> <\body> |Barashkov>|<\author-affiliation> Department of Mathematics and Statistics University of Helsinki |>>|Gubinelli>|<\author-affiliation> Hausdorff Center for Mathematics & Institute for Applied Mathematics University of Bonn |>>|<\doc-note> > > without cutoffs and a corresponding large deviation principle for any infinite volume limit. Any infinite volume measure is described via a novel kind of stochastic equations. Similar considerations apply to more general |)>> theories. We then turn to the \|)>> model for \8\> (the so called full > regime) and prove uniqueness of the infinite volume limit and a variational characterization of the unique infinite volume measure. The corresponding characterization for |)>> theories is lacking due to the difficulty of studying the stability of the Euler\ULagrange equations against local perturbations.>||> The aim of this paper is to develop further the variational approach to Euclidean quantum fields initiated in> by studying the infinite volume limit of the variational description and some of its consequences. For simplicity we will focus on the two dimensional setting, also because we will highlight some unsolved issues which would have to been worked out in this simpler setting before moving to more challenging models. We believe that also the infinite volume limit in the > theory considered in> can be attacked along the lines suggested here.\ The key tool is a novel kind of stochastic equations for Euclidean fields which involve an additional one dimensional parameter taking values in a subset of >. This parameter already appeared in> linked to a continuous scale decomposition of the random fields. Here the additional parameter can be chosen to take values in > with no particular association to the scale of fluctuations thanks to the milder irregularity of the two dimensional Gaussian free field and to the simpler renormalization (i.e. Wick ordering) needed. In a general perspective one can interpret the variational method and the associated stochastic equations as a way to a given field theory, alternative to the classical approach of Parisi\UWu via a Langevin equation and to the elliptic stochastic quantization recently introduced in via the supersymmetric dimensional reduction of Parisi\USourlas. The variational method we introduced in> has been recently used in various contexts to provide novel results: Bringmann used it to construct singular Gibbs measures and study their invariance under dispersive dynamics, Chandra, Gunaratnam and Weber used it to show the existence of phase transition in the > model, Oh, Okamoto and Tolomeo to prove existence and non-existence of certain Gibbs measures. Moreover in we provide another approach to describe the most singular part of the > measure by approximating the optimal drift of the variational description. The PhD thesis of the first author describes also how to study the infinite volume limit of the Sine\UGordon model via the variational method including a full verification of the Osterwalder\USchrader axioms.\ Similar ideas relying on a coupling between the Gaussian free field and an interacting field have been used by Bauerschmidt and Hofstetter to control the maximum of the Sine\UGordon model. More generally, mapping a Brownian motion into a measure of interest has revealed itself an useful method also in the context of functional inequalities, see e.g. the work of Borell, Lehec and the recent paper of Mikulincer and Shenfeld on the \PBrownian transport map\Q. \; Let us now give an informal and heuristic picture of the results of this paper. The basic object of interest is a two dimensional Euclidean quantum field, namely a probability measure > on the Schwarz space |)>> of tempered distributions given by the expression <\equation*> \|)>=|)>>|Z>\|)>, where > is the Gaussian free field (GFF) on > with covariance -\|)>> and is a local interaction term given by <\equation*> V|)>=>\p|)>\x, with :\\\> a compactly supported function to make the spatial integral finite.\ We interpret the GFF > as the terminal value of a Brownian martingale |)>>> with values in |)>> and covariance given by <\equation*> \W|]>=s|)>-\|)>,x,y\\,t,s\. If we denote by > the law on the path space ,\|)>|)>> of this Brownian motion. The measure > can be then interpreted as the law of > under the Gibbs measure <\equation*> \>W|)>=e|)>>\W|)>. It is well known that the measure >> is the minimizer of a Gibbs variational problem and that <\equation*> -log Z=-log \|)>>|]>=inf> |\>|)>|]>+H\|\|)>|>, where >> denotes the expectation with respect to a probability measure > on the path space ,\|)>|)>> and \|\|)>> the relative entropy of > with respect to >. A formula of Boué\UDupuis allows to rewrite the variational problem as an optimal control problem, namely <\equation> -log Z=-log \|)>>|]>=inf \|V+Z|)>+\|>, where the infimum is now taken over all processes |)>>> with values in |)>> and adapted to the filtration of and <\equation*> \=\=>>-\|)>>\s is a bilinear form giving the \Penergy\Q of the . Here >> denotes the derivative of > with respect to the parameter >. As we show in this paper, the meaning of an optimal drift is that of providing the law of the \Pinteracting field\Q via the equality <\equation> Law>>=Law> valid for any minimizer of the problem and in particular <\equation*> \=Law>+Z|)>. The existence of such minimizers will be guaranteed, in the rigorous discussion below, by suitably relaxing the variational problem into a problem on the space of probability measures. But let for the moment continue our heuristic discussion ignoring these fine points. \; The variational description loose meaning in infinite volume, i.e. when \1>, since the cost functional is always infinite, or at best not well-defined. However the main idea in this paper is to note that any minimizer of satisfy the Euler\ULagrange (EL) equation (first variation) <\equation> \|\V+Z|)>\K+2 \|>=0 for all \Ptest controls\Q , where V+Z|)>\K> denotes the derivative of \V|)>> at +Z> and in the direction of >. Provided the test drift decay sufficiently rapidly, this equation makes sense also in infinite volume and can be used, very much like the weak formulation of a usual elliptic problem, to get estimates on the minimizer , e.g. by testing with Z> where :\\\> is some weight function decaying at spatial infinity. In particular we will obtain estimates of the form <\equation*> |\Z|\<\|\|\>>H>\\, uniformly in the cutoff >. This estimate together with the identification is at the core of the proof of tightness of the family of measures >|)>>> with \1>. Any accumulation point > represents an infinite volume Euclidean field, moreover the corresponding drift gives, via the formula an explicit coupling of > and the GFF given by >. The point of view which we advocate in this paper is that the EL equation should provide a complete control over the drift describing the infinite volume Euclidean field, for which a naive variational description is meaningless. Moreover intuitively we should be able to introduce a variational description of the measure > with drift via a renormalized BD formula of the form <\equation> -log \|)>>|]>=inf \|f+K|)>+V+K|)>-V|)>+\-\|> with =W+Z> being the \Pinfinite volume interacting field\Q with potential and |)>> a test function which depends only locally on the field > and where the infimum is taken among all drifts which satisfy suitable decay at infinity. Given this decay one expects that the quantities +K|)>-V|)>> and -\> are finite even if each summand is infinite. \; Leaving this heuristic account of the ideas investigated in this paper, we describe now the plan for the rigorous proofs. We will start in Section by introducing the general setting for the variational description of two dimensional Euclidean quantum fields as outlined above. We prove certain useful bounds and also that the minimizers of this variational problem describe in a precise sense the Euclidean field and moreover that they satisfy \ Euler\ULagrange (EL) equations in weak form.