Ghost Conjecture: A Multifaceted Concept

• Ghost Conjecture is a multifaceted term that defines explicit ghost series in p-adic modular forms, mechanisms for ghost elimination in gravity, and a disproven graph theory conjecture on k-ghost-edges.
• In modular forms, the Bergdall–Pollack ghost series recovers Uₚ-slope data and unifies Buzzard’s slope algorithm with spectral-halo phenomena, supported by exact and computational evidence.
• Across disciplines, the term highlights distinct approaches: rigorous algorithmic predictions in arithmetic geometry, conjectural ghost-removal in gravitational theories, and a refuted converse in graph theory.

“Ghost Conjecture” is not a single universally standardized conjecture. In the literature surveyed here, the term designates several distinct programs united only by the role of a “ghost”: in $p$-adic modular forms, an explicit “ghost series” is conjectured to reproduce $U_p$-slope data; in gravitational physics, the phrase labels conjectural mechanisms for eliminating, quarantining, or reinterpreting ghost degrees of freedom; and in graph theory it appears in the now-refuted conjecture on $k$-ghost-edges. The most formal and sustained usage is the Bergdall–Pollack conjecture on slopes of overconvergent modular forms, together with its local, corrected, and algorithmic descendants (Bergdall et al., 2016).

1. Terminological scope

The surveyed literature uses the expression in several technically unrelated settings.

Domain	Core content	Status in surveyed papers
$p$-adic modular forms	Ghost series $G$ is conjectured to satisfy $NP(G_\kappa)=NP(P_\kappa)$	Central named usage
Local modular representation theory	Local ghost series attached to $\mathcal O[K_p]$-projective modules	Formalized, with structural theorems
Massive gravity	Physical sector may be protected from ghostlike branches by a barrier	Conjectural
Massive graviton interactions	Pseudo-linear ghost-free derivative terms may admit full nonlinear counterparts	Conjectural
Non-local or higher-derivative gravity	Competing claims about whether ghosts can be removed, hidden, or reinterpreted	Mixed, often controversial
Graph theory	Converse to Hickingbotham’s $k$-ghost-edge criterion	Disproved by counterexample

In arithmetic geometry, “Ghost Conjecture” is a precise technical label. In gravitational physics, by contrast, the expression is used more loosely for several different ghost-freedom hypotheses or anti-hypotheses. In graph theory, the term is attached to $k$-ghost-edges rather than to spectral or field-theoretic ghosts (Chen, 3 Feb 2026).

2. Bergdall–Pollack ghost series for modular slopes

In the Bergdall–Pollack formulation, the problem is to recover the $U_p$-slopes of overconvergent $U_p$0-adic cuspforms from an explicit combinatorial model. For a $U_p$1-adic weight $U_p$2, the relevant Fredholm series is

$U_p$3

and its Newton polygon $U_p$4 encodes the slopes. On a weight-space component $U_p$5, Coleman’s analytic variation gives a two-variable series

$U_p$6

with $U_p$7 (Bergdall et al., 2016).

The ghost series is built from classical dimension data

$U_p$8

using the observation that every $U_p$9-new eigenform in $k$0 has slope $k$1. The coefficients

$k$2

are defined so that

$k$3

and the multiplicities of the zero $k$4 across the consecutive coefficients follow the symmetric pattern

$k$5

Equivalently, the ghost series is engineered so that repeated $k$6-new slopes appear as repeated slopes in $k$7 (Bergdall et al., 2016).

The conjecture states that if $k$8 is an odd $k$9-regular prime, or $p$0 and $p$1, then for every $p$2-adic weight $p$3,

$p$4

This unifies Buzzard’s conjectural algorithm for classical slopes with spectral-halo expectations near the boundary of weight space (Bergdall et al., 2016). The unabridged version develops the same construction in greater detail, proves that the ghost series is entire after specialization, and analyzes its asymptotic and halo behavior (Bergdall et al., 2016).

The evidence is both exact and computational. Exact agreement is proved in several previously computed cases, including $p$5, $p$6, $p$7; $p$8, $p$9, $G$0; $G$1, $G$2, $G$3; and certain $G$4 and $G$5 boundary components (Bergdall et al., 2016). The same paper reports agreement with Buzzard’s algorithm for $G$6, $G$7, $G$8, and for $G$9, $NP(G_\kappa)=NP(P_\kappa)$0, $NP(G_\kappa)=NP(P_\kappa)$1 (Bergdall et al., 2016). The unabridged version also proves that the ghost series exhibits spectral-halo scaling and that, for odd $NP(G_\kappa)=NP(P_\kappa)$2, the slopes of the $NP(G_\kappa)=NP(P_\kappa)$3-adic Newton polygon are a finite union of arithmetic progressions with common difference $NP(G_\kappa)=NP(P_\kappa)$4 up to finitely many exceptions (Bergdall et al., 2016).

