Centre-Vortex Surfaces in Lattice Gauge Theory

• Centre-vortex surfaces are closed two-dimensional structures in lattice SU(3) gauge fields marking nontrivial centre flux.
• They underpin key phenomena including confinement, topological charge generation, and dynamical chiral symmetry breaking with quantifiable density measures.
• 3D visualizations reveal their 4D connectivity, highlighting branching behavior, monopole ambiguity, and infrared dynamics.

Searching arXiv for recent and foundational papers on centre-vortex surfaces. Centre-vortex surfaces are the closed two-dimensional world-surfaces that arise when nontrivial centre flux is identified on centre-projected plaquettes of lattice $SU(3)$ gauge fields. In this formulation, the projected objects are thin $Z_3$ representatives of thicker infrared vortex structure in the underlying gauge field: in four-dimensional Euclidean space-time they are surfaces, while a three-dimensional slice shows only line-like intersections and indicators of temporal orientation. Within the lattice centre-vortex program, these surfaces organize the discussion of confinement, topological charge, dynamical chiral symmetry breaking, and finite-temperature phase structure (Biddle et al., 2019, Mickley et al., 23 Jul 2025).

1. Lattice definition and centre projection

For $SU(3)$, the relevant centre is

$$Z_3=\left\{\exp\!\left(2\pi i\,\frac{m}{3}\right)\mathbf I \;\middle|\; m=-1,0,1\right\}.$$

The standard identification procedure first fixes the gauge field to Maximal Centre Gauge by seeking a gauge transformation $\Omega$ that maximizes

$$\sum_{x,\mu}\left|\mathrm{tr}\,U_\mu^\Omega(x)\right|^2 \stackrel{\Omega}{\to}\mathrm{max}.$$

This drives each link as close as possible to a centre element. After gauge fixing, each link is projected to the nearest centre element,

$$U_\mu(x)\to Z_\mu(x),\qquad Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m_\mu(x)}{3}\right)\mathbf I,\qquad m_\mu(x)=-1,0,1,$$

and the projected plaquette centre charge is

$$z=\prod_\Box Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m}{3}\right).$$

A plaquette with $z=1$ is trivial, while $z=e^{\pm 2\pi i/3}$ signals a thin centre vortex piercing that plaquette. Equivalently, a Wilson loop linked with such flux acquires a nontrivial centre phase (Biddle et al., 2019, Leinweber et al., 2022).

This lattice construction is routinely expressed as a decomposition

$Z_3$0

where $Z_3$1 is the centre-projected part and $Z_3$2 is the remainder. The corresponding vortex-removed field is

$Z_3$3

or equivalently

$Z_3$4

In this language, the vortex-only ensemble isolates the thin projected vortex content, whereas the vortex-removed ensemble removes the identified centre degrees of freedom from the original configuration (Biddle et al., 2018, Trewartha et al., 2014).

The thin projected objects are not presented as literal physical gauge fields. Rather, they are treated as plaquette-level markers of thick physical vortices in the original Monte Carlo configurations. This distinction between thin projected vortices and thick physical vortices is central to the interpretation of centre-vortex surfaces across the lattice literature (Biddle et al., 2018, Kamleh et al., 2022).

2. World-sheet geometry in four dimensions

The nontrivial projected plaquettes assemble into closed two-dimensional surfaces in four-dimensional Euclidean space-time. This closure is the lattice expression of centre-flux conservation modulo the centre, and in dual-lattice language these plaquettes tile the vortex world-sheet. In a fixed three-dimensional time slice, the same object appears only through its one-dimensional intersection with that slice, producing closed lines or line segments rather than the full surface (Biddle et al., 2019, Biddle et al., 2020).

