Consider the Klein-Gordon equation: \begin{equation} \frac{\partial^2 \psi}{\partial t^2} = c^2 \Delta \psi - \frac{m^2 c^4}{\hbar^2} \psi, \end{equation} and define for each one of its solutions $\psi$ the quantity:

\begin{equation} P(\mathbf{x},t)= \hbar^2 \frac{\partial \psi}{ \partial t} \frac{\partial \psi^{*}}{ \partial t} + \hbar^2 c^2 \nabla \psi \cdot \nabla \psi^{*} + m^2 c^4 \psi \psi^{*}, \end{equation} Let us adopt the convention in which the generic point of the Minkowski space-time is $(x,y,z,ct)$. In Section (4.6) of his wonderful treatise Quantum Theory Bohm states that under a Lorentz transformation

(i) $P(\mathbf{x},t)$ transforms as the (4,4)-coordinate a rank two tensor,

(ii) $\int P(\mathbf{x},t) d\mathbf{x}$ transforms as the fourth component of a four-vector.

Could someone give me a proof of these two statements, please?

NOTE (1). All that I know about the Klein-Gordon equation is that $\psi$ is invariant under Lorentz transformations, that is if $\psi(\mathbf{x},t)$ is a solution of the Klein-Gordon equation, then the new function $\phi(\mathbf{x'},t')$ obtained by replacing the equations of a Lorentz boost $(\mathbf{x},ct) \rightarrow (\mathbf{x'},ct')$ in $\psi$ is again a solution of the Klein-Gordon equation.

NOTE (2). Bohm justifies assertion (i) by considering the particular solution $\psi= \exp i \left( \frac{Et-\mathbf{p} \cdot \mathbf{x} } {\hbar} \right)$, for which we get \begin{equation} P=E^2+p^2c^2+m^2c^4=2E^2, \end{equation} so that $P$ transforms actually as the square of an energy.

  • 1
    $\begingroup$ Double-check your signs, then see if you can spot a tensor $A_{\mu\nu}$ for which $P=A_{tt}$. $\endgroup$
    – J.G.
    Oct 16, 2018 at 18:04
  • $\begingroup$ Sorry, what do you mean? $\endgroup$ Oct 16, 2018 at 20:20

1 Answer 1


Actually, the $P(\mathbf{x},t)$ you mention is the 4,4-component of the stress-tensor $T_{ik}$ of the Klein-Gordon (K-G) field. In the following instead I will use the metric tensor $\eta_{ik}=diag(1,-1,-1,-1)$ and identify $P(\mathbf{x},t)$ with the 0,0-component of $T_{ik}$. Bohm aparently uses the other metric $diag(-1,-1,-1,1)$ convention. Moreover $c=1=\hbar$ is assumed.

One starts best off from the Lagrange density of the complex K-G: (Double appearing indices is summed over, i.e. Einstein summation convention):

$$ L = \partial_i \phi^\star \partial^i \phi - m^2 \phi^\star \phi$$

For a complex field $\phi$ and $\phi^\star$ are considered as independent variables. The definition of the stress tensor is given by:

$$T_{ik}=\sum_{\varphi}\varphi_{,k}\frac{\partial L}{\partial \varphi^{,i}} -L\eta_{ik}$$

with $\varphi = (\phi, \phi^\star)$.

Upon inserting the expression for the Lagrange density of the K-G field in the stress tensor's definition we get:

$$T_{ik} = \partial_i\phi^\star \partial_k\phi + \partial_k\phi^\star \partial_i\phi -L\eta_{ik}$$

The tensor property of $T_{ik}$ is rather obvious, as the partial derivatives $\partial_i$ respectively $\partial_k$ transform like covariant vectors and $\eta_{ik}$ is also a tensor. This is in particular true for the $(i,k)=(0,0)$-component:

$$T_{00} = 2\frac{\partial\phi^\star}{\partial t}\frac{\partial\phi}{\partial t} -L=\frac{\partial\phi^\star}{\partial t}\frac{\partial\phi}{\partial t}+\nabla\phi^\star\nabla\phi + m^2\phi^\star \phi\equiv P(\mathbf{x},t) $$

The 4-momentum vector $P_i$ yields:

$$P_i = \int_{\partial\Omega} T_{ik} d\sigma^k$$

where $d\sigma^k$ is the vectorial hypersurface element which is parametrized by values $(u,v,w)$:

$$d\sigma_i =\epsilon_{ijkm}\frac{\partial x^j}{\partial u}\frac{\partial x^k}{\partial v}\frac{\partial x^m}{\partial w}du dv dw$$

If now the hypersurface $t=const$ is chosen, as parameters $(u,v,w)=(x,y,z)\equiv(x^1,x^2,x^3)$ can be used:

$$d\sigma_i =\epsilon_{ijkm}\delta^j_1\delta^k_2\delta^m_3 dx^1 dx^2 dx^3= \epsilon_{i,1,2,3}d^3x = (d^3x,\mathbf{0})$$

so we get for this particular hypersurface $t=const$:

$$P_i = \int_{t=const} T_{i0} d^3x $$

It can be shown that $P_i$ considered at another hypersurface $\partial\Omega$ which is Lorentz-transformed with respect to the original hypersurface $t=const$ has the same value if it is computed by the more general formula:

$$P_i = \int_{\partial\Omega} T_{ik} d\sigma^k$$

But due to the covariant way of writing it is clear that $P_i$ is a 4-vector (but this would not be longer true in curved space-time) and in particular

$$P_0 = \int_{t=const} T_{00} d^3x \equiv \int_{t=const}P(\mathbf{x},t) d^3x $$

the 0-component of the 4-vector $P_i$ (the momentum 4-vector), the energy of the K-G field.

  • $\begingroup$ Dear Frederic thank you very very ... much for having answered my question: I could never have done it by myself! I add here the only missing step needed to justify the claim that $P_i$ can be computed by the general formula you have given above. Let us note that $T_{ik}$ has zero divergence, that is $\partial^{k} T_{ik}=0$, as it is immediately checked by using the Klein-Gordon equation for $\phi$ and $\phi^{*}$. Then your claim follows from the general argument given in Møller, Theory of Relativity, $\S 63$ or Pauli, Theory of Relativity, $\S 21$. $\endgroup$ Oct 17, 2018 at 9:37
  • $\begingroup$ You are welcome. The claim you refer to in your comment is it that $P_i$ has the same value if the hypersurface $t=const$ is Lorentz-transformed to another hypersurface $\partial\Omega$ when computed with the more general formula? $\endgroup$ Oct 19, 2018 at 13:50
  • $\begingroup$ It is exactly that, Frederic. $\endgroup$ Oct 19, 2018 at 15:17

Your Answer

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

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