A simple construction of the sine-Gordon model via stochastic quantization
Massimiliano Gubinelli , Martin Hairer , Tadahiro Oh , and Younes Zine · 2024
Read A simple construction of the sine-Gordon model via stochastic quantization by Massimiliano Gubinelli , Martin Hairer , Tadahiro Oh , and Younes Zine (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-2412-16404.txt.
Introduction
Massimiliano Gubinelli, Martin Hairer, Tadahiro Oh, and Younes Zine construct a quantum field model in a 2024 math paper. The model is the sine-Gordon model, a classic of mathematical physics. Their route is stochastic quantization, which builds the model from a noisy diffusion equation. They give a construction of the sine-Gordon measure below the first threshold value, in finite and infinite volume settings. They prove the wave-like version has global solutions that behave well for all time, below a second, lower threshold in finite volume.
Catalog record
| CREATOR | Massimiliano Gubinelli , Martin Hairer , Tadahiro Oh , and Younes Zine |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 77 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-16404
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A simple construction of the sine-Gordon model via stochastic quantization Massimiliano Gubinelli , Martin Hairer , Tadahiro Oh , and Younes Zine Abstract . We present a simple PDE construction of the sine-Gordon measure below the first threshold ( β 2 < 4 π \beta^{2}<4\pi ), in both the finite and infinite volume settings, by studying the corresponding parabolic sine-Gordon model. We also establish pathwise global well-posedness of the hyperbolic sine-Gordon model in finite volume for β 2 < 2 π \beta^{2}<2\pi . 2020 Mathematics Subject Classification 81T08, 60H15, 35K05, 35L71 1. Introduction 1.1. Sine-Gordon model The Euclidean sine-Gordon field theory (with unit mass) on the periodic box \mathbb{T}_{L}^{2}=(\mathbb{R}/2\pi L\mathbb{Z})^{2} , L\geq 1 , is defined by the sine-Gordon measure on \mathcal{D}^{\prime}(\mathbb{T}_{L}^{2}) with a formal density: (1.1) \displaystyle\begin{split}d\rho_{L}^{\beta}(u)&=``Z_{L}^{-1}\exp\bigg(\frac{\gamma}{\beta}\int_{\mathbb{T}_{L}^{2}}\cos\big(\beta u(x)\big)dx\\ &\hphantom{XXXXXX}-\frac{1}{2}\int_{\mathbb{T}_{L}^{2}}u(x)^{2}dx-\frac{1}{2}\int_{\mathbb{T}_{L}^{2}}|\nabla u(x)|^{2}dx\bigg)du",\end{split} where \beta,\gamma\in\mathbb{R}\setminus\{0\} and Z_{L}=Z(\beta,\gamma,L)>0 is a normalization constant. In the current finite volume setting, the measure ( 1.1 ) has been rigorously constructed in the mathematical physics literature for the full subcritical range 0<\beta^{2}<8\pi [ 16 , 17 , 18 ] ; see also [ 19 , 20 , 6 ] . Our main goal in this note is to present a simple construction of the sine-Gordon measure \rho^{\beta} via a stochastic quantization approach in the restricted range 0<\beta^{2}<4\pi . A naive attempt to define the sine-Gordon measure \rho^{\beta}_{L} is to view it as a weighted Gaussian measure: \displaystyle d\rho^{\beta}_{L}(u)=``Z_{L}^{-1}\exp\bigg(\frac{\gamma}{\beta}\int_{\mathbb{T}^{2}_{L}}\cos\big(\beta u(x)\big)dx\bigg)d\mu_{L}(u)", where \mu_{L} denotes the massive Gaussian free field on \mathbb{T}^{2}_{L} . Namely, \mu_{L} is a Gaussian measure on L -periodic distributions with the covariance operator (1-\Delta)^{-1} . Unfortunately, \mu_{L} on \mathbb{T}^{2}_{L} is supported on distributions of negative regularity [ 22 ] and thus we cannot make sense of \cos(\beta u) for general u\in\supp(\mu_{L}) . Let us make