\ In Section, as a first example, we address the problem of the infinite volume limit of the > measure >> on a bounded domain \\>. From the EL equations we will derive a-priori bounds for minimizers strong enough to give tightness of the family >|)>>> as \\>. Unfortunately, establishing rigorously the form of the limiting variational problem requires certain locality properties of the solutions to the EL equation in infinite volume. This is usually possible when is convex. Ultraviolet renormalization in > however spoils any \Peasy\Q convexity of the potential and it is not clear how to gain sufficient control of in this case. Similar issues appear in other methods of stochastic quantization as soon as the infinite volume limit is considered for |)>> Euclidean quantum fields. Despite these difficulties, in Section we show that the EL equation is still a useful tool which can be used effectively to prove the large deviations of the > measure in the semiclassical limits. Let us remark that Lacoin Rhodes and Vargas obtained Large Deviations results in the case of the Liouville model Finally, in Section, we discuss the case of the exponential interaction, that is the \|)>> model or Høegh\UKrohn model introduced in. See e.g. for a discussion of its stochastic quantization in the elliptic setting. For this model we repeat the construction of the variational description of the infinite volume measure. At variance with what happens in |)>> theories, the renormalization of this model does not spoil the convexity properties of the potential and therefore the study of the related EL equations can go further and in particular we will be able to prove the necessary locality of the perturbed EL equations. This in turn will allow us to prove the >-convergence of the variational problem along with the infinite volume limit and establish eq. in its rigorous form. As a consequence, we are able to obtain a concise and well-behaved description of an infinite volume Gibbs measure related to an Euclidean quantum field. \; MG would like to thank Ronan Herry for pointing out the reference on the Brownian transport map. MG is supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) through the Hausdorff Center for Mathematics under Germany's Excellence Strategy \U EXC-2047/1 \U 390685813 and through CRC 1060 - project number 211504053. NB is supported by the ERC Advanced Grant 741487 \PQuantum Fields and Probability\Q. \; Let |a|\>\|)>>. Let |)>> the Schwarz space of test functions on > and |)>> the corresponding space of tempered distributions. For any \\> we define fractional derivatives |\|\>>=|)>/2>> on |)>> by\ <\equation*> |\|\>>f=\|\|\>/2>f|)>|)>|)>,f\\|)> where > is the Fourier transform on >. All along the paper we need to consider Banach subspaces of |)>> associated with a positive spatial weight >. Let :\\\0>> \ a smooth positive function such that for some sufficiently small \0> we have <\equation> *\|\|>+\|\|>\\*\,\\|\|\>. The number > can depend on various \ parameters but not on the volume cutoffs we introduce to regularize the Euclidean QFTs. For some \0>, \>, |]>>, we introduce the usual weighted Sobolev spaces >|)>> and weighted Besov spaces >|)>> with norms <\equation*> |f|\<\|\|\>>>|)>>=|\>|D|\>f|\<\|\|\>>|)>>,|f|\<\|\|\>>>|)>>=|\>f|\<\|\|\>>>, and let >|)>=W>|)>> and >|)>=B,\>>|)>>. Here > is the unweighted Besov norm given by\ <\equation*> |f|\<\|\|\>>>=-1>2|\f|\<\|\|\>>|)>>, where the Littlewood\UPaley operators |)>-1>> are defined by =\/2|)>> for 0> and some function :\\\0>> supported in an annulus centered at the origin and \ =|~>|)>> for |~>:\\\0>> supported on a ball centered at the origin and such that -1>\=1>. In general we avoid to indicate the weight when dealing with the unweighted (=0>) version of the above spaces. We denote by > the set of Borel measures on the topological space , for \> we denote > the corresponding expectation and for any random variable Y> on the topological space we let \\> the corresponding law. Consider the Gibbs measure\ <\equation> \\|)>\|)>>*\\|)>|e|)>>*>\|)>>*,\\\|)>, where > is the law of the Gaussian field > (GFF) on > with covariance -\|)>> for some mass 0> (fixed throughout the paper) and where > denotes the Laplacian on >. The function |)>\\> represent an interaction potential and for now we will only assume that it is measurable and with respect to > according to the following definition. <\definition> Let ,\|)>> be a probability space. A measurable function ,\|]>> is called tame with respect to > if there exists 1> with +=1> and\ <\equation*> \>|]>+\>|]>\\. Tameness has been introduced by Üstunel as a convenient condition for the validity of the Boué\UDupuis variational formula for Brownian functionals. It has been recently improved by Hariya and Watanabe to allow for . To introduce this formula we need to enlarge our probability space and realize the GFF as a marginal of a Brownian motion, following the approach initiated in>. Here we are able to use a simpler embedding thanks to the milder nature of the Gaussian free field in two dimensions compared to its three dimensional counterpart.\ \; Let > the law of a cylindrical Brownian motion |)>>> on |)>>. It can be realized on ,C>>|)>|)>> where \0> can be taken arbitrarily small. Let\ <\equation*> W\-\|)>*X,t\, and note that indeed =Law>|)>>. We will consider as a random variable in \C,C>>|)>|)>>.\ Let >|)>=HH*>|)>>, =|)>>. Denote with >|)>> the space of all processes \\\> progressively measurable with respect to the filtration generated by |)>>>, starting from and equipped with the norm of ,>|)>|)>>. Moreover define :\L|)>> as <\equation*> \\*|-\|)>*>|\<\|\|\>>>\s=*|\>|\<\|\|\>>>+m|>|\<\|\|\>>>|)>\s where >> denotes the time-derivative of . With these definitions, the Boué\UDupuis formula reads as follows. <\lemma> For any tame functional \\> we have <\equation*> -log \>|)>>|]>=inf> \>+Z>|)>+\|]>. <\proof> Theorem7 from Üstunel specialized in our context (see also>) gives <\equation*> -log \>|)>>|]>=inf\> \>+Z>|)>+|u|\<\|\|\>>>\s|]>, where > is the space of all progressively measurable processes starting from equipped with the norm of ,L L|)>> and <\equation*> Z\-\|)>*u\s. Observing that =-\|)>*>> we obtain the claim. Note that the functional >+Z>|)>> makes sense for any > since by Theorem2 from Üstunel we have that the push-forward measure =>+Z>|)>>\> has finite relative entropy with respect to > and therefore is absolutely continuous, i.e. \\>. \; We call a random variable with values in > a (or ) and we will understand that > and > whenever this does not lead to confusion. \; To simplify certain technical considerations later is convenient to augment the process with some additional informations. Consider a Polish space > endowed with a continuous projection :\> and let :\> a measurable map such that \=W>, i.e. |W>> is a fibered topological space and > a measurable section. We will understand > as an . The kind of enhancements needed will be different in different models. We will always consider the space > endowed with the measure given by the law of > under >. A functional on > is tame if it is tame for this measure. \; For any measurable |)>\\> (non necessarily linear) define > as\ <\equation*> e>\e|)>>*\\|)>. As a direct consequence of Lemma, the \ functional > has the variational representation <\equation*> \=inf> F-inf> F. for any such that is tame and where for a tame functional on > we let <\equation*> F\\+\|]>,Z\. To proceed we want to find a minimizer of >. It is however not clear if we can find one in >. Following the approach we used in>, we relax our problem and instead minimize over the law of the pair > with fixed and varying over >. In order to ensure compactness we need to complete the space of trial laws for a suitable topology for which our target functional is coercive. \; \ Let >|)>> to be >|)>> equipped with the weak topology and let |^>\|)>>, |^>\|)>> and |^>\|)>>. With a slight abuse of notation we denote ,Z|)>> the canonical coordinates on \|^>>. <\definition> Let <\equation*> \|\=Law>,Z|)>\\|^>|)>\Z\|^>,\>|Z|\<\|\|\>>|^>>|]>\\|>, and\ <\equation*> |\>\\\|^>|)>\\\\,\\\ ,sup\> \>|Z|\<\|\|\>>|^>>|]>\\|}>. We equip |\>> with the following notion of convergence: a sequence |)>> in |\>> converges to > if <\enumerate-alpha> \\> weakly on \|^>>, \> \>|Z|\<\|\|\>>|^>>|]>\\>. We will need to realize couplings of two elements of |\>>. These are constructed by the next lemma which needs the following preliminary definition of the space hosting the coupling. Denote the canonical coordinates on \|^>\|^>> by ,Z,K|)>>. <\definition> \ For \|\>> let |\>|)>\\|^>\|^>|)>> the set of all > such that >,Z|)>=\> and for which there exists sequences |)>,|)>\>> with <\equation*> Law>,Z,K|)>\\ sup\>\|Z|\<\|\|\>>|^>>+|K|\<\|\|\>>|^>>|]>\\. <\lemma> For any ,\\|\>> there exists \|\>|)>> such that\ <\equation*> Law>,Z+K|)>=Law>,Z|)>. <\proof> Since ,\> are in |\>> we can find sequences ,>\|^>> such that :=Law>,Z|)>\\> and \Law>,>|)>\\> weakly in |^>> and furthermore \|Z|\<\|\|\>>|^>>+|>|\<\|\|\>>|^>>|]>\\>. Now define\ <\equation*> K=>-Z, and set\ <\equation*> \=Law>,Z,K|)>. We have that \|K|\<\|\|\>>>>|]>\sup\|Z|\<\|\|\>>>>+|>|\<\|\|\>>>>|]>\\> so > is tight on \>\>>. Taking a convergent subsequence (not relabeled) we can assume that \\>. We then have <\eqnarray*> >,Z+K|)>>||\>Law>,Z+K|)>>>|||\>Law>,Z+K|)>>>|||\>Law>,>|)>=\>>>> which proves the claim.