This distinction between four-dimensional connectivity and three-dimensional appearance is structurally important. A single curved 4D sheet can intersect a 3D slice in several disconnected loops, so loops that look separate in a slice need not correspond to distinct vortex objects. An explicit four-dimensional reconstruction in pure $Z_3$5 gauge theory identifies the largest connected 4D object as the primary sheet and shows that many secondary loops in ordinary 3D visualisations are merely different cross-sections of that same sheet. In that study, the primary sheet contains above $Z_3$6 of vortex plaquettes at all temperatures except for a pronounced dip just above the deconfinement transition, and the normalized maximal separation of vortices in the primary sheet satisfies $Z_3$7 at every temperature studied, indicating full 4D percolation (Mickley et al., 23 Jul 2025).

The same reconstruction clarifies the status of secondary structures. Below the deconfinement temperature, most secondary loops in temporal slices are connected to the primary 4D sheet rather than being independent sheets; above the transition, connected secondary loops become vanishingly rare because the dominant sheet aligns with the temporal direction. The density of disconnected secondary sheets is found to be roughly

$Z_3$8

apart from a spike just above the transition (Mickley et al., 23 Jul 2025).

The world-sheet description therefore has two complementary levels. At the microscopic projected level, vortex plaquettes form closed 2D surfaces in 4D. At the observational level of slices and plots, one sees evolving 1D traces whose apparent fragmentation can be a slicing artifact rather than a statement about actual disconnectedness of the underlying surface (Biddle et al., 2020, Mickley et al., 23 Jul 2025).

3. Slice visualisation and orientation conventions

The modern visualisation program for centre-vortex surfaces represents the 4D world-sheet by combining spatial plaquettes and mixed space-time plaquettes in a single 3D display. Purely spatial plaquettes are rendered directly as jets. For example, for a plaquette in the $Z_3$9-$SU(3)$0 plane, an $SU(3)$1 vortex with

$SU(3)$2

is drawn as a blue jet pointing in the $SU(3)$3 direction, whereas an $SU(3)$4 vortex with

$SU(3)$5

is drawn as a red jet pointing in the $SU(3)$6 direction. The same rule applies cyclically to the other spatial plaquettes, with the right-handed ordering fixing the orientation (Biddle et al., 2019, Leinweber et al., 2022).

Plaquettes involving the Euclidean time direction cannot be drawn literally in the hidden fourth dimension, so they are encoded by decorating spatial links. If a space-time plaquette carries nontrivial centre charge, the corresponding spatial link is rendered in the 3D slice: cyan for $SU(3)$7, orange for $SU(3)$8, with the arrow indicating forward or backward time orientation. In the compact visual language of the 2020 proceedings study, blue and red jets denote spatial plaquettes with $SU(3)$9, cyan and orange arrows denote mixed space-time plaquettes with $$Z_3=\left\{\exp\!\left(2\pi i\,\frac{m}{3}\right)\mathbf I \;\middle|\; m=-1,0,1\right\}.$$0, and spheres denote singular points (Biddle et al., 2020).

This combined rendering makes visible how the vortex surface moves from one slice to the next. The space-time indicator links reveal ladder-like structures, sheets approximately parallel to the time direction, and cases where a spatial vortex line shifts by several lattice spacings between adjacent time slices. The addition of time-oriented plaquettes is therefore not a cosmetic supplement to the spatial plot; it is what turns a collection of local piercings into a readable world-sheet geometry (Biddle et al., 2019, Biddle et al., 2019).

These visualisations also establish several qualitative geometric facts that are difficult to infer from bulk observables alone. Vortex lines in a slice are continuous, closure of loops is directly visible, large connected structures dominate typical configurations, and branch points and singular points occur as localized features of larger coherent surfaces rather than as isolated random defects (Biddle et al., 2020, Leinweber et al., 2022).

