ABJM, M2 branes, and the M-theory/type-IIA regimes

Increasing the ABJM rank does not by itself tell us whether eleven-dimensional supergravity or ten-dimensional string theory is the useful description. The integer level changes a compact circle as well as the field-theory coupling. We will calculate the relevant dimensionless parameters and then use the resulting dictionary in a gravitational observable.

Prepare with the one-brane vacuum quotient, the two large-rank limits, and the first AdS/CFT dictionary. The field theory is three-dimensional compact $U(N)_k\times U(N)_{-k}$, rank $N$ at each node, positive integer $k$, with no additional common-center quotient. Put $\lambda=N/k$. This is dimensionless; it is not a Yang–Mills coupling. We use the conformal vacuum, with no real masses, FI deformation or Higgs expectation value.

The generic theory has twelve real Poincaré and twelve conformal supercharges, while its $\mathcal N=2$ action displays four Poincaré charges. The standard $k=1,2$ theories have the enhancement taught in the enhancement lesson. The proposed bulk background preserves twenty-four supersymmetries generically and thirty-two at those exceptional levels with the standard supersymmetric lift. The approximation tests below do not count supercharges or derive the duality.

From the vacuum cone to the compact factor

An M2 brane is a membrane with two spatial worldvolume directions. Its low-energy worldvolume theory therefore has three spacetime dimensions; eight directions are transverse in eleven-dimensional spacetime. The rank-one quotient already derived in the field theory is $\mathbb C^4/\mathbb Z_k$, where the four complex coordinates have a common phase identification $z_A\sim e^{2\pi i/k}z_A$.

Separate a radial coordinate by $r^2=\sum_A|z_A|^2$. At fixed nonzero $r$, the angular space is $S^7/\mathbb Z_k$. The finite action has no fixed points on the sphere: a nonzero vector cannot equal itself times a nontrivial common phase. The singularity is at the tip of the cone. The proposed near-horizon geometry is consequently $\mathrm{AdS}_4\times S^7/\mathbb Z_k$, with the radii and supporting flux of AB (4.1)–(4.4). This geometric agreement is evidence for the brane interpretation, not a proof that matching a vacuum space determines a complete quantum theory.

Let $R$ be the radius of the covering seven-sphere and $\ell_p$ the eleven-dimensional Planck length. In the leading two-derivative supergravity dictionary,

$$ds_{11}^2=R^2\left(\frac14ds^2_{\mathrm{AdS}_4,\mathrm{unit}} +ds^2_{S^7/\mathbb Z_k,\mathrm{unit}}\right), \qquad L=\frac R2. \tag{X03.1}$$

$L$ is the physical AdS radius. The factor $1/4$ cannot be omitted when converting a mass or Newton constant into AdS units. We use the GR curvature convention: $\mathcal R_{\mathrm{AdS}_4}=-12/L^2$ and $\mathcal R_{S^7}=42/R^2$. The target curved-SUSY paper uses the opposite curvature sign.

On the covering space, the sourced radius relation is $R^6=32\pi^2N'\ell_p^6$, where $N'$ is the covering flux. The quotient has $1/k$ of the covering volume, with the same local flux density. Its flux integer is therefore $N=N'/k$. Substitution gives

$$\frac R{\ell_p}=(32\pi^2kN)^{1/6},\qquad \frac L{\ell_p}=\frac12(32\pi^2kN)^{1/6}. \tag{X03.2}$$

This explains why both rank and level enter the curvature scale. The $N$ here is the brane/flux integer in the leading dictionary, not a dimension of a Lie algebra. Quantum-corrected charge definitions are discussed below.

A finite quotient leaves a circle

The Hopf description writes a unit seven-sphere as a circle over $\mathbb{CP}^3$. The base $\mathbb{CP}^3$ is the set of complex lines through the origin in $\mathbb C^4$: vectors of unit norm differing by any common phase represent the same base point. It has six real dimensions. A local base connection $\omega$ records how the fiber phase changes between patches.

With an angle $\varphi$ of period $2\pi$, the quotient metric is

$$ds^2_{S^7/\mathbb Z_k,\mathrm{unit}} =\frac1{k^2}(d\varphi+k\omega)^2+ds^2_{\mathbb{CP}^3}, \qquad R_{11}=\frac Rk. \tag{X03.3}$$

$R_{11}$ is the physical fiber radius; its circumference is $2\pi R_{11}$. The normalized connection has first Chern number $k$ on a positively oriented $\mathbb{CP}^1$: in this normalization $\int_{\mathbb{CP}^1}d\omega=2\pi$. This is the geometric origin of the type-IIA two-form flux integer $k$ in AB (4.5). Dividing by the finite group shortens the circle. Reducing along the remaining continuous circle is a further operation.

