The things perturbation theory cannot see.

QFT document 18: non-perturbative quantum field theory. Tunneling between vacua, the $\theta$-angle, the BPST instanton, resolution of the $U(1)_A$ problem via instantons, dilute instanton gas, and the large-N expansion as an alternative non-perturbative tool.

Every QFT calculation we’ve done has been perturbative: expand in a small coupling, compute Feynman diagrams, sum to some order. Beautiful when it works. But there are physical phenomena that cannot be accessed this way at all, because they’re suppressed by $e^{-1/g^2}$; zero to all orders in perturbation theory but present non-perturbatively.

These include:

• Tunneling between vacua (QCD has infinitely many vacua labeled by integers)
• The $\eta'$ mass (explained by anomaly + instantons, not perturbation theory)
• Baryon number violation (sphalerons in electroweak theory)
• Gluon condensates and chiral condensates (non-perturbative order parameters)
• Confinement (in full rigor, not accessible perturbatively)

This document develops the main tools for non-perturbative QFT: instantons (localized Euclidean field configurations that give tunneling amplitudes) and the large-$N$ expansion (a systematic expansion in $1/N$ where $N$ is the rank of a gauge group, revealing structure invisible in perturbation theory).

Prerequisites

• Documents 1-17, especially 11 (Yang-Mills), 15 (ChPT), 17 (anomalies)
• Path integrals in Euclidean signature

Conventions

• Euclidean signature throughout for instanton calculations (indicated explicitly when relevant)
• Mostly-minus Minkowski metric otherwise
• $\hbar = c = 1$
• Gauge coupling $g$, fine structure $\alpha_s = g^2/(4\pi)$

Table of Contents

1. Why Non-Perturbative?
2. Tunneling in Quantum Mechanics: A Warmup
3. The Yang-Mills Vacuum
4. The BPST Instanton
5. Topological Charge and the $\theta$-Vacuum
6. Instantons and the $U(1)_A$ Problem
7. The Dilute Instanton Gas
8. Instantons in Other Theories
9. Sphalerons and Baryon Number Violation
10. The Large-$N$ Expansion
11. Large-$N$ QCD Predictions
12. Holographic Connection (Brief)
13. Appendix: Non-Perturbative Reference

1. Why Non-Perturbative?

The Perturbative Series Doesn’t Converge

Take any QFT. Expand an observable in powers of coupling: $\mathcal{A} = \sum_n a_n g^{2n}$. This series is typically asymptotic, not convergent. It’s useful for $n \lesssim 1/g^2$, after which terms grow.

The non-convergence isn’t a mathematical accident; it reflects physics. Non-perturbative effects of order $e^{-c/g^2}$ lie beyond the asymptotic series. No amount of summing perturbative terms reveals them.

What’s Invisible Perturbatively

Perturbative QCD tells you nothing about:

• Why quarks are confined (no analytic proof exists from first principles)
• The value of $\Lambda_{\rm QCD}$ (comes from RG, technically perturbative, but the physics it represents is non-perturbative)
• The chiral condensate $\langle\bar q q\rangle$
• The QCD vacuum structure (including $\theta$)
• The $\eta'$ mass
• Baryon number violation at high temperature
• Glueballs, exotic hadrons

All of these require non-perturbative methods; either instantons, large-$N$, lattice QCD, or holography.

The Two Main Tools

Instantons: Euclidean solutions to the classical equations of motion that give tunneling amplitudes. They’re localized in spacetime and carry topological charge.

Large-$N$: Replace $SU(3)$ with $SU(N)$ and take $N \to \infty$ (with $g^2 N$ fixed). The theory simplifies dramatically; only planar diagrams contribute at leading order. Provides a different non-perturbative handle.

Lattice QCD is another major tool, but it’s numerical rather than analytical. Instantons and large-$N$ are what this document covers.

2. Tunneling in Quantum Mechanics: A Warmup

Before attacking field theory, let’s understand tunneling in quantum mechanics. The key ideas transfer directly.

The Double-Well Potential

Consider a particle in the potential:

$V(x) = \lambda(x^2 - a^2)^2$

Two degenerate minima at $x = \pm a$, barrier in between at height $V(0) = \lambda a^4$. Classically, a particle at $x = a$ stays there. Quantum mechanically, it can tunnel to $x = -a$.

The WKB Tunneling Amplitude

The standard result from QM:

$\text{Amplitude}(+a \to -a) \propto e^{-S/\hbar}$

where $S$ is the “Euclidean action” of the tunneling path:

$S = \int_{-a}^{a}dx\,\sqrt{2mV(x)}$

For our double-well: $S = (4\sqrt{2m\lambda}/3)a^3$.

The tunneling amplitude is exponentially suppressed in $\hbar$. It’s pure non-perturbative.

