Circular Half-BPS Wilson Loop

• Circular half-BPS Wilson loop is a supersymmetric observable constructed on a circular contour with precise gauge and scalar couplings, preserving half of the supercharges.
• It enables non-perturbative computations via localization and matrix models in four-dimensional and three-dimensional gauge theories, yielding exact expectation values.
• Acting as a defect CFT probe, the loop bridges field theory with holographic duals, revealing intricate features like framing effects and non-planar corrections.

The circular half-BPS Wilson loop is a supersymmetric line operator supported on a round circle whose couplings are tuned so that the operator preserves half of the supercharges of the ambient theory. In four-dimensional $\mathcal N=4$ and $\mathcal N=2$ gauge theories it is built from the gauge connection and adjoint scalars, while in three-dimensional Chern–Simons–matter theories such as ABJ(M) and $\mathcal N=4$ SCSM it is naturally formulated through a superconnection that may also contain fermionic couplings. Because the operator is simultaneously a non-local observable, a BPS defect, a localization observable, and a holographic probe, it has become a standard testing ground for perturbation theory, matrix models, defect CFT kinematics, and precision AdS/CFT (Andree et al., 2010, Griguolo et al., 2013, Botao et al., 2020).

1. Definition and supersymmetric structure

In four-dimensional supersymmetric Yang–Mills theories with gauge group $SU(N)$, the relevant class of supersymmetric Wilson loops is

$$W = \frac{1}{N}\,\text{Tr}\,P\exp\oint_C d\tau\, \Bigl( i\,\dot x^\mu(\tau) A_\mu(x(\tau)) + |\dot x(\tau)|\,\Theta^I(\tau)\,\Phi_I(x(\tau)) \Bigr),$$

with $\Phi_I$ the adjoint scalars available in the theory. For the circular half-BPS loop one takes the contour to be a unit circle,

$$x^\mu(\tau) = (\cos\tau,\sin\tau,0,0),\qquad 0\le \tau<2\pi,$$

so that $|\dot x|=1$, and the scalar coupling is constant, for example $\Theta^I(\tau)=\delta^{I1}$. In $\mathcal N=4$ SYM there are six real adjoint scalars, whereas in superconformal $\mathcal N=2$0 SYM only two real adjoint scalars belong to the vector multiplet; the loop has the same structure when restricted to those two scalars (Andree et al., 2010).

The half-BPS property means that the supersymmetry variation of the combination $\mathcal N=2$1 is a total derivative along the contour, so for a closed loop the operator is invariant. For the circle this is tied to conformal symmetry: the infinite straight line is trivially half-BPS, and in a superconformal theory the circle is related to it by conformal inversion. In $\mathcal N=2$2 the circular operator preserves 16 of the 32 supercharges; in $\mathcal N=2$3 it preserves 8 supercharges (Andree et al., 2010, Buchbinder et al., 2012).

In ABJ(M), the circular half-BPS operator is not purely bosonic. It is defined through a $\mathcal N=2$4 superconnection whose diagonal blocks are bosonic connections and whose off-diagonal blocks couple to bi-fundamental fermions. For the unit circle in the plane, the scalar coupling matrices are

$\mathcal N=2$5

and the fermionic couplings factorize with explicit $\mathcal N=2$6-dependent spinors. The operator preserves half of the 24 supercharges because supersymmetry acts on the superconnection as a super-gauge transformation rather than by strict invariance of each matrix entry (Griguolo et al., 2013).

2. Exact and perturbative results in four dimensions

For the planar circular half-BPS loop in $\mathcal N=2$7 SYM, localization gives the exact result

$\mathcal N=2$8

and the large-$\mathcal N=2$9 asymptotics

$\mathcal N=4$0

Equivalently,

$\mathcal N=4$1

A special latitude loop in the same family has trivial expectation value, $\mathcal N=4$2, and the ratio $\mathcal N=4$3 is therefore identical to $\mathcal N=4$4 itself (Botao et al., 2020).

