CPT: Hypotheses, Content, and Limits

CPT follows only after a definite package of assumptions has been fixed. In one standard four-dimensional Wightman setting, proper-orthochronous Poincaré covariance, a vacuum and forward spectrum, physical Hilbert-space positivity, controlled fields and domains, and spacelike locality imply the existence of one antiunitary operator that reverses spacetime and sends each field to the appropriate adjoint or conjugate field. The theorem does not assert separate charge-conjugation, parity, or time-reversal symmetries. This page states a deliberately non-minimal sufficient package, explains the proof architecture, and checks a fixed convention on free scalar and Dirac fields.

Required background. Microcausality and Relativistic Compatibility supplies physical-observable locality, graded field locality on a common domain, and the gauge-fixed/physical distinction. Poincaré Covariance and the Spectrum Condition supplies the unitary proper-orthochronous action, invariant vacuum, forward joint spectrum, and the site convention $U(a)=e^{+iP\cdot a}$. The Dirac Field supplies the four-dimensional Clifford algebra, Dirac adjoint, action, spinor representation, and positive- and negative-frequency solution branches; it does not assume any discrete symmetry.

CPT is one antiunitary symmetry

Let $\Theta$ denote the combined CPT operator. Antiunitarity means

$$\Theta(cA+dB)\Theta^{-1} =c^*\Theta A\Theta^{-1} +d^*\Theta B\Theta^{-1}, \qquad \Theta i\Theta^{-1}=-i.$$

Fix the vacuum phase by $\Theta\Omega=\Omega$. For a finite field multiplet, the theorem’s conclusion can be written schematically as

$$\Theta F_r(x)\Theta^{-1} =\sum_s \mathsf C_{rs}F_s^\dagger(-x).$$

Here $\mathsf C$ contains the Lorentz-index intertwiners, species mixing, and conventional phases appropriate to the declared field basis. It is not the matrix of a separately assumed charge-conjugation symmetry. For an irreducible Lorentz field of type $(k/2,l/2)$, one common suppressed-index convention is

$$\Theta F^{(k,l)}(x)\Theta^{-1} =(-1)^l(\pm i)^f F^{(k,l)\dagger}(-x), \qquad f\equiv k+l\pmod 2,$$

where $f=0$ for an even field and $f=1$ for an odd field. The sign and species phases can change with convention; the antiunitary spacetime inversion and adjoint map are the invariant content. Greenberg 2006, §§ 5–6, arXiv manuscript pp. 10–13 (PDF) derives this field law from the Wightman-function reversal identity and records its particle and scattering interpretations.

The site translation convention provides an important sign check. Geometry requires

$$\Theta U(a)\Theta^{-1}=U(-a).$$

Because $\Theta$ is antiunitary,

$$\Theta e^{+iP\cdot a}\Theta^{-1} =e^{-i(\Theta P\Theta^{-1})\cdot a} =e^{-iP\cdot a},$$

and therefore

$$\boxed{\Theta P^\mu\Theta^{-1}=P^\mu.}$$

Thus $x\mapsto-x$ does not turn a positive-energy state into a negative-energy state. In the standard particle interpretation, energy and momentum remain the same, particle and antiparticle labels are exchanged, and spin projection and helicity are reversed. Antiunitarity is precisely what reconciles spacetime inversion with the forward spectrum.

A representative four-dimensional theorem map

The following is a sufficient physical package, not a claim that every row is logically minimal or that every CPT theorem uses the same primitives.

1. Spacetime and field class. Work in 3+1-dimensional Minkowski space with a finite family of fields, closed under adjoint, transforming in finite-dimensional representations of the proper-orthochronous Lorentz cover.
2. Distributions and domains. The fields are operator-valued tempered distributions acting on a common invariant dense domain, with products and adjoints controlled in the theorem’s stated sense.
3. Physical positivity. The fields act in a Hilbert space with a positive-definite physical inner product.
4. Covariance. A strongly continuous unitary representation implements the proper-orthochronous Poincaré group and the declared field transformation laws.
5. Vacuum. There is a cyclic invariant vacuum, taken to be unique up to phase in this representative package.
6. Spectrum. The joint translation spectrum lies in the closed forward cone.
7. Local ordering. Graded locality holds for spacelike-separated smeared fields. The proof actually needs the weaker vacuum statement called weak local commutativity in a neighborhood of a Jost point.

