Skip to content

Relativistic Phase Space

The invariant momentum measure for a free particle is proportional to d3p/(2Ep)d^3p/(2E_{\mathbf p}), not d3pd^3p alone. The energy denominator is the Jacobian of restricting four-momentum to its mass shell. This page derives that one-particle measure and explains why it appears in state normalization; it does not identify it with a full six-dimensional classical phase-space volume or with a scattering probability by itself.

Required background. The Energy–Momentum Relation supplies the mass shell, and Four-Vectors supplies momentum transformations. The calculation also uses delta distributions at simple roots.

Restricting the invariant four-momentum measure

Section titled “Restricting the invariant four-momentum measure”

First retain constants. Write p0=E/cp^0=E/c and p2=(p0)2−p2p^2=(p^0)^2-\mathbf p^2. For a positive-energy particle of mass mm, consider a scalar test function FF and the integral

I[F]=∫d4p θ(p0) δ(p2−m2c2)F(p).I[F]=\int d^4p\,\theta(p^0)\, \delta(p^2-m^2c^2)F(p).

The four-volume measure and delta argument are Lorentz invariant. The step function is invariant on the massive or nonzero massless shell under proper orthochronous Lorentz transformations. It selects a time orientation; time reflection does not preserve that choice.

Let ωp=p2+m2c2=Ep/c\omega_{\mathbf p}=\sqrt{\mathbf p^2+m^2c^2}=E_{\mathbf p}/c. The roots in p0p^0 are ±ωp\pm\omega_{\mathbf p}, so

δ((p0)2−ωp2)=δ(p0−ωp)+δ(p0+ωp)2ωp.\delta\big((p^0)^2-\omega_{\mathbf p}^2\big) =\frac{\delta(p^0-\omega_{\mathbf p})+ \delta(p^0+\omega_{\mathbf p})}{2\omega_{\mathbf p}}.

Performing the p0p^0 integral gives

I[F]=∫c d3p2EpF(Ep/c,p).I[F]=\int\frac{c\,d^3p}{2E_{\mathbf p}} F(E_{\mathbf p}/c,\mathbf p).

The factor cc occurs because the integration coordinate was p0p^0, not EE. Multiplying an invariant measure by a fixed constant preserves invariance, so d3p/(2Ep)d^3p/(2E_{\mathbf p}) is also invariant. Numerical normalization must still be declared when using the measure in Fourier transforms.

For a passive boost along xx,

px′=γ(px−vEp/c2),py′=py,pz′=pz.p'_x=\gamma(p_x-vE_{\mathbf p}/c^2), \qquad p'_y=p_y,\quad p'_z=p_z.

On the mass shell, ∂Ep/∂px=c2px/Ep\partial E_{\mathbf p}/\partial p_x=c^2p_x/E_{\mathbf p}. The Jacobian is triangular in the transverse rows, giving

det⁡∂p′∂p=γ(1−vpxEp)=Ep′Ep.\det\frac{\partial\mathbf p'}{\partial\mathbf p} =\gamma\left(1-\frac{vp_x}{E_{\mathbf p}}\right) =\frac{E_{\mathbf p'}}{E_{\mathbf p}}.

It follows immediately that d3p′/Ep′=d3p/Epd^3p'/E_{\mathbf p'}=d^3p/E_{\mathbf p}. This check also shows why boosting a distribution while keeping its momentum volume element fixed gives the wrong normalization.

For the remainder of this page use c=ℏ=1c=\hbar=1 and Ep=p2+m2E_{\mathbf p}=\sqrt{\mathbf p^2+m^2}. Define

dΠp=d3p(2π)3 2Ep.d\Pi_p=\frac{d^3p}{(2\pi)^3\,2E_{\mathbf p}}.

A convenient scalar momentum basis obeys

⟨p′∣p⟩=(2π)3 2Epδ(3)(p′−p),I1=∫dΠp ∣p⟩⟨p∣.\begin{aligned} \langle\mathbf p'|\mathbf p\rangle &=(2\pi)^3\,2E_{\mathbf p} \delta^{(3)}(\mathbf p'-\mathbf p),\\ I_1&=\int d\Pi_p\,|\mathbf p\rangle\langle\mathbf p|. \end{aligned}

The right-hand side of the overlap is invariant because the delta distribution transforms with the inverse momentum Jacobian. A normalized wave packet ∣f⟩=∫dΠpf(p)∣p⟩|f\rangle=\int d\Pi_p f(\mathbf p)|\mathbf p\rangle has

⟨f∣f⟩=∫dΠp ∣f(p)∣2=1.\langle f|f\rangle=\int d\Pi_p\,|f(\mathbf p)|^2=1.