In superconformal $\mathcal N=4$5 SYM with gauge group $\mathcal N=4$6 and $\mathcal N=4$7 fundamental hypermultiplets, a three-loop comparison with $\mathcal N=4$8 shows that the expectation values agree through $\mathcal N=4$9. At $SU(N)$0, the difference is finite, proportional to the bare propagator structure, and entirely generated by two-matter-loop corrections. For the circular half-BPS loop one obtains

$SU(N)$1

in exact agreement with Pestun’s matrix model. The same calculation also yields a shape-independent propagator dressing for closed loops of the same supersymmetric type, so the circle is the simplest instance of a broader closed-loop statement (Andree et al., 2010).

The non-conformal extension displays a further structural refinement. For four-dimensional $SU(N)$2, $SU(N)$3 SYM with massless hypermultiplets and non-vanishing $SU(N)$4-function, localization on $SU(N)$5 leads to an interacting matrix model with

$SU(N)$6

In the regime

$SU(N)$7

the matrix-model prediction matches the flat-space circular half-BPS Wilson loop through $SU(N)$8. The three-loop answer can be written as

$SU(N)$9

so the conformal $$W = \frac{1}{N}\,\text{Tr}\,P\exp\oint_C d\tau\, \Bigl( i\,\dot x^\mu(\tau) A_\mu(x(\tau)) + |\dot x(\tau)|\,\Theta^I(\tau)\,\Phi_I(x(\tau)) \Bigr),$$0 structure survives, while an additional $$W = \frac{1}{N}\,\text{Tr}\,P\exp\oint_C d\tau\, \Bigl( i\,\dot x^\mu(\tau) A_\mu(x(\tau)) + |\dot x(\tau)|\,\Theta^I(\tau)\,\Phi_I(x(\tau)) \Bigr),$$1-dependent term appears through interference between evanescent contributions and UV renormalization (Billo' et al., 2024).

3. Quiver generalizations and large-$$W = \frac{1}{N}\,\text{Tr}\,P\exp\oint_C d\tau\, \Bigl( i\,\dot x^\mu(\tau) A_\mu(x(\tau)) + |\dot x(\tau)|\,\Theta^I(\tau)\,\Phi_I(x(\tau)) \Bigr),$$2 limits

In the $$W = \frac{1}{N}\,\text{Tr}\,P\exp\oint_C d\tau\, \Bigl( i\,\dot x^\mu(\tau) A_\mu(x(\tau)) + |\dot x(\tau)|\,\Theta^I(\tau)\,\Phi_I(x(\tau)) \Bigr),$$3 superconformal $$W = \frac{1}{N}\,\text{Tr}\,P\exp\oint_C d\tau\, \Bigl( i\,\dot x^\mu(\tau) A_\mu(x(\tau)) + |\dot x(\tau)|\,\Theta^I(\tau)\,\Phi_I(x(\tau)) \Bigr),$$4 quiver with two bifundamental hypermultiplets, each node admits a half-BPS circular loop

$$W = \frac{1}{N}\,\text{Tr}\,P\exp\oint_C d\tau\, \Bigl( i\,\dot x^\mu(\tau) A_\mu(x(\tau)) + |\dot x(\tau)|\,\Theta^I(\tau)\,\Phi_I(x(\tau)) \Bigr),$$5

At the symmetric orbifold point $$W = \frac{1}{N}\,\text{Tr}\,P\exp\oint_C d\tau\, \Bigl( i\,\dot x^\mu(\tau) A_\mu(x(\tau)) + |\dot x(\tau)|\,\Theta^I(\tau)\,\Phi_I(x(\tau)) \Bigr),$$6, planar equivalence implies

$$W = \frac{1}{N}\,\text{Tr}\,P\exp\oint_C d\tau\, \Bigl( i\,\dot x^\mu(\tau) A_\mu(x(\tau)) + |\dot x(\tau)|\,\Theta^I(\tau)\,\Phi_I(x(\tau)) \Bigr),$$7

The first non-planar correction is parameterized by

$$W = \frac{1}{N}\,\text{Tr}\,P\exp\oint_C d\tau\, \Bigl( i\,\dot x^\mu(\tau) A_\mu(x(\tau)) + |\dot x(\tau)|\,\Theta^I(\tau)\,\Phi_I(x(\tau)) \Bigr),$$8

For the orbifold theory, the strong-coupling analysis and numerical study support

$$W = \frac{1}{N}\,\text{Tr}\,P\exp\oint_C d\tau\, \Bigl( i\,\dot x^\mu(\tau) A_\mu(x(\tau)) + |\dot x(\tau)|\,\Theta^I(\tau)\,\Phi_I(x(\tau)) \Bigr),$$9

with the same $\Phi_I$0 exponent as in $\Phi_I$1, although the coefficient differs. The same study also derives

$\Phi_I$2

linking the non-planar Wilson loop correction to the derivative of the $\Phi_I$3 free energy (Beccaria et al., 2021).