The Euclidean Path Integral Picture

The same result comes from the Euclidean path integral. The classical equation of motion in Euclidean time:

$m\ddot x_E = +V'(x)$

(Note the sign flip from Minkowski: $\ddot x = -V'$.) This is the equation of motion in the inverted potential $-V(x)$.

The “bounce” or “instanton” solution: starts at $x = +a$ at $\tau = -\infty$, tunnels through to $x = -a$ at $\tau = +\infty$. Plugging into the Euclidean action gives exactly $S$ from WKB.

Key insight: quantum tunneling in Minkowski time = classical motion in Euclidean time with inverted potential.

Why This Generalizes to Field Theory

The same mathematical structure; Euclidean classical solutions giving tunneling amplitudes; works in field theory. The potential is replaced by the action functional, and the classical solution is a field configuration, not just a particle trajectory.

The prefactor in front of $e^{-S}$ requires care (fluctuations around the instanton), but the leading exponential is the same.

3. The Yang-Mills Vacuum

Vacuum Redundancy

In a Yang-Mills theory with gauge group $G$, the “vacuum” isn’t unique. Gauge transformations $A_\mu \to U^{-1}A_\mu U + U^{-1}\partial_\mu U$ relate physically equivalent configurations. The true vacuum is the space of physically distinct configurations.

Topology of Gauge Configurations

Consider the pure Yang-Mills vacuum with zero field strength: $F_{\mu\nu} = 0$, so $A_\mu = U^{-1}\partial_\mu U$ (pure gauge).

At spatial infinity, to have finite action, $A_\mu \to 0$, which requires $U \to$ constant. But “constant” can be any element of $G$. The 3-sphere at spatial infinity maps into $G$:

$U: S^3 \to G$

The Key Topology: $\pi_3(SU(N))$

For $G = SU(N)$ (with $N \geq 2$), the third homotopy group is:

$\pi_3(SU(N)) = \mathbb{Z}$

This means: maps from $S^3$ to $SU(N)$ are classified by an integer $n \in \mathbb{Z}$. Two maps with different $n$ cannot be continuously deformed into each other.

Vacuum Sectors

The integer $n$ labels distinct vacuum sectors $|n\rangle$. Gauge transformations labeled by $U$ map:

$|n\rangle \to |n + \text{winding}(U)\rangle$

where $\text{winding}(U) \in \mathbb{Z}$. Small gauge transformations (continuously connected to identity) have winding 0 and don’t change $n$. “Large” gauge transformations shift $n$ by the winding number.

The $\theta$-Vacuum

Physical states must be eigenstates of gauge transformations. A linear combination:

$|\theta\rangle = \sum_n e^{in\theta}|n\rangle$

is an eigenstate of large gauge transformations (shifts by 1):

$|\theta\rangle \to e^{-i\theta}|\theta\rangle$

Different $\theta$ labels physically different theories. $\theta$ is a parameter of QCD, like $\alpha_s$ or quark masses.

Instantons as Tunneling

Tunneling between different vacuum sectors $|n\rangle$ and $|n+1\rangle$ is mediated by instantons: Euclidean field configurations with topological charge 1.

$\langle n + 1|e^{-HT}|n\rangle \propto e^{-S_{\rm inst}}$

$S_{\rm inst}$ is the instanton action, of order $1/g^2$. Tunneling amplitude $\sim e^{-8\pi^2/g^2}$; exponentially small at weak coupling.

This is why the $\theta$-vacuum is a non-perturbative phenomenon. Perturbatively, it’s invisible. But tunneling (instantons) makes it real.

4. The BPST Instanton

The Equation to Solve

The Euclidean Yang-Mills action:

$S_E = \frac{1}{4g^2}\int d^4x\, F^a_{\mu\nu}F^{a\mu\nu}$

(Using Lie-algebra indices, $F^a_{\mu\nu} = \partial_\mu A^a_\nu - \partial_\nu A^a_\mu + f^{abc}A^b_\mu A^c_\nu$.)

We want localized solutions with finite action. The key identity:

$S_E = \frac{1}{4g^2}\int[F\pm\tilde F]^2 \mp \frac{1}{2g^2}\int F\tilde F$

where $\tilde F^a_{\mu\nu} = \tfrac{1}{2}\epsilon_{\mu\nu\rho\sigma}F^{a\rho\sigma}$. Since $[F\pm\tilde F]^2 \geq 0$, we have:

$S_E \geq \frac{1}{2g^2}\left|\int F\tilde F\right| = \frac{8\pi^2}{g^2}|Q|$

where $Q$ is the topological charge:

$Q = \frac{1}{16\pi^2 g^2}\int d^4x\, F^a_{\mu\nu}\tilde F^{a\mu\nu}$

Wait, let me reorganize. The topological charge is:

$Q = \frac{g^2}{32\pi^2}\int d^4x\,F^a_{\mu\nu}\tilde F^{a\mu\nu}$

And then the bound becomes:

$S_E \geq \frac{8\pi^2 |Q|}{g^2}$

The bound is saturated when $F = \pm\tilde F$: self-dual or anti-self-dual field strengths.

