What an Interacting Lagrangian Does and Does Not Specify

An interacting local Lagrangian specifies proposed fields, couplings, classical equations, candidate symmetries, and a classical stability test. It does not, by itself, choose a quantum construction, representation or measure, state, composite-operator prescription, renormalized parameters, observable class, regulator-removal procedure, or standard of evidence. A bounded-below scalar quartic potential makes this distinction especially clear: it supplies sound classical input, but not a defined continuum quantum field theory.

Required background. The Action Principle and Field Equations supplies the variational derivation and boundary hypothesis used below.

Helpful background. Wick’s Theorem and Free Gaussian Factorization supplies the free-state expansion used only to contrast formal perturbative coefficients with an exact interacting definition.

The quartic action and its classical content

In $d$ spacetime dimensions, consider the candidate bare density

$$\mathcal L_0 =\frac12\,\partial_\mu\phi\,\partial^\mu\phi -\frac12m_0^2\phi^2 -\frac{\lambda_0}{4!}\phi^4, \qquad \lambda_0>0.$$

Varying the action with compactly supported variations, or with a boundary condition that removes the surface term, gives

$$(\Box+m_0^2)\phi+\frac{\lambda_0}{3!}\phi^3=0.$$

The density is invariant under the discrete transformation $\phi\mapsto-\phi$. Power counting at the classical level gives

$$[\phi]=\frac{d-2}{2}, \qquad [m_0]=1, \qquad [\lambda_0]=4-d.$$

These are real pieces of information: the field equation, candidate symmetry, and engineering dimensions follow from the written density Schwartz 2014, §§ 3.1–3.3, printed pp. 29–34, and Appendix A.1, printed pp. 815–816.

With canonical momentum $\pi=\dot\phi$, the formal classical Hamiltonian density is

$$\mathcal H_0 =\frac12\pi^2+\frac12\lvert\boldsymbol\nabla\phi\rvert^2+V(\phi), \qquad V(\phi)=\frac12m_0^2\phi^2+\frac{\lambda_0}{4!}\phi^4.$$

Because $\lambda_0>0$, the quartic term dominates at large $\lvert\phi\rvert$, so $V$ is bounded below and coercive. It is not a bounded function. If $m_0^2\geq0$, its minimum is at $\phi=0$. If $m_0^2<0$, the two classical minima are

$$\phi_\star=\pm\sqrt{-\frac{6m_0^2}{\lambda_0}}.$$

This proves a classical large-field stability property. It does not choose a quantum vacuum or phase, prove that a Hamiltonian operator is self-adjoint and bounded below, define products such as $\phi^4(x)$, or establish an infinite-volume continuum limit.

What quantization must add

Continuum fields are distributions, so coincident products are not ordinary pointwise multiplication. Defining time-ordered and composite products requires a prescription and renormalization freedom beyond the classical polynomial Brunetti and Fredenhagen 2000, §§ 5.1–5.2, manuscript printed pp. 21–25 (PDF). More generally, every interacting claim must say which mathematical object is being constructed.

Data not fixed by the local interaction density alone

Required choice	Examples	Why it matters
Quantum construction	Hamiltonian and domains, operator algebra, Euclidean measure, lattice family, or formal perturbative algebra	These are different object categories; the same polynomial notation does not make them identical.
Regulator or formal meaning	Finite lattice, momentum cutoff, finite volume, mode truncation, switching function, or formal series	It makes intermediate products and calculations meaningful and states what is not yet a continuum claim.
State and boundary data	Vacuum sector, initial density operator, thermal state, Euclidean boundary condition, or in/out prescription	Correlation functions and particle interpretations depend on the state and boundary question.
Composite observables	Renormalized field products, currents, stress tensor, detector observables, or local algebra	The action's monomials do not automatically define continuum operators at coincident points.
Renormalized inputs	Mass, coupling, field normalization, scale, scheme, and matching conditions	Bare coefficients are regulator coordinates, not automatically measured parameters.
Limits and convergence	Infinite volume, regulator removal, continuum scaling, adiabatic limit, and the topology or observable class	A limit is a new mathematical or evidential claim, not an operation supplied by the density.
Validation and error control	Perturbative remainder, numerical uncertainty, symmetry tests, universality evidence, or existence proof	It determines the strength and domain of the final physical statement.

