Researchers used a mix of techniques to put tight numerical upper bounds on the dimension of the lowest scalar single-trace operator in two four-dimensional holographic quantum field theories. They worked in the planar large N limit (N large and called the “planar” limit) and for all values of the 't Hooft coupling λ, the parameter that controls interaction strength. One theory is N=4 SU(N) super‑Yang–Mills (SYM), which is dual to closed string scattering; the other is a 4d N=2 USp(2N) gauge theory with SO(8) flavor symmetry, which is dual to open string scattering on D7 branes. For N=4 SYM the bounds lie close to the precise values known from integrability. For the N=2 theory, where integrability results are not available, the bounds agree with known weak‑ and strong‑coupling expectations in the regimes where those predictions exist.

What the authors actually did was study four‑point correlation functions. For SYM they analyzed the four‑point function of operators in the stress tensor multiplet; for the N=2 theory they studied the four‑point function of flavor multiplet operators. They combined numerical conformal bootstrap ideas with dispersive sum rules (derived from dispersion relations and the theory’s good high‑energy, or Regge, behavior) and exact integrated constraints coming from supersymmetric localization (a method that gives certain sphere integrals or derivatives exactly). Using this toolbox they computed upper bounds on the scaling dimension of the lowest scalar single‑trace long multiplet — in SYM this is the Konishi operator — as a function of λ.

Why this combination helps needs a bit of explanation. The conformal bootstrap enforces basic consistency conditions on correlation functions, but in the planar limit the usual positivity assumptions fail for double‑trace contributions. To avoid that problem the authors used dispersion relations to rewrite the correlator in so‑called Polyakov‑Regge blocks that isolate single‑trace contributions. The localization inputs give additional exact equations for integrated versions of the correlators. Together these ingredients let the authors derive numerical constraints that are sensitive to single‑trace data at finite λ.