3. Local, corrected, and algorithmic descendants

The sequel “Slopes of modular forms and the ghost conjecture, II” abstracts the construction away from the full space of overconvergent forms. It associates an abstract ghost series $NP(G_\kappa)=NP(P_\kappa)$5 to dimension data $NP(G_\kappa)=NP(P_\kappa)$6 satisfying growth, monotonicity, and quasi-linearity conditions, and then specializes this framework to $NP(G_\kappa)=NP(P_\kappa)$7-isotypic spaces of modular forms. In that setting the generalized ghost conjecture predicts equality of Newton polygons for fixed residual Galois representations $NP(G_\kappa)=NP(P_\kappa)$8 that are locally reducible at $NP(G_\kappa)=NP(P_\kappa)$9 (Bergdall et al., 2017). The same paper proves asymptotic slope-distribution results at classical weights and shows that at non-integral weights the slopes form finite unions of arithmetic progressions (Bergdall et al., 2017).

A further localization was carried out in “A local analogue of the ghost conjecture of Bergdall-Pollack,” where the conjecture is reformulated in terms of primitive $\mathcal O[K_p]$0-projective augmented modules $\mathcal O[K_p]$1 for $\mathcal O[K_p]$2. The local characteristic series $\mathcal O[K_p]$3 of the abstract $\mathcal O[K_p]$4-operator is compared with a local ghost series

$\mathcal O[K_p]$5

and the conjecture becomes

$\mathcal O[K_p]$6

Even before proving that equality, the paper establishes explicit dimension formulas, compatibility with theta, Atkin–Lehner, and $\mathcal O[K_p]$7-stabilization symmetries, a characterization of Newton-polygon vertices via near-Steinberg ranges, and an integrality statement: when $\mathcal O[K_p]$8 is even the ghost slopes are integral, and when $\mathcal O[K_p]$9 is odd they lie in $k$0 (Liu et al., 2022).

The original $k$1-ghost conjecture was then corrected in “Slopes of modular forms and reducible Galois representations: an oversight in the ghost conjecture.” The issue was that the condition “$k$2 reducible” is too broad. The corrected dichotomy is between local representations that are “regular” and “irregular.” Irregular means, equivalently, either irreducible local type or a twist of an unramified representation with Frobenius trace zero. The salvaged conjecture keeps the same ghost series $k$3 but restricts the predicted equality of Newton polygons to regular $k$4 (Bergdall et al., 2021).

Two later papers sharpen the structural picture. “The slope-invariant of local ghost series under direct sum” proves that the Newton polygon of the product of local ghost series coincides with the Newton polygon of the ghost series attached to the direct sum only in a rigid situation: the local types must all coincide, up to the involution

$k$5

so the naïve direct-sum compatibility is false in general (Ren, 2022). “An explicit computation of the Hecke operator and the ghost conjecture” computes a definite quaternionic $k$6-matrix, shows that upper-left principal minors have nonzero corank with the same unimodal pattern as ghost multiplicities, proves the divisibility

$k$7

and conjectures a quaternionic variant of the ghost series (Truong, 2020).

More recently, “Slopes of modular forms and the Ghost conjecture” turns proved cases of the ghost program into a recursive slope algorithm on fixed Galois components. Under the non-split generic local hypothesis

$k$8

with $k$9, the paper gives an algorithm that computes the slope sequence from its first few entries and proves that it equals the increasing slope sequence of $k$0 (Hur, 3 Aug 2025).

4. Massive-gravity usages of the term

In massive gravity, “Ghost Conjecture” refers not to ghost series but to the persistence or quarantine of ghostlike degrees of freedom. In “Energy in ghost-free massive gravity theory,” the issue is whether dRGT massive gravity, although free of the Boulware–Deser mode at the level of degree-of-freedom counting, could still possess a ghost instability because the reduced Hamiltonian is not positive. In the spherically symmetric sector Volkov derives explicit nonlinear constraints and finds that the energy is positive for globally regular, asymptotically flat data in the physical sector, while negative-energy and even unbounded-below branches do exist elsewhere. The paper conjectures that the physical sector may nevertheless be protected from these branches by a potential barrier, because all known negative-energy solutions are singular, non-asymptotically flat, or require singular lapse functions (Volkov, 2014).

A different massive-gravity usage appears in “Ghost-Free Derivative Interactions for a Massive Graviton.” There the conjecture is that pseudo-linear ghost-free derivative interactions $k$1, constructed around flat space and shown to preserve the Hamiltonian constraint structure, should have fully nonlinear counterparts in massive gravity. In particular, in $k$2 the paper conjectures a missing two-derivative, non-diffeomorphism-invariant, ghost-free interaction whose lowest-order expansion is the cubic term

$k$3

and states that standard $k$4 dRGT, as usually studied, would then be incomplete (Hinterbichler, 2013).