For a spin-zero particle the amplitude can transform by pullback, f′(p′)=f(p)f'(p')=f(p), with no additional square-root Jacobian. In a basis normalized instead by δ(3)(p′−p)\delta^{(3)}(\mathbf p'-\mathbf p), such factors move into the state transformation. These are equivalent descriptions after consistent rescaling. Spin adds a momentum-dependent rotation on the spin indices; it does not change the mass-shell measure.

This construction is the one-particle part of the normalization used in Tong’s free-field notes and Weinberg’s representation treatment. It neither defines a local position probability density nor performs field quantization.

For ∣p∣≪m|\mathbf p|\ll m,

dΠp=d3p(2π)3 2m(1−p22m2+⋯ ).d\Pi_p=\frac{d^3p}{(2\pi)^3\,2m} \left(1-\frac{\mathbf p^2}{2m^2}+\cdots\right).

The leading constant can be absorbed into a nonrelativistic wavefunction’s normalization. The momentum-dependent correction cannot be discarded when working to the corresponding relativistic order.

For a massless particle in spherical momentum coordinates,

dΠp=p dp dΩ2(2π)3.d\Pi_p=\frac{p\,dp\,d\Omega}{2(2\pi)^3}.

The apparent 1/E1/E singularity does not make a bounded radial integral diverge at p=0p=0: ∫0ϵp dp\int_0^\epsilon p\,dp is finite. Singular amplitudes or additional propagator factors can still produce infrared divergences. The measure alone does not decide the convergence of a scattering calculation.

Multiparticle kinematics uses products of these measures together with a four-momentum conservation delta distribution. A cross section additionally requires an incident flux and a dynamical transition amplitude. Counting available final momenta is only one part of the prediction.

  1. With constants restored, integrate the mass-shell delta distribution over p0p^0 without the θ(p0)\theta(p^0) factor. What changes?
Solution

Both roots contribute with positive Jacobian weights:

∫c d3p2Ep[F(Ep/c,p)+F(−Ep/c,p)].\int\frac{c\,d^3p}{2E_{\mathbf p}} \left[F(E_{\mathbf p}/c,\mathbf p)+F(-E_{\mathbf p}/c,\mathbf p)\right].

This is a sum over two sheets, not a signed charge or KG norm.

  1. In natural units compute the invariant measure of the massless momentum ball 0≤p≤K0\leq p\leq K in one fixed frame. Is the ball itself invariant?
Solution

The integral is 4π∫0Kp dp/[2(2π)3]=K2/(8π2)4\pi\int_0^K p\,dp/[2(2\pi)^3]=K^2/(8\pi^2). A boost maps the ball to a different region. Invariance equates the integral over the original region to the integral over its transformed image, not to an independently imposed p′≤Kp'\leq K cutoff.

  1. Define ∣p⟩N=∣p⟩/2Ep|\mathbf p\rangle_N=|\mathbf p\rangle/\sqrt{2E_{\mathbf p}}. Find its overlap and completeness relation.
Solution

The overlap is (2π)3δ(3)(p′−p)(2\pi)^3\delta^{(3)}(\mathbf p'-\mathbf p), and I1=∫d3p ∣p⟩NN⟨p∣/(2π)3I_1=\int d^3p\,|\mathbf p\rangle_N{}_N\langle\mathbf p|/(2\pi)^3. Factors of 2E2E have moved to the basis vectors, not disappeared physically.

  • M. D. Schwartz, Quantum Field Theory and the Standard Model, Cambridge University Press, 2014, doi:10.1017/9781139540940 — state normalization and scattering phase space.
  • D. Tong, Lectures on Quantum Field Theory, University of Cambridge, 2006, section 2.4.1 — relativistic normalization of one-particle states.
  • S. Weinberg, The Quantum Theory of Fields, Volume I: Foundations, Cambridge University Press, 1995, chapter 2 — invariant one-particle representations and measures.