\ On functionals on |\>> we can define a first and second order derivatives as follows. Let \|\>> and let \>\>> a measurable process adapted to the filtration generated by ,Z|)>>>, continuous in the second component and such that L,>|)>>. We call such an admissible control for >. For all \\> let K>> the law of ,Z+\K|)>> under >. We have K>\|\>>. A functional >:|\>\\> is differentiable in |\>> in the direction of the admissible control if the map \>K>|)>> is differentiable in =0> and in this case we define <\equation*> |\>|)>=|\\>|\|>=0>>K>|)>. Second order derivatives can be similarly defined by taking two admissible controls > and letting <\equation*> \>|)>|)>=||\\>|\|>=0>||\\>|\|>=0>>K+\K>|)>, whenever ,\|)>\>K+\K>|)>> is > in ,\|)>=0>. \ We are ready to relax the variational problem and prove an alternative representation of >>. For a tame \\> define the functional >:|\>\\> by <\equation*> >|)>\\>>|)>+\|]>,\\|\>. <\definition> We say that a functional >:|\>\\> is if it is lower semicontinuous and such that <\equation> >|)>\1+\>|]>, this ensures in particular that the functional >:|\>\\> is , that is for all \> the level sets >\L|}>> are compact in |\>>. \; For a tame \\> define the functional >:|\>\\> by <\equation*> >|)>\\>>|)>+\|]>,\\|\>. We say that is admissible if is tame and >> is admissible. The following lemma is an obvious consequence of this relaxed variational setting. <\lemma> Assume that \\> is admissible, then <\equation> inf> F=inf|^>> F=inf\> >|)>=min\|\>> >|)>. Let \|\>> any minimizer of >>. If >> is differentiable at > it satisfies the EL equation >|)>=0> for all admissible controls . Moreover if it is twice differentiable at > we also have the positivity condition >|)>|)>\0> for all admissible controls >. <\proof> The equality > F=inf|^>> F> follows from the property of the admissible functional >>. The equality |^>> F=inf\> >|)>> follows directly from the definition of >, while the equality \> >|)>=min\|\>> >|)>> again from the admissibility of which ensures that sequences |)>\\> of approximate minimizers of >> converge in |\>> and that any limit > is a minimizer. The first and second order conditions are then consequences of the differentiability of >>. <\remark> Note that <\equation*> >K>|)>->|)>=\>K>|)>-G+\K|)>-\|]> <\equation*> =\>K>|)>-G+2\\+\\|]>. If is differentiable in the direction of > we have <\equation*> \>|)>=\>G>|)>\K+2\|]>. If is twice differentiable we have also <\equation*> \>|)>|)>=\>G>|)>\|)>+2\|)>|]>. The consequences of the existence of this minimizer are clarified by the following lemmas. <\lemma> If is admissible and > a minimizer of >> then we have <\equation> \>|]>\\\\>|]>. <\proof> If is admissible we have proven that <\equation> \=inf\|\>> >|)>-inf\|\>> >|)>=>|)>->|)>. Eq. implies in turn <\equation*> >|)>->|)>\\\>|)>->|)> and since <\equation*> >|)>->|)>=\>+V+\|]> <\equation*> -\>+\|]>=\>|]> for any , we obtain eq.. <\lemma> Let \\> be admissible. Then for all \\> continuous and bounded is admissible. Let > a minimizer for >>, then for all \\> we have <\equation> |\\>\f|)>=\f+V>>|]>. <\proof> The admissibility of is easy to obtain given that is bounded and continuous and is admissible. It is clear that \\f|)>> is differentiable for bounded , we need only to establish the representation of this derivative via the variational minimizer f>>. For any \0> we have, by direct computation <\equation*> \+\|)>f|)>-\f|)>\>+\|)>f+V>f,\>|)>->f+V>f+V>|)>=\\f+V>>|]>, and <\equation*> \f|)>-\-\|)>f|)>\>f+V>f+V>|)>->-\|)>f+V>f+V>|)>=\\f+V>>|]>. By differentiability of f|)>> we have <\equation*> \f+V>>|]>\lim\0+>+\|)>f|)>-\f|)>|\>=|\\>\f|)> and <\equation*> |\\>\f|)>=lim\0+>f|)>-\-\|)>f|)>|\>\\f+V>>|]> from which we can conclude. There are several interesting consequences of this formula. First of all, integrating in >, we have <\equation> \=|\\>\f|)>\\=\f+V>>|]>\\, which is a representation of > alternative to the variational formula. On the other hand, taking =0> in we obtain <\equation*> f|)>\\|)>=||\\>|\|>=0>\f|)>=\>+Z|)>|]>. <\theorem> The law of +Z> under > is indeed >. In particular for all \\> continuous and bounded <\equation*> \=-loge|)>>\\|)>=-log \>+Z|)>>|]>. This last formula naturally calls for a variational description of the average >+Z|)>>|]>> with respect to the base measure >. This extension of the BD formula to the interacting field is the subject of the next lemma. \; Let > the symmetric bilinear form associated to >, in such a way that <\equation*> \=\+\+2\. <\lemma> Let be admissible and denote by \|\>> a minimizer of >>. Then we have the following direct variational description of >:\ <\equation*> \=-log \>>|]>=inf\|\>|)>> >|)> with <\equation*> >|)>\\>|]>+>|)>, and <\equation*> >|)>\\>-V+2\+\|]>. Moreover for any \|\>> we have >|)>\0> for all \|\>|)>> iff > is a minimizer of >>. In this case and if is differentiable in the directions of >, >|)>> takes the alternative form <\equation*> >|)>=\>-V-\V\K+\|]>. <\proof> \ Recall that\ <\eqnarray*> >||\|\>>>-inf\|\>>>>>|||\|\>>\>+V+\|]>-\>+\|]>>>|||\|\>*|)>>\>+V+\|]>>>|||>+\|]>>>>> where in the last line we used Lemma to justify the change of variables in the minimization problem. Note that for any \|\>*|)>> we have that <\equation*> \>+\|]>=\>+\|]> so <\equation*> \=inf\*|)>>|\>|]>+>|)>|>. The fact that >>|)>\0> iff >> is a minimizer of >> follows from its definition, indeed note that <\equation*> >|)>=>>|)>->>|)>. If is differentiable in the direction of > we have that for any minimizer > of >> and any \|\>|)>>: <\equation*> 0=\>|)>=\>V\K+2\|]> so we can simplify the expression of >>. <\remark> The interest of this formulation is that >> has a well defined meaning also in the infinite volume provided the drift has spatial decay. The variational description of > for localized is then well-posed if we can restrict it to sufficiently decaying test measures > sufficiently localized in the component. We will demonstrate this in Section in the case of the exponential model. In this section we will consider the above general setting in the case where the interaction potential is given by <\equation*> V>|)>\\>|\|\>, where \\> is a measurable bounded set (of sufficient regularity, see below) and |\|\>> denotes the -th Wick power of the Gaussian free field > and \0> an arbitrary fixed constant. Let >\\>>> and >\\>>>. The measure >> on |)>> is the > Euclidean quantum field with a finite volume cutoff.\ \; We will prove that for a sufficiently large class of functions the family >|)>>> is uniformly bounded as \\>. This implies that the family of measures >|)>>> is tight on |)>>. We will describe any accumulation point via a formula for its Laplace transform involving the Euler\ULagrange equations obtained via the variational representation introduced in the previous section. <\remark> Let us note that all the discussion in this section can be extended to general |)>> models with a positive coefficient in the highest order polynomial. Moreover also > invariant vector models could be handled in similar ways. Convexity of the classical potential (i.e. before Wick ordering) is only needed in Section below to prove large deviations. We will need the following enhancement > of the GFF . Let <\equation*> \\|W|\>,\\|W|\>, where here |\|\>> denotes the Wick powers with respect to the Gaussian field . Then <\equation*> \\,\|)>\L,|)> for all 1>, where \C,,\>>|)>|)>|)>> for a fixed \0> which can be chosen as small as we like. By abuse of notation we will denote also =,\|)>> the canonical process on >. \; \ In the following we will always consider a family of |)>|\|>\\>> given by =|\|> S> for some fixed set of volume with >-boundary. This allow to prove that >> is sufficiently regular to make sense of quantities like >|\|\>>, >|W|\>> and in general of products with Wick monomials of the GFF. <\lemma> For any 1>, > and \> we have that >\B>|)>>, furthermore\ <\equation*> sup|\|>>|>|\<\|\|\>>>|)>>\\. <\proof> We have that \ >=\\\> where \> is the surface measure on > and > is the outer unit normal. This implies that for any 1>, |\|>>|>|\<\|\|\>>>|)>>\\>. Since clearly \ |\|>>|>|\<\|\|\>>>>|)>>\\>, we obtain the result by interpolation. <\lemma> The functional >>\>>>> is admissible and for all \|\>> <\equation*> >>|)>=>>>|)>=\>*>,Z>|)>|]>, with\ <\equation> \>,Z>|)>\4\>\Z+6\>\|)>+4\>W|)>+\>|)>. \; <\proof> It is known that\ <\equation*> \ \>|W|\>>|]>\\, for any 0> (see e.g or>) and\ <\equation*> \>|\|>|]>\\||W|\>|\<\|\|\>>>,\>>|)>>|\|>|]>\\, for some \0>, which shows that is a tame functional. Now it remains to prove the equality in. For this it is enough to show that for such that |Z|\<\|\|\>>|)>>\\> >-almost surely, we have <\eqnarray*> >|+Z|)>|\>>||>|W|\>+4>|W|\>Z+6>|W|\>|)>+4>W|)>+>|)>.