It is generally believed that in a two-dimensional metal, whose ground state is antiferromagnetically ordered with $\mathbf{Q}=(\pi ,\pi )$, thermal (static) magnetic fluctuations give rise to precursor behavior above $T_N$ in which the spectral function of a hot fermion (the one for which $\mathbf{k}$ and $\mathbf{k}+\mathbf{Q}$ are Fermi surface points) contains two peaks, separated by roughly the same energy as in the antiferromagnetically ordered state. The two peaks persist in some range of $T>T_N$ and eventually merge into a single peak at zero frequency. This behavior is obtained theoretically by departing from free fermions in a paramagnet and evaluating the dressed fermionic Green's function by summing up infinite series of diagrams with contributions from thermal magnetic fluctuations. We show, following [Y. M. Vilk and A.-M. S. Tremblay, J. Phys. I (France) 7, 1309 (1997)] that keeping vertex renormalization diagrams in these series is crucial as other terms only broaden the spectral function of a hot fermion but do not shift its maximum away from zero frequency. As the consequence, the magnetic pseudogap should be treated as an input for theories that neglect vertex corrections, such as, e.g., Eliashberg theory for magnetically mediated superconductivity. We also analyze the potential pseudogap behavior at $T=0$. We argue that it may exist, but only at a finite correlation length, and not as a precursor to antiferromagnetism.

• Received 12 June 2023
• Accepted 26 July 2023

DOI:https://doi.org/10.1103/PhysRevB.108.L081118