Even a manifest classical symmetry is only a candidate quantum symmetry until the construction, measure or representation, state, and renormalized operators are checked. Anomalies, spontaneous symmetry breaking, and phase selection are separate questions.

A finite-regulator test

The distinction can be made completely concrete before taking any continuum limit. Put the scalar on a finite Euclidean lattice with spacing $a$, finite set of sites $\Lambda_L$, specified boundary conditions, and real variables $\phi_n$. One possible regulated action is

$$S_{a,L}(\phi) =a^d\sum_{n\in\Lambda_L} \left[ \frac12\sum_{\mu=1}^{d} \left(\frac{\phi_{n+\hat\mu}-\phi_n}{a}\right)^2 +\frac12m_0^2\phi_n^2 +\frac{\lambda_0}{4!}\phi_n^4 \right].$$

For finitely many sites and $\lambda_0>0$, the positive quartic term controls every large-field direction. Therefore

$$Z_{a,L}[J] =\int_{\mathbb R^{\lvert\Lambda_L\rvert}} \left(\prod_{n\in\Lambda_L}\mathrm d\phi_n\right) \exp\!\left[ -S_{a,L}(\phi)+a^d\sum_{n\in\Lambda_L}J_n\phi_n \right]$$

is an ordinary finite-dimensional integral for finite real sources. Normalizing the source-free weight by $Z_{a,L}[0]$ defines a regulated Euclidean probability measure, while $Z_{a,L}[J]/Z_{a,L}[0]$ generates its moments. This is already more than the bare Lorentzian symbol $\int\mathcal D\phi\,e^{iS}$: the variables, measure, volume, regulator, and convergence are explicit.

It is still not the continuum QFT. One must identify renormalized observables $O_{a,L}$, tune the bare data along a declared trajectory, control finite-volume effects, and establish a limit such as

$$\lim_{\substack{a\to0\\L\to\infty}} \bigl\langle O_{a,L}\bigr\rangle_{a,L},$$

including the order and sense of convergence. Different phases may require different state-selection limits. Universality requires showing that distinct regulator descriptions approach the same declared continuum data, not merely that they share the same classical polynomial.

The constructive $\phi^4_2$ example makes the missing labor visible: a rigorous construction uses Wick ordering, cutoffs, a controlled Hamiltonian, cutoff and volume limits, a new representation obtained from the limiting state, and verification of QFT properties Summers 2016, § 2, printed pp. 5–7 (PDF). Its existence is a theorem about that model and dimension, not a consequence of writing $\lambda\phi^4$ in any dimension.

Formal perturbation is a different claim

A perturbative treatment first declares a free covariance or state and writes a formal expansion such as

$$\langle O\rangle \sim\sum_{n=0}^{\infty}\lambda_R^n\, c_n[O;\mu,\text{scheme}].$$

Wick’s theorem organizes the free Gaussian contractions used to calculate the coefficients. Renormalization then defines the time-ordered products, composite operators, and input conditions. A prediction truncated at order $N$ can be physically controlled when its parameters, scale, scheme, observable, regime, and uncertainty are stated.

That statement is not the same as convergence of the infinite series or construction of a nonperturbative continuum model. Formal perturbative algebraic QFT makes this distinction explicit by treating interacting observables as formal series while still imposing locality and causal-factorization conditions Fredenhagen and Rejzner 2015, §§ 5–7, arXiv v2 printed pp. 21–34 (PDF). Conversely, lack of a convergence theorem does not make every finite-order prediction meaningless; it fixes the claim’s scope.

