betaFeedback

WorksPhysics and Astronomy

Lagrangian averaging of singular stochastic actions for fluid dynamics

Abstract

We construct sub-grid scale models of incompressible fluids by considering expectations of semi-martingale Lagrangian particle trajectories. Our construction is based on the Lagrangian decomposition of flow maps into mean and fluctuation parts, and it is separated into the following steps. First, through Magnus expansion, the fluid velocity field is expressed in terms of fluctuation vector fields whose dynamics are assumed to be stochastic. Second, we use Malliavin calculus to give a regularised interpretation of the product of white noise when inserting the stochastic velocity field into the Lagrangian for Euler's fluid. Lastly, we consider closures of the mean velocity by making stochastic analogues of Talyor's frozen-in turbulence hypothesis to derive a version of the anisotropic Lagrangian averaged Euler equation.

Introduction

The modelling of multi-scale fluid flows is a well-known problem in fluid dynamics. In previously established work, fluid flows exhibiting multiple spatial scales have typically been modelled by considering averaging techniques and closures to capture the effects of the unresolvable scales on large-scale dynamics. These models are typically referred to as sub-grid scale models. One such model is the Generalised Lagrangian Mean (GLM) model introduced by Andrews & McIntyre [1] that considered the application of an arbitrary averaging operator to the Lagrangian expansion of ideal incompressible fluid equations and it is later closed in [10]. To preserve the geometric structure of ideal fluid equations during averaging, one option is to apply averaging to the variational formulation of ideal fluid flows, instead of applying to the equations directly. This approach resulted in several models such as the Lagrangian averaged Navier Stokes model [9]. More recently, fluid flows that exhibit multiple temporal scales have been modelled by stochastic partial differential equations where the evolution of the fast dynamics is represented by integration against driving stochastic processes such as Brownian motion and geometric rough paths. One class of structure-preserving stochastic perturbation, known as transport noise, is constructed by assuming a semi-martingale decomposition of the Lagrangian fluid flow map gt∈Diff⁡(M)g_t \in\operatorname{Diff}(\mathcal{M}) (for fluid domain M\mathcal{M}) in a stochastic variational principle, see, e.g., [13, 8].

Stochastic fluid flows naturally induce a type of ensemble averaging operator. This is the expectation operator E(⋅)\mathbb{E}(\cdot) over the probability space (Ω,F,P)(\Omega,\mathcal{F},\mathbb{P}) that supports the driving stochastic processes. E\mathbb{E} is linear, a projection and commutes with spatial derivative operations. Consequently, the expectation satisfies the required axioms of an average seen in GLM theory [1]. Additionally, one has the analogue of the zero mean fluctuations under expectation operator as martingales. More precisely, for a Rd\mathbb{R}^{d}-martingale YtY_t, the martingale property of Itô integral implies that

E[∫0tζsi dYsi]=0,\mathbb{E}\left[\int_{0}^{t}\zeta_{s}^{i}\,\mathrm{d}Y_{s}^{i}\right]=0,

for any Ft\mathcal{F}_{t}-adapted vector fields ζt=(ζti)i=1d∈X(M)\zeta_t=(\zeta_t^{i})_{i=1}^{d}\in\mathfrak{X}(\mathcal{M}).

In this work, we consider the averaging of stochastic fluid models using the expectation operator to construct probabilistic variants of sub-grid scale models for ideal fluids. Focusing on incompressible flows, the starting assumption is the semi-martingale Lagrangian fluid flow map gt∈SDiff⁡(M)g_t \in\operatorname{SDiff}(\mathcal{M}) similar to the derivation of stochastic transport fluid models in [13, 16]. The main contributions of this work are as follows. In section 2, we define the mean velocity by assuming a factorisation of the full flow map into the fluctuation and mean maps. Then, we use Magnus expansion of the velocity field of the composite flow such that it is expressed by vector fields and their actions only, in a similar fashion to [15, 18]. This way, the expectation operator acts on vector space quantities only and can be easily defined. In section 3, using tools from Malliavin calculus, we consider a normalised KE Lagrangian functional such that it is well-defined when the input velocity contains white noise terms. To obtain a dynamical equation of the mean flow, we first make dynamical assumptions on the fluctuations vector fields in assumptions 11 and 22. Then, we explicitly computed the averaged Lagrangian and derived a stochastically averaged version of the anisotropic Lagrangian averaged Euler’s equation [15] through an Euler-Poincaré variational principle.

