Hilbert energy-momentum tensor is generally defined as
$$T_{\mu\nu} \delta g^{\mu\nu} = -2\delta\mathcal{L} + g_{\mu\nu} \mathcal{L} \delta g^{\mu\nu}$$
where $\mathcal{L}$ is the Lagrangian density of the field. Now if the field is "on-shell", then according to this Physics SE answer [comment below Eq. 3] and this paper [Non-Existence of Black Hole Solutions for a Spherically Symmetric, Static Einstein-Dirac-Maxwell System, Eq. 3.8], in case of a "on-shell" Dirac field, the second term in the above expression, $g_{\mu\nu} \mathcal{L} \delta g^{\mu\nu}$ will be zero.
But in these papers [Cosmological model with non-minimally coupled fermionic field (Eq. 8) and Spinors, Inflation, and Non-Singular Cyclic Cosmologies (Eq. 12)], the aforementioned $g_{\mu\nu} \mathcal{L} \delta g^{\mu\nu}$ term for "on-shell" Dirac field has been kept as it is, without making it zero!
Therefore,
- Which one is correct? Will the aforementioned term be zero?
- If so, then why?