Taming NSPT fluctuations in O ( N ) O(N) Non-Linear Sigma Model: simulations in the large N N regime
Francesco Di Renzo · 2024
Read Taming NSPT fluctuations in O ( N ) O(N) Non-Linear Sigma Model: simulations in the large N N regime by Francesco Di Renzo (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-02624.txt.
Introduction
Paolo Baglioni and Francesco Di Renzo investigate, in a 2024 paper, noise troubles in a computational method for a classic field theory. The theory, a nonlinear sigma model, lives on a curved target space and shows asymptotic freedom. High-order calculations for it rely on a stochastic perturbation method.
That method's random fluctuations grow with each order, above all in low dimensions. In some models the noise can drown the signal entirely. The authors show the trouble depends strongly on the model's size parameter. Larger values make higher orders feasible again. Their simulations map where the method can be trusted. Taming such noise decides which theories can be computed perturbatively at all.
Catalog record
| CREATOR | Francesco Di Renzo |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-02624
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Taming NSPT fluctuations in O ( N ) O(N) Non-Linear Sigma Model: simulations in the large N N regime Francesco Di Renzo Abstract The Non-Linear Sigma Model (NLSM) is an example of a field theory on a target space exhibiting intricate geometry. One remarkable characteristic of the NLSM is asymptotic freedom, which triggers interest in perturbative calculations. In the lattice formulation of NLSM, one would naturally rely on Numerical Stochastic Perturbation Theory (NSPT) to conduct high-order computations. However, when dealing with low-dimensional systems, NSPT reveals increasing statistical fluctuations with higher and higher orders. This of course does not come as a surprise and one is ready to live with this, as long as the noise is not going to completely kill the signal, which unfortunately in some models does take place. We investigate how, in the O ( N ) O(N) context, this behaviour strongly depends on N N . As expected, larger N N values make higher-order computations feasible. [a,b]Paolo Baglioni 1 Introduction and motivations Lattice Gauge Theories (LGT) are an example of regularized field theories designed for a non-perturbative approach. Through Monte Carlo simulations, LGTs remain one of the primary tools for numerically investigating non-perturbative physics. In this context, perturbation theory is known to be particularly cumbersome. Nevertheless, perturbation theory holds its significance even in lattice theories. Numerical Stochastic Perturbation Theory (NSPT) [ 1 ] stands out as a renowned computational tool for generating perturbative expansions in lattice theories, sharing striking similarities with non-perturbative Monte Carlo methods [ 2 ] . NSPT is a fully automated perturbative Monte Carlo algorithm that involves order-by-order numerical integration of stochastic differential equations. It has gained considerable interest among lattice practitioners mainly because of the agility of its implementation compared to its significant potential. Since it has been introduced, NSPT simulations have allowed the calculation of high-order loop corrections in lattice QCD [ 3 , 4 , 5 , 6 , 7 ] and also across various other theories [ 8 , 9 , 10 ] . Currently, NSPT has been implemented in different forms [ 11 , 7 ] and remains one of the main distinctive tools for investigating asymptotic perturbative behavior in lattice QCD, probing fundamental properties such as infrared renormalons [ 12 , 13 , 14 , 15 ] . One interesting aspect of NSPT is the freedom to expand around