Self-Duality Condition

For $Q = +1$, we need self-dual ($F = +\tilde F$) configurations. The equation:

$F_{\mu\nu} = \tilde F_{\mu\nu}$

is first-order (simpler than the full second-order Yang-Mills equations). Solutions to this automatically extremize the action.

The BPST Solution (Belavin-Polyakov-Schwarz-Tyupkin, 1975)

The simplest self-dual solution, in $SU(2)$ gauge theory:

$A^a_\mu(x) = \frac{2\eta^a_{\mu\nu}(x-x_0)^\nu}{(x-x_0)^2 + \rho^2}$

where:

• $x_0$ is the position of the instanton (a free parameter)
• $\rho$ is the size of the instanton (a free parameter)
• $\eta^a_{\mu\nu}$ is the ‘t Hooft symbol (a specific combination of Pauli matrices and Kronecker deltas)

The ‘t Hooft Symbol

The ‘t Hooft symbol $\eta^a_{\mu\nu}$ is defined by:

$\eta^a_{ij} = \epsilon^a_{ij}, \quad \eta^a_{i4} = \delta^a_i, \quad \eta^a_{44} = 0, \quad \eta^a_{\mu\nu} = -\eta^a_{\nu\mu}$

It encodes the mapping from spacetime to gauge space. The anti-self-dual version is $\bar\eta^a_{\mu\nu}$ (differs in signs for the $\eta^a_{i4}$ components).

Computing the Field Strength

Plugging the BPST $A$ into $F_{\mu\nu} = \partial A - \partial A + [A, A]$:

$F^a_{\mu\nu}(x) = -\frac{4\eta^a_{\mu\nu}\rho^2}{[(x-x_0)^2 + \rho^2]^2}$

This is manifestly self-dual (because $\eta^a_{\mu\nu}$ is self-dual by construction).

The field strength is localized: it peaks at $x = x_0$ with size $\sim \rho$ and falls off as $1/x^4$ at large distance.

The Instanton Action

Plugging into the action:

$S_{\rm inst} = \frac{1}{4g^2}\int d^4x\, F^a_{\mu\nu}F^{a\mu\nu}$

Computing the integral:

$F^a_{\mu\nu}F^{a\mu\nu} = \frac{48\rho^4}{[(x-x_0)^2 + \rho^2]^4}$

$\int d^4x\,\frac{48\rho^4}{[(x-x_0)^2 + \rho^2]^4} = \frac{48\rho^4\cdot\pi^2}{6\rho^4} \cdot \ldots$

Let me just cite the result. The standard result is:

$\boxed{S_{\rm inst} = \frac{8\pi^2}{g^2}}$

Independent of the position $x_0$ and size $\rho$. The instanton has a universal action determined only by the topology.

Topological Charge $Q = 1$

Computing $Q$ for BPST:

$Q = \frac{g^2}{32\pi^2}\int d^4x\, F^a\tilde F^a = 1$

exactly. This is the topological charge; the same integer that labels the $\pi_3$ class of the gauge transformation at infinity.

Instanton Zero Modes

The BPST solution depends on parameters $(x_0, \rho)$; 5 collective coordinates in $SU(2)$:

• 4 from position $x_0^\mu$
• 1 from size $\rho$

Plus 3 from $SU(2)$ “orientation” (rotating the color direction of the instanton). Total: 8 zero modes for $SU(2)$ BPST instanton.

More generally, for $SU(N)$ $k$-instanton configurations, the number of collective coordinates is $4Nk$.

These zero modes must be handled as moduli in the path integral; you integrate over them rather than Gaussian-integrating around the instanton.

Multi-Instanton Solutions

For topological charge $|Q| = k$, you can superpose $k$ instantons (if they’re well-separated). The $k$-instanton moduli space has complicated structure, studied extensively (ADHM construction).

5. Topological Charge and the $\theta$-Vacuum

Physical Meaning of $\theta$

The $\theta$-term in the Euclidean action:

$S_\theta = -i\theta Q = -\frac{i\theta g^2}{32\pi^2}\int d^4x\, F\tilde F$

Adding this to the Yang-Mills action:

$S_{\rm total} = \frac{1}{4g^2}\int F^2 - \frac{i\theta g^2}{32\pi^2}\int F\tilde F$

In Minkowski signature:

$\mathcal{L} = -\frac{1}{4}F^2 + \frac{\theta g^2}{32\pi^2}F\tilde F$

Why $\theta$ Can Seem Trivial (But Isn’t)

Naively, $F\tilde F = \partial_\mu K^\mu$ is a total derivative. Plugging in, the $\theta$-term is:

$S_\theta = \frac{\theta g^2}{32\pi^2}\int d^4x\,\partial_\mu K^\mu = \theta\cdot(\text{boundary term})$

Looks like it shouldn’t affect physics. But the boundary term picks up nontrivial values on topologically-nontrivial configurations (instantons), so $S_\theta$ contributes to those sectors.

