Columbia Plot: QCD Phase Diagram

• Columbia Plot is the QCD phase diagram that uses light and strange quark masses to demarcate first-order, second-order, and crossover thermal transition regions.
• It employs lattice QCD techniques with observables like the Polyakov loop and Binder cumulants to extract critical scaling and phase boundaries.
• The diagram provides practical insights into chiral symmetry restoration, center symmetry breaking, and the continuum challenges in modern QCD research.

Searching arXiv for recent and foundational papers on the Columbia plot to ground the article in the supplied literature. The Columbia plot is the finite-temperature QCD phase diagram in the plane of the degenerate light-quark mass $m_{u,d}$ and the strange-quark mass $m_s$ at $\mu=0$. It classifies where the thermal transition is first order, second order, or an analytic crossover, and it organizes two symmetry-dominated limits: the heavy-quark upper-right corner, connected to pure $SU(3)$ gauge theory and center symmetry, and the light-quark lower-left corner, governed by chiral symmetry restoration. In contemporary usage, the term also encompasses extended constructions that add chemical potential as a third axis, as well as alternative parameterizations such as a mass–$N_f$ plane (Cuteri et al., 2017, Bernhardt et al., 29 Jul 2025).

1. Canonical structure and symmetry logic

At $\mu=0$, the standard Columbia plot places $m_{u,d}$ and $m_s$ on its axes and partitions the plane into first-order, second-order, and crossover regions. The physical point lies in the crossover region. In the upper-right corner, all quarks are very heavy and QCD approaches pure $SU(3)$ gauge theory, where the thermal transition is first order because the theory has an exact global center symmetry $\mathcal Z_3$, spontaneously broken at deconfinement. In the lower-left corner, quarks are light and chiral symmetry becomes the organizing principle; the classical expectation for three massless flavors is a first-order chiral transition, terminated at finite quark mass by a second-order boundary in the $m_s$0 Ising universality class, beyond which the transition becomes a crossover (Forcrand et al., 2017).

The light-quark corner is nonetheless unsettled. The standard scenario and its alternatives differ mainly in the chiral corner, especially for $m_s$1, and finite-lattice studies have historically produced markedly different critical masses for Wilson and staggered fermions. This has turned the Columbia plot from a purely schematic diagram into a precision problem in universality, discretization effects, and continuum extrapolation (Cuteri et al., 2017).

2. Heavy-quark boundary and deconfinement criticality

In the heavy-quark regime, the natural observable is the Polyakov loop. For three degenerate staggered flavors at temporal extent $m_s$2, the heavy endpoint on the $m_s$3 diagonal was studied through the spatially averaged Polyakov loop

$m_s$4

and its susceptibility

$m_s$5

Because dynamical quarks explicitly break $m_s$6, $m_s$7 is never exactly zero, but its rapid rise across the transition and the finite-size scaling of $m_s$8 remain the central diagnostics. The proceedings analysis emphasized $m_s$9 rather than Binder cumulants; despite the phrase “Polyakov loop and its moments,” no explicit $\mu=0$0 or $\mu=0$1 analysis was presented (Kara et al., 2021).

The infinite-volume extrapolation was organized through

$\mu=0$2

followed by the Ising-motivated mass fit

$\mu=0$3

For $\mu=0$4 staggered fermions at $\mu=0$5, the heavy critical mass was reported as $\mu=0$6 in one extrapolation strategy and $\mu=0$7 in another, with the spread reflecting finite-size/extrapolation systematics. The same study quoted

$\mu=0$8

indicating that the first-order region in the heavy corner begins only at very large masses (Kara et al., 2021).

A complementary continuum description of the same corner was developed in the center-symmetric Curci-Ferrari model, where the one-loop gluon average potential in a center-symmetric Landau gauge serves as a proxy for the Polyakov-loop potential. In that framework, the heavy-quark critical boundary at $\mu=0$9, the universal $SU(3)$0-flavor relation

$SU(3)$1

and the tricritical scaling near the Roberge-Weiss boundary are all reproduced, with results agreeing quantitatively with simulations to $SU(3)$2 accuracy, which the authors identify as the expected precision of the one-loop approximation (Surkau et al., 28 Mar 2025).