Jost’s formulation makes the last distinction exact: within the analytic Wightman package, weak local commutativity near a Jost point is necessary and sufficient for the CPT reversal identity Jost 1957, pp. 409–412 (open scan). Full operator locality is a physically transparent sufficient input, but it is stronger than that minimal order-reversal condition.

What each input contributes to the displayed four-dimensional CPT theorem.

Input	Role in the argument	If it is changed or removed
Four-dimensional point-local field setting	Places total inversion in the relevant proper complex-Lorentz continuation and fixes the field intertwiners.	Other dimensions, localization classes, or exchange topologies require a different theorem.
Tempered fields and common domain	Makes the correlation functions distributional boundary values and controls products, adjoints, and reconstruction.	Formal pointwise fields or uncontrolled products do not license the argument.
Positive physical Hilbert space	Supplies scalar-product conjugation and an antiunitary physical-space implementation.	An indefinite auxiliary space needs a separate physical quotient or algebraic formulation.
Poincaré covariance	Extends the real proper Lorentz action to the complex domain containing total inversion.	Lorentz covariance of only selected formulas is not enough.
Invariant cyclic vacuum	Defines the vacuum hierarchy and lets the reversal identity determine an operator on a dense set.	Degenerate vacua may be permuted; a chosen vacuum need not be fixed.
Forward spectrum	Produces the tube analyticity of Wightman boundary values.	Without it, the analytic route used here does not start.
Weak local commutativity	Reverses field order at real Jost configurations with the graded sign.	Covariance alone cannot supply the missing order reversal.

This package does not assume a mass gap, asymptotic particles, asymptotic completeness, or separate $C$, $P$, or $T$ invariance. Those are different inputs and conclusions.

The proof needs analyticity and order reversal

The full proof is a theorem in distribution theory and several complex variables. Its architecture is nevertheless physically legible:

1. Translation invariance rewrites an $n$-point vacuum distribution in coordinate differences.
2. The forward spectrum makes that distribution the boundary value of a function holomorphic in a tube domain.
3. Proper Lorentz covariance and the Bargmann–Hall–Wightman result extend the function to an enlarged tube acted on by the proper complex Lorentz group. In four dimensions, total inversion belongs to its identity component.
4. The enlarged domain contains real Jost configurations. Weak local commutativity reverses the field order there, including the required fermionic sign.
5. Analytic uniqueness propagates the identity through the complex domain and back to distributional boundary values.
6. Conjugation of the Hilbert-space scalar product and cyclic reconstruction realize the resulting identity by the antiunitary $\Theta$.

For a neutral Hermitian scalar with phase chosen to be one, the resulting distributional identity has the especially transparent form

$$\langle\Omega|\phi(x_1)\cdots\phi(x_n)|\Omega\rangle = \langle\Omega| \phi(-x_n)\cdots\phi(-x_1)|\Omega\rangle.$$

The reversed order is essential. Total coordinate inversion without it would put the two boundary values in opposite spectral tubes. More generally, antiunitarity gives the safe operator identity

$$\langle\Omega|A|\Omega\rangle = \langle\Omega|\Theta A\Theta^{-1}|\Omega\rangle^*;$$

taking the adjoint inside the complex conjugate reverses operator order. Greenberg traces the tube, complex-Lorentz, Jost-point, and order-reversal steps explicitly in Greenberg 2006, §§ 2–5, arXiv manuscript pp. 3–12 (PDF). Weinberg summarizes the same axiomatic route beside the local-field argument in Weinberg 1995, § 5.8, printed pp. 244–246.