Two eleven-dimensional length tests now appear:

$$\frac{\ell_p^2}{L^2}=\frac4{(32\pi^2kN)^{1/3}},\qquad \frac{R_{11}}{\ell_p}= \left(\frac{32\pi^2N}{k^5}\right)^{1/6}. \tag{X03.4}$$

The first is small when local curvature is small in Planck units. The second asks whether the compact circle is itself large in those units. Making $Nk$ large can improve local curvature while making the circle small. A simple eleven-dimensional supergravity treatment with all these lengths large requires both tests; parametrically, $N\gg k^5$ suffices at positive integer $k$. Failure of that condition does not mean that M-theory ceases to exist.

Reduce to type IIA and keep two expansion parameters

Type IIA string theory describes the small-circle limit of M-theory. Let $\ell_s=\sqrt{\alpha'}$ be the string length and $g_s=e^\phi$ the constant string coupling. The standard circle dictionary has $R_{11}=g_s\ell_s$ and $\ell_p=g_s^{1/3}\ell_s$. Thus $g_s=(R_{11}/\ell_p)^{3/2}$.

The string-frame metric uses a different constant rescaling from the eleven-dimensional metric. Define $R_{\rm str}$ by

$$ds^2_{10,\rm str}=R_{\rm str}^2 \left(\frac14ds^2_{\mathrm{AdS}_4,\mathrm{unit}} +ds^2_{\mathbb{CP}^3}\right). \tag{X03.5}$$

In dimensionless metrics, the reduction convention is

$$\frac{ds_{11}^2}{\ell_p^2} =e^{-2\phi/3}\frac{ds_{10,\rm str}^2}{\ell_s^2} +e^{4\phi/3}(d\varphi+k\omega)^2.$$

The fiber coefficient gives $e^{4\phi/3}=(R/(k\ell_p))^2$. Matching the base then gives $R_{\rm str}^2/\alpha'=e^{2\phi/3}(R/\ell_p)^2=(R/\ell_p)^3/k$. Together with (X03.2), this is the restored-unit version of AB (4.5)–(4.7):

$$\begin{aligned} \frac{R_{\rm str}^2}{\alpha'}&=4\pi\sqrt{2\lambda},& \frac{L_{\rm str}^2}{\alpha'}&=\pi\sqrt{2\lambda},\\ g_s^2&=4\pi\sqrt2\,\frac{N^{1/2}}{k^{5/2}} =4\pi\sqrt2\,\frac{\lambda^{5/2}}{N^2},& L_{\rm str}&=R_{\rm str}/2. \end{aligned} \tag{X03.6}$$

The two useful small parameters are $\alpha'/L_{\rm str}^2$, controlling string-scale curvature, and $g_s^2$, controlling the ordinary string-loop expansion at fixed geometry. Numerical factors are explicit in (X03.6); parametric weakly curved and weakly coupled IIA requires

$$1\ll\lambda\ll N^{4/5}, \qquad\hbox{equivalently}\qquad N^{1/5}\ll k\ll N. \tag{X03.7}$$

These inequalities define approximation regimes, not sharp phase boundaries. Near $k\sim N^{1/5}$ the circle is of Planck size and the string coupling is of order one. Neither of the two elementary perturbative descriptions is parametrically justified there. For $k\gg N$, $\lambda\ll1$ and field-theory perturbation theory is useful, while the string geometry is strongly curved.

Limit or region	What happens	Useful approximation
Fixed positive $k$, $N\to\infty$	Both eleven-dimensional radii grow in Planck units; the formal IIA coupling grows	Eleven-dimensional supergravity at leading order
Fixed large $\lambda$, $N\to\infty$	String loops vanish; string curvature stays fixed and small	Planar string theory; supergravity also needs the chosen large $\lambda$
$1\ll\lambda\ll N^{4/5}$	Both string parameters are small	Weakly curved, weakly coupled type IIA
Fixed small $\lambda$	Large $N$ suppresses loops but does not make the geometry weakly curved	Field-theory methods; a supergravity truncation is uncontrolled

For a concrete sequence, choose $(N,k)=(n^8,n^2)$. Then $\lambda=n^6$, $\alpha'/L_{\rm str}^2\propto n^{-3}$, $g_s^2\propto n^{-1}$ and $R_{11}/\ell_p\propto n^{-1/3}$. Local eleven-dimensional curvature also decreases, but the circle shrinks in Planck units. The two simultaneous IIA tests identify the useful description more precisely than “large rank.”