4. Branching, monopole ambiguity, and topological structure

A defining $$Z_3=\left\{\exp\!\left(2\pi i\,\frac{m}{3}\right)\mathbf I \;\middle|\; m=-1,0,1\right\}.$$1 feature of centre-vortex surfaces is branching. Because centre flux is conserved only modulo $$Z_3=\left\{\exp\!\left(2\pi i\,\frac{m}{3}\right)\mathbf I \;\middle|\; m=-1,0,1\right\}.$$2, nontrivial surface elements can split or join while maintaining consistent net centre flux. In slice visualisations this appears as points where three jets emerge from or converge on a point. The same local structure can also be interpreted as monopole or anti-monopole content, because

$$Z_3=\left\{\exp\!\left(2\pi i\,\frac{m}{3}\right)\mathbf I \;\middle|\; m=-1,0,1\right\}.$$3

may be read either as oppositely directed $$Z_3=\left\{\exp\!\left(2\pi i\,\frac{m}{3}\right)\mathbf I \;\middle|\; m=-1,0,1\right\}.$$4 flux or as two units of $$Z_3=\left\{\exp\!\left(2\pi i\,\frac{m}{3}\right)\mathbf I \;\middle|\; m=-1,0,1\right\}.$$5 flux modulo the centre. The literature therefore treats branching and monopole–antimonopole dynamics as closely intertwined in the projected picture (Biddle et al., 2019, Leinweber et al., 2022).

In three-dimensional slice language, branchings can be classified by the number $$Z_3=\left\{\exp\!\left(2\pi i\,\frac{m}{3}\right)\mathbf I \;\middle|\; m=-1,0,1\right\}.$$6 of vortex lines piercing the faces of an elementary cube. The possibilities include $$Z_3=\left\{\exp\!\left(2\pi i\,\frac{m}{3}\right)\mathbf I \;\middle|\; m=-1,0,1\right\}.$$7 for ordinary unbranched flow, $$Z_3=\left\{\exp\!\left(2\pi i\,\frac{m}{3}\right)\mathbf I \;\middle|\; m=-1,0,1\right\}.$$8 for simple branching, $$Z_3=\left\{\exp\!\left(2\pi i\,\frac{m}{3}\right)\mathbf I \;\middle|\; m=-1,0,1\right\}.$$9 for self-intersection or touching, $\Omega$0 for more complex branching, and $\Omega$1 for self-intersection or double branching. In a sample of $\Omega$2 $\Omega$3 configurations, branchings are common: the average is $\Omega$4 branching points per 3D slice, corresponding to a density $\Omega$5 (Biddle et al., 2019).

The topological interpretation is organized around singular points. The continuum topological charge density is written as

$\Omega$6

so nonzero $\Omega$7 requires field-strength components spanning all four dimensions. In vortex-surface language, singular points are those at which the tangent vectors to the vortex surface or surfaces span all four space-time directions. Following the terminology cited in the visualisation work, these include intersection points, touching points, and writhing points (Biddle et al., 2019).

On the lattice, the practical projected signature is concrete: a singular point occurs at a site where there is both a spatial-plaquette jet and a space-time plaquette indicator link parallel to that jet. A blue jet associated with $\Omega$8-$\Omega$9 flux together with an orange link associated with $$\sum_{x,\mu}\left|\mathrm{tr}\,U_\mu^\Omega(x)\right|^2 \stackrel{\Omega}{\to}\mathrm{max}.$$0-$$\sum_{x,\mu}\left|\mathrm{tr}\,U_\mu^\Omega(x)\right|^2 \stackrel{\Omega}{\to}\mathrm{max}.$$1 flux is the standard example; taken together, the local surface spans $$\sum_{x,\mu}\left|\mathrm{tr}\,U_\mu^\Omega(x)\right|^2 \stackrel{\Omega}{\to}\mathrm{max}.$$2, $$\sum_{x,\mu}\left|\mathrm{tr}\,U_\mu^\Omega(x)\right|^2 \stackrel{\Omega}{\to}\mathrm{max}.$$3, $$\sum_{x,\mu}\left|\mathrm{tr}\,U_\mu^\Omega(x)\right|^2 \stackrel{\Omega}{\to}\mathrm{max}.$$4, and $$\sum_{x,\mu}\left|\mathrm{tr}\,U_\mu^\Omega(x)\right|^2 \stackrel{\Omega}{\to}\mathrm{max}.$$5 (Biddle et al., 2019, Biddle et al., 2020).