this last observation more precise. Fix L\geq 1 . Given N\in\mathbb{N} , let \mathbf{\Pi}_{N}=\mathbf{\Pi}_{N}^{L} be a smooth frequency projector onto the frequencies \{n\in\mathbb{Z}_{L}^{2}:|n|\lesssim N\} defined by (1.2) \displaystyle\widehat{\mathbf{\Pi}_{N}f}(n)=\chi_{N}(n)\widehat{f}(n),\quad n\in\mathbb{Z}_{L}^{2}:=(\mathbb{Z}/L)^{2}, where \chi_{N}(n)=\chi(N^{-1}n) for a fixed function \chi\in C^{\infty}(\mathbb{R}^{2};[0,1]) such that \widecheck{\chi}\equiv 1 on \{x\in\mathbb{R}^{2}:|x|\leq\tfrac{1}{2}\} and \supp(\widecheck{\chi})\subset\{x\in\mathbb{R}^{2}:|x|\leq 1\} . Then, a variant of Proposition 2.5 below (see [ 31 , Theorem 3.5] ), we have, for any test function \phi\in C^{\infty}(\mathbb{T}^{2}_{L}) , \displaystyle\mathbb{E}_{\mu_{L}}[\langle\cos(\mathbf{\Pi}_{N}u),\phi\rangle^{2}]\lesssim N^{-\frac{\beta^{2}}{4\pi}}\longrightarrow 0, as N\to\infty , where \langle\cdot,\cdot\rangle denotes the usual duality pairing between distribution and test function. Thus, we expect that the frequency-truncated sine-Gordon measure: \displaystyle Z_{L,N}^{-1}\exp\bigg(\frac{\gamma_{L}}{\beta}\int_{\mathbb{T}^{2}_{L}}\cos\big(\beta\mathbf{\Pi}_{N}u(x)\big)dx\bigg)d\mu_{L}(u) converges weakly to the Gaussian measure \mu_{L} as N\to\infty . In order to circumvent this issue, we allow the coupling constant \gamma_{L} to depend on N\in\mathbb{N} and to diverge as N\to\infty . This procedure is referred to as renormalization in the literature on Euclidean quantum field theory and singular stochastic PDEs; see for instance [ 22 , 29 ] . For our purpose, define \gamma_{L,N}=\gamma_{L,N}(\beta) , N\in\mathbb{N} , by setting (1.3) \gamma_{L,N}=e^{\frac{\beta^{2}}{2}\sigma_{L,N}}, where \sigma_{L,N} is defined by 1 1 1 Here, we use a Riemann sum approximation and the decay of \chi in both physical and Fourier spaces. σ _L, N = E _ μ _L [( Π _Nu(x) ) ^2 ] = 1 4 π 2 ∑_n∈ Z _L^2 χ N 2 (n) ⟨ n ⟩ 2 1 L 2 = 1 4 π 2 ∫ _ R ^2 χ N (z)dz 1 + —z— 2 + O(1) = 1 2 π logN + O(1) as N\to\infty , independent of x\in\mathbb{T}_{L}^{2} , where \langle n\rangle=(1+|n|^{2})^{\frac{1}{2}} and the remainder O(1) is bounded uniformly in the parameter L\geq 1 ; see [ 30 , Lemma 3.2] for the L=1 case. From ( 1.3 ) and ( ), we have \gamma_{L,N}\sim N^{\frac{\beta^{2}}{4\pi}}\to\infty as N\to\infty , independently of L\geq 1 . Then, we define the renormalized truncated sine-Gordon measure \rho_{L,N}^{\beta} by (1.4) \displaystyle d\rho_{L,N}^{\beta}(u)=Z_{L,N}^{-1}\exp\bigg(\frac{\gamma_{L,N}}{\beta}\int_{\mathbb{T}_{L}^{2}}\cos\big(\beta\mathbf{\Pi}_{N}u(x)\big)dx\bigg)d\mu_{L}(u). Inheriting the regularity property of the Gaussian free field \mu_{L} , the truncated sine-Gordon measure \rho_{L,N}^{\beta} is supported on \mathcal{C}^{-\delta}(\mathbb{T}_{L}^{2}) for any \delta>0 . We now state our main results. Theorem 1.1 . Let 0<\beta^{2}<4\pi , L>0 , and \delta>0 . Then, there exists a subsequence \{\rho_{L,N_{j}}^{\beta}\}_{j\in\mathbb{N}} converging weakly to a limiting probability measure \rho^{\beta}_{L} on \mathcal{C}^{-\delta}(\mathbb{T}_{L}^{2}) as j\to\infty . As mentioned above, the construction of the sine-Gordon measure \rho_{L}^{\beta} has been completed for the full subcritical range 0<\beta^{2}<8\pi . See also [ 36 , Theorem 1.1] for the construction of the sine-Gordon measure \rho_{L}^{\beta} for 0<\beta^{2}<4\pi via a straightforward application of the variational approach [ 5 ] . Our main goal in this note is to present a simple proof of Theorem 1.1 by studying the associated stochastic