path: root/stokes.tex

\documentclass[aps,reprint,longbibliography,floatfix]{revtex4-2}

\usepackage[utf8]{inputenc} % why not type "Stokes" with unicode?
\usepackage[T1]{fontenc} % vector fonts plz
\usepackage{newtxtext,newtxmath} % Times for PR
\usepackage[
  colorlinks=true,
  urlcolor=purple,
  citecolor=purple,
  filecolor=purple,
  linkcolor=purple
]{hyperref} % ref and cite links with pretty colors
\usepackage{amsmath, graphicx, xcolor} % standard packages

\begin{document}

\title{}

\author{Jaron Kent-Dobias}
\author{Jorge Kurchan}

\affiliation{Laboratoire de Physique de l'Ecole Normale Supérieure, Paris, France}

\date\today

\begin{abstract}
\end{abstract}

\maketitle

Consider a thermodynamic calculation involving the (real) $p$-spin model for a
particular instantiation of the coupling tensor $J$
\begin{equation} \label{eq:partition.function}
  Z_J(\beta)=\int_{S^{N-1}}ds\,e^{-\beta H_J(s)}
\end{equation}
where $S^{N-1}$ is the $N-1$-sphere of radius $N$. Quantities of interest are
usually related to the quenched free energy, produced by averaging over the
$J$s the sample free energy $F_J$
\begin{equation}
  \overline F(\beta,\kappa)=-\beta^{-1}\int dJ\,p_\kappa(J)\log Z_J
\end{equation}
which can depend in general on the inverse temperature $\beta$ and on some
parameter $\kappa$ which governs the distribution of $J$s. For most
applications, $\beta$ is taken to be real and positive, and the distribution
$p_\kappa$ is taken to be Gaussian or discrete on $\pm1$.

We are interested in analytically continuing expressions like $\overline F$
into the region of complex $\beta$ or distributions $p_\kappa$ involving
complex $J$. The former has been considered extensively for the Gaussian
$p$-spin in the past \cite{complex_energy}, while the latter is largely
unexplored.

When the argument of the exponential integrand in \eqref{eq:partition.function}
acquires an imaginary component, various numeric and perturbative schemes for
approximating its value can face immediate difficulties due to the emergence of
a sign problem, resulting from rapid oscillations coinciding with saddles.

The surface $M\subset\mathbb C^N$ defined by $z^2=N$ is an $N-1$ dimensional
\emph{Stein manifold}, a type of complex manifold defined by the level set of a
holomorphic function \cite{Forstneric_2017_Stein}. Suppose that $z:\mathbb
C^{N-1}\to M$ is a holomorphic map. The Jacobian $J$ of the map is
\begin{equation}
  J_{i\alpha}=\frac{\partial z_i}{\partial u_\alpha}=\partial_\alpha z_i
\end{equation}
where Greek coefficients run from $1$ to $N-1$ and Latin coefficients from $1$
to $N$. The hermitian metric is $g=J^\dagger J$. For any smooth function
$\phi:M\to\mathbb R$, its gradient $\nabla\phi$ is a holomorphic vector field
given by
\begin{equation}
  \nabla\phi=(\partial^*\phi)^\sharp=(\partial^*\phi)g^{-1}
\end{equation}

For
coordinates $u\in\mathbb C^{N-1}$, dynamics consists of gradient descent on
$\operatorname{Re}H$, or
\begin{equation}
  \dot u=-\nabla\operatorname{Re}H=-\tfrac12(\partial H)^\dagger g^{-1}
\end{equation}
These dynamics preserve $\operatorname{Im}H$ and in fact correspond to
Hamiltonian dynamics, with the real and imaginary parts of the coordinates
taking the role of conjugate variables. \cite{Morrow_2006_Complex}

