Six-Dimensional Origins, Compactification, and Duality Frames

Compactifying a specified six-dimensional $(2,0)$ theory on a marked torus explains why the four-dimensional coupling is modular: the coupling is the torus complex structure, and the modular map underlying S-duality is a change of homology basis. This is a transport statement conditional on the six-dimensional input, its global completion, and a controlled Kaluza–Klein limit. It does not manufacture that input, replace the intrinsic status tests developed on the preceding page, or by itself prove equivalence of every four-dimensional observable.

Required background. Five- and six-dimensional fixed-point status supplies the intrinsic $(2,0)$ input and its string lattice. Deformations, compactification, and duality flows supplies the scale-separation and endpoint criteria.

Helpful background. The $\mathcal N=4$ theory card fixes $\tau$, while duality groupoids and walls explains why changing polarization can change the four-dimensional object.

The compactification input

Before transporting anything, specify:

1. the ADE type and whether the free center-of-mass tensor is included;
2. the integral self-dual-string lattice and its discriminant group;
3. the anomaly-polynomial convention and the tensor-branch scale;
4. the marked, oriented compactification surface, including punctures or defects;
5. a polarization selecting an absolute lower-dimensional theory; and
6. a scale hierarchy that decouples Kaluza–Klein and other unwanted sectors.

These data have different logical roles. The first three are properties of the six-dimensional input; the last three define the compactification operation. A successful four-dimensional check can test their compatibility, but it cannot be recycled as independent evidence for the intrinsic existence of the input.

Reduction of a self-dual tensor on a torus

Let $T^2$ have oriented one-cycles $A$ and $B$ with $A\cdot B=+1$, area $\mathcal A$, and complex structure $\tau_T$ with $\operatorname{Im}\tau_T>0$. Let $x_A,x_B$ be unit-period coordinates dual to this marking. A convenient metric is

$$ds_{T^2}^{2} =\frac{\mathcal A}{\operatorname{Im}\tau_T} \left\lvert dx_A+\tau_T\,dx_B\right\rvert^2.$$

For an Abelian $(2,0)$ tensor field, expand the two-form zero mode as

$$B_2=A_A\wedge dx_A+A_B\wedge dx_B+\cdots .$$

Its self-dual field strength $H_3=*_{6}H_3$ relates the two four-dimensional curvatures $F_A=dA_A$ and $F_B=dA_B$. They are not two independent photons: after choosing one electric polarization, the other is its magnetic dual. A Hamiltonian reduction, or a reduction of a self-dual-field pseudo-action followed by imposition of self-duality, gives four-dimensional Maxwell theory with

$$\tau_{\rm 4d}=\tau_T$$

in the orientation and theta convention used throughout this chapter. Reversing the torus orientation changes that convention; exchanging electric and magnetic cycles implements the $S$ transformation. This is the Abelian calculation behind the geometric modular action, not a six-dimensional non-Abelian Lagrangian derivation.

The non-Abelian $(2,0)$ theory has no ordinary two-form Lagrangian to reduce. Nevertheless, its tensor branch, BPS strings, anomalies, and circle reductions consistently identify the small-torus limit with maximally supersymmetric four-dimensional Yang–Mills. The geometric origin of the modular action is developed in Vafa 1997, pp. 158–163, while protected higher-derivative terms in toroidal reductions are analyzed in Córdova, Dumitrescu, and Yin 2019, §§2–4.

Scale separation through five dimensions

Take a rectangular limit first, with circle circumferences $2\pi R_A$ and $2\pi R_B$. Reduction on the $A$ circle gives five-dimensional maximally supersymmetric Yang–Mills with

$$g_5^2\propto R_A.$$

The omitted constant depends on trace normalization. Reducing that theory on the $B$ circle gives

$$g_4^2\propto\frac{R_A}{R_B},$$

while an off-diagonal torus metric supplies the four-dimensional theta angle. More explicitly, $g_4^2=g_5^2/(2\pi R_B)$ in the usual circle-reduction normalization. Constants in $g_5^2\propto R_A$ depend on generator and instanton conventions; the invariant statement is the equality of the complexified four-dimensional coupling with the torus modulus.

The controlled four-dimensional regime is

$$E\ll \min\left(R_A^{-1},R_B^{-1}\right).$$

Equivalently, take $\mathcal A\to0$ at fixed $\tau_T$ while holding four-dimensional energies fixed. Kaluza–Klein modes then decouple. If $R_A\ll R_B$, there is instead an intermediate window

$$R_B^{-1}\ll E\ll R_A^{-1}$$

in which the $A$ circle has reduced but the $B$ circle has not. The correct description there is five-dimensional, not pure four-dimensional $\mathcal N=4$ SYM.

This sequential argument contains a logical asymmetry: five-dimensional maximally supersymmetric Yang–Mills is itself an effective description whose proposed ultraviolet completion is the six-dimensional theory. Douglas analyzes the circle relation and the conjectured recovery of the Kaluza–Klein tower in Douglas 2011, §§2–3. That relation checks the compactification dictionary but cannot serve as an independent construction of the six-dimensional input.