different non-trivial vacua. In this respect, low-dimensional models emerge as the optimal candidates for exploratory analysis, but apart from first-order loop computations in quantum mechanics [ 16 ] , perturbative expansions on top of instanton-like vacua still remain mostly unexplored. As a matter of fact, it is well-known that significant fluctuations arise in low-dimensional models even at relatively low perturbative orders, precluding straightforward NSPT implementation for such computations [ 17 ] . A reasonable explanation is that fluctuations emerge due to the limited number of degrees of freedom. This hypothesis is supported by the absence of such fluctuations in more complex systems like lattice QCD. In this study, we consider a class of highly interesting low-dimensional lattice models, the O(N) Non-Linear Sigma Model (NLSM) in 2D . Besides its theoretical and phenomenological significance, the O(N) NLSM provides an ideal testing ground for the aforementioned conjecture. In fact, we can change the number of local degrees of freedom by tuning the value of N . On very general grounds, we expect that fluctuations are less severe in the large N regime, where the theory includes a large number of degrees of freedom. In this manuscript we report the current status of our simulations, which strongly suggests that the signals become increasingly free of fluctuations as the value of N grows (see also [ 18 , 19 ] for more details). 2 Numerical Stochastic Perturbation Theory NSPT has been inspired by the famous work of Parisi and Wu on Stochastic Quantization (SQ) [ 20 ] . The main idea is to translate the stochastic perturbation theory in a set of order-by-order differential equations that must be solved numerically, formulating a strategy for a perturbative Monte Carlo algorithm. In the SQ framework, we consider the degrees of freedom as function of an extra parameter, the stochastic time : \displaystyle\phi(\boldsymbol{x})\rightarrow\phi(\boldsymbol{x},\tau)\,. (1) The evolution takes place in the stochastic time following the Langevin equation \displaystyle\frac{d\phi(\boldsymbol{x},\tau)}{d\tau}=-\frac{\partial\mathcal{S}_{E}[\phi]}{\partial\phi(\boldsymbol{x},\tau)}+\eta(\boldsymbol{x},\tau)\,, (2) where \mathcal{S}_{E}[\phi] is the Euclidean action, and \eta(\boldsymbol{x},\tau) is a white noise normalized such that \displaystyle\langle\eta(\boldsymbol{x},\tau)\rangle_{\eta}=0\,,\quad\langle\eta(\boldsymbol{x},\tau)\eta(\boldsymbol{x}^{\prime},\tau^{\prime})\rangle_{\eta}=2\delta(\boldsymbol{x}-\boldsymbol{x}^{\prime})\delta(\tau-\tau^{\prime})\,. (3) In Eq. ( 3 ) the subscript \langle\ldots\rangle_{\eta} means the average over all possible white noise realizations, namely \displaystyle\langle\dots\rangle_{\eta}=\frac{\int\mathcal{D}\eta\ \dots\ e^{-\frac{1}{4}\int d\boldsymbol{x}d\tau\eta^{2}(\boldsymbol{x},\tau)}}{\int\mathcal{D}\eta\ e^{-\frac{1}{4}\int d\boldsymbol{x}d\tau\eta^{2}(\boldsymbol{x},\tau)}}\,. (4) The central assertion of SQ is that the averaging over noise realizations at the limit of infinite stochastic time results in the expectation value of the corresponding Euclidean field theory, namely \displaystyle\lim_{\tau\to\infty}\langle O[\phi(\boldsymbol{x}_{1},\tau),\phi(\boldsymbol{x}_{2},\tau),\dots,\phi(\boldsymbol{x}_{n},\tau)]\rangle_{\eta}=\langle O[\phi(\boldsymbol{x}_{1}),\phi(\boldsymbol{x}_{2}),\dots,\phi(\boldsymbol{x}_{n})]\rangle\,, (5) where \displaystyle\langle O[\phi(\boldsymbol{x}_{1}),\phi(\boldsymbol{x}_{2}),\dots,\phi(\boldsymbol{x}_{n})]\rangle\ =\frac{\int\mathcal{D}\phi\ O[\phi(\boldsymbol{x}_{1}),\phi(\boldsymbol{x}_{2}),\dots,\phi(\boldsymbol{x}_{n})]\ e^{-\mathcal{S}_{E}[\phi]}}{\int\mathcal{D}\phi\ e^{-\mathcal{S}_{E}[\phi]}}\,. (6) NSPT involves formal expansion in the field’s coupling constant, namely \displaystyle\phi(\boldsymbol{x},\tau)=\sum_{n=0}^{\infty}g^{n}\phi^{(n)}(\boldsymbol{x},\tau)\,, (7) and the order-by-order numerical integration of the set of equations that are obtained by inserting the previous definition in Eq. ( 2 ), i.e. \displaystyle\begin{split}\frac{d\phi^{(0)}(\boldsymbol{x},\tau)}{d\tau}&=-G_{0}^{-1}\phi^{(0)}(\boldsymbol{x},\tau)+\eta(\boldsymbol{x},\tau)\,,\\ \frac{d\phi^{(1)}(\boldsymbol{x},\tau)}{d\tau}&=-G_{0}^{-1}\phi^{(1)}(\boldsymbol{x},\tau)+D_{1}(\phi^{(0)})\,,\\ \dots{\\ }\frac{d\phi^{(n)}(\boldsymbol{x},\tau)}{d\tau}&=-G_{0}^{-1}\phi^{(n)}(\boldsymbol{x},\tau)+D_{n}(\phi^{(0)},\phi^{(1)},\dots,\phi^{(n-1)})\,.\end{split} (8) In the above equation G_{0} represents the free propagator, while D_{n} denotes functions of the perturbative fields that couple different perturbative orders. It is important to emphasize that the NSPT code only needs to handle the Langevin dynamics in Eq. ( 2 ) as it is : all the order-by-order operations are automatically decoded using polynomial operations in the coupling constant (a feature that has recently gained popularity as Automatic Differentiation [ 21 ] ). In addition, the observables are implemented in their non-perturbative form, with the corresponding perturbative expansion being automatically generated as \displaystyle\begin{split}O[\phi]&=O^{(0)}[\phi^{(0)}]+gO^{(1)}[\phi^{(0)},\phi^{(1)}]+g^{2}O^{(2)}[\phi^{(0)},\phi^{(1)},\phi^{(2)}]\,,\\ &=\sum_{n=0}^{\infty}g^{n}O^{(n)}[\phi^{(0)},\ldots,\phi^{(n)}]\,.\end{split} (9) The numerical integration of Eqs. ( 8 ) requires a designated small time step \Delta\tau=\tau/n_{\text{steps}} . The resulting discrete dynamics no longer satisfy Eq. ( 5 ), requiring the continuum stochastic time extrapolation \Delta\tau\to 0 (a detailed presentation concerning the continuum stochastic time extrapolation is reported in [ 19 ] ). 3 O(N) NLSM: Lattice perturbation theory setup In this work, we study the 2D O(N) Non-Linear Sigma Model defined by the action \displaystyle S=-\frac{1}{g}\sum_{x,\mu}\boldsymbol{s}_{x}\cdot\boldsymbol{s}_{x+\mu}\,, (10) where \boldsymbol{s}_{x} is a N -component real scalar field constrained by \boldsymbol{s}_{x}\cdot\boldsymbol{s}_{x}=1 , and g is the coupling constant. NLSMs are a very popular laboratory for studying particle and field properties in various contexts (we refer to [ 22 ] for a complete introduction to this topic). Here we are not interested in such features, but we will study statistical proprieties of the NSPT distributions as a function of N , which represents the number of local degrees of freedom. Perturbation theory is formulated in terms of the unconstrained field \boldsymbol{\pi}_{x} [ 23 ] , defined by \displaystyle\boldsymbol{s}_{x}=(\sqrt{g}\boldsymbol{\pi}_{x},\sigma_{x})\,, (11) where the last component \sigma_{x} is integrated out using the local constraint. Neglecting irrelevant contributions in perturbation theory [ 23 ] , the partition function is then expressed as \displaystyle Z=\lim_{\lambda\to 0}\int\prod_{x}\ d\boldsymbol{\pi}_{x}\ e^{-\frac{1}{2}\sum_{x,\mu}\Bigl[(\Delta_{\mu}\boldsymbol{\pi}_{x})^{2}+\lambda^{2}\boldsymbol{\pi}_{x}^{2}-\frac{1}{g}(\Delta_{\mu}\sqrt{1-g\boldsymbol{\pi}_{x}^{2}})^{2}\Bigr]-\frac{1}{2}\sum_{x}\log{(1-g\boldsymbol{\pi}_{x}^{2})}}\,. (12) In Eq. ( 12 ), we introduced the infrared regulator \lambda , while \Delta_{\mu} represents the usual lattice-discretized derivative. It is important to note that the action now involves logarithmic and square root interactions that need to be managed in the framework of perturbation theory. In this case, we have computed the