The constants still matter at finite rank. At $n=10$, the classical formulas give $g_s^2\simeq1.78$, even though $\alpha'/L_{\rm str}^2\simeq2.25\times10^{-4}$. At $n=1000$ these become approximately $0.0178$ and $2.25\times10^{-10}$. A favorable asymptotic power is not by itself a finite-input error estimate.

The compact factor is still dynamically relevant. Its curvature scale is comparable to the AdS scale, not arbitrarily shorter. Only the circle becomes parametrically short in (X03.3); the six-dimensional base does not disappear from all bulk processes. The mode exercise below quantifies this distinction.

Use the dictionary in a gravitational observable

A four-dimensional gravitational coupling provides another useful measure of bulk quantum effects. Let $G_4$ be Newton’s constant in the same metric frame as the AdS radius. The leading reduction gives

$$\frac{L^2}{G_4}=\frac{2\sqrt2}{3}\sqrt{k}\,N^{3/2}, \qquad \frac{G_4}{L^2}=\frac3{2\sqrt2\sqrt{k}\,N^{3/2}}. \tag{X03.8}$$

This is the leading unshifted form of DMP (5.44). A constant change of metric frame changes $L^2$ and $G_4$ by the same factor, leaving their ratio invariant. Small $G_4/L^2$ is a bulk gravity test at the AdS scale; it does not by itself bound the string curvature or string coupling in (X03.6).

For a round $S^3$ conformal boundary, the renormalized Euclidean AdS action is $I_{\rm ren}=\pi L^2/(2G_4)$. Since the site’s convention is $F=-\log Z$ and $Z\simeq e^{-I_{\rm ren}}$, the leading prediction is

$$F_{S^3}^{\rm grav}\simeq\frac{\pi\sqrt{2k}}3N^{3/2} =\frac{\pi\sqrt2}{3}\frac{N^2}{\sqrt\lambda}. \tag{X03.9}$$

The source calls $\log Z$ its free energy, so its displayed leading sign is opposite. The last exercise reconstructs the finite gravitational action rather than discarding an infinite volume. This is the sphere free energy, not the thermal free energy on flat space. The separate matrix-integral asymptotics in the large-rank lesson test the same leading coefficient with different methods and stated orders of limits.

All radii and couplings in (X03.2)–(X03.9) are leading classical relations. Integer brane charge, quantized Page charge and the flux including induced contributions need not agree once higher-derivative and flux effects are included. For example, DMP (5.6)–(5.10) distinguishes the shifted planar variable $\widehat\lambda=\lambda-1/24$ on the equal-rank slice. This is not an instruction to replace $N$ by a shifted number in every exact finite-rank observable. Nor does inserting this one shift include every quantum or nonperturbative correction.

Exercises

1. Test a family near the change of description

Let $N\to\infty$ along integers and choose integer $k$ with $k\sim cN^a$, fixed $c>0$ and $0\leq a\leq1$; for $a=0$, take $c$ to be a positive integer. Determine when the eleven-dimensional lengths are parametrically large and when both IIA parameters vanish. Inspect the boundary cases $a=1/5$ and $a=1$ separately. Does small $G_4/L^2$ settle either boundary case?

Hint 1

Compute the powers of $N$ in $R_{11}/\ell_p$, $\alpha'/L_{\rm str}^2$ and $g_s^2$ before interpreting them.

Hint 2

A constant can be chosen small or large, but its size does not improve merely by increasing $N$.

Solution

The circle scales as $N^{(1-5a)/6}$, the string curvature parameter as $N^{(a-1)/2}$, and the squared string coupling as $N^{(1-5a)/2}$. Local eleven-dimensional curvature decreases as $N^{-(1+a)/3}$. Thus all the eleven-dimensional length tests improve for $0\leq a<1/5$, while both string parameters vanish for $1/5<a<1$.

At $a=1/5$, the limiting squared coupling is $4\pi\sqrt2\,c^{-5/2}$ and the circle radius is $(32\pi^2/c^5)^{1/6}$ in Planck units. Increasing $N$ changes neither. A separate large-$c$ choice can make IIA weakly coupled; a small-$c$ choice makes the circle large. At $a=1$, $\lambda\to1/c$, so the string curvature is fixed: it is small only if $c$ is separately small. The string coupling still vanishes. Rounding $k$ to nearby positive integers changes none of these leading powers.

In every listed case $G_4/L^2\propto N^{-(3+a)/2}$ tends to zero. It therefore cannot discriminate between these different curvature and coupling outcomes. The equality cases require the additional fixed-constant tests just stated.