For long circular quivers $\Phi_I$4 with a smooth coupling profile $\Phi_I$5, the half-BPS circular loop at node $\Phi_I$6 localizes to

$\Phi_I$7

In the large-$\Phi_I$8 limit, the coupled saddle-point system simplifies and the leading solution is a local Wigner semicircle for any smooth $\Phi_I$9: $$x^\mu(\tau) = (\cos\tau,\sin\tau,0,0),\qquad 0\le \tau<2\pi,$$0 Higher corrections are organized in powers of $$x^\mu(\tau) = (\cos\tau,\sin\tau,0,0),\qquad 0\le \tau<2\pi,$$1. Combining strong coupling with $$x^\mu(\tau) = (\cos\tau,\sin\tau,0,0),\qquad 0\le \tau<2\pi,$$2 defines a double-scaling limit in which an integro-differential equation with a $$x^\mu(\tau) = (\cos\tau,\sin\tau,0,0),\qquad 0\le \tau<2\pi,$$3 term emerges, and the paper interprets this as dynamics along a fifth dimension generated by the quiver direction (Sobko, 2024).

4. Three-dimensional realizations: superconnections, framing, and cohomology

For the circular half-BPS Wilson loop in $$x^\mu(\tau) = (\cos\tau,\sin\tau,0,0),\qquad 0\le \tau<2\pi,$$4 ABJ(M), the complete two-loop perturbative calculation at framing zero gives

$$x^\mu(\tau) = (\cos\tau,\sin\tau,0,0),\qquad 0\le \tau<2\pi,$$5

Localization computes the same observable at framing one, and the relation is

$$x^\mu(\tau) = (\cos\tau,\sin\tau,0,0),\qquad 0\le \tau<2\pi,$$6

A central subtlety is that, although the half-BPS loop is classically cohomologically equivalent to a combination of bosonic $$x^\mu(\tau) = (\cos\tau,\sin\tau,0,0),\qquad 0\le \tau<2\pi,$$7-BPS loops, this equivalence is violated at framing zero by quantum effects. The missing contributions come from fermionic diagrams, and their finite values are controlled by “pivotal evanescent terms” in DRED (Griguolo et al., 2013).

The multiply wound great-circle generalization makes the framing structure more explicit. At framing zero, the $$x^\mu(\tau) = (\cos\tau,\sin\tau,0,0),\qquad 0\le \tau<2\pi,$$8-wound half-BPS loop obeys

$$x^\mu(\tau) = (\cos\tau,\sin\tau,0,0),\qquad 0\le \tau<2\pi,$$9

while the framing-one localization result satisfies

$|\dot x|=1$0

The comparison isolates the framing dependence of the fermionic sector and suggests that the special cohomological relation at framing one is sharper than at framing zero (Bianchi, 2016).

Three-dimensional $|\dot x|=1$1 super Chern–Simons–matter theories obtained by orbifolding ABJM admit circular Drukker–Trancanelli-type half-BPS loops built from superconnections with both scalar and fermionic couplings. In one explicit construction, the circular contour is

$|\dot x|=1$2

the scalar matrices are

$|\dot x|=1$3

and the fermionic couplings carry $|\dot x|=1$4 phases. The difference between the circular $|\dot x|=1$5-BPS and $|\dot x|=1$6-BPS loops is $|\dot x|=1$7-exact,

$|\dot x|=1$8

which is the key input for localization. In the M-theory regime, the dual probe M2-brane has on-shell action

$|\dot x|=1$9

More general circular quiver $\Theta^I(\tau)=\delta^{I1}$0 theories exhibit a large family of fermionic $\Theta^I(\tau)=\delta^{I1}$1-BPS loops, and special values of the couplings recover the known $\Theta^I(\tau)=\delta^{I1}$2-BPS circular loops as block-diagonal limits (Ouyang et al., 2015, Mauri et al., 2017).