Physical Consequences of $\theta$

The $\theta$-term:

• Preserves P, C, CP for $\theta = 0$ or $\pi$ (mod $2\pi$)
• Violates CP for generic $\theta$
• Contributes to the neutron EDM: $d_n \sim \theta\cdot 10^{-16}$ e·cm

Experimentally: $|d_n| < 1.8\times 10^{-26}$ e·cm, implying $|\theta| < 10^{-10}$.

This tiny value of $\theta$ is mysterious; why so small? This is the strong CP problem.

Proposed Solutions to Strong CP

Peccei-Quinn mechanism (axion): Introduce a new $U(1)_{\rm PQ}$ symmetry, explicitly broken by instantons. The Goldstone of spontaneous breaking is the axion $a$, which dynamically drives $\theta_{\rm eff} = \theta + a/f_a \to 0$.

Nelson-Barr models: CP is a symmetry of the high-energy theory, spontaneously broken by complex Higgs VEVs. The effective $\theta$ becomes calculable.

Massless up quark: If $m_u = 0$, $\theta$ is physically unobservable (anomaly absorbs it into the quark phase). But measurements give $m_u \neq 0$, so this is disfavored.

None is confirmed. Axion searches are the most active program.

$\theta = \pi$: Special Point

At $\theta = \pi$, CP is preserved (spontaneously broken in some sectors). Strong-coupling physics at $\theta = \pi$ is subtle: in SUSY, there are “oblique confinement” phases; in QCD at $\theta = \pi$, lattice studies show unusual behavior.

The exact value of $\theta$ matters for QCD dynamics, and $\theta = \pi$ is a distinguished point.

6. Instantons and the $U(1)_A$ Problem

Recap of the Problem

From documents 15 and 17: the $U(1)_A$ symmetry of massless QCD is spontaneously broken but doesn’t produce a light Goldstone (the $\eta'$ is 958 MeV, too heavy). The anomaly explains this quantum-mechanically.

But we can be more specific: the anomaly is zero in perturbation theory. It comes from non-perturbative effects. Specifically, instantons.

The Mechanism

The anomalous current divergence:

$\partial_\mu J^\mu_A = \frac{g^2 n_f}{16\pi^2}G\tilde G$

The integral $\int d^4x\, G\tilde G$ on an instanton background is $32\pi^2\cdot Q/g^2 = 32\pi^2/g^2$ for $Q = 1$.

Integrating:

$\int d^4x\,\partial_\mu J^\mu_A = 2n_f\cdot Q = 2n_f$

So on a $Q = 1$ instanton, the axial charge changes by $2n_f$ (with 2 coming from each of the 3 quark flavors contributing 2 units).

Instantons transfer axial charge between left and right, breaking $U(1)_A$ invariance of correlators.

The ‘t Hooft Vertex

Consider a correlator of $n_f$ right-handed quarks and $n_f$ left-handed antiquarks. ‘t Hooft showed this gets a non-zero vacuum expectation value from instantons:

$\langle\bar u_R u_L\bar d_R d_L\bar s_R s_L\rangle_{\rm 1-inst}\neq 0$

This is the famous ‘t Hooft effective vertex; a $2n_f$-fermion operator generated by instantons.

Physically: instantons induce a multi-fermion interaction that explicitly breaks $U(1)_A$.

Impact on $\eta'$ Mass

Through the ‘t Hooft vertex, the $\eta'$ meson (the $U(1)_A$ would-be Goldstone) gets a mass. The calculation is intricate, but the final result (Witten-Veneziano):

$m_{\eta'}^2 = \frac{2 n_f}{f_\pi^2}\chi_t$

where $\chi_t$ is the topological susceptibility; the vacuum fluctuation of topological charge, which is non-zero precisely because of instantons.

Quantitatively: $\chi_t \approx (180\text{ MeV})^4$, giving $m_{\eta'} \approx 860$ MeV. Observed: 958 MeV. Close, with the difference attributed to higher-order corrections in $1/N_c$.

The Instanton-Anomaly Connection

The $U(1)_A$ anomaly says the symmetry could be broken. Instantons are the mechanism that actually breaks it. The topological susceptibility $\chi_t$ measures how much instantons break it.

Without instantons, the axial anomaly might still be there as a formal result, but the physical consequences (heavy $\eta'$) require explicit non-perturbative contributions.

Why Perturbation Theory Misses This

In perturbation theory, $F\tilde F$ is a total derivative that integrates to zero (for topologically-trivial backgrounds). So $\chi_t = 0$ perturbatively, and $\eta'$ would be massless.