Mapping class group and charge transport

The orientation-preserving mapping class group of the marked torus is $SL(2,\mathbb Z)$. Let

$$M=\begin{pmatrix}a&b\\c&d\end{pmatrix}.$$

Use the passive basis convention

$$A'=dA+cB, \qquad B'=bA+aB.$$

Because $A'\cdot B'=ad-bc=1$, the new basis has the same orientation. Normalizing the $A'$ period to one sends

$$\tau_T\longmapsto\frac{a\tau_T+b}{c\tau_T+d}.$$

A six-dimensional self-dual string whose spatial direction wraps a primitive torus cycle becomes a four-dimensional BPS particle. Likewise, a six-dimensional surface defect with one direction wrapped on a torus cycle becomes a line operator. Write the same cycle in the two markings as

$$\gamma=eA+mB=e'A'+m'B'.$$

Identifying the wrapping numbers $(e,m)$ with four-dimensional electric and magnetic charges gives the passive relabeling

$$(e',m')=(ae-bm,-ce+dm),$$

and hence

$$e'+\tau_T'm' =\frac{e+\tau_Tm}{c\tau_T+d}.$$

For two cycles $\gamma=eA+mB$ and $\widetilde\gamma=\widetilde eA+\widetilde mB$, the intersection form is the antisymmetric Dirac pairing:

$$\gamma\cdot\widetilde\gamma =e\widetilde m-m\widetilde e.$$

Since an orientation-preserving mapping class preserves intersection number, it preserves mutual locality automatically. The displayed basis transformation also fixes all signs in the charge formula; stating only “$M$ acts on charges” would leave an active/passive ambiguity.

Polarization determines the global theory

The type-$G$ $(2,0)$ theory carries a finite defect group $\mathcal C=\Lambda_{\rm string}^*/\Lambda_{\rm string}$. For $A_1$ it is $\mathbb Z_2$. Compactification on $T^2$ produces electric and magnetic defect charges in $H^1(T^2,\mathcal C)$, but the relative six-dimensional theory does not by itself select a maximal mutually local subset.

A polarization is a maximal isotropic subgroup $L\subset H^1(T^2,\mathcal C)$ and chooses which charge classes are genuine. For $A_1$, the three subgroups generated by $(1,0)$, $(0,1)$, and $(1,1)$ reproduce the $SU(2)$, $SO(3)_+$, and $SO(3)_-$ line lattices in the chapter convention. A mapping class transports $L$ and may therefore map one global four-dimensional theory to another. Tachikawa derives this maximal-isotropic dictionary and its partition-vector origin in Tachikawa 2014, §§2–4. The geometric modular group acts naturally on the family; the stabilizer of one polarization is the necessary line-lattice candidate for an internal modular subgroup, while the full quantum-duality claim still requires compatible backgrounds, counterterms, and observables.

This also explains why the six-dimensional theory is naturally relative before global completion. Its partition function is a vector with components labeled by discrete flux, not a single canonical number on every closed six-manifold. Choosing a polarization and a consistent sum over flux sectors extracts an absolute four-dimensional theory. Discarding this step reproduces the false claim that every theory with a given gauge algebra is separately invariant under all of $SL(2,\mathbb Z)$. The global-theory orbit figure places this marked-torus construction beside the three $A_1$ polarizations and their modular stabilizers.

Wrapped objects and the BPS dictionary

Let a self-dual string of six-dimensional tension $T_{\rm str}$ wrap a torus cycle $\gamma=eA+mB$. Its four-dimensional mass is

$$M_\gamma=T_{\rm str}\,\ell(\gamma),$$

where $\ell(\gamma)$ is the length in the torus metric. From the displayed metric,

$$\ell(\gamma)^2 \propto \frac{\mathcal A}{\operatorname{Im}\tau_T} \,\lvert e+\tau_Tm\rvert^2.$$

After the tensor-branch scalar is rescaled to the canonically normalized four-dimensional expectation value, this becomes the modular BPS factor

$$M_\gamma^2 \propto \lvert v_{\rm can}\rvert^2 \,\frac{\lvert e+\tau m\rvert^2}{\operatorname{Im}\tau}.$$

The proportionality constant depends on the string-tension, root, and four-dimensional field normalization. The factor $\lvert e+\tau m\rvert^2/\operatorname{Im}\tau$ and the integral pairing do not. Under the passive transformation above, the numerator acquires $\lvert c\tau+d\rvert^{-2}$ while $\operatorname{Im}\tau$ acquires the same factor, so the mass is invariant.

For $N$ parallel M5-branes, the center-of-mass tensor descends to a free $U(1)$ multiplet. The interacting $A_{N-1}$ relative theory descends to the $\mathfrak{su}(N)$ sector. A comparison with $U(N)$ observables must retain the free factor; a comparison with $SU(N)$ must remove it consistently.

Interfaces and varying torus data

A torus mapping cylinder for $M$ reduces to a four-dimensional duality wall. If the torus varies over spacetime, branch cuts of $\tau$ support three-dimensional walls and degeneration loci support lower-dimensional defects. Assel and Schäfer-Nameki construct this hierarchy for elliptically fibered compactifications in Assel and Schäfer-Nameki 2016, §§2–4.

