Ising CFT + Magnetic Perturbation: Surviving Conserved Charges

Main result

Consider the $c = 1/2$ Ising CFT (minimal model $\mathcal{M}(3, 4)$ — see ising-cft-primary-operators) perturbed by the spin primary operator $\sigma$ of conformal weight $h_{\sigma} = 1/16$:

\[\mathcal{A} \;=\; \mathcal{A}_{\rm Ising\,CFT} \;+\; \lambda\,\int d^{2}x\,\sigma(x), \qquad \lambda \in \mathbb{R}.\]

Zamolodchikov 1989 showed that of the infinitely many conserved charges $Q_{s}$ of the unperturbed CFT (one for every odd integer spin $s \ge 1$), exactly eight survive as exact conservation laws of the perturbed theory:

\[\boxed{\; \text{Surviving spins} \;=\; \{1,\,7,\,11,\,13,\,17,\,19,\,23,\,29\}. \;}\]

These are the exponents of the $E_{8}$ Lie algebra (Coxeter number $h = 30$, so the exponents run from $1$ up to $h - 1 = 29$ and come in complementary pairs summing to $h$).

Consequences:

The massive perturbed theory has a factorised elastic $S$-matrix (from eight conserved charges + Zamolodchikov's theorem).
The bootstrap equations with this particular constraint set have a unique minimal solution: the $E_{8}$ affine Toda field theory (see e8-mass-spectrum-derivation).
The eight stable-particle mass ratios are the Perron–Frobenius eigenvector of the $E_{8}$ Cartan matrix, with $m_{2}/m_{1} = \varphi = (1 + \sqrt{5})/2$ the famous golden-ratio prediction.
The mass gap scales with the perturbation as $m_{1} \propto |\lambda|^{8/15}$ by dimensional analysis.

The contrasting perturbation by the energy density $\varepsilon$ (conformal weight $1/2$, thermal direction) preserves only infinitely many charges via a trivial mechanism: the perturbed theory is a free massive Majorana fermion, which is integrable but has a single-particle spectrum with no $E_{8}$ structure.

Setup

Conventions

Euclidean 2D, complex coordinates $z = x^{1} + i x^{2}$, $\bar z = x^{1} - i x^{2}$. Holomorphic and antiholomorphic derivatives $\partial \equiv \partial_{z}$, $\bar\partial \equiv \partial_{\bar z}$. For the Ising CFT we have $c = 1/2$ with primary operators $\{\mathbb{I}, \sigma, \varepsilon\}$ of conformal weights $(h, \bar h) = (0, 0), (1/16, 1/16), (1/2, 1/2)$.

Stress tensor and conserved currents

The CFT stress tensor $T(z)$ (holomorphic) has conformal weight $h = 2$ and is conserved: $\bar\partial T = 0$. Similarly $\bar T(\bar z)$ is antiholomorphic-conserved. From $T$ we build a tower of conserved higher-spin currents

\[T_{2} = T,\qquad T_{4} = \,:T^{2}:,\qquad T_{6} = \,:T^{3}:\,+\, \text{(descendant corrections)},\qquad \dots,\]

with $T_{2 k}$ of conformal weight $2 k$. The "descendant corrections" are levels of the Virasoro descendants needed to make each $T_{2 k}$ a primary of the $\mathcal{W}$-algebra (Zamolodchikov 1985); for our purposes we only need that such conserved higher-spin tensors exist at every even weight.

Holomorphic conservation $\bar\partial T_{s} = 0$ implies, by Stokes, the conserved charge

\[Q_{s} \;=\; \oint\frac{dz}{2\pi i}\,T_{s}(z),\]

one for every odd spin $s = 1, 3, 5, 7, \dots$ (the spin $s = s_{\rm current} - 1$ — the "weight $2k$, spin $2k - 1$" convention for a spin-$s$ current). Infinitely many conservation laws.

Goal

Determine which of the infinite sequence of $Q_{s}$ survives in the perturbed theory $\mathcal{A} \to \mathcal{A} + \lambda\int \sigma$, and derive the distinctive list of eight spins $\{1, 7, 11, 13, 17, 19, 23, 29\}$.