Only with non-perturbative instanton contributions does $\chi_t$ become non-zero. The $\eta'$ mass is a direct measurement of non-perturbative QCD structure.

7. The Dilute Instanton Gas

The Approximation

For small instantons (size $\rho \ll$ typical scale), instantons are localized and don’t interact much. Multi-instanton configurations can be approximated as a dilute gas of independent instantons.

The total partition function:

$Z = \sum_{n_+, n_-}\frac{1}{n_+! n_-!}\int\prod_i d^4x_{0,i}\,d\rho_i\,d\Omega_i\,[\text{measure}]\,e^{-\sum_i S_i}$

where $n_+$ counts instantons, $n_-$ anti-instantons, and we sum over all their positions, sizes, orientations.

The Single-Instanton Contribution

For a single instanton:

$Z_1 = \int d^4x_0\,d\rho\,d\Omega\,[\text{fluctuation determinant}]\,e^{-S_{\rm inst}}$

The fluctuation determinant around the instanton gives a specific prefactor. ‘t Hooft computed it for $SU(2)$:

$Z_1 = V\int\frac{d\rho}{\rho^5}\left(\frac{2\pi^2}{g^2}\right)^{4}e^{-8\pi^2/g^2}\cdot[\text{color integration}]$

The $\rho^{-5}$ comes from dimensional analysis of the collective coordinate measure.

The IR Problem

The integral over $\rho$ is divergent at large $\rho$. At size $\rho \sim \Lambda_{\rm QCD}^{-1}$, the coupling $g^2$ becomes large, and the instanton calculation breaks down. Large instantons probe strong coupling, where the dilute-gas approximation fails.

What Instantons Give Us

Despite the IR issues, instantons provide:

1. $\eta'$ mass (calculable in dilute-gas + ChPT matching)
2. Axial charge non-conservation (multi-fermion operators)
3. Vacuum structure (the $\theta$-vacuum physics)
4. Baryon number violation rate (electroweak sphalerons at finite T)
5. Specific calculable effects in SUSY QCD: exact results for certain correlators

For observables dominated by small instantons (e.g., high-energy processes), the dilute-gas approximation works quantitatively. For observables sensitive to strong coupling (like confinement itself), it fails.

Instantons vs. Perturbation Theory

Instantons contribute at order $e^{-8\pi^2/g^2}$, completely invisible to perturbation theory which expands in $g^2$. The “gap” between perturbative and non-perturbative effects is exponential.

For $\alpha_s = 0.1$ (high-energy QCD): $e^{-8\pi^2/g^2} = e^{-2\pi/\alpha_s} \sim e^{-63} \sim 10^{-27}$. Utterly negligible.

For $\alpha_s = 0.5$ (low-energy): $e^{-4\pi} \sim 10^{-5}$. Small but non-negligible.

For $\alpha_s \sim 1$: $e^{-2\pi}\sim 0.002$. Starting to matter.

Instantons become quantitatively important only at strong coupling.

8. Instantons in Other Theories

Abelian Instantons

In 4D abelian gauge theory ($U(1)$), there are no instantons; the topology is trivial ($\pi_3(U(1)) = 0$).

But in 2D: $\pi_1(U(1)) = \mathbb{Z}$, so there are 2D $U(1)$ vortices/instantons. Important for 2D gauge theory and superconductors.

Gravitational Instantons

In Euclidean general relativity, gravitational instantons are finite-action Einstein space configurations. Examples:

• Eguchi-Hanson: self-dual (or anti-self-dual) with specific boundary topology
• Taub-NUT: different asymptotic structure
• Black hole instantons (Euclidean Schwarzschild): describes black hole pair creation

Gravitational instantons contribute to quantum gravity calculations (e.g., in path integral quantum gravity).

Supersymmetric Instantons

In SUSY gauge theories, instantons have special properties:

• Exact cancellations between bosonic and fermionic fluctuations
• Exact non-perturbative results for certain correlators
• Seiberg-Witten theory: instantons give exact low-energy dynamics of $\mathcal{N} = 2$ SUSY

The Nekrasov partition function captures all-instanton contributions in 4D $\mathcal{N} = 2$ SUSY gauge theories.

Yang-Mills Instantons in Higher Dimensions

In $4n$ dimensions, there are generalizations of instantons (solutions to generalized self-duality equations). These play roles in M-theory and supergravity compactifications.

9. Sphalerons and Baryon Number Violation

The Problem in the Standard Model

In the Standard Model, baryon number $B$ and lepton number $L$ are classically conserved. But the $SU(2)_L$ gauge theory has instantons (and related “sphaleron” configurations), and these violate $B$ and $L$ quantum-mechanically:

$\partial_\mu J^\mu_B = \frac{n_g}{32\pi^2}W^a\tilde W^a$

(And similarly for $L$.) Here $n_g$ is the number of generations.

