Point-limit triviality of local four-fermion interactions: the unanalyzed multi-field, regulator-class, gauge-consistent, and finite-mass cases

In 1958 Abrikosov, Galanin, Gorkov, Landau, Pomeranchuk, and Ter-Martirosyan asked whether a theory of strongly interacting fermions can be built from a local (point) four-fermion interaction. Modeling the point interaction as the limit of a smeared one with two cutoffs — a fermion momentum cutoff $\Lambda$ and a momentum-transfer cutoff $\lambda$ — and summing the dominant chain-of-closed-loop graphs through an integral equation for the vertex, they found that in the point limit the physical interaction, the product $\alpha\beta^2$ of the vertex form factor $\alpha$ and the square of the fermion wavefunction renormalization $\beta$, vanishes in every case they treated: all spin channels of a single fermion field, and two-field interactions. The paper explicitly left open the more complex cases — interactions of three or more distinct fields, including beta-decay-type mixed bilinears — together with the question of whether the conclusions are independent of the form factor (regulator), the finite-mass refinement near $p^2 \to 0$, and the use of gauge-consistent, Ward-identity-preserving nonlocal form factors for charged fields. The later literature establishes the physics by consensus — four-fermion operators are irrelevant at the Gaussian fixed point in four dimensions, and lattice studies confirm logarithmic triviality for Nambu–Jona-Lasinio-type models — but no proof or counterexample covers the unanalyzed classes, and the graph-class-limited character of the original summation was flagged as a methodological gap already in 1964 and never closed. The problem: decide whether any Hermitian Fermi-type local four-fermion interaction of the kind studied in the paper — including multi-field and mixed-bilinear couplings beyond the analyzed one- and two-field cases — retains a finite nonzero physical interaction $\alpha\beta^2$ in the point limit taken in the paper's scheme: $\lambda \to \infty$ and $\Lambda \to \infty$ with $\lambda^2/\ln(\Lambda/\lambda)$ arbitrarily small and the Hermitian bare couplings of the Lagrangian held fixed.

Background

In the mid-1950s, Landau and collaborators mounted an attack on the foundations of quantum field theory through the behavior of renormalized couplings as ultraviolet cutoffs are removed. The renormalized charge of a theory is the effective coupling measured at finite momenta after all short-distance structure has been integrated out; the Moscow school's 'zero-charge' finding was that in quantum electrodynamics and in point-coupled meson theories the renormalized charge is driven to zero as the cutoff goes to infinity, so that the point-interaction limit is a free (trivial) theory [Landau & Pomeranchuk 1955; Pomeranchuk, Sudakov & Ter-Martirosyan 1956]. Their technical device was to smear the point interaction into an extended one that vanishes when momenta exceed a cutoff, and to ask what survives the limit in which the smearing is removed. Abrikosov, Galanin, Gorkov, Landau, Pomeranchuk and Ter-Martirosyan (1958) extended this program to four-fermion interactions — contact interactions of the Fermi type, of the form $V = 2\pi^2 g_0 (\bar\psi O_j \psi)(\bar\psi O_j \psi)$, where each $\bar\psi O_j \psi$ is a fermion bilinear and $O_j$ runs over spin (and isospin) matrices, giving the scalar, pseudoscalar, vector, pseudovector, and tensor coupling channels. In four spacetime dimensions the coupling $g_0$ has mass dimension $-2$, so the interaction is strong at short distances and perturbation theory in $g_0$ fails exactly where the point limit probes. The question was whether the strong nuclear interaction could be described this way, with Fermi's weak interaction as the prototype.

The method uses two cutoffs: the fermion momenta are bounded by $\Lambda$, and the momentum transferred through the interaction vertex is bounded by $\lambda$, with $\lambda \ll \Lambda$; the point limit takes $\lambda \to \infty$ and $\Lambda \to \infty$ while keeping $\lambda/\Lambda \to 0$ and $\lambda^2/\ln(\Lambda/\lambda) \to 0$, which renders dominant a specific chain of closed fermion loops. Summing this chain converts the perturbation series into an integral equation for the vertex operator, in direct analogy with the logarithmic vacuum-polarization equation of quantum electrodynamics. Its solution yields two renormalized quantities: the vertex form factor $\alpha(p)$, the strength of the interaction at momentum transfer $p$ after summing the chain, and the fermion wavefunction renormalization $\beta$, the residue of the fermion propagator pole; the physical interaction is the product $\alpha\beta^2$.

