Conformal Manifolds and Duality Actions

A conformal manifold is a space of inequivalent CFTs connected by exactly marginal couplings. In four-dimensional $\mathcal N=1$ theories, holomorphy turns the local problem into marginal chiral couplings modulo complexified continuous global symmetries. This gives a powerful dimension count, but global existence, singular cusps, accidental currents, and duality identifications require additional analysis.

Required background. The SQCD fixed-point page supplies interacting endpoints, and linearized RG flow distinguishes marginal from exactly marginal directions. Helpful background. Exact, infrared, and emergent equivalence clarifies how dual frames can cover one manifold.

Marginal operators and broken currents

In a four-dimensional $\mathcal N=1$ SCFT, a supersymmetric marginal deformation has the form

$$\delta S =\int\mathrm d^4x\,\mathrm d^2\theta\, \lambda^i\mathcal O_i+\text{h.c.},$$

where $\mathcal O_i$ is a chiral primary with

$$\Delta(\mathcal O_i)=3, \qquad R(\mathcal O_i)=2.$$

Superpotential monomials of R-charge two provide candidates, but R-charge alone does not make them independent chiral primaries. Gauge couplings require particular care: a variation of a holomorphic gauge coupling is represented formally by $\operatorname{Tr}W^\alpha W_\alpha$, yet at an interacting fixed point that operator can be a Konishi descendant or be related to superpotential operators. Only independent chiral-primary combinations belong in $V$. For example, $\operatorname{Tr}W^2$ is not by itself a marginal chiral primary in ordinary interacting SQCD; a free-gauge cusp must be analyzed separately Green et al. 2010, §2.4 and §4.3, PDF pp. 8–10 and 14–15.

A candidate marginal operator need not be exactly marginal. If its coupling breaks a continuous flavor current $J_a$, the current multiplet can recombine with the chiral operator into a long multiplet. The corresponding direction becomes marginally irrelevant. This recombination is the physical origin of quotienting by broken global symmetries.

The local quotient

Let $V$ be the vector space of marginal chiral couplings and let a continuous global group $G$ act on it. Locally, the conformal manifold is

$$\mathcal M_c\simeq\{\lambda\in V:D^a(\lambda,\bar\lambda)=0\}/G \simeq V\mathbin{/\!\!/}G_{\mathbb C},$$

near the reference SCFT and under the usual regularity assumptions. The double slash denotes the polystable complex quotient; writing an ordinary orbit space would miss closed-orbit and singular-locus qualifications. The real functions $D^a$ begin quadratically,

$$D^a\propto \lambda^i(T^a)_i{}^j\bar\lambda_j+\cdots,$$

and encode the beta functions associated with broken currents.

At a regular point of fixed orbit type, where the beta-function constraints are independent and the quotient map has locally constant orbit dimension, one may use

$$\dim_{\mathbb C}\mathcal M_c =\dim_{\mathbb C}V-\dim G+\dim H,$$

where $H$ is the stabilizer. Under the same regularity assumptions, the tangent space can be computed by linearizing the constraints and redundancies:

$$T_{[\lambda]}\mathcal M_c \simeq \frac{\ker(D\beta)_\lambda} {T_\lambda(G_{\mathbb C}\!\cdot\!\lambda)}.$$

Neither formula is universal at an enhanced or singular point. There the linearized quotient is only a first-order deformation space and can overcount directions obstructed by quadratic or higher constraints. One must compute the quotient of the slice representation or its invariant ring. The component dimension, orbit-type stratum dimension, first-order deformation dimension, and Zariski-tangent dimension can differ and should be reported separately.

This count is local. It does not prove that every polystable orbit extends to finite coupling, that the metric is complete, or that distinct patches are globally connected. The broken-current criterion and local quotient construction are developed in Green, Komargodski, Seiberg, Tachikawa, and Wecht 2010, §§2–4, PDF pp. 4–15.

A one-charge quotient

Take two candidate couplings $u,v\in\mathbb C$ with $U(1)$ charges $+1$ and $-1$. The leading moment map is

$$D=|u|^2-|v|^2.$$

The complex invariant is $z=uv$, so

$$\mathbb C^2\mathbin{/\!\!/}\mathbb C^*\simeq\mathbb C_z.$$

For $z\ne0$ the stabilizer is trivial. At $u=v=0$ the full $U(1)$ stabilizes the representative, but the quotient still has one local complex coordinate, not the two predicted by blindly inserting $\dim H=1$ into $2-1+\dim H$. Moreover, the derivative of the quadratic moment map and the orbit tangent both vanish at the origin, so the naive linearized quotient also has dimension two. The nonlinear slice invariant $z=uv$, rather than stabilizer or first-order data alone, resolves the special orbit.