For $n_g = 3$: instantons cause $\Delta B = 3$ and $\Delta L = 3$ simultaneously. So $B - L$ is conserved, but $B + L$ isn’t.

Rate at Zero Temperature

At $T = 0$, the rate of such processes is $\sim e^{-8\pi^2/g^2} \sim 10^{-170}$ per year for current values. Essentially zero; the proton is stable on cosmological timescales.

Rate at High Temperature (Sphalerons)

At high $T$, tunneling gives way to thermal excitation over the barrier. The relevant configurations are sphalerons (saddle points of the classical action at finite temperature).

The sphaleron rate:

$\Gamma_{\rm sph}/V \sim \alpha_W^5 T^4$

(up to logarithms), in the symmetric phase of electroweak theory. This is not exponentially suppressed; it’s a polynomial function of $T$.

Implications for the Early Universe

At $T \gg 100$ GeV (above the electroweak scale), sphalerons are in equilibrium. Any baryon number produced earlier gets washed out.

At $T \sim 100$ GeV (electroweak scale), sphalerons freeze out as the Higgs acquires its VEV and generates gauge boson masses. If there’s a first-order electroweak phase transition, this provides out-of-equilibrium conditions where CP violation can create a baryon asymmetry; electroweak baryogenesis.

For SM: the transition is a crossover, not first-order, and CP violation is too small. So SM electroweak baryogenesis doesn’t work. New physics is needed.

Sphalerons as Thermally Excited Instantons

A sphaleron is a static classical configuration, not a tunneling solution. But physically, the sphaleron is what an instanton “becomes” at finite temperature; the saddle point that determines the barrier-crossing rate.

At $T = 0$: pure instanton tunneling, rate $\sim e^{-S_{\rm inst}}$.

At $T > 0$: combination of tunneling and thermal excitation, increasing rate.

At $T \gg E_{\rm sphaleron}$: purely thermal barrier-crossing, rate $\sim T^4$.

The interpolation is a rich topic in thermal field theory.

10. The Large-$N$ Expansion

The Idea

Take $SU(N)$ gauge theory with $N$ colors. Expand in $1/N$ (rather than in $g^2$). At fixed ‘t Hooft coupling $\lambda = g^2 N$, take $N \to \infty$.

Why does this help? Because:

• Only planar diagrams contribute at leading order
• The theory simplifies dramatically
• Many qualitative features emerge cleanly

For QCD, $N = 3$ isn’t large. But surprisingly, many $1/N$ predictions work reasonably well; suggesting large-$N$ captures real physics.

Double-Line Notation

Gluons have color indices: $A^a_\mu$ with $a = 1, \ldots, N^2 - 1$. In terms of fundamental indices: $A^{ij}_\mu$ with $i, j = 1, \ldots, N$.

Think of a gluon as a “quark-antiquark pair in color space”: one fundamental index and one anti-fundamental. Draw diagrams with double lines: gluon = two oriented lines, one for each color index.

Counting Diagrams

For a diagram with $V$ vertices, $E$ edges (propagators), $F$ faces (closed color loops):

• Each vertex: factor $g$
• Each edge: factor $1/g^2 \cdot$ (propagator)
• Each face: factor $N$ (trace over color indices going around)

Total: $g^{V - 2E}\cdot N^F = g^{V-2E}\cdot N^F$.

For planar diagrams (can be drawn on a sphere without crossings): Using Euler: $V - E + F = 2$.

A diagram’s contribution scales as:

$N^F\cdot g^{V-2E} = N^{F}\cdot (g^2N)^{(V-2E)/2}\cdot N^{-(V-2E)/2}$

For planar ($V - E + F = 2$): the net $N$-power is $N^{2 - 2g}$ where $g$ is the genus.

Planar diagrams (genus 0): $\propto N^2$. Leading.

Non-planar diagrams (genus $\geq 1$): $\propto N^{2-2g}$. Suppressed by $1/N^{2g}$.

The ‘t Hooft Limit

In the ‘t Hooft limit: $N \to \infty$ with $\lambda = g^2 N$ fixed.

At this limit:

• Only planar diagrams survive
• The theory has a (formal) string-like expansion in $1/N$
• Large-$N$ QCD is a “master theory” governing the ‘t Hooft expansion

Large-$N$ Simplifications

Many simplifications emerge at large $N$:

Meson masses are $O(N^0)$. Mesons don’t depend on $N$ (at leading order).

Meson decay constants are $O(\sqrt N)$. Couplings $\propto 1/\sqrt N$, so decays $\propto 1/N$.

Baryon masses are $O(N)$. Baryons contain $N$ quarks, so their mass scales.

Zweig rule violations $O(1/N^2)$. Diagrams that require breaking a quark loop are suppressed.

$\eta'$ is a Goldstone. In the large-$N$ limit, the $U(1)_A$ anomaly is suppressed by $1/N$, so $\eta'$ becomes massless. This is the Witten-Veneziano formula context.