5. Holographic descriptions and precision tests

For the $\Theta^I(\tau)=\delta^{I1}$3 circular half-BPS loop in the planar limit, the dual string worldsheet in $\Theta^I(\tau)=\delta^{I1}$4 has the same $\Theta^I(\tau)=\delta^{I1}$5 embedding as the special latitude, but the latitude carries three bosonic zero modes associated with an $\Theta^I(\tau)=\delta^{I1}$6 family of degenerate embeddings. The ratio of string partition functions is

$\Theta^I(\tau)=\delta^{I1}$7

which matches the exact field-theory result because $\Theta^I(\tau)=\delta^{I1}$8. In the Gel’fand–Yaglom analysis, all non-zero-mode contributions cancel in the ratio, and the entire next-to-leading prefactor comes from the zero-mode degeneracy of the latitude solution (Botao et al., 2020).

The ABJM counterpart in $\Theta^I(\tau)=\delta^{I1}$9 is obtained by comparing the $\mathcal N=4$0-BPS circle with a fermionic latitude whose expectation value becomes trivial at a special point. The classical worldsheet action gives

$\mathcal N=4$1

and the full semiclassical ratio is

$\mathcal N=4$2

Taking the trivial-latitude limit yields

$\mathcal N=4$3

A crucial technical point is that exact agreement requires mixed boundary conditions for the massless fermions and a diffeomorphism-invariant IR cutoff (Medina-Rincon, 2019).

Precision tests around the circular background are not uniformly trivial. A Gel’fand–Yaglom computation of $\mathcal N=4$4-BPS latitude-to-circle ratios in $\mathcal N=4$5 found a mismatch with localization encoded in the remainder

$\mathcal N=4$6

This does not alter the exact status of the half-BPS circle itself, but it shows that one-loop normalization, measure factors, or boundary-condition choices around the circular background are technically delicate in strong-coupling string calculations (Forini et al., 2015).

6. Defect-CFT viewpoint and related structures

The circular half-BPS loop defines a conformal defect. In planar $\mathcal N=4$7 SYM, the subgroup of the conformal group preserving the circle is $\mathcal N=4$8, and the normalized correlator with two local operators is fixed to

$\mathcal N=4$9

where the two invariants are

$\mathcal N=2$00

For the simplest choice of dimension-two chiral primaries, weak coupling gives

$\mathcal N=2$01

At strong coupling, when both insertions are light, the leading semiclassical contribution factorizes into the product of the two one-point functions in the circular defect background (Buchbinder et al., 2012).

A recurrent misconception is that the straight line and the circle are completely equivalent. Conformal symmetry relates them, but their expectation values are not the same: $\mathcal N=2$02 What does survive the map is the equality of normalized correlators with local operator insertions, once the operator positions are transformed consistently (Buchbinder et al., 2012).

The circle also sits inside wider families of supersymmetric loops. In superconformal $\mathcal N=2$03 SYM, the three-loop “difference” with $\mathcal N=2$04 is valid for closed Wilson loops of general shape, because the effect reduces to a finite multiplicative dressing of the common-sector propagator. For planar Zarembo loops, the combined gauge–scalar exchange kernel vanishes identically, so the trivial expectation value is preserved at least up to $\mathcal N=2$05. For longitudes on $\mathcal N=2$06, the same propagator-level dressing implies that the $\mathcal N=2$07 correction remains proportional to the enclosed area (Andree et al., 2010).

Across these realizations, the circular half-BPS Wilson loop functions as a unifying observable: a protected non-local operator whose exact matrix-model description, perturbative structure, defect-CFT kinematics, and string or M-brane duals can all be confronted quantitatively. The common pattern is that the circle is simple enough to be computable, yet rigid enough to expose subtle issues such as framing, evanescent anomalies, zero-mode measures, and non-planar or non-conformal deformations.