The results: when the quadratically divergent loop integral $J_0 \sim \Lambda^2$ is present, the combination $\alpha\beta^2\Lambda^2$ stays finite, so $\alpha\beta^2 \sim 1/\Lambda^2 \to 0$. When the quadratic divergence is cancelled — for example by an oscillating form factor (a non-monotone cutoff function), or automatically in the vector and tensor channels where current conservation removes it — one finds $\alpha\beta^2 \sim 1/(p^2 L)$ with $L = \ln(\Lambda/\lambda)$, which again vanishes, only logarithmically. For two interacting fields with couplings $g_1, g_2, g_3$, the chain summation produces vertex denominators of the form $\Delta = 1 + 2gI + f^2 I^2$ with $I \simeq \Lambda^2 - p^2 L$; removing the quadratic and logarithmic infinities from these denominators — the only route to a nonvanishing limit — is possible only for an imaginary value of $g_3$, so for Hermitian Lagrangians the physical interaction vanishes in this case as well. For charged fields the authors note that introducing the form factor as a naive momentum-space cutoff contradicts Ward identities (the identities expressing gauge invariance that relate the vertex to the propagator; their violation would let a neutral particle acquire a charge) when $\lambda^2 \ll \Lambda^2$, and indicate that a gauge-covariant form factor built from the electromagnetic potential is needed, without carrying out the analysis. Their summary conclusion is deliberately restricted: 'In all of the simplest cases of local interaction considered up to the present, physical interaction between the fields is absent', while the weak interaction retains a perturbation theory valid at sufficiently small energies.

The paper itself lists what remains: the more complex multi-field cases, such as interactions of three different fields, which admit beta-decay-type mixed bilinears $(\bar\psi\psi)(\bar\chi\varphi)$ and were set aside as 'more unwieldy' with 'no apparent physical reasons to expect other results'; the dependence of the quadratically divergent integral, and hence possibly of the conclusions, on the form factor employed; the refinement of the near-threshold behavior at $p^2 \sim 0$ where the massless formula must be corrected to $\alpha\beta^2 \sim 1/(m^2 L)$; and the gauge-consistent treatment of charged fields.

Later developments frame rather than settle these items. Two-dimensional four-fermion theories (the Gross–Neveu model) are nontrivial and asymptotically free — free at short distances — showing that dimensionality, not fermionicity, is decisive [Gross & Neveu 1974]; the $d = 2+\epsilon$ analysis of the same integral equations by Makeenko, Ter-Martirosyan & Zamolodchikov [1976] found a nontrivial solution only for an SU(2)-symmetric interaction. Feinberg & Pais [1964] extended the sum-the-most-divergent-graphs approach ('peratization') to weak-interaction field theories and explicitly cautioned that conclusions from a restricted graph class cannot settle the fate of a Fermi-type theory. On the lattice, the Nambu–Jona-Lasinio model — the prototype four-fermion theory, in which the interaction dynamically generates a fermion mass by spontaneously breaking chiral symmetry (the symmetry between left- and right-handed fermions) — and its gauged variants were confirmed to be logarithmically trivial, their continuum limits realizing the zero-charge scenario [Hands 1998; Kim, Kogut & Lombardo 2002; Kogut & Strouthos 2005]. A parallel program kept the interaction fundamentally nonlocal rather than taking the point limit, constructing gauge-invariant nonlocal weak-interaction theories [Efimov & Seltser 1971]. The modern renormalization-group view — at the Gaussian fixed point (the free theory) in four dimensions, four-fermion operators are irrelevant, meaning their coupling fades away at long distances as the ultraviolet cutoff is removed — underlies both the triviality consensus and the electroweak completion of Fermi theory [Buoninfante 2025], while the asymptotic-safety program (the search for interacting theories whose couplings approach a finite fixed point at short distances instead of vanishing) has revived the question of whether four-fermion couplings can sit at a non-Gaussian ultraviolet fixed point [de Brito, Eichhorn & Schiffer 2021]. What does not exist anywhere in this literature is the completion of the original program: a proof that the vanishing extends to the unanalyzed classes, or a construction showing it does not.