3. Chiral boundary, universality, and the continuum puzzle

The lower-left corner is the most controversial part of the Columbia plot. A key intervention studied the $SU(3)$3 theory with standard staggered fermions precisely to eliminate rooting as an explanation for the dramatic lattice-spacing dependence seen in $SU(3)$4. The central observable was the Binder cumulant of the chiral condensate,

$SU(3)$5

with $SU(3)$6. Assuming $SU(3)$7 Ising universality, the analysis used

$SU(3)$8

and converted the resulting critical bare masses into the dimensionless quantity

$SU(3)$9

The main conclusion was that the dramatic reduction of $N_f$0 as the lattice spacing is reduced remains present in unrooted $N_f$1 staggered QCD, and is therefore not related to rooting (Forcrand et al., 2017).

This sharpened a broader universality problem. At finite lattice spacing, Wilson and staggered studies had produced violently different critical pseudoscalar masses in the three-flavor chiral corner, while staggered results showed a particularly steep downward trend as the lattice spacing was reduced. Continuum extrapolations were described as compatible with a zero value, and the paper’s conclusion was that “extremely light quarks are needed to obtain a first-order thermal transition in the continuum theory.” A plausible implication is that many schematic versions of the lower-left first-order region are too large, and that much of the finite-$N_f$2 signal is dominated by cutoff effects rather than continuum thermodynamics (Forcrand et al., 2017).

4. Extended, alternative, and three-dimensional Columbia plots

One extension adds chemical potential as a third axis. In a lattice roadmap for the chiral phase transition in the $N_f$3 Columbia plot, the relevant sign-problem-free parameter space was written as

$N_f$4

with the lower boundary identified as the Roberge-Weiss plane. The standard $N_f$5 Columbia plot is then the slice $N_f$6. That work did not present a completed determination; it formulated a numerical programme based on chiral-condensate kurtosis, finite-size scaling, Ferrenberg-Swendsen reweighting, and a released software stack including CL$N_f$7QCD, BaHaMAS, script utilities, and MCC++ (Sciarra, 20 Dec 2025).

A related status report distinguished three variants: the standard plot at $N_f$8, the extended plot at imaginary chemical potential, and an alternative plot in the plane of quark mass and a formally continuous flavor number $N_f$9. In the extended version at the Roberge-Weiss value, the $\mu=0$0 critical lines of the standard plot are replaced by tricritical lines, and the first-order regions are wider. In the alternative version, the key scaling relation is

$\mu=0$1

used to infer the order of the $\mu=0$2 chiral transition from non-integer $\mu=0$3 simulations (Cuteri et al., 2017).

A functional study using a hybrid of lattice Yang-Mills theory and truncated Dyson-Schwinger equations pushed the three-dimensional construction further by examining the slice

$\mu=0$4

For $\mu=0$5 MeV, the critical endpoint coordinates were reported as

$\mu=0$6

so that lowering the light mass moves the CEP toward smaller $\mu=0$7 and slightly larger $\mu=0$8. The same study parameterized the crossover line as

$\mu=0$9

with $m_{u,d}$0 for $m_{u,d}$1 MeV, respectively. It also reported only crossover behavior at imaginary chemical potential for all nonzero light masses considered, and a rough linear extrapolation toward an indicative tricritical location $m_{u,d}$2 MeV, while stressing that this estimate is not quantitatively reliable (Bernhardt et al., 29 Jul 2025).