>>>> To see this, recall that if >\\>\\>, is a standard mollifier, >=\>\W> and >=6\>|)>|]>>, >=3\>|)>|]>> <\eqnarray*> >|+Z|)>|\>>||\0>>>\+Z|)>|)>-a>>>\+Z|)>|)>-b>>>|||\0>>>|)>-a>>|)>-b>+4>>|)>->|2>W>|)>Z>>>|||>>|)>->|6>|)>>|)>+4>W>>|)>+4>>|)>>>>> where we have denoted >=\>\Z>. Now it is well known that almost surely in >>|)>> we have >\W> and\ <\equation*> >|)>-a>>|)>\|W|\>,>|)>->|2>W>|)>\|W|\>,>|)>->|6>W>|)>\|W|\>. Furthermore >\Z> in > which implies >|)>\Z,>|)>\>|)>> in ,1>> for some \0>, which gives the statement.\ To check the lower semicontinuity of >>>> recall from the Section3 of that <\equation*> \>,Z>|)>+\\-Q|)> with |)>=C|\|\<\|\|\>>>+1|)>> for some 1>. Furthermore observe that the function <\equation*> ,Z|)>\\>,Z>|)>+\ is lower semicontinuous on \>. Since the first marginal in |\>> is fixed, the Portmanteau theorem gives, for any \\> in |\>>: <\eqnarray*> \> >>>|)>>||\> \>>,Z>|)>+\+Q|)>|]>-\>|)>|]>>>||>|>>,Z>|)>+\+Q|)>|]>-\>|)>|]>>>|||>>>|)>.>>>> Finally, by a slight adaptation of the proof of Lemma below we see that there exists \0> and ,|\|>>\\> such that <\equation*> >>>|)>\\*\>+\|\>Z|\<\|\|\>>>|]>-C,|\|>>, for all \|\>>, from which the strong coercivity condition follows. The admissibility of >>> implies the variational representation for >> and the existence of minimizers >> of >>\>>>> with the corresponding EL equation <\lemma> For any bounded and continuous \\> there exists a minimizer >\|\>> of >>\>>>>. Moreover if is differentiable in the directions of > any minimizer > satisfies the Euler\ULagrange (EL) equations <\equation> |>>Z*K+2\|]>>>||=-\>f\K+\>*+12*\*Z+12*W Z|)>*K|]>>>>>> for any admissible control . <\proof> As we already remarked the existence of >> follows from the admissibility of >>>. The functional \>,Z>|)>> is differentiable in > and <\equation*> \\>,Z>|)>\K=4\>\K+12\>\ZK+12\>W|)>K+4\>|)>K. Provided f> is also differentiable, we have <\equation*> 0=\>>|)>=\>f\K+\\>,Z>|)>\K+2\|]> which is. <\remark> Note that \>,Z>|)>> is twice differentiable, with <\equation*> \\>,Z>|)>|)>=12*\*>\KK+24*\*>WZKK+4*\*>|)>KK. Therefore if is twice differentiable then we have also the second order condition: <\equation*> 0\\>>|)>|)>=\>f|)>+\\>,Z>|)>|)>+2\,K|)>|]> for any minimizer > and any pair > of admissible controls. We record also the normalized variational representation for >>. <\lemma> Denote by >\|\>> a minimizer of >>>. Then we have\ <\equation*> \>=inf\|\>>|)>> >>f|]>+>>|)>|}> with <\equation*> >>|)>\\>>>-\|)>*>\s+>>-\|)>*>\s|]> <\equation*> +\>>||\>*K+6\>||\>*K+4\>||\>*K+\>K|]>. <\proof> This follows from Lemma by direct computation. We now show that the EL equation gives us an a-priori bound uniform in |\|>> on the minimizer in terms of weighted norms of >. <\theorem> There exists a constant independent of |\|>> such that, for any minimizer >\|\>> of >>|)>> we have \; <\equation> \>>>\*Z+\Z|)>|]>\C. <\proof> Note that the map \*Z> is continuous from > to >. Take *Z> in to get that any minimizer >\|\>> satisfies <\equation> |>>>\*Z+>-\|)>*\*>|)>\s|]>>>||>>>>-\|)>,\|]>*\*>\s|]>>>||>>f*|)>+\>\*Z+12\*Z+12*W Z|)>|]>.>>>>> From Lemma and Lemma below we deduce that the r.h.s. is controlled by a small fraction of the l.h.s. and obtain the claim. <\lemma> > <\equation*> >>-\|)>,\|]>*\*>\s|\|>\\|>|\<\|\|\>>|)>>\s|]> <\proof> We observe that > obviously commutes with >, and <\equation*> ,\|]>=*\|)>+\\\\. Using the assumptions on the weight > we deduce <\equation*> |,\|]>*f|\<\|\|\>>>\\|\*f|\<\|\|\>>>+|\\f|\<\|\|\>>>|)>. <\lemma> Assume that, for \0> small enough, either <\enumerate-alpha> f|)>\\> and <\equation*> |\f|\<\|\|\>>|)>>\\|>|\<\|\|\>>,4>|)>>+C, <\equation*> |\f|\<\|\|\>>|)>>\\|W+Z|\<\|\|\>>>|)>>. Then there exists |)>> such that \ <\eqnarray*> ||f\|)>+\>*\*Z+3*\*\*Z+12*W \*Z|)>|\|>>>||>|*\>\*Z+\Z|)>|)>+C.>>>> \; <\proof> First observe that we have that >\*Z=|\>\*Z|\<\|\|\>>>>, and Z|)>\C|\*Z|\<\|\|\>>>>, for instance by fundamental theorem of calculus. By our assumptions on we have, in the first case\ <\eqnarray*> f\|)>|\|>>|>||f|\<\|\|\>>|)>>|>Z|\<\|\|\>>|)>>>>||>||>|\<\|\|\>>,4>|)>>|>Z|\<\|\|\>>|)>>>>||>||>Z|\<\|\|\>>|)>>+|>W|\<\|\|\>>,4>|)>>|>Z|\<\|\|\>>|)>>|)>>>||>||>Z|\<\|\|\>>|)>>+C|>W|\<\|\|\>>,4>|)>>.>>>> While in the second case\ <\equation*> f\|)>|\|>>|>||\f|\<\|\|\>>|)>>|Z|\<\|\|\>>|)>>>>||>||W+Z|\<\|\|\>>>|)>>|Z|\<\|\|\>>|)>>>>||>||W|\<\|\|\>>>|)>>|Z|\<\|\|\>>|)>>+\|Z|\<\|\|\>>|)>>>>||>||W|\<\|\|\>>>|)>>+2\|Z|\<\|\|\>>|)>>.>>>>> For the second term, by Cauchy\USchwarz we obtain\ <\equation*> >\*\*Z|\|>\C>|>\|\<\|\|\>>|)>>+\|Z|\<\|\|\>>|)>>. For the third term, we use LemmaA.7 from to get <\eqnarray*> >\*\*Z|\|>>|>||\*\|\<\|\|\>>>>|\*>Z|\<\|\|\>>>>>>||>||\*\|\<\|\|\>>>>|\*>>|\<\|\|\>>>|\>|\<\|\|\>>>>>>||>||\*\|\<\|\|\>>>>|\>*>|\<\|\|\>>>|\>|\<\|\|\>>>>>>||>|>|\*\|\<\|\|\>>>>+\|\*>>|\<\|\|\>>>+|\>|\<\|\|\>>>>|)>.>>>> For the final term we proceed similarly. By Lemma we have |\|>>|>|\<\|\|\>>>|)>>\\> \ for any 1/p> and any \1/16>. Then with \1/8> and for any \1/16> we have\ <\eqnarray*> >\*W Z|\|>>|>||\>*W|\<\|\|\>>>>|\>>Z|\<\|\|\>>>>|\*>Z|\<\|\|\>>>>>>||>||\>*W|\<\|\|\>>>>|\*Z|\<\|\|\>>>|\*>Z|\<\|\|\>>>>>>||>||\>*W|\<\|\|\>>>>|\*Z|\<\|\|\>>>|\*>Z|\<\|\|\>>>>>||>|>|\>*W|\<\|\|\>>>>+\|\*Z|\<\|\|\>>>+|\>*Z|\<\|\|\>>>|)>.>>>> and we can choose =1/8-\\1/16> to get the claim. \ Let us now show that these estimates give tightness of >|)>>> as |\|>\\> on a suitable weighted Sobolev space.\ <\lemma> Consider >|)>>> as a family of measures on >|)>>. We have for some \0>\ <\equation*> sup>exp|>\|\<\|\|\>>,4>|)>>|)>\\>+exp|\|\<\|\|\>>>|)>>|)>\\>|]>\\ In particular >|)>>> is exponentially tight and any accumulation point > satisfies\ <\equation*> exp|\|\<\|\|\>>,4>|)>>|)>\\\\. <\proof> For |)>=\|\|\<\|\|\>>>|)>>> we have f|)>=\|D|\>>\> so\ <\equation*> |\f|\<\|\|\>>>>\\|W+Z|\<\|\|\>>>>, which implies by Lemma that\ <\equation*> sup> \>>|Z|\<\|\|\>>>>|]>\\, and\ <\equation*> logexp\\>\\>>|]>+\>>|]>\C\+\, a fact that implies exponential tightness of >|)>>>. For the second statement we proceed similarly. Indeed if > is an accumulation point of >|)>>> then\ <\equation*> exp|\|\<\|\|\>>,4>|)>>|)>\\>|>||~>>exp||~>>\|\<\|\|\>>,4>|)>>|)>\\>>||>||~>> limsup\\>exp||~>>\|\<\|\|\>>,4>|)>>|)>\\>>>||>||~>> sup\|~>>exp||~>>\|\<\|\|\>>,4>|)>>|)>\\>.>>>>> This gives the claim provided that for |)>=\||~>>\|\<\|\|\>>,4>|)>>> we have\ <\equation*> logexp\\>\C. Indeed we compute f|)>=\\>|\|\>>|\|\>>\|)>>, from which we can check that\ <\equation*> |\f|)>|\<\|\|\>>|)>>\C\|\>|\|\>>\|)>|\<\|\|\>>>\C\|\|\|\>>\|\<\|\|\>>>\C\|\|\<\|\|\>>/2,4>>, and f|)>|)>\\>. Applying Lemma we get that the minimizer >> satisfies\ <\equation*> sup|~>> sup\|~>> \>>||~>>Z|\<\|\|\>>|)>>|]>\sup> \>>|>Z|\<\|\|\>>|)>>|]>=:C\\. As a consequence, eq. gives <\equation*> logexp\\>\\>>|]>+\>>|]>\2C\+\, which implies the claim. We have actually a more refined result: let >> be a minimizer for >>>. By Theorem the family >|)>>> is tight in \|^>|)>>. <\lemma> Denote by \\\|^>|)>> any accumulation point of >|)>>> as |\|>\\>. Then the following equation holds: <\equation> \>f\K+4*\>||\>*K+2*\|]>=0. for any \|^>\|^>> continuous, adapted to the filtration generated by ,Z|)>>> and compactly supported. <\proof> Let the support of be contained in > for some 0>. Let \C>|)>> such that =1> on > so that K>. Then Lemma implies that <\equation*> \>>|\*Z|\<\|\|\>>>+|\*Z|\<\|\|\>>>|]>\C, where the constant depends on but not on > and we have <\eqnarray*> ||>f\K+4*\>||\>*K+2*\|]>>>|||>f\K|)>+4*\>||\>*\*K+2*\K|)>|]>>>>> Observe first that by interpolation <\equation*> ||f\K|)>+4*\*>||\>*\*K+2*\K|)>|\|>>>||>|||~>|\<\|\|\>>>>|\*K|\<\|\|\>>>+|\*>|\<\|\|\>>>\t|)>|\*>|\<\|\|\>>>\t|)>>>|||||~>\|\<\|\|\>>>|\*Z|\<\|\|\>>>|\*Z|\<\|\|\>>>|)>|\*K|\<\|\|\>>>>>||||\*Z|\<\|\|\>>>|\*K|\<\|\|\>>>>>>>> where |~>\C>|)>> and |~>=1> on the support of >. The r.h.s. of this last equation is integrable by assumptions on and bounds on . From Lemmas12 and14 of we know that there exist a family |)>1>> adapted to the filtration generated by |)>> with <\equation> |>|\K|\<\|\|\>>>-|\K|\<\|\|\>>>|]>>|>|>|>|\*>|)>|\<\|\|\>>>\t|)>-|\*>|)>|\<\|\|\>>>\t|]>>|>|>>>> as \>. After possibly passing to a subsequence, we can assume that K\\K> in ,L|)>|)>> where > denotes > with the weak topology, and in ,L\\|)>|)>>. This together with the bound implies that \ <\eqnarray*> ||\> \>f\K|)>+4*\>||\>*\*K+2*\K|)>|]>>>|||>f\K|)>+4*\>||\>*\*K+2*\K|)>|]>.