What the paper is about. The authors revisit the long-standing idea of extending quantum field theories (QFTs) from integer space-time dimensions to fractional or complex values of the dimension d. This operation—called analytic continuation in d—underlies common tools in theoretical physics, such as dimensional regularization and the ϵ expansion. The main finding reported here is that familiar theories like quantum chromodynamics (QCD) and quantum electrodynamics (QED) develop branch cuts and discrete jumps when viewed as functions of complex d. In plain terms, some basic properties of these theories change abruptly at certain integer dimensions, so there is no obvious single, smooth way to continue them to all complex d.

What the researchers did. The paper inspects how correlation functions and operator data behave when one tries to treat d as a continuous complex parameter. The authors study general symmetry constraints coming from rotations (the group SO(d)) and from reflections (the larger group O(d)). They point out concrete mechanisms that force discontinuous behavior as d crosses certain integer thresholds. They also revisit standard continuation tricks used in the literature—such as ’t Hooft–Veltman type prescriptions that split space into parallel and transverse subspaces—and argue that those tricks work only in a tiny neighborhood around a given integer dimension, not across a wide region of the complex d plane.

A simple example that illustrates the problem. The paper gives a concrete example using a free, massless Dirac fermion and the scalar operator ¯ψψ (psi-bar psi). Low-order correlation functions look harmless and admit an obvious continuation in d. But the five-point function behaves differently: it vanishes for all integers d ≥ 4, yet it is nonzero in d = 3. That produces a discrete sequence whose values jump with d and which has no canonical analytic interpolation to complex d. The authors trace this kind of jump to a more general geometric fact. When the number of operator insertions becomes comparable to the space-time dimension, an extra invariant—the orientation of a set of vectors, represented by an epsilon sign—can appear. That orientation-dependent piece is allowed in theories invariant only under rotations (SO(d)), and it forces a qualitative change in correlation functions as d crosses an integer. The change is milder in theories invariant under reflections (O(d)), because parity-odd pieces are forbidden there.