cos t)/t³. The second line writes the same thing with the pair correlation function: multiplying the hole in which g↑↑ rises from 0 to 1 (the Fermi hole, which integrates to −1 electron) by the Coulomb 1/r and integrating gives exactly −(3/4π)kF (in atomic units) of Eq. (2.54) of the thesis (−4.7 eV per electron for the conduction electrons of Cu). The figure at the lower right of tab ④ of the simulator is this Fermi hole.
The answer to the question “does exchanging π mesons stabilize atomic nuclei?” is “yes, in the sense of (A).” It is the attractive potential produced by π exchange (−105 MeV from one-pion exchange for ⁴He in §3.4) that binds the nucleons, and this is a separate matter from the exchange energy of (B) due to swapping identical fermions (between nucleons too there is a Fock term of type (B), which accounts for part of the −136 MeV in nuclear Hartree–Fock calculations, but it is not the main actor). However, the nuclear force has the character of an “exchange force” in yet another sense. Before Yukawa, Heisenberg (1932) likened the force binding the neutron and the proton to the covalent bond of the hydrogen molecular ion H2+ (two protons bond by “exchanging” an electron), and formulated it as an “exchange force” in which the proton and the neutron exchange charge (Majorana 1933 used exchange of coordinates). That the name tags of the proton and the neutron are swapped by the exchange of a charged π meson (the nucleus view of simulator ①) is exactly this, and τ1·τ2 = 2Pτ − 1 (the isospin exchange operator) and σ1·σ2 = 2Pσ − 1 (the spin exchange operator) in Eq. (26) are the modern version of Heisenberg’s exchange force. Through the antisymmetry PrPσPτ = −1 this is tied to the exchange of coordinates Pr, and yields the property of the nuclear force that it is “attractive for pairs with even L, weak for odd L” (one of the reasons the nucleus saturates and has a constant density).
| “Exchange” | What is exchanged | What comes out | Examples |
|---|---|---|---|
| (A) Exchange of particles | The carrier of the force (a boson) | Potential V(r) ∝ e−μr/r (Eq. (25)) | Photon → Coulomb force, π → nuclear force, W → β decay, gluon → confinement |
| (B) Swapping of identical particles | The coordinates (labels) of two particles | Exchange energy (Fock term, Eq. (27)) | Hartree–Fock, LDA exchange −(3/4)(3/π)1/3∫ρ4/3, Hund’s rule |
| (C) Exchange force (Heisenberg–Majorana) | The charge, spin, and coordinates of two nucleons | Potential with the operators Pτ, Pσ, Pr | p ⇄ n by charged-π exchange, the (T, S) dependence of the nuclear force, nuclear saturation |
In tab ① of the simulator, in the “Atom” and “Electron in an atom” views, a blue blur (the position uncertainty Δx) and a bundle of orange arrows (the velocity uncertainty Δv) are drawn around the blue dot. Here we put together the physics behind them: ΔxΔp of the 1s state itself, why Nelson’s path has no velocity, and how Δx and Δv are fixed by the “resolution of the measurement” chosen with the slider, in that order.
The second moment of |ψ1s|² ∝ e−2Z′r/a0 is ⟨r²⟩ = 3a0²/Z′², so the standard deviation of the position along each axis is Δx = a0/Z′ (31 pm for He). From the virial theorem ⟨p²⟩ = 2m⟨T⟩ = Z′²ħ²/a0², the momentum gives Δpx = Z′ħ/√3a0. The product, independent of Z′, is
only 15% above the minimum ħ/2 — the 1s state is quite close to a Gaussian wave packet (a minimum-uncertainty state). On the right is the wave function in the momentum representation; the momentum distribution of the 1s state is not Gaussian but a Lorentzian-type function raised to the fourth power (width Z′ħ/a0). “The speed of the electron vrms = Z′αc” (Eq. (3)) is the rms of this distribution. In other words, if we do not measure, the uncertainty in the position of the 1s electron is the size of the atom itself, and the uncertainty in its velocity is Z′αc, and both are the “fluctuation” we saw in §2.
The dot in the simulator is a realization of the stochastic process dX = u dt + √(ħ/m) dW of Eq. (7). Over a short time Δt diffusion wins over drift, and ⟨|ΔX|²⟩ = 3(ħ/m)Δt. The “apparent speed” seen when the motion is advanced frame by frame is
One frame at the playback speed “Normal” is Δt = 4 × 0.0032 atomic units = 0.31 as, which gives vapp ≈ 3.3×10⁴ km/s (11% c, 9 times the vrms of He). If Δt is made 100 times smaller, the speed becomes 10 times larger, and in the limit it diverges. A Brownian path is nowhere differentiable, and in a picture where the position is definite at every instant, the velocity cannot be defined — this is the Nelsonian way the uncertainty relation shows up: from Δx ≈ √(ħΔt/m) and Δv ≈ √(ħ/mΔt) we get Δx·mΔv ≈ ħ. This is why the dot appears to move “at an enormous speed.”