Beta functions in Lagrangian variables

For a superpotential

$$W=\sum_i\lambda^i\mathcal O_i,$$

holomorphy makes its beta functions proportional to wavefunction anomalous dimensions. For a monomial $\mathcal O_i=\prod_A\Phi_A^{n_{iA}}$,

$$\beta_{\lambda^i} =\lambda^i\left[ -3+\sum_A n_{iA}\left(1+\frac12\gamma_A\right) \right].$$

The gauge beta function supplies another constraint, for example the NSVZ numerator

$$3T(G)-\sum_A T(r_A)(1-\gamma_A)=0.$$

Here $\gamma_A=2(\Delta_A-1)$, matching the convention on the fixed-point page; a convention with $\widehat\gamma_A=\Delta_A-1$ replaces $1-\gamma_A$ by $1-2\widehat\gamma_A$. If the number of couplings exceeds the number of independent anomalous-dimension constraints, the remaining combinations can be exactly marginal. Redundancies among beta functions often reflect nonanomalous flavor currents; the quotient formulation makes that structure intrinsic.

Solving these equations at one perturbative order suggests local directions but does not establish exact marginality unless symmetry and holomorphy control higher orders or an all-order argument is available.

Example: the N=4 locus in N=1 language

Write $\mathcal N=4$ SYM as an $\mathcal N=1$ vector multiplet plus three adjoint chirals $\Phi_1,\Phi_2,\Phi_3$ with

$$W=h\,\operatorname{Tr}\Phi_1[\Phi_2,\Phi_3].$$

The complex gauge coupling $\tau$ and $h$ are both classically marginal. Extended supersymmetry relates their beta functions, leaving a complex one-dimensional locus with the full $\mathcal N=4$ symmetry. In an appropriate normalization, $h$ is tied to the gauge coupling; changing conventions changes the written relation but not the one-complex-dimensional family.

Viewed purely with $\mathcal N=1$ supersymmetry, additional cubic couplings can produce Leigh–Strassler deformations. Counting them requires quotienting by the complexified flavor transformations of the three adjoints and checking which symmetries survive. The familiar $\mathcal N=4$ line is a sublocus, not automatically the entire $\mathcal N=1$ conformal manifold; the original all-order beta-function analysis is Leigh and Strassler 1995, §§III and VII, PDF pp. 6–7 and 18–23.

The coupling $\tau$ is further related by electric–magnetic duality, but the global statement depends on which theory is being held fixed. The full modular action can permute distinct absolute theories with different global forms or genuine line-operator lattices. For one chosen absolute theory, only its stabilizer subgroup is quotiented; for the whole family, the safer object is a duality groupoid relating its members Aharony, Seiberg, and Tachikawa 2013, §§1–2, especially §2.3, PDF pp. 17–20. Weak-coupling cusps can then represent different duality frames, or different absolute theories related inside that groupoid.

Example: quartic SQCD at Nf = 2Nc

At the SQCD fixed point with $N_f=2N_c$,

$$R(Q)=R(\widetilde Q)=\frac12, \qquad R(M)=1.$$

Quartic meson operators therefore have R-charge two. One standard contraction is

$$\begin{aligned} W_{\mathrm{quartic}}=\kappa\bigg[& (Q^r_{\alpha}\widetilde Q^{\alpha}_{s}) (Q^s_{\beta}\widetilde Q^{\beta}_{r})\\ &-\frac{1}{N_c} (Q^r_{\alpha}\widetilde Q^{\alpha}_{r}) (Q^s_{\beta}\widetilde Q^{\beta}_{s})\bigg]. \end{aligned}$$

Here $r,s$ are flavor indices and $\alpha,\beta$ are color indices. The contraction preserves the diagonal $SU(N_f)$ and $U(1)_B$, not the full chiral flavor group $SU(N_f)_L\times SU(N_f)_R$. Together with the gauge coupling, $\kappa$ supplies a candidate tangent direction: the gauge and quartic Lagrangian couplings must co-vary because their beta-function constraints are dependent. This is not the claim that $\operatorname{Tr}W^2$ is independently marginal.

In the magnetic self-rank description, the quartic can be represented by a mass term for the singlet meson and, after integrating it out, by an inverse quartic magnetic coupling. Leigh and Strassler proposed that this produces a nontrivial duality action on the coupling coordinate, but they explicitly described the global argument as speculative and listed unresolved assumptions Leigh and Strassler 1995, §VIII, especially PDF pp. 23–30. The defensible conclusion is therefore a duality-supported local identification under those assumptions, not a proved global action on a completely constructed manifold.

The example illustrates the difference between local and global statements: R-charge identifies marginal candidates; quotient and beta functions test local exactly marginal directions; a separately established duality action may then relate distant coordinate patches.

Singular loci and cusps

A conformal manifold can have:

• orbifold points, where a discrete duality stabilizes a coupling;
• symmetry-enhanced strata, where additional conserved currents appear and the quotient stabilizer grows;
• weak-coupling cusps, at infinite Zamolodchikov distance or a boundary in a chosen coordinate;
• degeneration loci, where extra operators become free or a different effective description is needed.

At an enhanced-symmetry point, a marginal direction can recombine with the new current and cease to be exactly marginal. Orbit type and Zariski-tangent dimension can therefore change between strata, although the dimension of a smooth connected component is locally constant. A coordinate singularity in $\tau$ may be removed by a duality transformation, while a genuine new massless sector changes the local CFT data.