2. Separate a short circle from the remaining compact modes

Use a massless minimally coupled scalar as a probe on the product metric (X03.1). It is a test field, not an assertion that the complete eleven-dimensional supergravity spectrum contains a decoupled scalar of this kind. A degree-$l$ harmonic polynomial on $\mathbb R^8$ restricts to a seven-sphere mode with eigenvalue $l(l+6)/R^2$.

The polynomial $z_1^k$ is invariant under the quotient. Determine its four-dimensional mass in AdS units and the larger scalar dimension root. Compare it with the quotient-invariant harmonic $|z_1|^2-|z_2|^2$. What becomes heavy as $k$ increases? Why should neither dimension be substituted for a protected ABJM chiral-primary dimension without further work?

Hint 1

Separate the product wave equation and use $L=R/2$. The boundary relation from the first dictionary is $m^2L^2=\Delta(\Delta-3)$.

Hint 2

The first polynomial has degree $k$ and a nonzero fiber charge. The second has degree two and zero common-phase charge; its flat Laplacian vanishes because the two terms have equal Laplacians.

Solution

The first mode has $m^2L^2=k(k+6)/4$, so $\Delta_+=[3+\sqrt{9+k(k+6)}]/2=3+k/2$. It becomes heavy as the fiber shrinks relative to the AdS radius. This is consistent with a vertical gradient contribution of order $k^2/R^2$, but the full sphere eigenvalue also includes the horizontal dependence needed for a globally defined mode.

The neutral quadratic has $m^2L^2=2(2+6)/4=4$, hence $\Delta_+=4$, independent of $k$. Its survival shows why circle reduction does not remove all compact-space excitations at the AdS scale. A globally valid four-dimensional truncation to a restricted set of fields requires its own consistency argument.

These dimensions belong to the specified minimally coupled probe. Physical supergravity perturbations can mix metric and flux components and have different mass operators. Identifying them with protected field-theory multiplets requires that coupled spectrum and an operator map. In particular, the probe’s result $3+k/2$ is not a derivation of a chiral dimension $k/2$.

3. Recover a finite sphere action

Take Euclidean AdS with $ds^2=L^2(du^2+\sinh^2u\,d\Omega_3^2)$, cut off at $u=u_0$. The outward normal gives $K=3\coth u_0/L$. Use the Einstein action, its boundary term and counterterms

$$\begin{aligned} I_{\rm bulk}&=-\frac1{16\pi G_4}\int\sqrt g\, (\mathcal R-2\Lambda),\qquad \Lambda=-3/L^2,\\ I_{\rm bdy}&=-\frac1{8\pi G_4}\int\sqrt h\,K,\\ I_{\rm ct}&=\frac1{8\pi G_4}\int\sqrt h\, \left(\frac2L+\frac L2\mathcal R[h]\right). \end{aligned}$$

Show that the sum tends to $\pi L^2/(2G_4)$. Then use (X03.8) and the sign convention for $F$ to recover (X03.9). Explain why keeping only the finite-looking term of an unrenormalized bulk integral would not specify this calculation.

Hint 1

$\operatorname{vol}(S^3)=2\pi^2$ and $\int_0^{u_0}\sinh^3u\,du=\cosh^3u_0/3-\cosh u_0+2/3$. The cutoff boundary has $\mathcal R[h]=6/(L^2\sinh^2u_0)$.

Hint 2

Factor out $A=\pi L^2/(4G_4)$ and write $s=\sinh u_0$, $c=\cosh u_0$. Keep all three terms until the powers of $s$ cancel.

Solution

The bulk has $\mathcal R-2\Lambda=-6/L^2$. Direct integration gives $I_{\rm bulk}/A=c^3-3c+2$, $I_{\rm bdy}/A=-3s^2c$, and $I_{\rm ct}/A=2s^3+3s$. Using $c^2=s^2+1$, their sum is $I/A=2-2c^3+2s^3+3s$. For large $s$, $c^3=(s^2+1)^{3/2}=s^3+3s/2+3/(8s)+O(s^{-3})$. Thus $I/A=2-3/(4s)+O(s^{-3})\to2$.

It follows that $F\simeq I_{\rm ren}=\pi L^2/(2G_4)=\pi\sqrt{2k}N^{3/2}/3$. The positive sign follows from $F=-\log Z$, independently of the source’s alternative free-energy name. The finite-cutoff error in this particular subtraction is not a string or quantum-gravity correction.

The surface term implements the metric variational problem; the local counterterms cancel divergences in a specified prescription. Omitting them and selecting a desired constant would not establish that prescription or its response to changed boundary data. This exercise reconstructs the supplied gravity action of DMP (5.38)–(5.43), not an exact finite-$N$ ABJM partition function.

The original ABJM paper guide locates this calculation in the source argument and gives a changed example to check while reading.