Fully covariant force in general relativity

Difficulty level:   ★ ★ ★

It is often said Newton was fortunate to define force on a particle as the change in momentum f:=dp/dt, not from the change in velocity m\,dv/dt, because the former generalises better. Here the momentum is p:=mv, and clearly the expressions for force coincide if the mass m is constant.

In relativity, the force (4-force) on a particle is usually defined as the change in 4-momentum over proper time as follows:

    \[\mathbf f := \frac{d\mathbf p}{d\tau}\qquad\qquad\textrm{(LIF)}\]

However this expression is only valid in a local inertial frame (LIF), as Hartle (2003 , §20.4) clearly qualifies. Recall, the 4-momentum is \mathbf p := m\mathbf u where \mathbf u is the 4-velocity of the particle. We can split the force into two orthogonal vectors:

    \[\mathbf f = \frac{d(m\mathbf u)}{d\tau} = m\frac{d\mathbf u}{d\tau} + \frac{dm}{d\tau}\mathbf u = m\mathbf a + \frac{dm}{d\tau}\mathbf u\]

where \mathbf a:=d\mathbf u/d\tau (LIF) is the 4-acceleration. The m\mathbf a term is called a “pure force”, because they “create motion in three-dimensional space and correspond to the Newtonian forces”, as Tsamparlis (2010 §11.2) describes, meaning motion in the instantaneous 3-space orthogonal to \mathbf u. The term containing \mathbf u is called a “thermal force”, at least by Tsamparlis. Examples which are at least partly thermal include a particle heated by an external source, or a rocket losing mass. Another example, considered by Einstein apparently, is an object which absorbs two photons with equal energies and opposite directions in the object’s frame, which results in a thermal force but no pure force. On relativistic force, see also Gourgoulhon (2013 §9.5). (Note if the mass does change over time, this is nothing to do with the old-fashioned “relativistic mass” m\gamma dependent on the Lorentz factor, rather we use the modern meaning of mass as “rest mass”.)

Now textbooks and webpages on relativistic mechanics typically assume special relativity, in particular inertial frames within Minkowski spacetime. So how should we generalise to arbitrary coordinates and curved spacetime? According to Hartle (§20.4), the derivative d/d\tau (LIF) generalises to the covariant derivative \nabla_{\mathbf u}. Hence, the fully covariant expression for 4-force is:

    \[\boxed{\mathbf f := \nabla_{\mathbf u}\mathbf p}\]

In words, this is the change of 4-momentum in the direction of the 4-velocity. But in the particle’s frame, its 4-velocity is precisely the “time” direction. So, we could say force is the change of momentum with time in the particle’s frame. So while the mathematics is more general, the concept has clear lineage from special relativity and even Newton!

Now we can repeat the above splitting:

    \[\mathbf f = \nabla_{\mathbf u}(m\mathbf u) = m\nabla_{\mathbf u}\mathbf u + (\nabla_{\mathbf u}m)\mathbf u  = m\mathbf a + \frac{dm}{d\tau}\mathbf u\]

since \mathbf a := \nabla_{\mathbf u}\mathbf u is the usual fully covariant expression, and \nabla_{\mathbf u} of a scalar is d\cdot/d\tau. This expression for force is the same as the specific LIF case above.