If we define spacetime as a Lorentzian manifold $(E^4,g)$, how can we define the property that $$\iiint_{E^3}\Psi^*\Psi\ d^3\mathbf{x}=1$$ for a wavefunction $\Psi$? Because there's no global sense of time.

With my approach, I started with a Lorentzian manifold $(M,\,g)$, the frame bundle $LM\xrightarrow{\pi_{LM}} M$, and the associated bundle $E\equiv (LM\times\mathbb{C})/GL(\mathrm{dim}\,M,\mathbb{R})\xrightarrow{\pi_E}M$. Then, a wavefunction is a section $\psi:M\to E$ of this $\mathbb{C}$-vector bundle $E\xrightarrow{\pi_E}M$. This is so $\psi(p)=[\mathbf{e},z]=\{(\mathbf{e}\lhd g,g^{-1}\rhd z)\in L_pM\times \mathbb{C}\ |\ g\in GL(\mathrm{dim}\,M,\mathbb{R})\}$ is independent of frame.

We can also look at a wavefunction as a function $\Psi:LM\to\mathbb{C}$ defined by $\psi(p)=[\mathbf{e},\Psi(\mathbf{e})]$. We can additionally pull $\Psi$ back using a local section $\sigma:U\to LU$ to view it as a local complex-valued function $\sigma^*\Psi:U\to\mathbb{C}$. Whichever global method we take, how do we parameterize $\Psi$ over time so that we can normalize? I tried using the volume form with respect to the metric but this integrates over spacetime instead of just space. $$\int_M \Psi^*\Psi\ vol_g\neq 1$$

How do we define this property of a wavefunction? Do we need quantum field theory for this task?

  • 1
    $\begingroup$ Integrating on the entire spacetime shouldn't work since the probability is only one at a given time. The thing usually done is to make it so that the wavefunction is of unit norm for any Cauchy surface, regardless of foliation. $\endgroup$
    – Slereah
    Jun 15, 2020 at 23:34
  • $\begingroup$ I believe that @Slereah just did express it mathematically $\endgroup$
    – WillO
    Jun 16, 2020 at 3:18
  • $\begingroup$ volume integrals can't be invariant, because volumes transform under lorentz transformations. Any definition of a volume is going to have a trailing timelike lorentz index. $\endgroup$ Apr 24, 2022 at 15:33

1 Answer 1


If you quantize the manifestly covariant Polyakov action for the point particle, written in its 1st-order form as $$ S[x^{\mu}(\tau), p_{\mu}(\tau), \lambda(\tau)] = \int d\tau \left( p_{\mu} \dot{x}^{\mu} - \lambda \left( g^{\mu \nu} p_{\mu} p_{\nu} - m^2 \right) \right), $$

you will end up with wavefunctions which are functions over the spacetime like you wanted, with $x$ and $p$ acting by $$ \hat{x}^{\mu} \Psi(x) = x^{\mu} \Psi(x), \quad \hat{p}_{\mu} \Psi(x) = - i \hbar \partial_{\mu} \Psi(x). $$

However, these functions aren't elements of the physical Hilbert space. You have to implement the constraint that is fixed by the Lagrange multiplier $\lambda$ as a quantum operator: $$ \left( \hbar^2 \Box + m^2 \right) \Psi(x) = 0. $$

This is of course just the Klein-Gordon equation.

Naively, to recover the physical Hilbert space you should look at the rigged Hilbert space of functions over space-time, and select the kernel of the constraint operator to be your physical Hilbert space. However, this operation is only well defined when $0$ is in the discrete spectrum of the contraint operator, and it is easy to see that $0$ is in the continuous spectrum of the Klein-Gordon operator. You therefore need additional input solve this theory.

The solution comes from noticing that the solutions of the Klein-Gordon equation are completely determined by $\Psi(t_0, \vec{x})$ and $\dot{\Psi}(t_0, \vec{x})$ for some $t_0$. Therefore your wavefunctions are two-component functions over the spatial slice parametrized by $\vec{x}$. The two components correspond to particles and anti-particles.

However, it isn't possible to define a positive definite inner product on the space of solutions.

That is not the case for Dirac's equation, where a positive definite inner product can be defined.

In either case, second quantization makes the problem of defining an inner product on the space of solutions to a relativistic field equation obsolete. The modern treatment is to start with the field Lagrangian instead of a particle Lagrangian and derive a positive Hilbert space of the QFT.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.