>>>> At this point we have reduced the problem to showing that\ <\eqnarray*> ||\\> \>>f\K|)>+4*\>||\>*\*K+2*\K|)>|]>>>||>|>f\K|)>+4*\>||\>*\*K+2*\K|)>|]>>>>> since by assumption\ <\equation*> \>>f\K|)>+4*\>||\>*\*K+2*\K|)>|]>=0. This follows from the weak convergence >\\> and the bound <\equation*> \>>|\*Z|\<\|\|\>>>+|\*Z|\<\|\|\>>>|]>\C, since we can approximate by bounded admissible controls |)>> defined as \ =ZT>> with the stopping time <\equation*> T\1\inf:|\*Z|\<\|\|\>>>+|\*Z|\<\|\|\>>>\L|}>, in such a way that <\equation*> sup>|\*Z|\<\|\|\>>>+|\*Z|\<\|\|\>>>\L, almost surely. Note that \1> as \> and therefore indeed \Z> as \>. To summarize the result of this section: we obtained tightness of the finite volume > measures and described any possible infinite volume limit via a novel kind of stochastic equation of the form. Note that it cannot be interpreted anymore as the first order condition of a functional. Unfortunately, due to the strongly non-convex character of the renormalized potential we have currently limited informations on its solutions. One expects that in the \Psingle phase region\Q (e.g. where is large compared to >) there should be uniqueness, and alternatively that two solutions should appear in \Ptwo phase region\Q. Despite these limitations \ eq. is useful and in the next section we show that it allows to prove large deviations for > in the semiclassical limit. In this section we want to investigate the large deviations of > in the infinite volume in the semiclassical limit \0> where \0> is a parameter which is inserted in the model as follows. For any |)>\\>, let >>> be defined by <\equation*> e>\>>>\e>f|)>>*\>>\|)>, where\ <\equation*> \>>\|)>\>V>|)>|)>\>\|)>|exp>V>|)>|)>\>\|)>>. Here >> is the Gaussian measure on |)>> with covariance -\|)>> and the Wick ordering is now taken with respect to >>. With some minor modifications to the derivation in the previous section we have the identity\ <\equation*> \>>=inf\|\>>>,\>|)>-inf\|\>>>,\>|)>, where\ <\equation*> >,\>|)>\\\|\>>W+Z|)>+V>W+Z>|)>+\|]>,\\|\>. Let ,\>> a minimizer for this admissible >,\>>. Reasoning as in Proposition above we have that the family ,\>|)>,\>> is exponentially tight for any bounded and continuous . For each \0> we pick a weakly convergent subsequence of ,\>|)>>> with limit >>. We have then also >>\\>=Law>>+Z|)>> along the same subsequence and clearly, for all bounded continuous we have <\equation*> \>>\\>, where >> is now defined as <\equation*> e>\>>\e>f|)>>*\>\|)>. In this setting where we do not have sufficient information to establish uniqueness of the limiting measures >|)>>>, nonetheless we are able to obtain a Laplace principle as \0> for the corresponding family >|)>>>. Note that the family >|)>>> is exponentially tight by Lemma. <\theorem> Any family of accumulation points >|)>>> of >>|)>,\>> satisfies a Laplace principle on |)>> with rate function\ <\equation*> J|)>=>\+>\-\|)>\*,>||\H|)>>;>>>|,>||>>>>> That is <\equation*> lim\0>\>=inf\\|)>>|)>+J|)>|}>. <\proof> Assume that 0> and \C>|)>> and that is of the form <\equation> f|)>=C\|)>-\|)>|)>. It is enough to prove the theorem for this particular family of functions since we already know that >|)>>> are exponentially tight, and the family given by is rate function determining in the sense of Definition3.15 of. \; From we have <\equation*> |\\>\>>f|)>=\f,\,\>>W+Z|)>|]>=f|)>e>\f|)>>*\>>\|)>|e>\f|)>>*\>>\|)>>. By taking a further subsequence we can also have that f,\,\>=\f,\,\>\\f,\>>. We can then pass to the limit and obtain <\equation*> |\\>\>f|)>=\f,\>>W+Z|)>|]>, for any > and arbitrarily chosen accumulation point f,\>>. Note that we do not have to do a diagonal argument over the uncountably many choices of ,f>: for any given ,f> we need just to take a subsequence of the sequence chosen to have >>|)>>> converge in order to establish the equality. \; Now recall that the minimizer >> satisfies an infinite volume Euler\ULagrange equation of the form \; <\equation> ||>>fW+Z|)>\H+\\\H+\\\Z H+|\>.>>||||+\\W H+\ZH+2*\|]>>>>>> for any with compact support. Furthermore we can carry over the proof of Lemma, and obtain that\ <\equation*> sup\1> \>>|Z|\<\|\|\>>*|)>>+|Z|\<\|\|\>>*|)>>|]>. This implies that the family >|)>>> is also tight and therefore we can consider any accumulation point > as \0> which will satisfy the same estimate. By passing to the limit in via by now standard estimates similar to those in Lemma we obtain <\equation> ||>f\H+\ZH+2*\|]>,>>>>> valid for any adapted and compactly supported . Note that this equation does not contain anymore the GFF. This equation has a unique solution. Indeed if > is another solution we can couple the two solutions in \|\>|)>> such that =Law>> and have <\equation*> 0=\>f-\f|]>\H+\-Z|]>H+2*\|]>. Now taking K> for \C>|)>> and \0>, we obtain\ <\equation*> 0=\>f-\f|]>\K|)>+\\-Z|]>K+2*\K|)>|]>. By convexity of and |)>> we have that\ <\equation*> \>f-\f|]>\K|)>+\\-Z|]>K|]>\0. Sending \1> we obtain that\ <\equation*> \>|]>\0\K=0\\=\. Therefore > is the unique minimizer of the admissible (convex) functional <\equation*> >|)>\\>+\Z+\|]>. Let >=Law>s\|)>> for \H> then <\equation*> >>|)>=f|)>+\\+\-\|)>\=f|)>+J|)>. Using the fact that the variational problem in the time variable for the functional <\equation*> \=>-\|)>>\s is minimized for >> constant, we have for any > <\equation*> >|)>=\>+\Z+\|]> <\equation*> \\>+\Z+Z-\|)>Z|]>=\>+J|]> <\equation*> \inf\H>|)>+J|)>|]>. As a consequence we obtain that <\equation*> inf\|\>>>|)>=inf\H>|)>+J|)>|]>, and that =\>> for the unique > which attain the convex deterministic minimization problem on the r.h.s.. \; To sum up, we have established that\ <\equation> lim\0>|\\>\>f|)>=lim\0>\f,\>>W+Z|)>|]>=ff>|)>. Integrating eq. and using dominated convergence <\equation*> lim\0>\>f|)>=>lim\0>|\\>\>f|)>\\=>lim\0>\f,\>>W+Z|)>|]>\\ <\equation*> =>ff>|)>\\\f|)>. This also implies that the derivative in \> of \f|)>> exists and <\equation*> |\\>f|)>=ff>|)>. To conclude it is enough to show that this coincides with <\equation*> |\\>inf\H*|)>>f|)>+J|)>|}>=|\\>|~>f|)>. Recall that f>> is the minimizer of \f|)>+J|)>> so the Euler\ULagrange equation gives <\equation*> \\ff>|)>\h+\Jf>|)>\h=0. At this point we note that <\equation*> 0=\f+\|)>f>|)>-\ff>|)>|]>\+\|)>f>-\f>|)> <\equation*> +J+\|)>f>|)>-\Jf>|)>|]>\+\|)>f>-\f>|)> <\equation*> +\\f+\|)>f>|)>\+\|)>f>-\f>|)>. By convexity of and the first two terms are positive and <\equation*> |\+\|)>f>-\f>|\<\|\|\>>>\\|\f+\|)>f>|)>|\<\|\|\>>>\\|\+\|)>f>|\<\|\|\>>>+C|)>\\. We can calculate\ <\equation*> |~>+\|)>f|)>-|~>f|)>>||f+\|)>f>|)>+\+\|)>f>|)>-ff>|)>|)>>>|||+\|)>f>|)>-Jf>|)>.>>>>> Given that |\+\|)>f>-\f>|\<\|\|\>>>\\> we have, since by Sobolev embedding |\|\<\|\|\>>>> is locally Lipschitz in the > norm\ <\eqnarray*> ||+\|)>f>|)>-ff>|)>|)>+J+\|)>f>|)>-Jf>|)>|)>>>|||ff>|)>+\|)>f>-\f>|)>+-\|)>f>|)>*+\|)>f>-\f>|)>+\f>|)>+\|)>f>-\f>|)>+O*|)>>>||||)>,>>>> where in the last line we used the Euler\ULagrange equation for f>>. This implies in turn\ <\equation*> |\\>|~>f|)>=lim\0>>|~>+\|)>f|)>-|~>f|)>|)>=lim\0>>\f+\|)>f>|)>=ff>|)> which concludes the proof. In this section we will discuss a variational description of the exponential interaction, that is the measure\ <\equation*> \,exp>\|)>=>|)>>*\\|)>|e>|)>>*\\|)>>, where now >> is given by\ <\equation*> V>|)>=\>\|exp\|)>|\>, with some \|)>> and where >\|exp\|)>|\>> is the Gaussian multiplicative chaos and \C>,\|)>> is a (smooth and compactly supported) spatial cutoff. We start with some preliminaries about the rigorous construction of the GMC. <\proposition> Let be a GFF and \ >=\>\W> where >\\> and >\|)>> and let \|)>>. The Borel measures ,\>|)>>> on > given by <\equation*> M,\>=expW>-\\>|]>|)>\x,A\\|)>, converges almost surely set-wise to a measure\ <\equation> M>=lim\0>M,\>,A\\|)>. We call >> the Gaussian multiplicative chaos (GMC) with parameter > associated to the GFF . It enjoys the following properties <\enumerate-alpha> For p\8\/\> and for any compact set \> <\equation*> sup> \,\>|\|>|]>\\. Consequently\ <\equation*> \>|\|>|]>\\. For p\8\/\> <\equation*> \,\>|B|)>|)>|\>|\|>|]>\C*\\/4\>*\,\>|B|)>|\>|\|>|]>, where \\> denotes the open ball centered in of radius . <\proof> For a proof of see. The property a) is due to Kahane, while the estimate in b) is due to Hoshino, Kawabi and Kusuoka. From these properties we can determine easily the Besov regularity of >> following(Section2). <\proposition> Assume that /|)>|)>>, p\8\/\>, \ then <\equation*> \|M>|\<\|\|\>>|)>>|]>\\. <\proof> We have <\equation*> |M,\>|\<\|\|\>>|)>>|]>>||0>2\|\M,\>|\<\|\|\>>|)>>>>||>|0>2\M,\>|\|>.