Large-N, Loop, and ℏ Approximation Hierarchies

Large-N, loop, and $\hbar$ expansions organize different approximations. With $N$ independent matter species, the mean stress grows as $N$ and the root-mean-square connected fluctuation grows as $N^{1/2}$, so the relative fluctuation falls as $N^{-1/2}$. Holding $NG_N$ fixed yields a controlled mean-metric saddle in ordinary regimes, but it does not make every matter loop small, remove stochastic corrections, or remain uniform near critical and secular growth.

Required background. The semiclassical Einstein equation fixes the mean saddle; one-loop matter effective actions fixes what “matter loop” means; and large-N normalizations supplies the species scaling.

Helpful background. Loops, renormalization, and EFT separates loop from derivative order, and nonuniform large-N limits supplies the failure diagnostics.

Species scaling of the mean equation

Let $\phi_a$, $a=1,\ldots,N$, be identical independent fields in the same state. Their matter functional and stress satisfy

$$\Gamma_{\rm m}^{(N)}[g]=N\Gamma_{\rm m}^{(1)}[g], \qquad \langle T_{\mu\nu}^{\rm tot}\rangle =N\langle T_{\mu\nu}^{(1)}\rangle.$$

Take

$$G_N=\frac{\bar G}{N}, \qquad N\to\infty$$

with $\bar G$ and the renormalized curvature couplings in the $O(N)$ effective action scaled consistently. The mean equation becomes

$$G_{\mu\nu}+\cdots =8\pi\bar G\, \langle T_{\mu\nu}^{(1)}\rangle_{\rm ren},$$

which is $O(1)$. Both the gravitational action $1/G_N$ and the total matter effective action scale as $N$, so the mean metric is their leading saddle.

For independent species,

$$\langle \hat t_{\mu\nu}^{\rm tot}(x) \hat t_{\rho\sigma}^{\rm tot}(y)\rangle_c =N\,C^{(1)}_{\mu\nu\rho\sigma}(x,y),$$

where $\hat t=\hat T-\langle\hat T\rangle$. The induced metric two-point function carries two powers of $G_N$ and therefore scales as

$$G_N^2\,N\,C^{(1)}=O(N^{-1}).$$

Thus mean geometry is leading, while connected metric fluctuations are subleading in this scaling. They are not absent and are not determined by the mean equation. Hu and Verdaguer give the large-N relation between semiclassical and stochastic descriptions with this hierarchy made explicit (Hu and Verdaguer 2020, §§ 4.2 and 8.1).

Loop counting is separate. Integrating $N$ free fields exactly resums their one-loop determinants at leading $N$; it does not truncate the determinant to one derivative order. Metric loops are suppressed by the gravitational saddle scaling, whereas matter self-interaction loops depend on their own coupling normalization. Restoring $\hbar$, a vacuum matter loop carries $\hbar$, but a large coherent occupation can make $\langle T\rangle$ effectively classical. Every calculation should state all three counts rather than calling them collectively “semiclassical order.”

The structure map places this bookkeeping before self-consistency and stability, because a missing nominally same-order term cannot be repaired by later convergence.

Large-N organization of mean backreaction. The diagram is schematic and not to scale; causal response and constraints are imposed at every order, and the leading mean does not include all connected fluctuations.

The failure map tests uniformity. If a retarded response eigenvalue becomes small, a formally $1/N$ correction can be amplified into an order-one effect.

Nonuniform-limit witness for large-N backreaction. This schematic, not-to-scale map requires response eigenvalues, observation time, and state occupation to remain within the regime where the counting is uniform.

Application: classify the leading and subleading terms

For $N$ free scalars with identical Hadamard state, write the CTP effective action as

$$\Gamma_{\rm CTP}[g_+,g_-] =N\left[ \bar S_g[g_+]-\bar S_g[g_-] +\Gamma_{\rm m}^{(1)}[g_+,g_-] \right].$$

The stationary point supplies:

• mean metric $g^{(0)}_{\mu\nu}=O(1)$;
• total mean stress $N\langle T^{(1)}_{\mu\nu}\rangle=O(N)$;
• finite backreaction $G_N\langle T^{\rm tot}\rangle=O(1)$;
• retarded matter polarization $N\Pi_{\rm R}^{(1)}$, which contributes at leading order to linear response;
• connected stress $NC^{(1)}$ and induced metric variance $O(N^{-1})$;
• mean-metric correction from the next saddle order, generically $O(N^{-1})$.

This list prevents a frequent mistake: because $\Pi_{\rm R}$ is a connected two-stress correlator, one might classify it as negligible. In the linearized mean equation it is multiplied by $G_NN=\bar G$ and is leading. The noise-driven metric variance contains a second $G_N$ and is subleading. Response and noise use related correlators but answer different questions.

To reproduce the hierarchy, record whether observables are per species or summed, whether $\bar G$ or $G_N$ is fixed, the matter interaction scaling, the state occupation per species, and the time interval. Verify that the mean curvature and every retained response eigenvalue remain finite as $N$ changes.

Adversarial test: critical and secular enhancement

Suppose a physical response mode has inverse propagator

$$D_N(\omega)=D_0(\omega)+\bar G\Pi_{\rm R}(\omega) +\frac{1}{N}\Sigma_1(\omega).$$

Away from zeros of the first two terms, $\Sigma_1/N$ is small. Near a frequency where

$$D_0+\bar G\Pi_{\rm R}=O(N^{-1}),$$

the correction shifts the pole by order one relative to its distance from criticality. Likewise a secular term $(t/N)\Sigma_{\rm sec}$ becomes order one for $t\sim N$. Formal counting at fixed frequency and time therefore does not justify the leading saddle in those double-scaling regimes.

The strongest surviving statement is large-N control on compact parameter and time domains separated from critical response poles. Beyond them, one must resum the enhanced sector or downgrade the result.

Domain and failure conditions

See the chapter domain and failure-conditions table. The $1/N$ hierarchy assumes independent or consistently scaled species, finite per-species state data, $NG_N$ fixed, and uniform response bounds. It does not establish small stress fluctuations in absolute units, and it says nothing about ultraviolet derivative control unless that expansion is stated separately.

Exercise

For independent species, compute the relative root-mean-square fluctuation of their summed stress smeared with one fixed test tensor.

Solution

If one species has mean $\mu$ and variance $\sigma^2$, the sum has mean $N\mu$ and variance $N\sigma^2$. Hence

$$\frac{\sqrt{\operatorname{Var}T_{\rm tot}}} {\langle T_{\rm tot}\rangle} =\frac{\sigma}{|\mu|\sqrt N}.$$

This relative suppression fails when $\mu=0$ and does not replace a calculation of the induced metric fluctuation.

Handoff

Self-consistent state–geometry solutions turn the leading mean equation into a fixed-point or coupled evolution problem with residual tests.

References

• Hu, Bei-Lok, and Enric Verdaguer. Semiclassical and Stochastic Gravity: Quantum Field Effects on Curved Spacetime. Cambridge: Cambridge University Press, 2020. doi:10.1017/9780511667497.
• Roura, Albert, and Enric Verdaguer. “Cosmological Perturbations from Stochastic Gravity.” Physical Review D 78 (2008): 064010. doi:10.1103/PhysRevD.78.064010.