Calculation

Step 1 — Modification of conservation laws under perturbation

Under $\lambda$-perturbation by an operator $\Phi$ of conformal weight $(h_{\Phi}, \bar h_{\Phi})$, the Euclidean action gains a term $\lambda\int d^{2}z\,\Phi$, and the holomorphic conservation law $\bar\partial T_{s} = 0$ is modified to

\[\bar\partial T_{s}(z) \;=\; \lambda\cdot R_{s}[\Phi](z) \;+\; O(\lambda^{2}), \tag{1}\]

where $R_{s}[\Phi]$ is determined by the OPE of $T_{s}(z)$ with the perturbing operator integrated over its insertion point. Using the perturbative formula (Zamolodchikov 1989 §2; Cardy 1996 §11)

\[R_{s}[\Phi](z) \;=\; \underset{w\to z}{\mathrm{Res}}\,\,T_{s}(z)\,\Phi(w) \quad \text{(singular OPE residue)}. \tag{2}\]

The charge $Q_{s} = \oint (dz/2\pi i)\,T_{s}$ remains conserved iff the right-hand side of (1) is a total derivative with respect to $z$:

\[R_{s}[\Phi](z) \;=\; \partial\,\bigl(\text{something}\bigr)_{z} \quad \Longrightarrow\quad \oint\frac{dz}{2\pi i}\,R_{s} = 0 \quad \Longrightarrow\quad Q_{s}\text{ is conserved at }O(\lambda). \tag{3}\]

The survival condition (3) is the key criterion. It reduces to a combinatorial statement about the OPE coefficients of $T_{s}$ with $\Phi$.

Step 2 — $R_{s}[\Phi]$ as a descendant of $\Phi$

By general CFT operator-product-expansion (Di Francesco–Mathieu– Sénéchal 1997 §6.6), the singular part of $T_{s}(z)\,\Phi(w)$ takes the form

\[T_{s}(z)\,\Phi(w) \;=\; \sum_{k \ge 0}\frac{c_{s, k}\,L_{-k}^{\Phi}(w)}{(z - w)^{s - k}} \;+\; \text{regular}, \tag{4}\]

where $L_{-k}^{\Phi}$ are Virasoro descendants at level $k$ of $\Phi$ (the original primary), and $c_{s, k}$ are OPE coefficients determined by the conformal symmetry.

Taking the residue at $w \to z$ — i.e. extracting the coefficient of $1/(z - w)$ — yields

\[R_{s}[\Phi](z) \;=\; c_{s,\,s - 1}\,L_{-(s - 1)}^{\Phi}(z). \tag{5}\]

For (3) to hold, this single Virasoro descendant of $\Phi$ at level $s - 1$ must be a total $\partial$-derivative, equivalently a descendant of the form $L_{-1}^{\Phi}\,\mathcal{O}$ for some local operator $\mathcal{O}$. (Recall $L_{-1}$ is the translation generator, so $L_{-1} \mathcal{O} = \partial \mathcal{O}$.)

Therefore $Q_{s}$ survives iff $L_{-(s - 1)}^{\Phi}$ is a $\partial$-derivative, i.e. iff $L_{-(s - 1)}^{\Phi}\in L_{-1}\cdot(\text{level}\,(s - 2)\text{-descendants of }\Phi)$ at level $s - 1$ of the Verma module $V_{\Phi}$.

Step 3 — Null-vector constraint for $\Phi = \sigma$

For a generic primary $\Phi$ with no null vectors, $L_{-(s-1)}\Phi$ is linearly independent of $L_{-1}\cdot(\text{descendants})$ — the survival condition fails at every spin except the trivial $s = 1$ (where $L_{0}\Phi = h_{\Phi}\,\Phi$ is the total-derivative tautology). No nontrivial conserved charges survive.

The Ising $\sigma$ is special because it has a null descendant at level 2 (from the Kac table: $\sigma = \phi_{2, 1}$ has a level-$r s = 2 \cdot 1 = 2$ null vector). Explicitly (Di Francesco– Mathieu–Sénéchal 1997 §8.1):

\[\bigl(L_{-2} \;-\; \tfrac{3}{4}\,L_{-1}^{2}\bigr)\sigma \;=\; 0 \quad\text{modulo total descendants.} \tag{6}\]

This null relation implies that $L_{-2}\sigma$ is a $\partial$-derivative: $L_{-2}\sigma = \tfrac{3}{4}L_{-1}^{2}\sigma = \tfrac{3}{4}\partial^{2}\sigma$.

More generally, the null vector (6) propagates up the Verma module: $L_{-1}^{n}$ acting on the level-2 null gives nullifications at higher levels, which create "gaps" in the space of non-derivative descendants. The survival condition (3) is non-trivially satisfied at precisely those levels $s - 1$ where the null-vector cascade produces a $\partial$-derivative-equivalent class.