Projected topological correlations can be quantified. In the 2019 visualisation study, singular points account exactly for the topological charge density of the projected configurations, with the normalized correlation

$$\sum_{x,\mu}\left|\mathrm{tr}\,U_\mu^\Omega(x)\right|^2 \stackrel{\Omega}{\to}\mathrm{max}.$$6

The normalized correlation between branching points and projected topological charge density is

$$\sum_{x,\mu}\left|\mathrm{tr}\,U_\mu^\Omega(x)\right|^2 \stackrel{\Omega}{\to}\mathrm{max}.$$7

and the corresponding projected vortex–topological-charge correlation is

$$\sum_{x,\mu}\left|\mathrm{tr}\,U_\mu^\Omega(x)\right|^2 \stackrel{\Omega}{\to}\mathrm{max}.$$8

The same study argues that nearby branchings provide a natural mechanism for generating high-$$\sum_{x,\mu}\left|\mathrm{tr}\,U_\mu^\Omega(x)\right|^2 \stackrel{\Omega}{\to}\mathrm{max}.$$9 regions by placing several same-sign vortex contributions into a clover arrangement (Biddle et al., 2019).

The broader interpretation advanced by these works is that centre-vortex surfaces are not only confinement-carrying objects but also geometrical carriers of nontrivial topology. Large, complex vortex structures and singular points tend to lie within regions of nontrivial topological charge density measured on the original Monte Carlo gauge fields, and vortex removal is argued to destroy nontrivial topology and destabilize would-be instantons under smoothing algorithms (Biddle et al., 2019).

5. Confinement, propagators, and dynamical chiral symmetry breaking

The centre-vortex surface program is partly defined by comparisons among untouched, vortex-only, and vortex-removed ensembles. The central physical question is whether the projected surface content carries the long-range information responsible for confinement and related infrared phenomena. In pure gauge $$U_\mu(x)\to Z_\mu(x),\qquad Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m_\mu(x)}{3}\right)\mathbf I,\qquad m_\mu(x)=-1,0,1,$$0, vortex-only configurations reproduce the long-distance linear static potential qualitatively, while vortex-removed configurations lose the long-range confining potential; however, the vortex-only string tension is only about $$U_\mu(x)\to Z_\mu(x),\qquad Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m_\mu(x)}{3}\right)\mathbf I,\qquad m_\mu(x)=-1,0,1,$$1 of the full one. In full QCD with dynamical fermions, by contrast, the vortex-only string tension is reported to be in excellent agreement with the untouched ensemble, while removing vortices again eliminates the long-distance potential (Leinweber et al., 2022, Kamleh et al., 2022).

The Landau-gauge gluon propagator provides a complementary diagnostic. On untouched configurations, $$U_\mu(x)\to Z_\mu(x),\qquad Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m_\mu(x)}{3}\right)\mathbf I,\qquad m_\mu(x)=-1,0,1,$$2 exhibits the expected nonperturbative infrared peak. On vortex-removed configurations, the infrared propagator is strongly suppressed and the ultraviolet is enhanced. On vortex-only configurations, the infrared peak remains substantial—about two thirds of the original peak strength—while most ultraviolet strength is absent. After smoothing and matching average actions between original and vortex-only ensembles, the smoothed vortex-only propagator reproduces the smoothed untouched propagator with remarkably good accuracy. This has been interpreted as evidence that projected thin vortices are skeletons or seeds of the thick infrared structures that dominate the Yang–Mills vacuum (Biddle et al., 2018).