>>>>> Now using translation invariance of the law of ,\>> and the fact that \> is a Schwarz function we estimate <\equation*> M,\>|\|>>||\>\|)>x|)>M,\>\x|\|>>>||>|\>|)>\>M,\>x|)>\x|\|>>>|||\>|)>2\|)>>M,\>\x|\|>>>||>|\i/|)>>.>>>>> This gives the statement, since by weak convergence of ,\>\M>> we have\ <\equation*> \|\M>|\<\|\|\>>|)>>\limsup\0> \|\M,\>|\<\|\|\>>|)>>. \; In this model we let \\\|)>> and =>|)>:\> will be the appropriate enhanced GFF. By the usual abuse of notation we will denote also =>|)>> the canonical process on >. \; We also need a slightly improved definition of the relaxed trial space |\>>. <\definition> Set <\equation*> \=Law>,Z|)>\\|^>|)>:Z\|^> \>|Z|\<\|\|\>>>>|]>\\|}>, and\ <\equation*> |\>\\\|^>|)>:\|)>\:\\\ ,sup\> \>|Z|\<\|\|\>>|^>>|]>\\|}> We equip |\>> with the following topology: a sequence |)>\|\>> converges to > if <\enumerate-alpha> \\> weakly in \|^>|)>>, \> \>|Z|\<\|\|\>>>*>|]>\\>. <\lemma> The potential >> is admissible and <\equation*> >>|)>\>>>|)>=\>>\expZ|)>\M>+\|]> for all \|\>>. <\proof> Note that >> is positive and that Proposition implies that for close enough to and +=1> <\equation*> \>|)>|]>+\>|\|>|]>\\, so >> is a tame functional. A result of Shamov gives <\equation*> >\|expW+Z|)>|\>=>\|expW|)>|\>expZ|)>=>\expZ|)>\M>. Therefore we have for all \|\>>: <\equation*> >>|)>\>>>|)>=\>>\expZ|)>\M>+\|]>\\>|]>\\>|Z|\<\|\|\>>|^>>|]>, from which also the coercivity of >>>> is clear. To conclude admissibility it remains to show that >>>> is lower semicontinuous on |\>>. We have for any \|\>> a sequence |)>> such that \\> in |\>> <\equation*> liminf \>|]>\\>|]>. Now let :\\\> be defined by =x> if \N>, =sgnN> if \N>. Then \N>, > is Lipschitz with constant and \x>. Set =g|)>> and =Law|)>>. Then <\eqnarray*> ||\>\>\expZ|)>\M>>>|||\>>0|}>>\expZ|)>\M>+0|}>>\expZ|)>\M>|]>>>|||>0|}>>\expZ|)>\M>+0|}>>\expZ|)>\M>>>|||>\expZ|)>\M>>>>> where in the second to last line we used monotone convergence. It remains to prove that <\eqnarray*> \>\>\expZ|)>\M>>||>\expZ|)>\M>,>>>> then the statement will follow from\ <\eqnarray*> >\expZ|)>\M>>||\>\>\expZ|)>\M>>>|||\>lim\>\>\expZ|)>\M>>>||>|\>\>\expZ|)>\M>.>>>> We have that |Z|\<\|\|\>>>>\N> and \>|Z|\<\|\|\>>>|]>\\>|Z|\<\|\|\>>>|]>>. By interpolation\ <\equation*> |Z|\<\|\|\>>,p>>|)>>\|Z|\<\|\|\>>|)>>>>|Z|\<\|\|\>>>|)>>, provided \\>. Note that we can choose > such that |\>\p> and |4\>\\> and \\> by taking\ <\equation*> \\\|4\>. By Besov embedding we can assume that \\> weakly on \B>|)>>. We have established that on \B>|)>> the functional \expZ|)>\M>> is continuous and since\ <\equation*> \*expZ|)>\M>\eN>, it is also bounded. By weak convergence <\equation*> lim\>\>\*expZ|)>\M>=\>\*expZ|)>\M>, and we can conclude. Our general framework now gives: <\lemma> The admissible functional >>> has a minimizer >\|\>> and <\equation*> ,exp>\-logexp|)>|)>\\>=min\|\>>>>|)>=>>>|)>. Furthermore for any minimizer \|\>> the following Euler\ULagrange equation <\equation*> \>\\*expZ|)>*K\M>+2\|]>=\>f*K|]> holds for any admissible control . <\proof> This follows as in Lemma. <\lemma> The functional >>> has a unique minimizer >> in |\>>. Furthermore\ <\equation*> Z\0, >> almost surely, and\ <\equation*> sup> \>>|Z|\<\|\|\>>|^>>|]>\\. <\proof> Observe that >> is convex in . Now let ,\> be two minimizers for >>>. Recall from Lemma that there exists \|\>|)>> such that >=\>. Now\ <\eqnarray*> ||>\expZ+K|)>\M>+\|]>-\>\expZ|)>\M>+\|]>>>|||\\+\>\\expZ|)>K\M>+2\|]>.>>>> By Lemma we have that\ <\equation*> \>\\expZ|)>K\M>+2\|]>=0, so\ <\equation*> \\>>|)>->>|)>=0, which implies and therefore =\>. For the second statement observe that the mapping -> leaves the energy *> invariant by chain rule, while\ <\equation*> \expA|)>\M>\\expZ|)>\M>, so if >> is a minimizer, so is >>|)>>. By uniqueness we can conclude. For the third claim we test the EL equation with Z> and obtain\ <\equation*> \>\\*expZ|)>\*Z\M>+2\*Z|)>|]>=\>f\*Z|]>. We can again estimate as in the proof of Lemma <\equation*> 2\*Z|)>\|Z|\<\|\|\>>|^>>, and\ <\equation*> f\*Z\|Z|\<\|\|\>>|^>>. Furthermore observe that\ <\equation*> >\*expZ|)>\*Z\M>\-C>\\*\M>, since Z|)>Z> is bounded below. Summing up, we have the estimate\ <\equation*> \>>|Z|\<\|\|\>>|^>>|]>\\|Z|\<\|\|\>>|^>>|]>+\>>\\*\M>|]>\\|Z|\<\|\|\>>|^>>|]>+1 where in the last bound we used Lemma. <\lemma> Let >\|\>>. For \|\>>|)>> we have <\equation*> >,exp>|)>\>>>|)>=\>+\|]> <\equation*> +\*\>\expZ+K|)>\M>-\expZ|)>\M>|]>. Then >,exp>|)>\0> for all \|\>>|)>> iff >> is a minimizer of >>>. Moreover in this case, thanks to the Euler\ULagrange equations for we have <\equation*> >,exp>|)>=\>+\\expZ|)>K|)>-1-\K|)>\M>|]>, and \ <\equation*> \>=inf\|\>>|)>> >,exp>|)>,>,exp>|)>\\>|]>+>,exp>|)>. Any minimizer \|\>>|)>> of >,exp>> satisfies the Euler\ULagrange equation\ \; <\equation> >+\\\expZ|)>K|)>-1|)>H\M>|]>>>|\>f\H|]>.>>>>> The proof of Lemma is completely analogous to Lemma and Lemma, thus we omit the details. Thanks to the convexity of the interaction we have a good control of the locality properties of the solutions to the Euler\ULagrange equation.\ <\lemma> Assume that 0> and\ <\equation*> sup\L|)>>|\f|)>|\<\|\|\>>|)>>\\. Then for any minimizer > of >,exp>|)>> we have\ <\equation*> \>|K|\<\|\|\>>|)>>|]>\\. <\proof> Testing equation with K> we have\ <\equation*> \; <\equation*> |>>\>-\|)>>\s|]>>>||*\\>\expZ|)>K|)>-1|)>\K\M>|]>>>||>\\fK|]>\C\>|K|\<\|\|\>>|)>>|]>>>>>> Now\ <\eqnarray*> ||>>\>-\|)>>\s|]>>>|||>>\-\|)>>|)>\s|]>>>|||>>-\|)>,\|]>>-\|)>>\s|]>>>>> and we can compute, as in the proof of Lemma, <\equation*> -\|)>,\|]>>\\\>. Furthermore observe that\ <\equation*> \>\expZ|)>K|)>-1|)>\K\M>|]>\0, Putting everything together we have\ <\eqnarray*> |)>\>|K|\<\|\|\>>|)>>>|||)>\>>\-\|)>>|)>\s|]>>>||>|>|K|\<\|\|\>>|)>>>>>> which implies the statement. <\definition> For 0> large (to be chosen suitably below) denote by |)>> the set |)>> of measure \\|^>\|)>|)>> such that <\enumerate-alpha> >>,Z|)>=\> |)>,|)>\>> for which <\equation*> Law,n>,Z,K|)>\\ sup\>\|Z|\<\|\|\>>|^>>+|K|\<\|\|\>>*|)>>|]>\2M. Let \\|\>>|)>>. We say that |)>\> converges to \\> if \\> weakly in \|^>\|)>|)>> and\ <\equation*> sup\>\>|Z|\<\|\|\>>|^>>+|K|\<\|\|\>>|)>>|]>\2M. <\remark> Note that any sequence in > has a convergent subsequence.\ <\corollary> There exists a sufficiently large > (not depending on >) such that\ <\equation*> inf\\>|)>> >,exp>|)>=inf\\>|)>> >,exp>|)>. <\proof> This is a direct consequence of the decay of the minimizer established in Lemma. <\definition> Define\ <\equation*> >,exp>|)>=>,exp>|)> \\>>|)>>> +\ <\lemma> Let |)>> be a sequence of spatial cutoffs such that =1> for \n>, \ and\ <\equation*> sup\>|\|\<\|\|\>>>>\+\,k\\. Pick a subsequence > (not relabeled) such that \\> weakly on \|^>> where > is the minimizer of >>> in |\>>. Then >,exp>> >-converges to >,exp>> on > where\ <\equation*> >,exp>|)>=\>+expZ|)>K|)>-1|)>\M>+|]> if \\|)>> and > otherwise. <\proof> We first prove the liminf inequality. Assume that \\> in >. By definition of > we can restrict ourselves to \\|)>>. From the definition of convergence in > it is not hard to see that\ <\equation*> \>+K|)>|]>\\>|]>, and\ <\equation*> \>|]>\liminf\> \>|)>|]>. by the Portmanteau theorem. Now we can invoke the Skhorohod's embedding theorem from and pass to a probability space |~>> and on it a sequence ,M,n>,Z,K|)>> convergent almost surely on \|^>\|)>> to >,Z,K|)>> such that ,M,n>,Z,K|)>=\>. \ Using the fact that\ <\equation*> sup \|~>>|Z-Z|\<\|\|\>>|)>>+|K-K|\<\|\|\>>|)>>|]>\\ and almost-sure convergence we can assume that\ <\equation*> \|~>>|M,n>-M>|\<\|\|\>>>>\0,\|~>>|Z-Z|\<\|\|\>>|)>>+|K-K|\<\|\|\>>|)>>|]>\0, for any ,p> satisfying the assumptions of Proposition and 2>. Let :\\\\\> be continuous such that <\equation*> \=z z,\N, =0 z\2N \2N> and moreover \\.> We will write in the following > for the function \\\,x|)>>. We claim that \ <\equation*> |\>\expZ|)>\K|)>-1|)>,x|)>\M,n>>>||expZ|)>\K|)>-1|)>,x|)>\M>.