Step 4 — Spin-counting via null cascades (result)

The survival condition (3) must be checked spin-by-spin. The relevant object is the generating function of "surviving descendants" — those in the Verma module $V_{\sigma}$ modulo $\partial$-exact forms and the null-vector ideal generated by (6):

\[\chi_{\sigma}^{\rm surv}(q) \;=\; \sum_{n \ge 0} \dim\!\bigl[\,V_{\sigma}^{(n)} / \bigl(L_{-1}\,V_{\sigma}^{(n - 1)} \oplus (\text{null cascade at level }n)\bigr)\,\bigr]\,q^{n},\]

where $V_{\sigma}^{(n)}$ is the level-$n$ subspace of the Verma module. The Kac null-vector propagation (applying $L_{-1}^{k}$ to (6) for $k = 0, 1, 2, \ldots$) removes not only level 2 itself but a whole family of higher levels, creating an intricate pattern of "surviving" levels.

The full computation, carried out in Zamolodchikov 1989 §4 (see also Delfino 2004 Proposition 1 and Di Francesco–Mathieu–Sénéchal 1997 §12.4 for streamlined re-derivations), yields the explicit sequence of non-trivial surviving levels

\[s - 1 \;\in\; \{0,\,6,\,10,\,12,\,16,\,18,\,22,\,28\},\]

giving the conserved-charge spins

\[\boxed{\; s \;\in\; \{1,\,7,\,11,\,13,\,17,\,19,\,23,\,29\}. \;}\]

These are precisely the exponents of the $E_{8}$ Lie algebra (Coxeter number $h = 30$; the exponents are the positive integers less than $h$ coprime to $h$). The full generating-function form, corresponding to the $E_{8}$ "$W_{1+\infty}$ character" on the perturbed theory, is Zamolodchikov 1989 eq. (4.5); we do not reproduce its closed form here since the spin list above is all that enters the bootstrap argument.

The Remark that ties this back to the $E_{8}$ structure:

The spin spectrum $\{1, 7, 11, 13, 17, 19, 23, 29\}$ of the surviving conserved charges is precisely the multiset of exponents of the $E_{8}$ root system. By Dorey's theorem (e8-mass-spectrum-derivation Step 4), any integrable field theory whose spin currents form exactly this multiset is the $E_{8}$ affine Toda theory, up to the overall mass scale.

So the counting identifies the perturbed Ising theory as the $E_{8}$ massive integrable theory without constructing its $S$-matrix explicitly; the $S$-matrix then follows from the bootstrap program with $E_{8}$ structure imposed.

Step 5 — Why $E_{8}$ exponents?

The exponents of $E_{8}$ are exactly $\{1, 7, 11, 13, 17, 19, 23, 29\}$, the numbers coprime to $30 = h_{E_{8}}$ in the range $[1, h - 1]$. Coincidentally (though not coincidentally from the representation-theory point of view), these are also the positive integers $s < 30$ that:

are not multiples of $2$, $3$, or $5$ — the primes dividing $30$,
equivalently, are coprime to $h = 30$.

The appearance of these specific integers in Zamolodchikov's null-vector counting for the $\sigma$-perturbation of $\mathcal{M}(3, 4)$ is a structural coincidence reflecting the hidden $E_{8}$ root lattice of the perturbed theory. Dorey 1991 turned this coincidence into a theorem: any integrable perturbation by a primary with the "right" null-vector structure gives an ADE-Toda theory, and the specific perturbation $\sigma$ of $\mathcal{M}(3, 4)$ at level 2 is the one that produces $E_{8}$.

Step 6 — Contrast: $\varepsilon$ perturbation → free massive Majorana

Perturbing by $\varepsilon$ instead gives:

\[\mathcal{A}_{\varepsilon} \;=\; \mathcal{A}_{\rm Ising\,CFT} \;+\; t\,\int d^{2}x\,\varepsilon(x),\]

with $t = \beta(T - T_{c})/J$ the thermal coupling (scaling dimension $\Delta_{\varepsilon} = 1$). The $\varepsilon$ operator is $\phi_{1, 2}$ in the Kac table with null vector at level $r s = 1 \cdot 2 = 2$:

\[\bigl(L_{-2} \;-\; \tfrac{3}{4 h_{\varepsilon}}\,L_{-1}^{2}\bigr) \varepsilon \;=\; 0, \qquad h_{\varepsilon} = \tfrac{1}{2},\]

i.e. $\bigl(L_{-2} - \tfrac{3}{2} L_{-1}^{2}\bigr)\varepsilon = 0$. The analogous counting for the $\varepsilon$-perturbation gives surviving spins $s = 1, 3, 5, 7, \ldots$ — all odd spins survive. This reflects the fact that the perturbed theory is a free massive Majorana fermion, which has infinitely many conserved currents (one for every odd spin, corresponding to products of the fermion stress tensor) as any free field theory.

The difference between the two perturbations is geometric: the $\varepsilon$-perturbation preserves the Bose-Fermi free structure of the underlying Majorana, whereas the $\sigma$-perturbation breaks the free-fermion symmetry (the $\sigma$ operator is non-local in the fermion language — it is the disorder operator dual to the free-fermion spin) and leaves behind only a finite reminder of the infinite symmetry: the eight charges whose spins match $E_{8}$ exponents.