Overlap-fermion studies sharpen the link to chiral dynamics because of their sensitivity to topology. On untouched configurations, the overlap quark propagator shows strong infrared enhancement of the mass function $$U_\mu(x)\to Z_\mu(x),\qquad Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m_\mu(x)}{3}\right)\mathbf I,\qquad m_\mu(x)=-1,0,1,$$3. After vortex removal, the propagator becomes far more tree-like and the infrared enhancement of $$U_\mu(x)\to Z_\mu(x),\qquad Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m_\mu(x)}{3}\right)\mathbf I,\qquad m_\mu(x)=-1,0,1,$$4 is almost gone, indicating a loss of dynamical mass generation. Raw vortex-only $$U_\mu(x)\to Z_\mu(x),\qquad Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m_\mu(x)}{3}\right)\mathbf I,\qquad m_\mu(x)=-1,0,1,$$5 configurations do not reproduce the mass function directly—the mass function sits on the input bare mass at all momenta—but after $$U_\mu(x)\to Z_\mu(x),\qquad Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m_\mu(x)}{3}\right)\mathbf I,\qquad m_\mu(x)=-1,0,1,$$6 sweeps of cooling, the vortex-only mass and renormalization functions become almost identical to those of cooled untouched configurations (Trewartha et al., 2014).

Hadron spectroscopy with the overlap action extends the same pattern. On vortex-removed fields, the low-lying hadron spectrum exhibits chiral partner degeneracies at light bare quark mass, consistent with chiral symmetry restoration, while at heavier masses it is consistent with weakly interacting constituent quarks. On vortex-only fields, $$U_\mu(x)\to Z_\mu(x),\qquad Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m_\mu(x)}{3}\right)\mathbf I,\qquad m_\mu(x)=-1,0,1,$$7 sweeps of cooling are sufficient to restore the smoothness required for overlap locality, and the resulting spectrum reproduces the salient features of the untouched low-lying hadron spectrum, including the pseudo-Goldstone pion and the $$U_\mu(x)\to Z_\mu(x),\qquad Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m_\mu(x)}{3}\right)\mathbf I,\qquad m_\mu(x)=-1,0,1,$$8-$$U_\mu(x)\to Z_\mu(x),\qquad Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m_\mu(x)}{3}\right)\mathbf I,\qquad m_\mu(x)=-1,0,1,$$9 splitting (Trewartha et al., 2017).

This infrared role persists in dynamical QCD. In $$z=\prod_\Box Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m}{3}\right).$$0-flavour dynamical ensembles, removal of centre vortices suppresses dynamical mass generation in the Landau-gauge overlap quark propagator, and the deep-infrared suppression of the quark renormalization function becomes more pronounced at lighter quark masses (Virgili et al., 2022).

6. Dynamical fermions and finite-temperature evolution

Dynamical fermions do not erase the centre-vortex surface picture; they substantially restructure it. Visual studies comparing pure-gauge and full $$z=\prod_\Box Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m}{3}\right).$$1-flavour QCD on matched $$z=\prod_\Box Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m}{3}\right).$$2 ensembles report a much denser and more branched vortex network in full QCD. Over $$z=\prod_\Box Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m}{3}\right).$$3 configurations and $$z=\prod_\Box Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m}{3}\right).$$4 spatial slices per configuration, the average number of vortices in the primary cluster is

$$z=\prod_\Box Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m}{3}\right).$$5

versus

$$z=\prod_\Box Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m}{3}\right).$$6

with $$z=\prod_\Box Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m}{3}\right).$$7 spatial plaquettes available per slice. The same work emphasizes increased nontrivial centre charge, more branching points, and more complex secondary loops in the presence of sea quarks (Leinweber et al., 2022, Leinweber et al., 2022).

A more quantitative geometric analysis at zero temperature confirms this increase in density and branching. For pure gauge theory, the vortex area density and branching density are

$$z=\prod_\Box Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m}{3}\right).$$8

$$z=\prod_\Box Z_\mu(x)=\exp\!\left(2\pi i\,\frac{m}{3}\right).$$9

For the $z=1$0 MeV dynamical ensemble, they become

$z=1$1

$z=1$2

In the same analysis, the fitted physical branching rate rises from

$z=1$3

in pure gauge theory to

$z=1$4

for the $z=1$5 MeV ensemble, and the average branch-to-branch separation drops from

$z=1$6

to

$z=1$7

The directed-graph representation introduced there interprets branching as an approximately binomial process with a short-distance enhancement reflecting branch-point clustering (Biddle et al., 2023).

