Fake Z

TL;DR

This paper investigates the distinction between genuine Narain CFT partition functions and fake solutions that satisfy modular invariance and positivity but do not correspond to actual CFTs, highlighting limitations of the modular bootstrap.

Contribution

The authors identify and analyze fake polynomials that meet bootstrap constraints but do not correspond to real Narain CFTs, demonstrating limitations of the bootstrap approach.

Findings

Six fake polynomials are shown not to be Narain CFT partition functions.

Four of the fake polynomials are not associated with any unitary 2D CFT.

The analysis reveals that not all bootstrap solutions correspond to actual CFTs.

Abstract

Recently introduced connections between quantum codes and Narain CFTs provide a simple ansatz to express a modular-invariant function $Z(\tau,\bar \tau)$ in terms of a multivariate polynomial satisfying certain additional properties. These properties include algebraic identities, which ensure modular invariance of $Z(\tau,\bar \tau)$, and positivity and integrality of coefficients, which imply positivity and integrality of the $\mathfrak{u}(1)^n \times \mathfrak{u}(1)^n$ character expansion of $Z(\tau,\bar \tau)$. Such polynomials come naturally from codes, in the sense that each code of a certain type gives rise to the so-called enumerator polynomial, which automatically satisfies all necessary properties, while the resulting $Z(\tau,\bar \tau)$ is the partition function of the code CFT -- the Narain theory unambiguously constructed from the code. Yet there are also ``fake'' polynomials satisfying all necessary properties, that are not associated with any code. They lead to $Z(\tau,\bar \tau)$ satisfying all modular bootstrap constraints (modular invariance and positivity and integrality of character expansion), but whether they are partition functions of any actual CFT is unclear. We consider the group of the six simplest fake polynomials and denounce the corresponding $Z$'s as fake: we show that none of them is the torus partition function of any Narain theory. Moreover, four of them are not partition functions of any unitary 2d CFT; our analysis for other two is inconclusive. Our findings point to an obvious limitation of the modular bootstrap approach: not every solution of the full set of torus modular bootstrap constraints is due to an actual CFT. In the paper we consider six simple examples, keeping in mind that thousands more can be constructed.