So in the visualization the 1s distribution is approximated by a Gaussian (center O = the nucleus, standard deviation σ0 = a0/Z′ along each axis), and we consider the state just after the position of the point P in the figure (a Nelson sample) has been “measured” with resolution σm. The product of the prior distribution and the measurement is again a Gaussian,
For σm → ∞ (no measurement), Δx → σ0 and the center returns to the nucleus; for σm ≪ σ0, Δx → σm and the center coincides with the measured value P. A true Gaussian wave packet is a minimum-uncertainty state, so the spreads of the momentum and the velocity are
The orange arrows in the figure are projections of velocity vectors drawn from a three-dimensional Gaussian distribution with standard deviation Δv in each component (the length is proportional to the velocity). When ⟨T⟩ greatly exceeds the 1s binding energy 13.6 Z′² eV, the wave packet cannot stay in the atom. When Δx approaches the reduced Compton wavelength ƛC = 0.39 pm, Δv → c, and we enter the pair-creation regime of §6.2, where the picture of “the position of a single electron” breaks down. If the packet is free, its width spreads on the time scale τ = 2mΔx²/ħ of Eq. (22).
| Resolution σm/σ0 (He, σ0 = 31.4 pm) | Δx | Δv = ħ/2mΔx | Δv/c | ⟨T⟩ | Comparison with ⟨T⟩ = 38.7 eV of the 1s state |
|---|---|---|---|---|---|
| ∞ (no measurement) | 31.4 pm | 1,850 km/s | 0.62% | 29 eV | Gaussian approximation of the 1s state (the exact value is ħ/√3) |
| 1 | 22.2 pm | 2,610 km/s | 0.87% | 58 eV | Still bound |
| 0.4 (initial value) | 11.6 pm | 4,970 km/s | 1.66% | 210 eV | 5 times the binding energy: flies out |
| 0.1 | 3.1 pm | 18,600 km/s | 6.2% | 2.9 keV | The energy of X-rays |
| 0.02 | 0.63 pm | 92,000 km/s | 31% | 73 keV | Relativistic. 1.6 times ƛC |
“Is there an equation that describes the properties of a vibrating string?” — there is. A relativistic string is a field theory on a two-dimensional worldsheet (τ, σ), whose equation of motion is the wave equation, and which becomes infinitely many harmonic oscillators when quantized. If you know the |n⟩ and En = (n + ½)ħω of a harmonic oscillator, the spectrum of the string is just the sum of them. Here we proceed, within what can be followed at the master’s-course level, through action → wave equation → mode expansion → mass formula → classification of particles → hadron strings → minimum length (textbooks: Zwiebach [43], Polchinski [44], Ohta [45]). Below we write ħ = c = 1 and restore the units at the end.
Just as the action of a point particle is the length of its worldline, −m∫ds, the action of a string is the area of its worldsheet multiplied by the tension T (Nambu 1970, Goto 1971 [30]). Embedding the worldsheet in spacetime by Xμ(τ, σ) and writing Ẋ = ∂τX and X′ = ∂σX, we have
α′ is a constant called the “Regge slope,” which fixes the length ℓs of the string and its tension (as we will see in §11.5, α′ ≈ 0.9 GeV−2 for hadron strings and α′ ≈ ℓP² in superstring theory). The square root is awkward to handle, so we rewrite the action in the Polyakov form by introducing an auxiliary worldsheet metric hab. Using the freedom of choosing the worldsheet coordinates (reparametrization) and the freedom of Weyl transformations to fix hab = ηab (conformal gauge), the action becomes that of a free field, and
That is, each point of the string obeys the equation of a wave traveling at the speed of light, and as a remnant of the gauge fixing there are constraints (the Virasoro constraints). The constraints express the facts that “motion along the string is not physical (only transverse vibrations are physical)” and give the relation between energy and momentum (the mass formula). As for boundary conditions, a closed string is periodic, X(τ, σ + 2π) = X(τ, σ), and an open string has either X′ = 0 at the ends (Neumann: the ends move freely) or Xi = constant (Dirichlet: the ends are fixed on a D-brane). That the speed of waves on the string is the speed of light can also be seen from the tension T and the linear mass density T/c², as √(T/μ) = c.