At finite temperature in pure $z=1$8, deconfinement is presented as an abrupt geometric reorganization rather than disappearance of vortex matter. Above $z=1$9, the vortex sheet aligns with the Euclidean temporal direction, so temporal slices remain dominated by a large connected structure while spatial slices show many short disconnected lines running parallel to time. Across the transition, the temporal-slice vortex density drops from

$z=e^{\pm 2\pi i/3}$0

the spatial-slice density drops from

$z=e^{\pm 2\pi i/3}$1

and the branching-point density falls from values approaching

$z=e^{\pm 2\pi i/3}$2

below $z=e^{\pm 2\pi i/3}$3 to

$z=e^{\pm 2\pi i/3}$4

above it. The same studies report that the zero-temperature tendency of branching points to cluster at short distances vanishes at high temperature, and they interpret the change as a first-order restructuring of centre-vortex geometry (Mickley et al., 2024, Mickley et al., 28 Mar 2025).

In full QCD at finite temperature, the reported pattern is more elaborate. A 2024 study on anisotropic FASTSUM ensembles states that vortex percolation persists through the chiral transition and ceases only near

$z=e^{\pm 2\pi i/3}$5

Using cluster extent, temporal correlations, vortex densities, branching densities, secondary-cluster counts, and branching-path statistics, it argues for evidence of a second thermal transition associated with deconfinement, distinct from the chiral crossover. In the paper’s interpretation, this yields three regimes: confined and chirally broken below $z=e^{\pm 2\pi i/3}$6, confined and chirally symmetric for $z=e^{\pm 2\pi i/3}$7, and deconfined and chirally symmetric above $z=e^{\pm 2\pi i/3}$8 (Mickley et al., 2024).

7. Limitations and interpretive issues

Centre-vortex surfaces, as studied on the lattice, are projected thin vortices rather than the full thick physical vortices of the underlying gauge field. Their identification is gauge dependent because it relies on Maximal Centre Gauge and nearest-centre projection. Several studies explicitly note that practical MCG fixing is not an ideal centre gauge, that some vortex matter may remain in vortex-removed configurations, and that Gribov-copy issues in vortex identification remain relevant (Biddle et al., 2018, Biddle et al., 2019).

Observable comparisons often require smoothing, which is physically suggestive but methodologically consequential. Raw vortex-only configurations are extremely rough, so overlap-fermion and gluon-propagator studies smooth them before quantitative comparison with untouched configurations. The same literature interprets smoothing as allowing thin projected vortices to relax toward thick physical vortices or instanton-like objects, but it also recognizes that smoothing can alter short-distance structure (Trewartha et al., 2014, Biddle et al., 2018).

Topological comparisons require further caution. Singular points are identified on projected configurations, whereas topological charge density is often measured on the original Monte Carlo fields after cooling or smearing. The resulting correlations are therefore physically suggestive rather than exact one-to-one equalities on the original ensemble, even though on the projected configurations singular points account exactly for projected topological charge density (Biddle et al., 2019, Biddle et al., 2019).

A further conceptual limitation concerns dimensional reduction in visualisation. Three-dimensional slices suppress one coordinate, and mixed space-time plaquettes must be represented indirectly. More importantly, disconnected loops in a 3D slice need not be disconnected in four dimensions. The 2025 reconstruction study shows that many such loops are parts of a single connected 4D sheet, so slice-based intuition about multiplicity and cluster structure can be misleading unless supplemented by explicit 4D connectivity analysis (Biddle et al., 2020, Mickley et al., 23 Jul 2025).

Finally, the $z=e^{\pm 2\pi i/3}$9 distinction between oppositely oriented flux and multiple units of centre flux means that branching and monopole content are not sharply separable in the projected picture. This ambiguity is not incidental; it is part of the group-theoretic structure that makes $Z_3$00 centre-vortex surfaces richer than their $Z_3$01 counterparts and complicates any attempt at a purely local classification of branching, monopoles, and topological support (Biddle et al., 2019, Leinweber et al., 2022).