>>>>> Since\ <\equation*> sup\>|expZ|)>|\<\|\|\>>>>\1,sup\>|\K|)>-1|)>,x|)>|\<\|\|\>>>>\N, we can apply Lemma below provided that, for any 2>, <\equation*> \|\expZ|)>\K|)>-1|)>,x|)>-expZ|)>\K|)>-1|)>,x|)>|\<\|\|\>>|)>>|]>\0, which in turn is ensured by Lemma and Lemma. Then, by monotone convergence\ <\eqnarray*> ||expZ|)>K|)>-1|)>\M>>>|||expZ|)>\K|)>-1|)>,x|)>\M>>>||| lim\>\expZ|)>\K|)>-1|)>,x|)>\M,n>>>||>| liminf\>\expZ|)>K|)>-1|)>\M,n>>>|||\>\expZ|)>K|)>-1|)>\M,n>.>>>> Let us now show that\ <\equation*> lim\>\expZ|)>K|)>-1|)>\M,n>=expZ|)>K|)>-1|)>\M>. To start, note that\ <\equation*> Z|)>|\|>\1,K|)>-1|)>|\|>\1, so again we have that from Lemmas , below, for any 2>, <\equation*> \|expZ|)>K|)>-1|)>-expZ|)>K|)>-1|)>|\<\|\|\>>|)>>\0. From this and using that =1> for \n>, it is easy to see that\ <\equation*> \|\expZ|)>K|)>-1|)>-expZ|)>K|)>-1|)>|\<\|\|\>>|)>>\0. we have that,\ <\equation*> \expZ|)>K|)>-1|)>\M,n>\expZ|)>K|)>-1|)>\M>. To construct the recovery sequence consider > such that >,exp>|)>\\>. Then \ <\equation*> \>expZ|)>K|)>-1|)>\M>\\,\>\M where the second inequality holds by definition of >. Again, after an application of Lemma and interpolation between >> and |)>> and provided 1/\>, we have that <\equation*> |>expZ|)>K|)>-1|)>\M>>>|>||K|)>-1|)>|\<\|\|\>>\>>,p>|)>>|]>\|M>|\<\|\|\>>>|)>>|]>>>|>||K|)>-1|)>|\<\|\|\>>|)>>|]>\|M>|\<\|\|\>>>|)>>|]>>>|>|.>>>>> So\ <\equation*> |>expZ|)>K|)>-1|)>\M>>>|>|>expZ|)>K|)>-1|)>\M>|\|>+>expZ|)>K|)>-1|)>\M>|\|>>>|>|.>>>>> Now set =\>, by monotone convergence\ <\equation*> lim\>\>expZ|)>K|)>-1|)>\M>\\>expZ|)>K|)>-1|)>\M>. Moreover, since K|)>-1|)>|\|>\K|)>-1|)>|\|>>, by dominated convergence also holds\ <\equation*> lim\>\>expZ|)>K|)>-1|)>\M>\\>expZ|)>K|)>-1|)>\M>. Invoking there exist a probability space |~>,|~>,|~>|)>> and random variables ,M>,Z|)>> on it such that ,M>,Z|)>\>,Z|)>> almost surely in \|)>> and |~>>,M>,Z|)>=\.>\ Let be a random variable |~>\|)>> such that =Law|~>>>. Set again =\> and =Law|~>>,Z,K|)>>. We claim that\ <\equation*> >,exp>|)>\>,exp>|)>. From Lemmas,, we get <\equation*> \|~>>\expZ|)>K|)>-1|)>\M,n>\\|~>>expZ|)>K|)>-1|)>\M>. Furthermore\ <\equation*> K|)>-1|)>\exp>,expZ|)>\1, and ,Z|)>\,Z|)>> in |)>>. So again from Lemmas ,, <\equation*> \|~>>expZ|)>K|)>-1|)>\M,n>\\|~>>expZ|)>K|)>-1|)>\M>. Finally\ <\equation*> lim\>\|~>>f+Z+K|)>\\|~>>f|)> follows by boundedness and continuity of . Taking a diagonal sequence we obtain the statement.\ <\corollary> If |)>> is as in Lemma we have\ \; <\equation*> lim\>-logexp|)>|)>\\,exp>=inf\\|)>>>,exp>|)> <\proof> This follows from >-convergence, the fact that > is sequentially compact and the fundamental theorem of >-convergence(see e.g.). We conclude with some technical results needed in the proofs above. <\lemma> Assume that \\> is continuously differentiable, > are bounded. Define >(the negative part of ). We have that the map\ <\equation*> \\f|)>, continuously sends >|)>> in itself with <\equation*> |f|)>|\<\|\|\>>>|)>>\1+|\|\<\|\|\>>>|)>>. If =0> we have in particular <\equation*> |f|)>|\<\|\|\>>>|)>>\|\|\<\|\|\>>>|)>>. <\proof> Assume that =0>. By definition of we have\ <\equation*> |)>|\|>\\x\|\|>\\x. Furthermore by p.291\U292, we have\ <\equation*> f|)>|\|>\\x=|)>\0>|)>\\|\|>\\x\\|\|>\\x. Now for weak continuity, let \\> weakly in >|)>>. We know that <\equation*> sup|\|\<\|\|\>>>|)>>\\sup|f|)>|\<\|\|\>>>|)>>. Now let > >be a subsequence such that >|)>\\>. We can extract a further subsequence (not relabeled) such that\ <\equation*> \>\\ almost surely. Then >|)>\f|)>> almost surely which in turn implies =f|)>>. From this we can conclude. The second statement follows from the first by writing +g> with =0>. \ <\lemma> Let ,n>> be a sequence of random measures in |)>> on a probability space ,\,\|)>>, such that ,n>\M>> >-almost surely in |)>> for some |4\>> and 8\/\> close enough to |4\>,8\/\> respectively. There exists 2> such that we have\ <\equation*> \f\M,n>\\f\M> whenever <\equation*> \|f-f|\<\|\|\>>|)>>|]>\0 and\ <\equation*> sup\>|f|\<\|\|\>>>>+|f|\<\|\|\>>>>=C\\. <\proof> By interpolation we have that for \q/p> and \> we have with <\equation*> |f-f|\<\|\|\>>,p>>|)>>\C|f-f|\<\|\|\>>>|)>>. Now recall that for any |\>\p\1> and \|4\>> we have\ <\equation*> sup\|M,n>|\<\|\|\>>>|)>>|]>\\. so provided =1>\ <\equation*> f\M,n>-f\M>|]>>||-f|)>\M,n>|]>+\f\M,n>-f\M>|]>>>||>||f-f|\<\|\|\>>,p>>|)>>|M,n>|\<\|\|\>>>|)>>|]>>>||||f|\<\|\|\>>,p>>|)>>|M,n>-M>|\<\|\|\>>>|)>>|]>>>||>||f-f|\<\|\|\>>\>|)>>|]>>>\|M,n>|\<\|\|\>>>|)>>|]>>>>||||f|\<\|\|\>>\>|)>>|]>>>\|M,n>-M>|\<\|\|\>>>|)>>|]>>.>>||>|>>>> since \\q>. Note that for any \8\> we can choose 2> such that <\equation*> q/p=\|4\> so if we choose > such that <\equation*> \\\|4\> we obtain as required\ <\equation*> |\>\p\1\\|4\>\\q/p=. <\lemma> For any 2> there exists an \> such that if\ <\equation*> |f|\<\|\|\>>>>,|g|\<\|\|\>>>>\C, then <\equation*> |f g|\<\|\|\>>|)>>\|f|\<\|\|\>>|)>>|g|\<\|\|\>>|)>>. Note that the >> norms are unweighted. If furthermore if \f> in |)>> and \g> in |)>> and |f|\<\|\|\>>>>+|g|\<\|\|\>>>>\C> then\ <\equation*> f*g\f*g|)>>>. <\proof> It is enough to prove the statement for smooth functions by density. By Hölder's inequality we have <\equation*> |f g|\<\|\|\>>|)>>\|f|\<\|\|\>>>>|g|\<\|\|\>>|)>>. By the Leibniz rule\ <\equation*> |\|\<\|\|\>>|)>>\|f|\<\|\|\>>>>|\g|\<\|\|\>>|)>>+|\f|\<\|\|\>>|)>>|g|\<\|\|\>>N>|)>>, provided that N\2>, and =+>. By interpolation and provided that /p=1/2>, we also have\ <\equation*> |g|\<\|\|\>>N>|)>>\|g|\<\|\|\>>|)>>>>|g|\<\|\|\>>>|)>>\C|g|\<\|\|\>>>|)>>. Choosing large enough gives the statement. For the second statement we rewrite\ <\equation*> f*g-f*g=g*|)>+f*|)> and proceed analogously. <\bibliography|bib|tm-abbrv|qft-drift> <\bib-list|34> S.Albeverio, F.C.De VecchiM.Gubinelli. The elliptic stochastic quantization of some two dimensional Euclidean QFTs. , 57(4), nov 2021. S.Albeverio, F.C.D.VecchiM.Gubinelli. Elliptic stochastic quantization. , 48(4):1693\U1741, jul 2020. N.BarashkovM.Gubinelli. A variational method for >. , 169(17):3339\U3415, nov 2020. N.Barashkov. . , Bonn, 2021. N.BarashkovM.Gubinelli. The > measure via Girsanov's theorem. , 26(none), jan 2021. R.BauerschmidtM.Hofstetter. Maximum and coupling of the sine-Gordon eld. , 49, 2022. N.Berestycki. An elementary approach to Gaussian multiplicative chaos. , 22:12, 2017. Id/No 27. C.Borell. Diffusion equations and geometric inequalities. , 12(1):49\U71, 2000. M.BouéP.Dupuis. A variational representation for certain functionals of Brownian motion. , 26(4):1641\U1659, oct 1998. 10.1214/aop/1022855876. B.Bringmann. Invariant Gibbs measures for the three-dimensional wave equation with a Hartree nonlinearity I: Measures. , 2020. B.Bringmann. Invariant Gibbs measures for the three-dimensional wave equation with a hartree nonlinearity II: Dynamics. , 2020. A.Chandra, T.S.GunaratnamH.Weber. Phase transitions for >. , aug 2020. ArXiv: 2006.15933. A.Dal Maso. >-Convergence>. Birkhäuser Boston, 1993. 10.1007/978-1-4612-0327-8. L.C.Evans. , 19. American Mathematical Society, Providence, RI, Second, 2010. J.FengT.G.Kurtz. . American Mathematical Society, 2015. J.GlimmA.Jaffe. . Springer-Verlag, New York, 2, 1987. M.GubinelliM.Hofmanová. Global Solutions to Elliptic and Parabolic > Models in Euclidean Space. , 368(3):1201\U1266, 2019. 10.1007/s00220-019-03398-4. M.GubinelliM.Hofmanová. A PDE Construction of the Euclidean > Quantum Field Theory. , apr 2021. Y.HariyaS.Watanabe. The Boué\UDupuis formula and the exponential hypercontractivity in the Gaussian space. , nov 2021. ArXiv: 2110.14852. R.Høegh-Krohn. A general class of quantum fields without cut-offs in two space-time dimensions. , 21(3):244\U255, sep 1971. J.v.d.Hoeven etal. GNU TeXmacs. , 1998. M.Hoshino, H.KawabiS.Kusuoka. Stochastic quantization associated with the |)>>-quantum field model driven by space-time white noise on the torus in the full >-regime. 2020. A.Jakubowski. The Almost Sure Skorokhod Representation for Subsequences in Nonmetric Spaces. , (42(1)):167\U174, 1998. J.-P.Kahane. Sur le chaos multiplicatif. (On multiplicative chaos). , 9:105\U150, 1985. H.Lacoin, R.RhodesV.Vargas. Semiclassical limit of liouville field theory. , 273:0, 05 2017. H.Lacoin, R.RhodesV.Vargas. The semiclassical limit of liouville conformal field theory. 