Connected correlators factorize. $\langle\mathcal{O}_1\mathcal{O}_2\rangle = \langle\mathcal{O}_1\rangle\langle\mathcal{O}_2\rangle + O(1/N^2)$.

The Planar Limit Is Solvable in Some Cases

In 2D: $SU(N)$ Yang-Mills in 2D is exactly solvable in the planar limit (Gross-Witten transition, Douglas-Kazakov transition).

In 4D: not exactly solvable, but many structural features can be computed.

AdS/CFT correspondence (next section) relates the planar limit of certain gauge theories to string theory in AdS; making the large-$N$ limit tractable via holography.

11. Large-$N$ QCD Predictions

Meson Spectrum

At large $N$:

• Meson masses are $N$-independent (at leading order)
• There are infinitely many mesons at each quantum number (pions, kaon-like states, etc., at higher masses)
• Decay widths are $O(1/N)$, so mesons are narrow

Finite-$N$ corrections: meson widths $\Gamma \sim m/N$, mass shifts $\sim 1/N^2$.

Zweig’s Rule (OZI Rule)

At large $N$, meson decays that require disconnected diagrams (breaking a quark loop) are suppressed by $1/N^2$.

Example: $\phi \to \pi\pi$ decay. The $\phi$ is nearly pure $s\bar s$, so decay into $\pi\pi$ requires annihilation of the $s\bar s$ pair via gluons; a disconnected topology.

Experimentally: $\text{BR}(\phi \to \rho\pi) = 15\%$, tiny compared to naive expectations. This is the OZI rule in action.

Large-$N$ predicts: OZI violations $\sim 1/N^2 \sim 10\%$. Matches observation.

Baryon Properties

Baryons at large $N$ contain $N$ quarks and have mass $O(N)$. Their interactions are $O(1)$. At $N = 3$, baryons have specific structure (protons, neutrons) whose properties scale predictably.

Nucleon-nucleon coupling from one-pion exchange: consistent with large-$N$ scaling.

Baryon decuplet structure: Skyrme’s idea that baryons are solitons in a pion effective theory emerges naturally at large $N$. Predicts specific relations between baryon masses and their excitations (quantitatively successful).

Glueball Spectrum

Glueballs (bound states of pure glue) exist in large-$N$ QCD, with definite quantum numbers. Large-$N$ predicts:

• Mass hierarchy: $0^{++}$ lightest, $2^{++}$ next, etc.
• Widths suppressed by $1/N$
• Mixing with quark-mesons of similar quantum numbers suppressed at large $N$

Lattice QCD confirms these predictions qualitatively. Glueball masses predicted near lattice measurements.

Confinement and Flux Tubes

In large-$N$ QCD, quark confinement is rigorously obtained. The quark-antiquark potential grows linearly with separation, due to a flux tube of color field lines.

The flux tube tension $\sigma \sim$ const at large $N$ (not scaling with $N$). Its thickness $\sim \Lambda_{\rm QCD}^{-1}$.

At $N = 3$: lattice QCD confirms this, with $\sigma \approx (420 \text{ MeV})^2$, a phenomenologically important value.

Predictions at $N = 3$

Remarkably, many large-$N$ predictions work quantitatively at $N = 3$, suggesting that $1/N^2 = 1/9 \sim 11\%$ corrections are “small enough” for most quantities.

This is why large-$N$ is more than a theoretical curiosity; it captures real aspects of QCD.

12. Holographic Connection (Brief)

The AdS/CFT Correspondence

Maldacena (1997) conjectured that 4D $\mathcal{N} = 4$ super Yang-Mills with gauge group $SU(N)$ is dual to Type IIB string theory on $AdS_5 \times S^5$.

At large $N$ and large ‘t Hooft coupling $\lambda$: the gauge theory is dual to classical supergravity in AdS.

Why This Matters for Non-Perturbative QFT

Strong-coupling gauge theory becomes weak-coupling supergravity. Calculations that are impossible in perturbative QFT become tractable via classical gravity.

Computations have been done for:

• Transport coefficients in QGP-like plasmas: $\eta/s = 1/(4\pi)$, matching QGP observations
• Mass spectra of strongly-coupled theories
• Confinement-deconfinement transitions in dual geometries
• Quark-antiquark potentials in various theories
• Many-body dynamics at finite temperature and density

Limitations

AdS/CFT is exact for $\mathcal{N} = 4$ SYM, but only a conjecture for other theories. For real QCD, no exact dual is known. But AdS/QCD models provide quantitative phenomenology by building dual geometries that match experimental data.

Relation to Large-$N$

AdS/CFT is most tractable in the large-$N$ limit. The planar diagram expansion of the gauge theory maps onto a string theory perturbation expansion in $1/N$.