A useful counterfactual shows why the conjunction matters. A covariant free or generalized-free construction can be arranged with unequal particle and antiparticle masses. Its weak local commutativity fails, so it is not a local CPT counterexample; it demonstrates that covariance of the Wightman fields alone does not imply CPT Greenberg 2006, Appendix D, arXiv manuscript p. 16 (PDF).

First application: a free scalar and Dirac field

We now choose one explicit convention. It is a calculation in free theory, not a proof of the general theorem. For both fields, $\Theta$ is antiunitary, $\Theta\Omega=\Omega$, and unshown species phases are set to one.

The scalar maps to its adjoint at the inverted point

For a complex scalar $\Phi$ choose

$$\Theta\Phi(x)\Theta^{-1}=\Phi^\dagger(-x).$$

A real scalar is the self-adjoint specialization

$$\Theta\phi(x)\Theta^{-1}=\phi(-x).$$

Because two derivatives remove the sign from $x\mapsto-x$,

$$\Theta\bigl[(\Box+m^2)\phi(x)\bigr]\Theta^{-1} = \bigl[(\Box+m^2)\phi\bigr](-x).$$

The free complex-scalar Lagrangian therefore maps to its value at $-x$, and the action is unchanged after the integration variable is inverted. In the mode expansion, $\Theta$ exchanges particle and antiparticle creation operators at the same four-momentum. A real scalar is self-conjugate.

The two-point check keeps both antiunitarity and the spectral convention visible. With

$$W_0(z) = \int\frac{\mathrm d^3\mathbf p}{(2\pi)^3 2E_{\mathbf p}} e^{-ip\cdot z}, \qquad p^0=E_{\mathbf p},$$

one has

$$W_0(-z)^*=W_0(z).$$

The future mass shell has not been sent to the past shell; complex conjugation compensates the coordinate inversion.

The Dirac map must declare its gamma basis

For the Dirac check, choose the chiral gamma basis, retain $\gamma_5=i\gamma^0\gamma^1\gamma^2\gamma^3$, and define $\psi^*:=(\psi^\dagger)^{\mathsf T}$. With a conventional overall phase, take

$$\boxed{ \Theta\psi(x)\Theta^{-1} =-\gamma_5\psi^*(-x) }.$$

The matrix written here is basis dependent; under a change of gamma basis, the intertwiner changes with the spinor components. This is why it should not be copied as a representation-independent identity or silently replaced by the separate charge-conjugation matrix $B$ from $\psi^c=B\psi^*$.

In the chiral basis, $\gamma_5$ is real and $\{\gamma_5,(\gamma^\mu)^*\}=0$. Writing $D=i\gamma^\mu\partial_\mu-m$ and $y=-x$ gives the direct equation check

$$\Theta[(D\psi)(x)]\Theta^{-1} = -\gamma_5[(D\psi)(y)]^*\big|_{y=-x}.$$

Thus the antiunitary image of $D\psi=0$ is its CPT-conjugate equation, which vanishes whenever the original does. At the density level, use vacuum Wick ordering to define the Hermitian free density

$$\mathopen{:}\!\mathcal L_D\!\mathclose{:} = \frac{i}{2}\mathopen{:}\!\bar\psi\gamma^\mu \overleftrightarrow{\partial_\mu}\psi\!\mathclose{:} -m\mathopen{:}\!\bar\psi\psi\!\mathclose{:}$$

This Wick-ordered density maps to $\mathopen{:}\!\mathcal L_D\!\mathclose{:}(-x)$ as an operator-valued distribution. Schwartz’s fixed chiral-basis calculation gives this combined transformation in Schwartz 2014, § 11.6.2, printed pp. 199–201, especially Eq. (11.90).