Problem Statement

Decide whether any Hermitian Fermi-type local four-fermion interaction of the kind studied by Abrikosov, Galanin, Gorkov, Landau, Pomeranchuk, and Ter-Martirosyan (1958) — including couplings beyond the single-field and two-field cases they analyzed — retains a finite, nonzero physical interaction in the point limit, completing their program for the classes they did not analyze. Those classes are the four further directions enumerated below — the multi-field and mixed-bilinear coupling classes, the form-factor (regulator) class, the gauge-consistent treatment of charged fields, and the finite-mass refinement near $p^2 \to 0$. Concretely: for the smeared interaction with fermion-momentum cutoff $\Lambda$ and momentum-transfer cutoff $\lambda$, in the paper's limiting scheme — $\lambda \to \infty$ and $\Lambda \to \infty$ with $\lambda^2/\ln(\Lambda/\lambda)$ arbitrarily small, the interaction smeared so that it vanishes above the cutoffs, and the Hermitian bare couplings of the Lagrangian held fixed — determine the behavior of the physical interaction $\alpha\beta^2$ (vertex form factor times the square of the fermion wavefunction renormalization). Read the point limit, in every class except the finite-mass class, at fixed momentum transfer $p^2 > 0$ and fixed fermion mass $m$ as the cutoffs are removed, $L = \ln(\Lambda/\lambda) \to \infty$; the finite-mass class is read instead as the joint limit $p^2 \to 0$ together with $L \to \infty$, with the order of limits stated. Call a choice admissible when it stays within the paper's framework: field content, Hermitian couplings held fixed, and a form factor (smearing regulator) consistent with the two-cutoff scheme set out above, preserving the Ward identities where the charged-field class requires it. Over these admissible choices the question is whether $\alpha\beta^2 \to 0$ for every one of them throughout all four classes, or whether some admissible choice, within one of the classes, keeps $\alpha\beta^2$ finite and nonzero — or makes it diverge or fail to have a limit. Vanishing must hold for every admissible choice in all four classes, whereas a single admissible choice in any one class with $\alpha\beta^2$ finite and nonzero, or divergent, or without a limit, settles the governing question in the affirmative.

The cases already settled in the source paper all yield vanishing: the single-field case in every spin channel (scalar, pseudoscalar, vector, pseudovector, and tensor), with $\alpha\beta^2 \sim 1/\Lambda^2$ when the quadratic divergence $J_0 \sim \Lambda^2$ is present and $\alpha\beta^2 \sim 1/(p^2 L)$, $L = \ln(\Lambda/\lambda)$, when it is cancelled or absent; and the two-field case of two distinct neutral fermion fields with Hermitian couplings $g_1, g_2, g_3$, where removing the quadratic and logarithmic infinities from the vertex denominators $\Delta = 1 + 2gI + f^2 I^2$ (with $I \simeq \Lambda^2 - p^2 L$ the loop integral entering the chain summation, $g = \tfrac{1}{2}(g_1 + g_2)$, and $f^2 = g_1 g_2 - g_3^2$) is possible only for an imaginary value of $g_3$ and is therefore excluded for Hermitian Lagrangians. The paper itself noted four further directions that its analysis left open:

1. Multi-field and mixed-channel cases. Interactions of three or more distinct fermion fields, including beta-decay-type mixed bilinears such as $(\bar\psi\psi)(\bar\chi\varphi)$, which the paper set aside as 'more unwieldy' with 'no apparent physical reasons to expect other results'; whether the vanishing persists for all Hermitian coupling sets, or whether some real-coupling multi-field structure evades the constraint that forced imaginary couplings in the two-field case, is open.

2. Form-factor (regulator) dependence. The paper noted that the quadratically divergent integral, and hence possibly the conclusions, may significantly depend on the form factor employed, worked out the oscillating-form-factor case, and expressed the view that the physical results should not depend on the form of the cutoff factor; whether the point-limit outcome in the paper's scheme is the same vanishing for other admissible form-factor families — including other non-monotone cutoff behaviors and gauge-covariant form factors — is open.

3. Gauge-consistent treatment of charged fields. The paper notes that introducing the form factor as a naive momentum-space cutoff contradicts Ward identities (the identities expressing gauge invariance, relating the interaction vertex to the propagator) when $\lambda^2 \ll \Lambda^2$ — in particular, a neutral particle acquires a charge — and indicates that a gauge-covariant form factor built from the electromagnetic potential is needed, without carrying out the analysis; the point-limit consequences in a Ward-identity-preserving setting are undetermined.

4. Finite-mass refinement. The near-threshold behavior at small momentum transfer, $p^2 \sim 0$, where a footnote of the paper corrects the massless formula to $\alpha\beta^2 \sim 1/(m^2 L)$ (with $m$ the fermion mass); the behavior under the combined limits $p^2 \to 0$ and $L \to \infty$ is not worked out.

Either answer settles the source paper's governing question: whether a strongly interacting fermion theory can be constructed on the basis of local four-fermion interactions.