Working with a particular map is inconvenient, and we would like to develop a map-independent dynamics. Using the chain rule, one finds
\begin{equation}
  \begin{aligned}
    \dot z_i
    &=\dot u_\alpha\partial_\alpha z_i
    =-\tfrac12(\partial_\beta H)^*g^{-1}_{\beta\alpha}\partial_\alpha z_i\\
    &=-\tfrac12(\partial_j H)^*(\partial_\beta z_j)^*g^{-1}_{\beta\alpha}\partial_\alpha z_i
    =-\tfrac12(\partial H)^\dagger(J^\dagger g^{-1}J)\\
  \end{aligned}
\end{equation}
where $J$ is the Jacobian of the coordinate map and $g$ is the metric. In stereographic coordinates this can be worked out directly.
Consider the coordinates $z_i=u_i$ for $1\leq i\leq N-1$ and
\begin{equation}
  z_N=\sqrt{N-u_\alpha u_\alpha}
\end{equation}
The Jacobian is
\begin{equation}
  J_{\alpha i}=\partial_\alpha z_i=\delta_{\alpha i}-\delta_{Ni}\frac{u_\alpha}{\sqrt{N-u_\beta u_\beta}}
\end{equation}
and the corresponding hermitian metric is
\begin{equation}
  g_{\alpha\beta}=J_{i\alpha}^*J_{i\beta}
  =\delta_{\alpha\beta}+\frac{u_\alpha^*u_\beta}{|N-u_\gamma u_\gamma|}
\end{equation}
The metric can be inverted explicitly:
\begin{equation}
  g^{-1}_{\alpha\beta}
  =\delta_{\alpha\beta}-\frac{u_\alpha^*u_\beta}{|N-u_\gamma u_\gamma|+|u|^2}.
\end{equation}
Putting these pieces together, we find
\begin{equation}
  (J^\dagger g^{-1}J)_{ij}
  =\delta_{ij}-\frac{z_iz_j^*}{|z|^2}
\end{equation}
which is just the projector onto the constraint manifold.

Therefore, a map-independent dynamics is given by
\begin{equation}
  \dot z
  =-\frac12(\partial H)^\dagger\left(I-\frac{zz^\dagger}{|z|^2}\right)
\end{equation}

Stokes lines are trajectories that approach distinct critical points as time
goes to $\pm\infty$. From the perspective of dynamics, these correspond to
\emph{heteroclinic orbits}. What are the conditions under which Stokes lines
appear? Because the dynamics conserves imaginary energy, two critical points
must have the same imaginary energy if they are to be connected by a Stokes line.

The level sets defined by $\operatorname{Im}H=c$ for $c\in\mathbb R$ are surfaces of
$2(N-1)-1$ real dimensions. They must be simply connected, since gradient
descent in $\operatorname{Re}H$ cannot pass the same point twice.

\section{2-spin}

\begin{equation}
  H_0=\frac12z^TJz
\end{equation}
$J$ is generically diagonalizable by a complex orthogonal matrix $P$. With
$z\mapsto Pz$, $J\mapsto D$ for diagonal $D$. Then $\partial_i H=d_iz_i$.
Suppose that two critical points have the same imaginary energy; without loss
of generality, assume these are the first and second components. Since the gradient is proportional to $z$, any components that are zero at some time will be zero at all times. The dynamics for the components of interest assuming all others are zero are
\begin{equation}
  \begin{aligned}
    \dot z_1
    &=-z_1^*\left(d_1^*-\frac{d_1^*z_1^*z_1+d_2^*z_2^*z_2}{|z_1|^2+|z_2|^2}\right) \\
    &=-(d_1-d_2)^*z_1^*\frac{|z_2|^2}{|z_1|^2+|z_2|^2}
  \end{aligned}
\end{equation}
and the same for $z_2$ with all indices swapped.  Since $\Delta=d_1-d_N$ is
real, if $z_1$ begins real it remains real, with the same for $z_2$. Since the critical points are at real $z$, we make this restriction, and find
\begin{equation}
  \begin{aligned}
    \frac d{dt}(z_1^2+z_2^2)=0\\
    \frac d{dt}\frac{z_2}{z_1}=\Delta\frac{z_2}{z_1}
  \end{aligned}
\end{equation}
Therefore $z_2/z_1=e^{\Delta t}$, with $z_1^2+z_2^2=N$ as necessary.  Depending on the sign of $\Delta$, $z$ flows
from one critical point to the other over infinite time. This is a Stokes line,
and establishes that any two critical points in the 2-spin model with the same
imaginary energy will possess one.

The critical points of the 2-spin model are all adjacent: no critical point is separated from another by the separatrix of a third. This means that when the imaginary energies of two critical points are brought to the same value, their surfaces of constant imaginary energy join.

\section{p-spin}

\section{(2 + 4)-spin}

\begin{acknowledgments}
  MIT mathematicians have been no help
\end{acknowledgments}

\bibliography{stokes}

\end{document}