The general solution of the wave equation is the sum of a right-moving wave and a left-moving wave. For a closed string, periodicity gives
xμ and pμ are the position and momentum of the center of mass, and αn and α̃n are the amplitudes of the n-th right-moving and left-moving modes (for an open string the left and right movers are reflected at the ends and merge into one, and only the cos nσ modes remain). Rewriting the canonical quantization [Xμ(σ), Pν(σ′)] = iημνδ(σ − σ′) in terms of the modes gives
This is the answer: for each mode n there is one harmonic oscillator with creation and annihilation operators an†, an (one for each transverse direction of spacetime), and its frequency is n times the fundamental (an overtone). As with a guitar, the state of the string is specified by “which overtones are excited, and how many quanta of each,” |N1, N2, …⟩ = Π(an†)Nn|0⟩, and the “level” corresponding to En = (n + ½)ħω of a harmonic oscillator is counted by the weighted number
Here D − 2 is the number of transverse directions (in light-cone gauge the two longitudinal directions are removed by the constraints). Adding up the ½ħω of every oscillator, the zero-point energy ½Σn = ½(1 + 2 + 3 + …) diverges, but with zeta-function regularization (or a Casimir-energy calculation that treats the end effects correctly) this sum becomes ζ(−1) = −1/12. “1 + 2 + 3 + … = −1/12” is not a number game; it is a physical quantity that actually appears in the mass formula of the string.
Imposing the zero mode of the Virasoro constraint (the Hamiltonian of the worldsheet) on the states links the momentum of the center of mass, p² = −M², to the level N. For the bosonic string (a string with only Xμ),
The condition for Lorentz symmetry to be preserved in light-cone gauge (Goddard–Goldstone–Rebbi–Thorn 1973 [31]) is a = 1, that is, D = 26. Then N = 0 is a tachyon with M² < 0 (a sign that the vacuum is unstable), and N = 1 is massless: for the open string it is α−1i|0⟩ (a transverse vector = a photon-type particle), and for the closed string α−1iα̃−1j|0⟩, whose symmetric traceless part is the graviton, whose antisymmetric part is the B field, and whose trace is the dilaton. This is the place where “quantizing a string makes gravity come out by itself” (Scherk–Schwarz [32], Yoneya 1974 [33]). The superstring adds fermions ψμ on the worldsheet (Ramond 1971, Neveu–Schwarz 1971); it has D = 10, the GSO projection [34] removes the tachyon and spacetime supersymmetry appears, and the zero-point energies cancel, so that
| Level N | Mass of the closed superstring MNc² (ℓs = ℓP) | Maximum spin (open: N + 1, closed: 2N + 2) | Contents |
|---|---|---|---|
| 0 | 0 | 1 (open), 2 (closed) | Photon, gluon, W, Z, graviton, and also quarks and leptons (their charges and masses are in §11.4) |
| 1 | 2.4×1019 GeV (44 µg) | 2 (open), 4 (closed) | The first “overtone.” 1015 times the LHC energy. Not yet observed |
| 2 | 3.5×1019 GeV | 3, 6 | The degeneracy of the overtones grows exponentially with N |
| 4 | 4.9×1019 GeV | 5, 10 | The mass goes as √N and the spin in proportion to N: J ≤ α′M² + 1 (the Regge trajectory of the string) |
Even in the ground state the string is not at rest. Adding up the zero-point motion of each mode, ⟨|Xn|²⟩ ∝ α′/n, gives ⟨ΔX²⟩ ∝ α′ Σ1/n ∝ α′ ln Nmax, which diverges logarithmically, so the string is “blurred” over a size of about ℓs (Karliner–Klebanov–Susskind 1988 [46]). The “Vibrating string” view of the simulator (ground state) draws this zero-point motion by making the amplitude of mode n proportional to 1/√n.
“The way a string vibrates determines the type of particle” — if we translate the metaphor familiar from documentaries into the language of superstring theory, it has four steps.
(1) Spin is determined by the type of vibration. The massless states of an open superstring are, in the NS sector (bosonic boundary conditions), ψi−1/2|0⟩ = a transverse vector (8 components), and in the R sector (fermionic boundary conditions), the ground state = a spinor (8 components). The former are gauge particles (spin 1) and the latter are fermions (spin ½). A closed string is the product of a left mover and a right mover, so (vector)⊗(vector) = 1 + 28 + 35 = dilaton + B field + graviton (spin 2), and (vector)⊗(spinor) = gravitino (spin 3/2). The ellipse that pulsates vertically and horizontally in the “Graviton” view of the simulator depicts the polarization of a symmetric traceless tensor (the + and × polarizations of gravitational waves).