On free Fock states the same convention has the structure

$$\begin{aligned} \Theta b_s^\dagger(\mathbf p)\Omega &=e^{i\beta_s}d_{-s}^\dagger(\mathbf p)\Omega,\\ \Theta d_s^\dagger(\mathbf p)\Omega &=e^{i\widetilde\beta_s}b_{-s}^\dagger(\mathbf p)\Omega, \end{aligned}$$

with conventional unit phases $e^{i\beta_s}$ and $e^{i\widetilde\beta_s}$. Particle and antiparticle are exchanged, spin is reversed, and $p^\mu$ is unchanged. The scalar and Dirac calculations agree with the bounded local-field transformations in Weinberg 1995, § 5.8, printed pp. 244–246. Their agreement is a sign and convention check, not the analytic proof above.

What CPT licenses—and what it does not

CPT is not separate $C$, $P$, or $T$. A theory can violate each of those operations, and can violate $CP$, while satisfying the combined antiunitary theorem. Multiplying three separately valid symmetry operators is therefore not the general definition of $\Theta$.

Particle consequences need particle hypotheses. When isolated conjugate one-particle states exist, CPT intertwining of the full Poincaré representation gives equal masses and spin magnitudes; $\Theta P^\mu\Theta^{-1}=P^\mu$ is the momentum part of that statement. If a unitary $S$-matrix and the relevant asymptotic states exist, time reversal interchanges in and out and yields

$$\Theta S\Theta^{-1}=S^\dagger=S^{-1}.$$

This relates a process to the inverse process with particles replaced by antiparticles and spin projections reversed. It does not by itself equate an arbitrary partial rate with the rate of the merely $CP$-conjugate process. Asymptotic completeness is not an assumption of the field theorem and cannot be recovered from it. Greenberg 2006, § 6, arXiv manuscript p. 13 (PDF) gives the scattering statement and its required inverse-process interpretation.

The square is convention- and field-content-sensitive. One common four-dimensional convention makes $\Theta^2$ act trivially on even fields and as fermion parity on odd fields. Internal symmetries or different field bases can alter the displayed implementation. The theorem is not a universal claim that $\Theta^2=1$ on every field or sector.

A local Lagrangian check is not the general proof. A Hermitian local Lorentz-scalar interaction built from the standard fields transforms into its value at $-x$, so its integrated action is invariant. This is a powerful model-building check, but it does not replace the Wightman-domain, analyticity, and locality argument for a theory not presented by such a Lagrangian.

Changes of setting require separate theorems

Other dimensions and exchange topology. In odd spacetime dimension, $\det(-\mathbf1)=-1$, so the literal four-dimensional step connecting total inversion to the identity inside the proper complex Lorentz group is not available. In 2+1 dimensions, braided sectors can also replace the ordinary two-sign exchange structure. Neither observation is a violation of the four-dimensional theorem.

Gauge-fixed and ghost fields. Covariant gauges can use an indefinite auxiliary form and Grassmann-odd scalar ghosts. The positive physical Hilbert-space theorem is applied only after physical states and gauge-invariant observables have been identified, or it is replaced by a BRST or algebraic variant. Auxiliary fields are not physical counterexamples.

Nonlocal fields and charged dressings. A string-, cone-, or infinity-localized field may fail the compact point-local order relation used above. Local gauge-invariant observables can still obey a suitable CPT theorem, but the actual localization class must appear among its hypotheses.

Degenerate vacua and sectors. If several invariant vacua or phases exist, $\Theta$ may permute them. The theory can possess a CPT map even though one selected vacuum is not fixed by it. A theorem within one vacuum sector then needs additional assumptions.

Curvature, boundaries, states, and open evolution. Generic curved spacetimes need not admit a global map $x\mapsto-x$; boundaries and defects can break that geometry; thermal or finite-density states are not the vacuum used in the proof; and reduced nonunitary evolution need not admit the same antiunitary implementation. Locally covariant, boundary, thermal, and open-system statements are distinct results with their own geometric and state hypotheses. Greaves and Thomas give a modern comparison of Lagrangian, Wightman, and dimension-sensitive CPT statements in Greaves and Thomas 2014, §§ 1–4 and 9–10. For one precise curved-spacetime analogue, Hollands relates local operator-product-expansion data on the same metric with the space and time orientations reversed; it does not assert invariance of a chosen state under a global antipodal map Hollands 2004, Theorem 5.1 and interpretation, arXiv manuscript pp. 16–18 (PDF).