quantization equation for the sine-Gordon model (namely, the parabolic sine-Gordon model ( 1.5 ) below) with techniques arising from the study of singular stochastic PDEs. Given L\geq 1 , let \rho^{\beta}_{L} be the sine-Gordon measure on \mathbb{T}^{2}_{L} constructed in Theorem 1.1 , which we view as a probability measure on \mathcal{D}^{\prime}(\mathbb{R}^{2}) , but supported on 2\pi L -periodic distributions. Then, by taking a large torus limit ( L\to\infty ) of the 2\pi L -periodic sine-Gordon measure \rho^{\beta}_{L} , our approach also allows us to construct the sine-Gordon measure on \mathbb{R}^{2} . Theorem 1.2 . Let 0<\beta^{2}<4\pi and \delta>0 . Then, there exists an increasing sequence \{L_{j}\}_{j\in\mathbb{N}} of positive numbers such that the sequence \{\rho_{L_{j}}^{\beta}\}_{j\in\mathbb{N}} of the 2\pi L_{j} -periodic sine-Gordon measures converges weakly to a limiting probability measure \rho^{\beta} on \mathcal{D}^{\prime}(\mathbb{R}^{2}) as j\to\infty . More precisely, the weak convergence of \{\rho_{L_{j}}^{\beta}\}_{j\in\mathbb{N}} takes places as measures on the weighted Hölder-Besov space \mathcal{C}^{-\delta}_{\lambda,M}(\mathbb{R}^{2}) defined in ( 4.4 ) for some \lambda,M\gg 1 . 1.2. Parabolic sine-Gordon model Let L\geq 1 and consider the following stochastic nonlinear heat equation (SNLH) on \mathbb{T}_{L}^{2} with a sine nonlinearity: (1.5) \displaystyle\partial_{t}u_{L}+(1-\Delta)u_{L}+\gamma\sin(\beta u_{L})=\sqrt{2}\xi_{L},\qquad(t,x)\in\mathbb{R}_{+}\times\mathbb{T}_{L}^{2}, where u_{L} is a real-valued unknown and \xi_{L} denotes a space-time white noise on \mathbb{R}_{+}\times\mathbb{T}_{L}^{2} . Namely, \xi_{L} is a distribution-valued Gaussian process with covariance given by (1.6) \displaystyle\mathbb{E}\big[\langle\xi_{L},\phi\rangle\langle\xi_{L},\psi\rangle\big]=\langle\phi,\psi\rangle_{L^{2}_{t,x}}, for any \phi,\psi\in\mathcal{D}(\mathbb{R}_{+}\times\mathbb{T}^{2}_{L}) . Here, \langle\cdot,\cdot\rangle denotes the \mathcal{D}(\mathbb{R}_{+}\times\mathbb{T}^{2}_{L})-\mathcal{D}^{\prime}(\mathbb{R}_{+}\times\mathbb{T}^{2}_{L}) duality pairing. The dynamical model ( 1.5 ) corresponds to the so-called stochastic quantization [ 35 ] of the quantum sine-Gordon model represented by the measure \rho_{L}^{\beta} in ( 1.1 ) in the sense that the sine-Gordon measure \rho_{L}^{\beta} is formally invariant under the dynamics generated by ( 1.5 ). For this reason, we refer to ( 1.5 ) as the parabolic sine-Gordon model in the following. In [ 31 , 13 ] , the second author with Chandra and Shen observed that the difficulty of the problem depends sensitively on the value of \beta^{2}>0 (see, for example, Proposition 2.5 ) and showed that there is an infinite number of thresholds: \beta^{2}=\frac{j}{j+1}8\pi , j\in\mathbb{N} , where one encounters new divergent stochastic objects, requiring further renormalizations. By adapting the theory of regularity structures [ 29 ] to the non-polynomial setting, they proved local well-posedness of ( 1.5 ) (for sufficiently smooth initial data) in the full subcritical range 0<\beta^{2}<8\pi . Let us describe how to study the dynamical problem ( 1.5 ), at least below the first threshold 0<\beta^{2}<4\pi . Due to the roughness of the space-time white noise \xi , a solution u(t) for any fixed t>0 is merely a distribution and thus we cannot make sense of the nonlinearity \sin(\beta u) in ( 1.5 ). We overcome this issue by transferring the renormalization at the level of the measure in ( 1.4 ) to a renormalization