Zamolodchikov metric and duality action

Exactly marginal operators define a Hermitian metric through their two-point functions,

$$\langle\mathcal O_i(x)\overline{\mathcal O}_{\bar j}(0)\rangle =\frac{g_{i\bar j}}{|x|^6}.$$

After removing redundant directions, $g_{i\bar j}$ is the Zamolodchikov metric on $\mathcal M_c$. Where the global duality action is established, it acts by an isometry together with an operator-basis transformation. Contact terms supply a connection on the bundle of operators over theory space, so parallel transport can have nontrivial holonomy Kutasov 1989. A complementary one-loop treatment with spacetime-dependent gauge couplings and $SL(2,\mathbb R)$ covariance is given in Osborn 2003, pp. 174–182; that perturbative calculation does not by itself establish the global duality quotient.

Local quotient data do not fix this metric globally. Localization or conformal perturbation theory may compute protected parts in special theories, with counterterm ambiguities treated explicitly.

A reliable workflow

1. List every independent marginal chiral primary, retaining gauge-kinetic combinations only after descendant and equation-of-motion relations are imposed.
2. Determine the faithful continuous global symmetry acting on the couplings.
3. Remove descendants, equations-of-motion operators, and redundant directions.
4. Solve the moment-map or beta-function constraints and take the polystable quotient by $G_{\mathbb C}$.
5. At regular points, check constraint rank and orbit dimension; at special points, compute the slice or invariant ring and treat the linearized quotient only as a first-order candidate.
6. Include accidental currents and recompute the local data on every affected stratum.
7. Identify discrete dualities, cusps, global forms, line-operator lattices, and stabilizers.
8. Distinguish a local tangent candidate, a finite patch, and a globally established conformal manifold.

The RG, extremization, and conformal-manifold map summarizes where accidental-current corrections end and coupling-space quotienting begins. Its machine-readable record keeps the tangent-space and global-existence qualifications separate.

Common pitfalls

Counting R-charge-two operators directly. Broken currents pair with some of them and make those directions marginally irrelevant.

Using the stabilizer as a universal dimension correction. The subtraction-plus-stabilizer formula is valid only on a regular orbit-type stratum. At an enhanced or singular orbit, compute the slice or invariant ring; a linearized quotient can still overcount obstructed directions.

Inferring a global manifold from local beta functions. Obstructions, accidental symmetries, and duality identifications can appear away from the reference point.

Exercises

Two candidate marginal couplings $u,v$ have charges $+1,-1$ under a $U(1)$ global symmetry, with leading moment map $D=|u|^2-|v|^2$.

1. Find a holomorphic invariant that labels the quotient.
2. Determine the local complex dimension at a generic point and at $u=v=0$.
3. Explain why $\dim V-\dim G+\dim H$ fails at the origin.
4. State what this local calculation does not establish globally.

Solution

The invariant ring is generated by

$$z=uv,$$

so the quotient is one complex-dimensional. For $z\ne0$, both $u$ and $v$ are nonzero and the stabilizer is trivial. At the origin the representative $u=v=0$ has stabilizer $H=U(1)$, but the invariant coordinate $z$ still gives one local quotient direction.

Blind substitution would instead give

$$\dim_{\mathbb C}V-\dim G+\dim H=2-1+1=2.$$

It fails because the origin is a special orbit at which the generic orbit-dimension argument is not a local dimension theorem; the slice invariants must be computed. Here the linearized moment map also vanishes at the origin and would overcount the same obstructed first-order directions. Finally, the coordinate $z$ establishes only the local quotient model. It does not prove continuation to arbitrary finite coupling, completeness of the metric, or the absence of further duality identifications and accidental currents.

References

• Aharony, Ofer, Nathan Seiberg, and Yuji Tachikawa. “Reading between the Lines of Four-Dimensional Gauge Theories.” Journal of High Energy Physics 08 (2013): 115. arXiv:1305.0318.
• Green, Daniel, Zohar Komargodski, Nathan Seiberg, Yuji Tachikawa, and Brian Wecht. “Exactly Marginal Deformations and Global Symmetries.” Journal of High Energy Physics 06 (2010): 106. arXiv:1005.3546.
• Kutasov, David. “Geometry on the Space of Conformal Field Theories and Contact Terms.” Physics Letters B 220 (1989): 153–158. doi:10.1016/0370-2693(89)90028-2.
• Leigh, Robert G., and Matthew J. Strassler. “Exactly Marginal Operators and Duality in Four-Dimensional $\mathcal N=1$ Supersymmetric Gauge Theory.” Nuclear Physics B 447 (1995): 95–136. arXiv:hep-th/9503121.
• Osborn, Hugh. “Local Couplings and $SL(2,\mathbb R)$ Invariance for Gauge Theories at One Loop.” Physics Letters B 561 (2003): 174–182. arXiv:hep-th/0302119.