Match the claim to the supplied data

What progressively stronger interaction claims require

Claim	Licensed by	Not licensed by
Classical field equation, symmetry, and stability test	The differentiable classical action plus its boundary and sign assumptions	The interaction monomial without a specified action or sign
Finite-regulator quantum model	An explicit regulated space, measure or operator construction, state, and domains	A formal path-integral symbol
Renormalized perturbative prediction	Renormalization conditions, measured inputs, observable definition, perturbative order, and uncertainty	Finite diagrams with unspecified scheme or physical inputs
Effective-theory prediction	Cutoff and regime, operator basis, matching, power counting, and truncation error	The demand that every useful cutoff be removed
Continuum or exact model	A defined limiting or nonperturbative object with the claimed observables and structural properties established	Classical boundedness, a few stable cutoff values, or all formal coefficients alone

Common failure modes

Bounded below is not bounded and is not existence. The quartic potential grows to positive infinity. That controls classical large-field behavior; it does not prove quantum self-adjointness or continuum convergence.

Bare coefficients are not automatically observables. Their regulator dependence can be essential. Physical masses, couplings, matrix elements, or dimensionless ratios require renormalized definitions and conditions.

A formal functional integral is not automatically a measure. Oscillatory Lorentzian notation and a normalized finite-dimensional Euclidean integral are different objects. A continuum Euclidean measure is a further construction.

The same local density does not fix the state or representation. Vacuum phases, thermal states, boundary conditions, and inequivalent representations can lead to different physical questions.

Engineering marginality is not a continuum theorem. The relation $[\lambda_0]=0$ in $d=4$ identifies a power-counting fact. It does not prove regulator removal, ultraviolet completion, universality, or nontriviality.

Check your understanding

1. Derive the field equation and coupling dimension.

Answer

Varying the kinetic term gives $-\Box\phi$ after integration by parts, while the potential derivative is $m_0^2\phi+\lambda_0\phi^3/3!$. The Euler–Lagrange equation is therefore $(\Box+m_0^2)\phi+\lambda_0\phi^3/3!=0$. Since the action is dimensionless, $[\phi]=(d-2)/2$ and $[\lambda_0]=d-4[\phi]=4-d$.

2. State exactly what $\lambda_0>0$ proves.

Answer

For the displayed real classical model, it makes the polynomial potential bounded below and coercive at large field. It does not define a quantum state, an operator Hamiltonian, a composite field, renormalized parameters, or a continuum limit.

3. A paper writes $Z[J]=\int\mathcal D\phi\,e^{iS+i\int J\phi}$. What should you ask next?

Answer

Ask whether the expression is formal, regulated, perturbative, Euclidean, or a genuine measure; what state or boundary prescription it represents; how products and observables are defined; which parameters are renormalized inputs; and what limit or approximation claim is made. The next page supplies the regulator-and-limit checklist.

Where to continue

• Regulators, Cutoffs, and Continuum Limits turns the missing regulator, tuning, convergence, universality, and error data into an explicit continuum-claim checklist.
• Ultraviolet Sensitivity and the Renormalization Problem develops why the bare interaction creates scale-dependent renormalization questions.
• Renormalization and Effective Field Theory develops running, matching, power counting, renormalized composite operators, and controlled effective descriptions.
• Existence, Construction, Reconstruction, and Continuum Claims gives theorem-level language for exact construction claims.
• Classical Observables and Poisson Factorization is a later rigorous continuation for the classical observable structure; it is not a prerequisite for this physical test.

References

• Brunetti, Romeo, and Klaus Fredenhagen. “Microlocal Analysis and Interacting Quantum Field Theories: Renormalization on Physical Backgrounds.” Communications in Mathematical Physics 208 (2000): 623–661. DOI. Open PDF.
• Fredenhagen, Klaus, and Katarzyna Rejzner. “Perturbative Algebraic Quantum Field Theory.” In Mathematical Aspects of Quantum Field Theories, edited by Damien Calaque and Thomas Strobl, 17–55. Mathematical Physics Studies. Springer, 2015. DOI. Open PDF, arXiv:1208.1428v2, revised 2013.
• Schwartz, Matthew D. Quantum Field Theory and the Standard Model. Cambridge University Press, 2014. DOI.
• Summers, Stephen J. “A Perspective on Constructive Quantum Field Theory.” arXiv:1203.3991v2 [math-ph], revised 2016 (originally submitted 2012). Stable record. Open PDF, v2.