The historical development also matters for scope. Lüders and Pauli’s local-Lagrangian arguments and Jost’s later axiomatic proof are related but not identical theorem statements. Blum and Martínez de Velasco 2022, p. 1 and §§ 3–5 reconstruct that progression without retroactively treating the later Wightman package as the premise of the earlier proofs.

Check your understanding

Check 1: resolve the momentum-sign puzzle

Given $U(a)=e^{+iP\cdot a}$, $\Theta U(a)\Theta^{-1}=U(-a)$, and antiunitary $\Theta$, determine $\Theta P^\mu\Theta^{-1}$.

Answer

Antiunitarity conjugates $i$:

$$\Theta e^{+iP\cdot a}\Theta^{-1} =e^{-i(\Theta P\Theta^{-1})\cdot a}.$$

Comparison with $U(-a)=e^{-iP\cdot a}$ gives $\Theta P^\mu\Theta^{-1}=P^\mu$. Coordinate inversion and momentum reversal must not be identified when the implementing operator is antiunitary.

Check 2: test the scalar boundary value

Show that the future-shell free scalar function satisfies $W_0(-z)^*=W_0(z)$.

Answer

Using the real positive mass-shell measure,

$$W_0(-z)^* = \left[ \int\mathrm d\Pi_{\mathbf p}\,e^{+ip\cdot z} \right]^* = \int\mathrm d\Pi_{\mathbf p}\,e^{-ip\cdot z} =W_0(z).$$

No change from the future to the past energy shell occurred.

Check 3: identify the failed theorem input

Why does a Lorentz-covariant generalized free field with unequal particle and antiparticle masses not refute the CPT theorem stated here?

Answer

Its weak local commutativity at Jost points fails. It therefore lacks the order-reversal input required by this theorem package, even if selected field transformation laws remain Lorentz covariant.

Where to continue

• Compare the neighboring structural theorem: The Spin–Statistics Connection uses related analytic, locality, and positivity inputs but reaches a different conclusion.
• Study rigorous CPT variants: CPT Theorem Variants and Their Hypotheses develops the field, domain, dimensional, gauge, and algebraic theorem statements.
• Follow the shared analytic machinery: Analyticity, CPT, and Spin–Statistics develops complex-Lorentz continuation, Jost configurations, and the places where the two proofs diverge.
• Vary one assumption at a time: Structural Hypotheses and Failure Modes compares what survives when positivity, locality, Lorentz covariance, vacuum structure, a gap, or asymptotic completeness is changed.

References

• Blum, Alexander S., and Andrés Martínez de Velasco. “The Genesis of the CPT Theorem.” The European Physical Journal H 47 (2022), article 5. DOI.
• Greaves, Hilary, and Teruji Thomas. “On the CPT Theorem.” Studies in History and Philosophy of Modern Physics 45 (2014), 46–65. DOI. Open manuscript.
• Greenberg, O. W. “Why Is CPT Fundamental?” Foundations of Physics 36 (2006), 1535–1553. DOI. Open PDF.
• Hollands, Stefan. “A General PCT Theorem for the Operator Product Expansion in Curved Spacetime.” Communications in Mathematical Physics 244 (2004), 209–244. DOI. Open PDF.
• Jost, Res. “Eine Bemerkung zum CTP-Theorem.” Helvetica Physica Acta 30 (1957), 409–416. DOI and open scan.
• Lüders, Gerhart. “Proof of the TCP Theorem.” Annals of Physics 2, no. 1 (1957), 1–15. DOI.
• Schwartz, Matthew D. Quantum Field Theory and the Standard Model. Cambridge University Press, 2014. DOI.
• Weinberg, Steven. The Quantum Theory of Fields. Vol. I, Foundations. Cambridge University Press, 1995. DOI.