Expansion of Kinetic Energy Lagrangian

Fix a smooth, compact, boundary-less Riemannian manifold (M,g)(\mathcal{M},\mathbf{g}) of dimension dd. Throughout this paper we equip (M,g)(\mathcal{M},\mathbf{g}) with the Levi-Civita connection ∇\nabla. Our aim is to derive deterministic dynamical equations for the mean flow velocity of Euler’s fluid flow from an averaged variational principle of the form,

S:=∫0TE[ℓ(vt)] dt,ℓ(vt):=12∫Mg(vt,vt) dμ,vt∈Xdiv⁡∇(M),(1)S := \int_{0}^{T} \mathbb{E}\left[\ell(v_t)\right]\,\mathrm{d}t,\qquad\ell(v_t) := \frac{1}{2}\int_{\mathcal{M}}\mathbf{g}(v_t,v_t)\,\mathrm{d}\mu,\qquad v_t \in\mathfrak{X}_{\operatorname{div}_{\nabla}}(\mathcal{M}), \tag*{(1)}

where ℓ\ell is the kinetic energy Lagrangian of divergence free vector fields. Here, we have denoted μ=vol⁡g\mu= \operatorname{vol}_{\mathbf{g}} the Riemannian volume. The covariant divergence div⁡∇\operatorname{div}_{\nabla} appearing in (1) and used throughout is defined for any tensor T∈T(M)T \in\mathcal{T}(\mathcal{M}) by div⁡∇(T):=tr⁡(∇T)\operatorname{div}_{\nabla}(T) := \operatorname{tr}(\nabla T) by contracting the covariant ∇i\nabla_i index with, by convention, the first (possibly after raising) contravariant index of TT. In the case of vector fields1, X∈Xdiv⁡∇(M)X \in\mathfrak{X}_{\operatorname{div}_{\nabla}}(\mathcal{M}) implies that

div⁡∇(X)=tr⁡(dxi⊗∂∂xj(∂Xj∂xi+XkΓikj))=∂Xi∂xi+XkΓiki=0.\operatorname{div}_{\nabla}(X) = \operatorname{tr}\left(\mathrm{d}x^i \otimes\frac{\partial}{\partial x^j}\left(\frac{\partial X^j}{\partial x^i}+X^k\Gamma^j_{ik}\right)\right) = \frac{\partial X^i}{\partial x^i}+X^k\Gamma^i_{ik}=0.

We assume the full Lagrangian flow map gt∈SDiff⁡(M)g_t \in\operatorname{SDiff}(\mathcal{M}) is a semi-martingale satisfying the stochastic differential equation,

dgt=ut(gt)dt+σt(gt)∘dWt,(2)\mathrm{d}g_t = u_t(g_t)\mathrm{d}t+\sigma_t(g_t)\circ\mathrm{d}W_t, \tag*{(2)}

such that vtv_t is formally associated with the vector field ddtgtgt−1∈Xdiv⁡∇(M)\frac{\mathrm{d}}{\mathrm{d}t}g_tg_t^{-1}\in\mathfrak{X}_{\operatorname{div}_{\nabla}}(\mathcal{M}). Throughout this paper, we denote by Wt=(Wt1,…,Wtn)W_t=(W_t^1,\ldots,W_t^n) as a nn-dimensional Brownian motion with identity covariance and abbreviate stochastic differentials such as σt(gt)∘dWt:=∑i=1nσti(gt)∘dWti\sigma_t(g_t)\circ\mathrm{d}W_t:=\sum_{i=1}^{n}\sigma_t^i(g_t)\circ\mathrm{d}W_t^i. To define the mean flow velocity, we assume the factorisation of the stochastic flow map gt∈SDiff⁡(M)g_t\in\operatorname{SDiff}(\mathcal{M}) into

gt=gtϵ=Ξtϵ∘gˉt,(3)g_t=g_t^\epsilon=\Xi_t^\epsilon\circ\bar{g}_t, \tag*{(3)}

where gˉ:[0,T]→SDiff⁡(M)\bar{g}:[0,T]\to\operatorname{SDiff}(\mathcal{M}) is the mean flow map and Ξˉtϵ\bar{\Xi}_t^\epsilon is a ϵ\epsilon-parametrised semi-martingale flow map Ξϵ:Ω×[0,T]×[0,1)→SDiff⁡(M)\Xi^\epsilon:\Omega\times[0,T]\times[0,1)\to\operatorname{SDiff}(\mathcal{M}), with following properties

