2
$\begingroup$

The 1D Ising model at criticality is given by the Hamiltonian $H=-\mathcal{N} \sum_i (\sigma^x_i \sigma^x_{i+1} + \sigma_i^z)$ in terms of Pauli operators and a normalization $\mathcal{N}$. In CFT language, the operator content of this theory is supposed to be described by the primary fields usually called $\mathbb{1}, \epsilon,\psi$ and $\bar{\psi}$ with scaling dimensions $\Delta_\mathbb{1}=0$, $\Delta_\epsilon = 1$ and $\Delta_\psi =\Delta_\bar{\psi}=1/2$. In addition, plenty of papers (such as this one) mention the fields $\sigma$ and $\mu$ with $\Delta_\sigma = \Delta_\mu = 1/8$.

Question: What physical (local) operators/observables do these fields correspond to? My understanding so far:

  • The identity field $\mathbb{1}$ seems to be equivalent to a local identity operator.

  • Is the "energy" $\epsilon$ at site $i$ simply $\sigma^x_i \sigma^x_{i+1} + \sigma_i^z$? The Ising CFT Wikipedia article also mentions a field called $\epsilon^\prime$ with $\Delta_{\epsilon^\prime}=4$, whose meaning is unexplained.

  • The fields $\psi$ and $\bar{\psi}$ seem to correspond to the Majorana modes $c_{2i-1}$ and $c_{2i}$ after Jordan-Wignerizing the Hamiltonian.

  • The "order" $\sigma$ appears to correspond to the Pauli operator $\sigma^x_ i$ at site $i$.

  • I have no idea what the "disorder" $\mu$ corresponds to, but it seems to be an order parameter on the dual lattice.

The only proper textbook I have currently available (Christe/Henkel's "Introduction to Conformal Invariance and Its Application to Critical Phenomena") talks about these fields but doesn't explain $\sigma$ and $\mu$, only mentioning that they can't be written locally in terms of $\epsilon$ and the $\psi,\bar{\psi}$.

|cite|improve this question
$\endgroup$
  • 1
    $\begingroup$ There's a whole chapter about this in the big yellow CFT book. $\endgroup$ – Ryan Thorngren Mar 10 '18 at 22:53
3
$\begingroup$

Lattice-continuum correspondence for the primary fields of the Ising CFT:

  1. Indeed, the identity field $1$ simply corresponds to the identity operator on the lattice. More generally, any lattice operator with a finite expectation value in the ground state has support on the identity field in the continuum.
  2. The energy field $\varepsilon$ corresponds to $\sigma^x_{n} \sigma^x_{n+1} - \sigma^z_n$. Notice the relative minus sign which was not in your formula! This is essential: without the minus sign, the operator would have a non-zero expectation value (of course, since it is the Hamiltonian density), meaning it would (partially) have support on $1$ in the continuum. But with the minus sign it is now odd under the Kramers-Wannier duality. Firstly, this means it is zero at the critical point, but moreover if one adds $g \; \varepsilon$ to the action/Hamiltonian, then depending on the sign of $g$, one lands into the symmetry-broken or paramagnetic phase.
  3. The order field $\sigma$: you are correct that it corresponds to the order parameter $\sigma^x_n$ on the lattice.
  4. The disorder field $\mu$ corresponds to $\sigma^z_1 \sigma^z_2 \cdots \sigma^z_n$ on the lattice. There are a few ways of interpreting this: (a) note that this is the operator that maps to $\sigma^x_n$ under the Kramers-Wannier duality, (b) more physically, this operator creates a domain wall between sites $n$ and $n+1$ (and also near site $1$, but it is convenient to imagine having open boundary conditions there). Hence the fact that the paramagnetic phase has $\langle \mu \rangle = \langle \sigma^z_1 \sigma^z_2 \cdots \sigma^z_n \rangle \neq 0$ is exactly telling us that domain walls have condensed into the ground state. Lastly, (c) from a more general point of view, this non-local operator arises as the string order parameter for the trivial symmetry-protected topological phase. (Hence, if you want to learn more about this latter perspective, just google for `string order parameter'.)
  5. With regards to the fermionic fields $\psi,\bar \psi$, from what I remember they don't exactly correspond to the lattice Majorana operators $\gamma_n$ and $\tilde \gamma_n$, but rather their linear combinations $\gamma_n \pm \tilde \gamma_n$. (One way of seeing this has to be so, is that in the continuum, one gets expressions like $\psi \partial \psi$, yet on the lattice $\gamma$ can only couple to $\tilde \gamma$.)
|cite|improve this answer
$\endgroup$
  • $\begingroup$ Since you identify $\epsilon$ with $\sigma^x_n \sigma^x_{n+1} - \sigma^z_n = i \tilde{\gamma}_n (\gamma_{n+1} + \gamma_n)$, does that mean we should identify $\psi$ with $\tilde{\gamma}_n$ and $\bar{\psi}$ with $\gamma_{n+1} + \gamma_n$, so $i \psi \bar{\psi}$ gives us a "mass term"? $\endgroup$ – Scytale May 10 '18 at 5:41
  • $\begingroup$ Following this script (page 28), one can transform from physical fermionic operators $c_i$ on a lattice of width $a$ to physical fields $\Psi(x_n) = c_n / \sqrt{a}$. Then the Hamiltonian term proportional to the gap is $c^\dagger_n c_n = 1 + i \gamma_n \tilde{\gamma}_n$, suggesting the identification $\epsilon \sim \gamma_n \tilde{\gamma}_n$ and $\psi, \bar{\psi} \sim c_n, c^\dagger_n$. This seems inconsistent with your reasoning. $\endgroup$ – Scytale May 10 '18 at 14:05
  • $\begingroup$ @Scytale I might have been wrong about the fermionic fields (my point (5)), that's the part I haven't thought about too carefully. But with regards to $\varepsilon$, the reason to prefer $XX-Z$ is that it is odd under Kramers-Wannier, which ensures that its expectation value is zero at the critical point. The suggestion $\varepsilon \sim Z$ (which is what you suggest) is not good because $\langle Z \rangle \neq 0$ at the critical point, which implies that $Z$ has support in the identity field $1$. $\endgroup$ – Ruben Verresen May 10 '18 at 14:22
  • $\begingroup$ In understand your point, but it seems to violate the usual definition $\epsilon = i \psi \bar{\psi}$ where $\psi$ and $\bar{\psi}$ are the continuous version of the fermionic operators on the lattice. Perhaps I am thinking of the wrong continuum limit? $\endgroup$ – Scytale May 10 '18 at 15:46
  • $\begingroup$ @Scytale I don't know why you say that that is the usual definition of $\varepsilon$. Are you confusing the $\varepsilon$-field with the Hamiltonian energy density? $\endgroup$ – Ruben Verresen May 10 '18 at 16:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.