References

1. A. A. Abrikosov, A. D. Galanin, L. P. Gorkov, L. D. Landau, I. Ya. Pomeranchuk, K. A. Ter-Martirosyan, 'Possibility of Formulation of a Theory of Strongly Interacting Fermions', Physical Review 111, 321-328 (1958). DOI: 10.1103/PhysRev.111.321, https://doi.org/10.1103/PhysRev.111.321
2. L. D. Landau, I. Ya. Pomeranchuk, 'On Point Interactions in Quantum Electrodynamics' [Doklady Akademii Nauk SSSR 102, 489 (1955)], reprinted in D. ter Haar (ed.), Collected Papers of L. D. Landau, pp. 654-658, Pergamon (1965). DOI: 10.1016/B978-0-08-010586-4.50091-2, https://doi.org/10.1016/B978-0-08-010586-4.50091-2
3. I. Ya. Pomeranchuk, V. V. Sudakov, K. A. Ter-Martirosyan, 'Vanishing of Renormalized Charges in Field Theories with Point Interaction', Physical Review 103, 784-802 (1956). DOI: 10.1103/PhysRev.103.784, https://doi.org/10.1103/PhysRev.103.784
4. G. Feinberg, A. Pais, 'A Field Theory of Weak Interactions. II', Physical Review 133, B477-B486 (1964). DOI: 10.1103/PhysRev.133.B477, https://doi.org/10.1103/PhysRev.133.B477
5. David J. Gross, André Neveu, 'Dynamical symmetry breaking in asymptotically free field theories', Physical Review D 10, 3235-3253 (1974). DOI: 10.1103/PhysRevD.10.3235, https://doi.org/10.1103/PhysRevD.10.3235
6. Yu. M. Makeenko, K. A. Ter-Martirosyan, A. B. Zamolodchikov, 'On the Theory of the Direct Four-Fermion Interaction', Zhurnal Eksperimental'noi i Teoreticheskoi Fiziki 71, 24-45 (1976) [Soviet Physics JETP 44, 12 (1976)]; English version in New Pathways in High-Energy Physics I, pp. 267-302, Plenum Press/Springer, Boston (1976). DOI: 10.1007/978-1-4684-2922-0_10, https://doi.org/10.1007/978-1-4684-2922-0_10
7. David J. E. Callaway, 'Triviality pursuit: Can elementary scalar particles exist?', Physics Reports 167, 241-320 (1988). DOI: 10.1016/0370-1573(88)90008-7, https://doi.org/10.1016/0370-1573(88)90008-7
8. T. Jaroszewicz, P. S. Kurzepa, 'Random Walk Representations and Four Fermion Interactions', Annals of Physics 230, 52-100 (1994). DOI: 10.1006/aphy.1994.1017, https://doi.org/10.1006/aphy.1994.1017
9. Simon Hands, 'Log triviality in the Nambu-Jona-Lasinio model', Nuclear Physics B (Proceedings Supplements) 63, 673-675 (1998). DOI: 10.1016/S0920-5632(97)00868-2, arXiv:hep-lat/9709041, https://doi.org/10.1016/S0920-5632(97)00868-2
10. S. Kim, John B. Kogut, Maria-Paola Lombardo, 'Gauged Nambu-Jona-Lasinio studies of the triviality of quantum electrodynamics', Physical Review D 65, 054015 (2002). DOI: 10.1103/PhysRevD.65.054015, arXiv:hep-lat/0112009, https://doi.org/10.1103/PhysRevD.65.054015
11. John B. Kogut, Costas G. Strouthos, 'The Logarithmic triviality of compact QED coupled to a four Fermi interaction', Physical Review D 71, 094012 (2005). DOI: 10.1103/PhysRevD.71.094012, arXiv:hep-lat/0501003, https://doi.org/10.1103/PhysRevD.71.094012
12. G. V. Efimov, Sh. Z. Seltser, 'Gauge invariant nonlocal theory of the weak interactions', Annals of Physics 67, 124-144 (1971). DOI: 10.1016/0003-4916(71)90007-8, https://doi.org/10.1016/0003-4916(71)90007-8
13. Gustavo P. de Brito, Astrid Eichhorn, Marc Schiffer, 'Light charged fermions in quantum gravity', Physics Letters B 815, 136128 (2021). DOI: 10.1016/j.physletb.2021.136128, https://doi.org/10.1016/j.physletb.2021.136128
14. Luca Buoninfante, 'Strict renormalizability as a paradigm for fundamental physics', Journal of High Energy Physics 2025(7), 175 (2025). DOI: 10.1007/JHEP07(2025)175, arXiv:2504.05900, https://doi.org/10.1007/JHEP07(2025)175
15. Diego Buccio, John F. Donoghue, Gabriel Menezes, Roberto Percacci, 'Renormalization and running in the 2D CP(1) model', Journal of High Energy Physics 2025(2), 146 (2025). DOI: 10.1007/JHEP02(2025)146, arXiv:2408.13142, https://doi.org/10.1007/JHEP02(2025)146