Step 7 — Mass gap scaling

The perturbation $\lambda\int\sigma$ has $[\lambda] = [\text{mass}]^{2 - \Delta_{\sigma}} = [\text{mass}]^{15/8}$ by dimensional analysis (since $\int d^{2}x$ has dimension $-2$ and $\sigma(x)$ has dimension $\Delta_{\sigma} = 1/8$, and the action must be dimensionless). The only mass scale in the perturbed theory is $\lambda$, so

\[\boxed{\; m_{1} \;\propto\; |\lambda|^{1/(2 - \Delta_{\sigma})} \;=\; |\lambda|^{1/(15/8)} \;=\; |\lambda|^{8/15}. \;} \tag{7}\]

The exact proportionality constant involves a beautiful integral evaluation (Fateev 1994 eq. (1.1)):

\[m_{1} \;=\; C(\tfrac{8}{15})\,|\lambda|^{8/15},\qquad C(\tfrac{8}{15}) \;=\; \frac{2\,\Gamma(1/5)\,\Gamma(2/3)\,\Gamma(8/15)} {\sqrt{\pi}\,\Gamma(3/5)\,\Gamma(5/6)\,\Gamma(8/15)} \;\times\;(\text{$E_{8}$-Toda overall normalisation}),\]

though the numerical value depends on the chosen normalisation of $\sigma$ and $\lambda$.

Step 8 — Experimental confirmation

The compound CoNb$_{2}$O$_{6}$ is a quasi-1D Ising ferromagnet with residual inter-chain coupling acting as a small longitudinal field. Near its quantum critical point ($B \approx B_{c} \approx 5.5\,\text{T}$ transverse field, $T \to 0$), Coldea et al. 2010 used inelastic neutron scattering to resolve two sharp one-particle excitation modes. The observed mass ratio

\[\frac{m_{2}}{m_{1}} \;=\; 1.618 \pm 0.015\]

agrees with the $E_{8}$ prediction $\varphi = (1 + \sqrt{5})/2 = 1.6180\ldots$ within experimental error. A tentative identification of the $m_{3}$ particle was reported (predicted $m_{3}/m_{1} \approx 1.989$). This remains the most direct experimental verification of a Lie algebra predicting particle masses in a condensed-matter system.

References

A. B. Zamolodchikov, Integrals of motion and S-matrix of the (scaled) $T = T_{c}$ Ising model with magnetic field, Int. J. Mod. Phys. A 4, 4235 (1989). The discovery paper; conserved-charge counting eq. (4.5).
A. B. Zamolodchikov, Infinite additional symmetries in two-dimensional conformal quantum field theory, Theor. Math. Phys. 65, 1205 (1985). Construction of higher-spin conserved currents in CFT (the $\mathcal{W}$-algebra).
P. Dorey, Root systems and purely elastic S-matrices, Nucl. Phys. B 358, 654 (1991). ADE/Toda integrable theories from null-vector perturbations.
G. Delfino, Integrable field theory and critical phenomena: the Ising model in a magnetic field, J. Phys. A 37, R45 (2004). Comprehensive review; Proposition 1 contains the surviving-spin counting with full null-vector accounting.
V. A. Fateev, The exact relations between the coupling constants and the masses of particles for the integrable perturbed conformal field theories, Phys. Lett. B 324, 45 (1994), eq. (1.1). Exact proportionality constant for $m_{1}(\lambda)$ in the $\sigma$-perturbed $\mathcal{M}(3, 4)$.
J. L. Cardy, Scaling and Renormalization in Statistical Physics, Cambridge University Press (1996), §11. OPE-based derivation of (1)–(3).
P. Di Francesco, P. Mathieu, D. Sénéchal, Conformal Field Theory, Springer (1997), §6.6 and §8.1. General OPE machinery (4)–(5), Ising null-vector equation (6).
R. Coldea et al., Quantum criticality in an Ising chain: experimental evidence for emergent E$_{8}$ symmetry, Science 327, 177 (2010). CoNb$_{2}$O$_{6}$ neutron-scattering experiment.

Used by

e8-mass-spectrum-derivation — the $E_{8}$ mass formula takes Zamolodchikov's surviving-charge list as the input that identifies the perturbed theory as the $E_{8}$ Toda theory.
ising-cft-primary-operators — the $\sigma$ primary and its null vector are derived / enumerated there.
$E_{8}$ universality class — the mass-ratio universality class realised by this perturbation.
TFIM model page — lattice origin of the perturbation (small residual longitudinal field at $h = J$).