5. Effective theories, anomaly realizations, and model dependence

Low-energy effective models have made the Columbia plot a testing ground for how the $m_{u,d}$3 anomaly is encoded. In an extended linear sigma model, the anomalous sector was written as

$m_{u,d}$4

with the effective anomaly coupling

$m_{u,d}$5

A central result of both the 2024 and 2026 analyses was that vacuum phenomenology constrains $m_{u,d}$6 much more directly than the individual couplings, whereas the Columbia plot is highly sensitive to which operator dominates. With only the linear ’t Hooft determinant, the lower-left corner exhibits the conventional first-order region; as the quadratic determinant gains weight, that region shrinks; and in the pure quadratic case it can disappear entirely in the first quadrant, leaving a second-order/crossover structure much closer to recent lattice indications (Giacosa et al., 2024, Giacosa et al., 7 Jan 2026).

These anomaly-based studies also extended the plot into the unphysical region of negative strange mass, identified with $m_{u,d}$7, where a $m_{u,d}$8-broken phase appears and the $m_{u,d}$9 mode becomes critical. The 2026 work further introduced the “polydeterminant” as a mathematical extension of the determinant for anomalous interactions among distinct heterochiral multiplets. This suggests a broader operator-level perspective: what the Columbia plot constrains is not merely whether $m_s$0 is broken, but which anomalous operators dominate near $m_s$1 (Giacosa et al., 7 Jan 2026).

A distinct conclusion emerged from the symmetry-improved CJT formalism in a three-flavor linear sigma model. There, enforcing the chiral Ward-Takahashi identities and generalized GOR relations yielded a robust first-order region in the light-mass corner, a single tricritical point on the strange-mass axis at

$m_s$2

and, in the three-flavor symmetric limit,

$m_s$3

The same work argued that much of the sigma-mass sensitivity seen in conventional Hartree-CJT studies is a truncation artifact produced by symmetry violation. This does not remove model dependence; it relocates it to the level of symmetry implementation, truncation, and anomaly input (Guan et al., 4 Aug 2025).

6. Observables, numerical methodology, and unresolved issues

Across lattice and continuum treatments, the Columbia plot is diagnosed through approximate order parameters and their moments. In the heavy corner, the Polyakov loop and its susceptibility dominate; in the chiral corner, the condensate and Binder cumulants dominate. A standard lattice definition of standardized moments is

$m_s$4

with $m_s$5 used to locate pseudocritical couplings and $m_s$6 to test first-order, $m_s$7-critical, or crossover behavior. Finite-size scaling, multi-histogram Ferrenberg-Swendsen reweighting, and, in difficult regions, parallel tempering in $m_s$8 are recurrent techniques (Cuteri et al., 2017, Kara et al., 2021).

Several misconceptions are therefore best avoided. The Columbia plot is not a single universally fixed diagram at accessible lattice spacing; it is a continuum object approached through discretization-dependent critical lines. It is also not a purely chiral diagram: the heavy corner is deconfinement-driven and center-symmetry-based. Nor do all studies use the same diagnostics: some heavy-quark determinations rely almost entirely on Polyakov-loop susceptibility scaling, whereas many chiral studies hinge on Binder-cumulant crossings and Ising scaling. Finally, recent three-dimensional extensions at imaginary chemical potential include both completed functional calculations and explicitly programmatic lattice proposals; not every plotted critical surface is a fully determined continuum result (Sciarra, 20 Dec 2025).

The main unresolved issues are consistent across the literature. In the heavy corner, direct staggered determinations exist only at fixed $m_s$9 without continuum extrapolation. In the chiral corner, the continuum size of the first-order region may be extremely small, or possibly compatible with zero within present uncertainties. In three dimensions, the fate of the chiral critical surface away from $SU(3)$0, its relation to Roberge-Weiss tricriticality, and its connection to any real-$SU(3)$1 critical endpoint remain open. The Columbia plot is therefore best understood not as a settled phase diagram, but as a compact representation of several intertwined limits—center symmetry, chiral symmetry, anomaly realization, and cutoff control—whose quantitative reconciliation remains an active problem.