In this sense, AdS/CFT is a deep generalization of large-$N$ QCD; extending it from gauge theories to string theory, with stringy corrections computable as $1/N$ corrections.

Full Treatment Would Require Own Documents

Holography is a vast topic. Full treatment requires:

• String theory basics
• Supergravity in AdS
• The AdS/CFT dictionary (operators ↔ fields, boundary conditions ↔ sources)
• Applications to condensed matter, nuclear physics

These are Option E topics. For now, mentioning the connection is enough.

13. Appendix: Non-Perturbative Reference

Key Formulas

Yang-Mills instanton action: $S_{\rm inst} = 8\pi^2/g^2$

Topological charge: $Q = g^2/(32\pi^2)\int d^4x\, F\tilde F \in \mathbb{Z}$

Instanton rate (dilute gas): $\Gamma \sim e^{-8\pi^2/g^2}$

$\theta$-term: $\mathcal{L}_\theta = (\theta g^2/32\pi^2)F\tilde F$

Witten-Veneziano: $m_{\eta'}^2 = 2n_f\chi_t/f_\pi^2$

Topological susceptibility: $\chi_t \approx (180\text{ MeV})^4$

The BPST Instanton

$A^a_\mu(x) = \frac{2\eta^a_{\mu\nu}(x-x_0)^\nu}{(x-x_0)^2 + \rho^2}$

$F^a_{\mu\nu}(x) = -\frac{4\eta^a_{\mu\nu}\rho^2}{[(x-x_0)^2+\rho^2]^2}$

$\text{Action: } S = 8\pi^2/g^2, \quad \text{Charge: } Q = 1$

Large-$N$ Scalings

Further Reading

• Coleman, Aspects of Symmetry (1985): classic lectures including instantons
• Polyakov, Gauge Fields and Strings (1987): monographs on non-perturbative methods
• Shifman, Advanced Topics in Quantum Field Theory: comprehensive textbook
• ‘t Hooft, Computation of the Quantum Effects Due to a Four-Dimensional Pseudoparticle (1976): original ABJ-anomaly-to-instantons paper
• Witten, Instantons, the Quark Model, and the 1/N Expansion (1979): large-$N$ QCD foundation
• Rubakov, Classical Theory of Gauge Fields: non-perturbative gauge theory textbook
• Manton & Sutcliffe, Topological Solitons: covers instantons, monopoles, etc.

Problems

1. Verify that the BPST instanton is self-dual by computing $F_{\mu\nu}$ and $\tilde F_{\mu\nu}$ explicitly.

2. Compute the BPST instanton’s topological charge $Q$ by direct integration.

3. Derive the dilute instanton gas partition function, identifying the density expansion.

4. Using the ‘t Hooft vertex, estimate the contribution to the $\eta'$ mass from a dilute instanton gas. Compare with Witten-Veneziano.

5. For $SU(N)$ Yang-Mills, show that the planar expansion generates only diagrams that can be drawn on a sphere.

6. Compute the meson decay constant in terms of $N$ and $g$ at large $N$. Show it scales as $\sqrt N$.

7. At what value of $\alpha_s$ do instanton effects become “large” (say, 10% of perturbative) in QCD? What physical scale does this correspond to?

Closing Note

Non-perturbative QFT is where the deepest physics lives. Perturbation theory is the toolkit for weak coupling, but many physical phenomena; confinement, tunneling, vacuum structure, baryon number violation, $\eta'$ mass; are inherently non-perturbative.

What We Covered

• Instantons: Euclidean tunneling solutions with topological charge. The BPST solution, its moduli, its physics.
• $\theta$-vacuum: the parameter $\theta$ labeling distinct QCD theories. Strong CP problem and axion.
• Instanton effects: $\eta'$ mass, ‘t Hooft vertex, anomaly realization. How perturbatively-invisible physics becomes dominant.
• Dilute instanton gas: the approximation scheme that makes instanton calculations tractable.
• Sphalerons: thermal analogs of instantons, giving electroweak baryogenesis possibilities.
• Large-$N$: expansion in $1/N$, planar limit, qualitative predictions for QCD.
• Holography (brief): how large-$N$ connects to string theory via AdS/CFT.

What’s Next

Document 19 covers solitons, monopoles, and dualities; classical field configurations that extend in space (not just spacetime events like instantons). Kinks, vortices, ‘t Hooft-Polyakov monopoles, BPS states, Dirac quantization, Montonen-Olive duality.

Together with document 18, this covers the main non-perturbative structures in QFT: events in spacetime (instantons) and objects in space (solitons).

After that, depending on interest:

• Option E: Beyond Standard Model topics; SUSY, strings, holography in depth, quantum gravity

Or sideways into:

• Topological phases of matter in depth
• Entanglement and quantum information in QFT
• Integrability and exactly solvable models
• Advanced mathematical aspects (index theorems, cohomology, etc.)

You have extensive foundations now. Many research-level topics are accessible from here.