2019. J.Lehec. Representation formula for the entropy and functional inequalities. , 49(3):885\U899, 2013. MR3112438. D.MikulincerY.Shenfeld. The Brownian transport map. , nov 2021. ArXiv: 2111.11521. T.Oh, M.OkamotoL.Tolomeo. Focusing >-model with a Hartree-type nonlinearity. , aug 2021. ArXiv: 2009.03251. G.ParisiN.Sourlas. Random Magnetic Fields, Supersymmetry, and Negative Dimensions. , 43(11):744\U745, sep 1979. G.ParisiY.S.Wu. Perturbation theory without gauge fixing. , 24(4):483\U496, 1981. A.Shamov. On Gaussian multiplicative chaos. , 270(9):3224\U3261, 2016. B.Simon. |)>> Euclidean (Quantum) Field Theory>. Princeton University Press, Princeton, N.J, 1974. MR489552. A.S.Üstünel. Variational calculation of Laplace transforms via entropy on Wiener space and applications. , 267(8):3058\U3083, 2014. 10.1016/j.jfa.2014.07.006. <\initial> <\collection> <\attachments> <\collection> <\associate|bib-bibliography> <\db-entry|+1CLJhZmi1u3N0Sc6|article|barashkov_gubinelli_variational> <|db-entry> M. > >> > <\db-entry|+1CLJhZmi1u3N0Sc8|article|parisi_perturbation_1981> <|db-entry> Y. S. > > <\db-entry|+HWynyN0X3XTwsR|article|albeverioEllipticStochasticQuantization2020> <|db-entry> Francesco C. De Massimiliano > > 2R2\\mathbb{R}\B{2}. This formula is the simplest instance of dimensional reduction, discovered in the physics literature by Parisi and Sourlas (Phys. Rev. Lett. 43 (1979) 744\U745), which links the law of an elliptic SPDE in d+2d+2d+2 dimension with a Gibbs measure in ddd dimensions. This phenomenon is similar to the relation between a \d+1Rd+1\\mathbb{R}\B{d+1} dimensional parabolic SPDE and its \dRd\\mathbb{R}\B{d} dimensional invariant measure. As such, dimensional reduction of elliptic SPDEs can be considered a sort of elliptic stochastic quantisation procedure in the sense of Nelson (Phys. Rev. 150 (1966) 1079\U1085) and Parisi and Wu (Sci. Sin. 24 (1981) 483\U496). Our proof uses in a fundamental way the representation of the law of the SPDE as a supersymmetric quantum field theory. Dimensional reduction for the supersymmetric theory was already established by Klein et al. (Comm. Math. Phys. 94 (1984) 459\U482). We fix a subtle gap in their proof and also complete the dimensional reduction picture by providing the link between the elliptic SPDE and the supersymmetric model. Even in our d=0d=0d=0 context the arguments are nontrivial and a nonsupersymmetric, elementary proof seems only to be available in the Gaussian case.> <\db-entry|+HWynyN0X3XTwss|article|albeverioEllipticStochasticQuantization2021> <|db-entry> Francesco C. Massimiliano > > <\db-entry|+HWynyN0X3XTwsQ|article|parisiRandomMagneticFields1979> <|db-entry> N. > > 2)-dimensional spin system in the absence of a magnetic field. This is due to the hidden supersymmetry of the associated stochastic differential equation. We identify a space with one anticommuting coordinate with a space having negative dimensions -2.> <\db-entry|+1CLJhZmi1u3N0ScG|article|Bringmann-20-1> <|db-entry> > <\db-entry|+1CLJhZmi1u3N0ScH|article|Bringmann-20-2> <|db-entry> > <\db-entry|+HWynyN0X3XTwsu|article|chandraPhaseTransitionsPhi2020> <|db-entry> Trishen S. Hendrik > > <\db-entry|+HWynyN0X3XTwsw|article|ohFocusingPhiModel2021> <|db-entry> Mamoru Leonardo > > 2$, while we prove its non-normalizability for $\\beta 2$. Furthermore, we establish the following phase transition at the critical value $\\beta = 2$: normalizability in the weakly nonlinear regime and non-normalizability in the strongly nonlinear regime. We then study the canonical stochastic quantization of the focusing Hartree $\\Phi\B4_3$-measure, namely, the three-dimensional stochastic damped nonlinear wave equation (SdNLW) with a cubic nonlinearity of Hartree-type, forced by an additive space-time white noise, and prove almost sure global well-posedness and invariance of the focusing Hartree $\\Phi\B4_3$-measure for $\\beta 2$ (and $\\beta = 2$ in the weakly nonlinear regime). In view of the non-normalizability result, our almost sure global well-posedness result is sharp. In Appendix, we also discuss the (parabolic) stochastic quantization for the focusing Hartree $\\Phi\B4_3$-measure. We also construct the defocusing Hartree $\\Phi\B4_3$-measure for $\\beta 0$.> <\db-entry|+HWynyN0X3XTwsy|article|barashkovPhiMeasureGirsanov2021> <|db-entry> Massimiliano > > 34 measure via Girsanov's theorem.pdf:application/pdf> <\db-entry|+HWynyN0X3XTwt0|phdthesis|barashkovVariationalApproachGibbs2021> <|db-entry> > <\db-entry|+HWynyN0X3XTwsv|article|bauerschmidtMaximumCouplingSineGordon2022> <|db-entry> Michael > > \ 6\, we prove that the distribution of the centred maximum of the \-regularised continuum sine-Gordon eld on the two-dimensional torus converges to a randomly shifted Gumbel distribution as \ \ 0. Our proof relies on a strong coupling at all scales of the sineGordon eld with the Gaussian free eld, of independent interest, and extensions of existing methods for the maximum of the lattice Gaussian free eld.> <\db-entry|+1CLJhZmi1u3N0Sc0|article|borell_diffusion_2000> <|db-entry> > > <\db-entry|+1CLJhZmi1u3N0SbP|article|lehec_representation_2013> <|db-entry> > MR3112438> > <\db-entry|+1JemQ8wu1MudQ9jI|article|mikulincerBrownianTransportMap2021> <|db-entry> Yair > > <\db-entry|+1CLJhZmi1u3N0Sc1|book|simon_po2_1974> <|db-entry> > |)>> Euclidean (Quantum) Field Theory> MR489552> > <\db-entry|+1H0fhwuD1uZ1SgV1|book|glimm_quantum_1987> <|db-entry> Arthur > <\db-entry|+1CLJhZmi1u3N0SbO|article|boue_variational_1998> <|db-entry> P. > 10.1214/aop/1022855876> > <\db-entry|+1CLJhZmi1u3N0SbS|article|ustunel_variational_2014> <|db-entry> > 10.1016/j.jfa.2014.07.006> > <\db-entry|+1CLJhZmi1u3N0Sbw|article|gubinelli_global_2018> <|db-entry> Martina > > Models in Euclidean Space> 10.1007/s00220-019-03398-4> > <\db-entry|+1H0fhwuD1uZ1SgV3|article|gubinelli_pde_2018> <|db-entry> Martina > > Quantum Field Theory> > <\db-entry|+1H0fhwuD1uZ1SgV0|article|hoegh-krohn_general_1971> <|db-entry> > > <\db-entry|+1JemQ8wu1MudQ9jG|article|hariyaBouDupuisFormula2021> <|db-entry> Sou > > <\db-entry|+1H0fhwuD1uZ1SgV6|book|Feng_2015> <|db-entry> Thomas G. > <\db-entry|+1rmuK6WO1pWjska2|article|Berestycki-2017> <|db-entry> > <\db-entry|+1rmuK6WO1pWjska4|article|Kahane-1985> <|db-entry> > <\db-entry|+HWynyN0X3XTwsq|article|Hoshino-2020> <|db-entry> Hiroshi Seiichiro > |)>>-quantum field model driven by space-time white noise on the torus in the full >-regime> <\db-entry|+1H0fhwuD1uZ1SgV4|article|Shamov-2016> <|db-entry> > > <\db-entry|+1CLJhZmi1u3N0Sbx|article|jakubowski> <|db-entry> > <\db-entry|+1CLJhZmi1u3N0Sbz|book|DalMaso1993> <|db-entry> > >-Convergence> 10.1007/978-1-4612-0327-8> > <\db-entry|+1H0fhwuD1uZ1SgV5|book|Evans-2010> <|db-entry> > > <\db-entry|+HWynyN0X3XTwsU|misc|TeXmacs:website> <|db-entry> > > <\references> <\collection> > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > <\auxiliary> <\collection> <\associate|bib> TeXmacs:website barashkov_gubinelli_variational barashkov_gubinelli_variational barashkov_gubinelli_variational parisi_perturbation_1981 albeverioEllipticStochasticQuantization2020 albeverioEllipticStochasticQuantization2021 parisiRandomMagneticFields1979 barashkov_gubinelli_variational Bringmann-20-1 Bringmann-20-2 chandraPhaseTransitionsPhi2020 ohFocusingPhiModel2021 barashkovPhiMeasureGirsanov2021 barashkovVariationalApproachGibbs2021 bauerschmidtMaximumCouplingSineGordon2022 borell_diffusion_2000 lehec_representation_2013 mikulincerBrownianTransportMap2021 simon_po2_1974 glimm_quantum_1987 boue_variational_1998 ustunel_variational_2014 gubinelli_global_2018 gubinelli_pde_2018 albeverioEllipticStochasticQuantization2021 LacoinSemiclassical2017 LacoinSemiclassical2019 hoegh-krohn_general_1971 albeverioEllipticStochasticQuantization2021 mikulincerBrownianTransportMap2021 ustunel_variational_2014 hariyaBouDupuisFormula2021 barashkov_gubinelli_variational ustunel_variational_2014 barashkov_gubinelli_variational ustunel_variational_2014 barashkov_gubinelli_variational glimm_quantum_1987 glimm_quantum_1987 barashkov_gubinelli_variational barashkov_gubinelli_variational gubinelli_pde_2018 barashkov_gubinelli_variational glimm_quantum_1987 Feng_2015 Berestycki-2017 Kahane-1985 Hoshino-2020 Hoshino-2020 Shamov-2016 jakubowski jakubowski DalMaso1993 Evans-2010 <\associate|toc> |Acknowledgments. |.>>>>|> > |Notations. |.>>>>|> > |math-font-series||1A general variational setting for 2d Euclidean fields> |.>>>>|> |math-font-series||2The quartic interaction> |.>>>>|> |math-font-series||3Large deviations> |.>>>>|> |math-font-series||4Exponential interaction > |.>>>>|> |math-font-series||Bibliography> |.>>>>|>