(2) Charge is determined by where the ends of the string attach. The ends of an open string lie on D-branes (Polchinski 1995 [35]), and on a stack of N coincident branes each end has the Chan–Paton degree of freedom i, j = 1…N, “which brane of the stack it attaches to.” A string with both ends on the same stack is in the adjoint representation of U(N) (N² gauge particles: for 3 branes, 8 gluons + 1), and a string connecting stack a (Na branes) and stack b (Nb branes) is in the bifundamental representation (Na, N̄b):
Quarks are strings with ends on both the color stack and the weak stack, and leptons are strings with no end on the color stack — this is the picture of intersecting D-brane models (review by Blumenhagen–Cvetič–Langacker–Shiu [36]). The charges +2/3, −1/3, and −1 come from mixtures of the U(1)s (hypercharge).
(3) The number of generations is the number of times the branes intersect. When brane stacks a and b wind at angles around a six-dimensional compact space (three tori), chiral fermions are localized at the intersection points. The number of intersection points is a topological integer determined by the winding numbers (nr, mr)
and three generations are realized by choosing windings such that Iab = 3. Ibáñez–Marchesano–Rabadán (2001) [37] were the first to construct an intersecting-brane configuration that gives exactly three generations of SU(3)×SU(2)×U(1) and nothing else (no extra fermions). In the older construction that wraps a heterotic string on a Calabi–Yau manifold (Candelas–Horowitz–Strominger–Witten 1985 [38]), the number of generations is half the Euler number χ of the manifold.
(4) The mass is zero in the ground state, and geometry fixes the Yukawa couplings. A fermion living at a brane intersection is massless (N = 0). Mass is born from the Higgs condensation and the Yukawa coupling Y, which depends exponentially on the area A of the worldsheet triangle whose vertices are the three intersection points (left-handed quark, right-handed quark, Higgs), as
(worldsheet instantons, Cremades–Ibáñez–Marchesano 2003 [39]). A slightly different area changes the mass by orders of magnitude — the picture is one of explaining the hierarchy from the electron’s 0.5 MeV to the top quark’s 173 GeV by geometry.
| Particle | Spin | Type of string (vibration pattern) | Origin of the charge | Origin of the mass |
|---|---|---|---|---|
| Photon, W, Z | 1 | One lowest transverse vibration of an open string (a vector). Both ends on the same stack | No charge for U(1); W± and Z when it spans an SU(2) stack | 0 (W and Z through the Higgs condensation) |
| Gluon | 1 | Same as above. The ends are on different branes of the color stack | Color × anticolor: 3×3 − 1 = 8 | 0 |
| Graviton | 2 | Closed string. Lowest vibration, left × right (symmetric traceless) | None | 0 |
| Quark | ½ | Open string. Ground state of the R sector. Intersection of the color stack and the weak stack | (3, 2): end 1 is color, end 2 is weak isospin | 0 → Yukawa coupling ∝ e−A/2πα′ |
| Electron, ν | ½ | Open string. Ground state of the R sector. Intersection of the weak stack and a U(1) brane | No color end. e or ν according to the brane of the weak stack | 0 → Yukawa coupling (ν by a different mechanism) |
| Higgs | 0 | Open string. Vibration perpendicular to the branes (the position of the brane) | (1, 2) | The shape of the potential |
| Excited states | ≤ 2N + 2 | Overtones N ≥ 1 | Same as the ground state | 2√N ħc/ℓs ≈ 1019 GeV |
It is for hadrons that “the vibration and rotation of a string classify particles” has been confirmed by experiment. As we saw in §4.4, the lines of color force between a quark and an antiquark are bundled into a string of tension σ ≈ 0.9 GeV/fm (15 tonnes-force) (lattice QCD). Consider the classical solution in which this string rotates with its ends moving at nearly the speed of light. A point at distance l from the center has speed v = 2l/L, the energy of a line element is σ dl/√(1 − v²), and its angular momentum is σlv dl/√(1 − v²), so
The spin is proportional to the square of the mass, and the slope is fixed by the tension alone — this is the Regge trajectory (Chew–Frautschi 1961 [40]). Putting in σ = 0.9 GeV/fm × ħc = 0.178 GeV² gives α′ = 0.90 GeV−2, which agrees with the measured slope ≈ 0.9 GeV−2. Quantization and the spin of the quarks at the ends add an intercept α0, giving J = α0 + α′M²:
| Meson (ρ series, α0 = 0.5) | J | Measured M (PDG 2024) | String √((J − α0)/α′) | String length 2M/πσ | Rotation period πL/c |
|---|---|---|---|---|---|
| ρ(770) | 1 | 775 MeV | 747 MeV | 0.55 fm | 5.7×10−24 s |
| a2(1320) | 2 | 1,318 MeV | 1,294 MeV | 0.93 fm | 9.8×10−24 s |
| ρ3(1690) | 3 | 1,689 MeV | 1,670 MeV | 1.19 fm | 1.3×10−23 s |
| a4(1970) | 4 | 1,967 MeV | 1,976 MeV | 1.39 fm | 1.5×10−23 s |
| ρ5(2350) | 5 | 2,330 MeV | 2,241 MeV | 1.65 fm | 1.7×10−23 s |
| a6(2450) | 6 | 2,450 MeV | 2,478 MeV | 1.73 fm | 1.8×10−23 s |
| Δ(1232) → Δ(1950) → Δ(2420) → Δ(2950) (baryons, α0 ≈ 0.15) | 3/2 … 15/2 | 1,232 … 2,950 MeV | Same slope, 0.90 GeV−2 | — | — |
With a single parameter σ, mesons and baryons alike lie on the same slope — this is one piece of evidence that “a string binding the quarks really exists.” There are also “daughter” trajectories in which the string is made to vibrate radially (ρ(1450), ρ(1700), …; the same slope in the (n, M²) plane), and for heavy quarks (c, b) the trajectory is bent by the masses at the ends (Sonnenschein–Weissman 2014 [41], review [42]). The Veneziano amplitude of 1968 [29] and the string of Nambu and Goto [30] were theories made precisely to explain these Regge trajectories. The hadron string is still alive today as the lines of force of QCD, and string theory (which later led to superstring theory) is what you get by shrinking it to 10−20 of its size and reinterpreting it in 1974 as a theory of gravity. Both satisfy the same formula J ≤ α′M² + 1; the only difference is the value of α′ — 0.9 GeV−2 = (0.19 fm)², or ℓP² = (1.6×10−35 m)².
For a point particle, by putting in momentum Δp one can see details down to Δx ≈ ħ/Δp without limit (§1.1). For a string, the energy put in stretches the string: in high-energy scattering the size of the string grows in proportion to α′E (Gross–Mende 1988 [47]), and together with Heisenberg’s term
is the “uncertainty relation of the string” (Veneziano 1986, Amati–Ciafaloni–Veneziano 1989 [48], Konishi–Paffuti–Provero 1990 [49]). Yoneya (1989) [50] proposed a “spacetime uncertainty relation,” ΔTΔX ≳ α′, for the product of the uncertainties of time and space. T-duality says the same thing from another side. The mass formula of a closed string compactified on a circle of radius R is
Here the momentum mode n (going around the circle) and the winding mode w (winding w times around the circle) are interchanged, and the space of radius R and the space of radius α′/R are physically identical. Therefore a geometry with a radius smaller than ℓs is merely another name for a geometry larger than ℓs. Even without assuming string theory, if the energy E is concentrated within Δx, the Schwarzschild radius 2GE/c⁴ exceeds Δx and a black hole forms, so
The two curves of the string (Eq. (44)) and of gravity (Eq. (46)) form the “floor” drawn in the plot on the right of the “Vibrating string” view of the simulator. Whichever road is taken, structure finer than 10−35 m cannot be measured in principle — this is the basis of the intuition that “space may not be continuous but may have a smallest graduation of about the Planck length.” String theory (spacetime arises from the vibrational state of the string called the graviton), matrix models (BFSS 1997 [51]: the spatial coordinates appear as eigenvalues of matrices), and loop quantum gravity (area is quantized in units of ℓP²) are different formulations of this intuition, and all of them await experimental tests.
Mochizuki Research Group, Tokyo University of Science. To build a bridge from materials science and condensed-matter physics to particle physics, we start from the hydrogen and helium atoms and reread nuclei, nucleons, quarks, and the four forces through the hierarchy of “size” and “fluctuation.” The numerical values rest on PDG 2024, CODATA 2022, and the papers above, and the models used for the visualization (Nelson diffusion, Malfliet–Tjon, Cornell) are stated in §12. If you would like to check things by hand, go to the simulator (elementary-particles-simulator-en.html). Related: nuclear fusion simulator (in Japanese), Lie groups and Lie algebras, and the hydrogen atom. The formulas are typeset with the browser’s standard MathML (no external libraries or network communication are used at all).