This construction has additional hypotheses: the fibration must preserve the stated supersymmetry, singular fibers require localized degrees of freedom, and monodromies must be compatible with both the charge lattice and polarization. A wall inferred from a branch cut is not fully specified until those localized sectors, global-form changes, and anomaly inflow are included. Its source and target are the polarized theories on the two sides, as required by the duality-groupoid composition law.

Compactification on a punctured higher-genus surface instead produces four-dimensional $\mathcal N=2$ class-$S$ systems. A pants decomposition selects a weak-coupling frame, while changes of decomposition generate duality transformations; puncture labels, twists, defects, and polarization data remain essential. Gaiotto introduced this duality-frame organization in Gaiotto 2012, §§2–3. This is a different theory family and is not evidence that the torus calculation alone proves all lower-dimensional dualities.

What the origin does and does not prove

The six-dimensional construction explains:

• why the four-dimensional coupling has a modular action;
• why electric and magnetic charges form an integral symplectic lattice;
• why duality transformations can change global form;
• how BPS particles and duality walls descend from wrapped objects and mapping cylinders; and
• why different weakly coupled descriptions can be coordinate frames on one compactification.

It does not by itself prove an intrinsic construction of the $(2,0)$ theory, the decoupling of every Kaluza–Klein or center-of-mass sector, or equality of all unprotected four-dimensional observables. Those are separate existence, limit, and duality claims. Brane and holographic constructions are developed on the M2/M5 page in Volume 15; Volume 16’s comparison page explains why a protected dictionary is weaker than a constructed equivalence of complete theories.

Exercises

1. Derive the passive charge map. Starting from $A'=dA+cB$, $B'=bA+aB$, and $eA+mB=e'A'+m'B'$, solve for $(e',m')$.

Solution

Equating the coefficients of $A$ and $B$ gives

$$\begin{pmatrix}e\\m\end{pmatrix} = \begin{pmatrix}d&b\\c&a\end{pmatrix} \begin{pmatrix}e'\\m'\end{pmatrix}.$$

The determinant is $ad-bc=1$, so inversion yields

$$\begin{pmatrix}e'\\m'\end{pmatrix} = \begin{pmatrix}a&-b\\-c&d\end{pmatrix} \begin{pmatrix}e\\m\end{pmatrix}.$$

2. Recover strong–weak exchange. For a rectangular torus with $\theta=0$, interchange $R_A$ and $R_B$. What happens to $g_4^2$?

Solution

Since $g_4^2\propto R_A/R_B$, the interchange sends it to a quantity proportional to $R_B/R_A$, the strong–weak inversion. With the standard $4\pi$ normalization and the oriented cycle exchange, this is $\tau\mapsto-1/\tau$.

3. Check the BPS invariant. Use the displayed transformation laws to verify that $\lvert e+\tau m\rvert^2/\operatorname{Im}\tau$ is unchanged.

Solution

The charge identity gives

$$e'+\tau'm'=\frac{e+\tau m}{c\tau+d},$$

while

$$\operatorname{Im}\tau' =\frac{\operatorname{Im}\tau}{\lvert c\tau+d\rvert^2}.$$

The same factor therefore cancels between numerator and denominator.

4. Find the missing assumption. A calculation retains a fixed torus area and studies energies comparable to $R_A^{-1}$. May it use pure four-dimensional $\mathcal N=4$ SYM?

Solution

No. Kaluza–Klein modes are then dynamical, so the four-dimensional truncation is invalid. One must use the five- or six-dimensional effective description appropriate to the hierarchy of radii. A modular charge identity does not repair a failed scale separation.

References

• Assel, Benjamin, and Sakura Schäfer-Nameki. “Six-Dimensional Origin of $\mathcal N=4$ SYM with Duality Defects.” Journal of High Energy Physics 12 (2016): 058. doi:10.1007/JHEP12(2016)058.
• Córdova, Clay, Thomas T. Dumitrescu, and Xi Yin. “Higher Derivative Terms, Toroidal Compactification, and Weyl Anomalies in Six-Dimensional $(2,0)$ Theories.” Journal of High Energy Physics 10 (2019): 128. doi:10.1007/JHEP10(2019)128.
• Douglas, Michael R. “On $D=5$ Super Yang–Mills Theory and $(2,0)$ Theory.” Journal of High Energy Physics 02 (2011): 011. doi:10.1007/JHEP02(2011)011.
• Gaiotto, Davide. “$\mathcal N=2$ Dualities.” Journal of High Energy Physics 08 (2012): 034. doi:10.1007/JHEP08(2012)034.
• Tachikawa, Yuji. “On the 6d Origin of Discrete Additional Data of 4d Gauge Theories.” Journal of High Energy Physics 05 (2014): 020. doi:10.1007/JHEP05(2014)020.
• Vafa, Cumrun. “Geometric Origin of Montonen–Olive Duality.” Advances in Theoretical and Mathematical Physics 1 (1997): 158–166. doi:10.4310/ATMP.1997.v1.n1.a6.