5. Non-local and higher-derivative gravity: competing ghost hypotheses

Several gravitational papers use “ghost conjecture” more loosely for claims about removing or reinterpreting ghost excitations. In “Counting ghosts in the ‘ghost-free’ non-local gravity,” the conjectural hope is that a special entire-function form factor can make the graviton propagator free of extra poles. The paper argues that this is only a tree-level statement: loop corrections generically spoil the exact exponential tuning and produce infinitely many massive unphysical poles, mostly complex, so the model is not truly ghost-free at the quantum level (Shapiro, 2015).

In “Ghosts in pure and hybrid formalisms of gravity theories: a unified analysis,” the question is which hybrid metric–Palatini higher-curvature theories avoid ghost or tachyonic excitations. The analysis singles out $k$5 as a viable class: around Minkowski, its propagator coincides with the GR propagator, whereas generic hybrid $k$6 theories contain two scalar modes that cannot both be healthy, and hybrid Ricci-square theories with $k$7 are “seriously haunted by ghosts” (Koivisto et al., 2013).

Other papers do not formulate a single named conjecture, but they advance ghost-elimination theses. “Solution to the ghost problem in higher-derivative gravity” argues, via $k$8-symmetric quantization of the Pais–Uhlenbeck system, that the usual negative-residue diagnosis is misleading: the correct Hilbert-space inner product is not the Dirac one, the generic unequal-frequency case has no negative-norm states, and the pure $k$9 case is Jordan-block with zero-norm rather than negative-norm states (Mannheim, 2021). “Exorcising ghosts in quantum gravity” argues instead that Ostrogradsky ghosts in finite-order higher-derivative gravity are artifacts of truncating a more complete theory with infinitely many curvature invariants, and that ghost poles can be projected out by a contour prescription without changing the $U_p$0 ultraviolet behavior responsible for renormalizability (Kuntz, 2019).

A distinct but related asymptotic-safety theme appears in “Ghost wave-function renormalization in Asymptotically Safe Quantum Gravity.” There the ghost sector is not removed but promoted to a nontrivial part of the RG flow: including the ghost wave-function renormalization changes the beta functions of Newton’s constant and the cosmological constant, preserves the non-Gaussian ultraviolet fixed point, and yields a four-dimensional ghost anomalous dimension

$U_p$1

which the authors interpret as supporting effective two-dimensional short-distance behavior (Groh et al., 2010).

6. Graph-theoretic ghost edges

In graph theory, the relevant notion is a $U_p$2-ghost-edge. For a connected graph $U_p$3 with $U_p$4, a non-edge $U_p$5 is a $U_p$6-ghost-edge if every tree decomposition of width at most $U_p$7 contains a bag with both $U_p$8 and $U_p$9. Hickingbotham proved that if there are at least $U_p$00 internally vertex-disjoint $U_p$01-paths, then $U_p$02 is a $U_p$03-ghost-edge, and conjectured the converse: if there are at most $U_p$04 such paths, then $U_p$05 is not a $U_p$06-ghost-edge (Chen, 3 Feb 2026).

“A counterexample to Hickingbotham’s conjecture about $U_p$07-ghost-edges” disproves that converse at $U_p$08. The paper constructs a graph $U_p$09 with

$U_p$10

such that $U_p$11 is a $U_p$12-ghost-edge even though there are only four internally vertex-disjoint $U_p$13-paths. A key lemma shows that any $U_p$14-$U_p$15 separator $U_p$16 of size at most $U_p$17 must equal a rigid set $U_p$18, and the contradiction argument then shows that no width-$U_p$19 tree decomposition can keep $U_p$20 and $U_p$21 separated in all bags (Chen, 3 Feb 2026). In this domain, the Ghost Conjecture is therefore not open but false.

7. Comparative status and significance

The surveyed literature shows that “Ghost Conjecture” is best understood as a family resemblance term rather than a single doctrine. In modular forms it names an explicit spectral model: a ghost series over weight space whose Newton polygons are conjectured, and in substantial settings proved, to recover $U_p$22-slope sequences. That program has developed a hierarchy of increasingly local statements, corrections of its admissible residual representations, direct-sum obstructions, and effective algorithms (Bergdall et al., 2017).

In gravitational physics, the phrase usually marks a consistency claim about ghosts rather than a fixed formal conjecture. Some papers are protective, asserting that pathological branches are dynamically isolated or that apparently ghostly poles are artifacts of truncation or quantization choices; others are negative, arguing that tree-level ghost-freedom is not radiatively stable. The net result is not convergence to a single theorem but a map of competing mechanisms: potential barriers in dRGT, conjectured nonlinear completions of pseudo-linear derivative interactions, loop-generated complex poles in non-local gravity, $U_p$23-symmetric redefinitions of the Hilbert space, and contour-based removal of truncation ghosts (Volkov, 2014).

The graph-theoretic usage is narrower and cleaner. There the conjecture was a precise converse statement about treewidth and ghost-edges, and a single $U_p$24 counterexample sufficed to refute it (Chen, 3 Feb 2026). A plausible implication is that, outside arithmetic geometry, the phrase “Ghost Conjecture” often serves more as an informal label for a ghost-related mechanism than as a settled item of standard nomenclature.