Ξˉt0=e∀t∈[0,T],Ξ0ϵ=e∀ϵ∈[0,1),\bar{\Xi}_t^0=e\quad\forall t\in[0,T],\qquad\Xi_0^\epsilon=e\quad\forall\epsilon\in[0,1),

such that Ξtϵ\Xi_t^\epsilon is a flow of near identity diffeomorphism. From the mean flow decomposition (2), the canonical definition of the mean flow velocity field is the velocity assumed with the mean flow map uˉt:=ddtgˉtgˉt−1\bar{u}_t:=\frac{\mathrm{d}}{\mathrm{d}t}\bar{g}_t\bar{g}_t^{-1}. By assumption, uˉt\bar{u}_t is a deterministic object and it is natural to seek the dynamics of uˉt\bar{u}_t through the variational principle (1) where the stochastic fluctuation induced by the fluctuation map Ξtϵ\Xi_t^\epsilon is averaged. Via direct computation, one finds the vector field generated by gtϵg_t^\epsilon can be written in terms of the mean velocity uˉt\bar{u}_t

∘dgtϵgtϵ;−1=∘dΞtϵΞtϵ;−1+Ad⁡Ξtϵuˉt dt=σt∘dWt+uˉt dt,(4)\circ\mathrm{d}g_t^\epsilon g_t^{\epsilon;-1}=\circ\mathrm{d}\Xi_t^\epsilon\Xi_t^{\epsilon;-1}+\operatorname{Ad}_{\Xi_t^\epsilon}\bar{u}_t\,\mathrm{d}t=\sigma_t\circ\mathrm{d}W_t+\bar{u}_t\,\mathrm{d}t, \tag*{(4)}

where the notation ∘dgtϵ\circ\mathrm{d}g_t^\epsilon is there to denote the Stratonovich interpretation of the stochastic differential dgtϵ\mathrm{d}g_t^\epsilon. We observe that SDiff⁡(M)\operatorname{SDiff}(\mathcal{M}) does not necessarily possess a vector space structure and the term Ad⁡Ξtϵuˉt\operatorname{Ad}_{\Xi_t^\epsilon}\bar{u}_t is non-linear in Ξtϵ\Xi_t^\epsilon. In order to express terms of the form E[Ad⁡Ξtϵuˉt]\mathbb{E}[\operatorname{Ad}_{\Xi_t^\epsilon}\bar{u}_t] in the averaged Lagrangian in a workable form, we rewrite the fluctuations of uˉt\bar{u}_t defined in powers of ϵ\epsilon via Taylor expansion as,

∘dgtϵgtϵ;−1=∘dgtϵgtϵ;−1∣ϵ=0+ϵddϵ(∘dgtϵgtϵ;−1)∣ϵ=0+ϵ22d2dϵ2(∘dgtϵgtϵ;−1)∣ϵ=0+O(ϵ3):=∘dv0+ϵ∘dv′+ϵ22∘dv′′+O(ϵ3),(5)\begin{aligned} \circ\mathrm{d}g_t^\epsilon g_t^{\epsilon;-1} &=\left.\circ\mathrm{d}g_t^\epsilon g_t^{\epsilon;-1}\right|_{\epsilon=0} +\left.\epsilon\frac{\mathrm{d}}{\mathrm{d}\epsilon}\left(\circ\mathrm{d}g_t^\epsilon g_t^{\epsilon;-1}\right)\right|_{\epsilon=0} +\left.\frac{\epsilon^2}{2}\frac{\mathrm{d}^2}{\mathrm{d}\epsilon^2}\left(\circ\mathrm{d}g_t^\epsilon g_t^{\epsilon;-1}\right)\right|_{\epsilon=0} +\mathcal{O}(\epsilon^3)\\ &:=\circ\mathrm{d}v^0+\epsilon\circ\mathrm{d}v'+\frac{\epsilon^2}{2}\circ\mathrm{d}v''+\mathcal{O}(\epsilon^3), \tag*{(5)} \end{aligned}

such that expectation is applied to ∘dgtϵgtϵ;−1∈X(M)\circ\mathrm{d}g_t^\epsilon g_t^{\epsilon;-1}\in\mathfrak{X}(\mathcal{M}). The (formal) Lie algebra of divergence free vector fields is a linear space where these operations can be performed unambiguously. In the expansion (5), we will express the first and second order fluctuation fields ∘dv′\circ\mathrm{d}v' and ∘dv′′\circ\mathrm{d}v'' from the deformations of Ξtϵ\Xi_t^\epsilon

ddϵ∣ϵ=0Ξtϵ=:wt,d2dϵ2∣ϵ=0Ξtϵ:=χt,wt,χt∈Xdiv⁡∇(M),∀t∈[0,T].(6)\left.\frac{\mathrm{d}}{\mathrm{d}\epsilon}\right|_{\epsilon=0}\Xi_t^\epsilon=:w_t,\qquad\left.\frac{\mathrm{d}^2}{\mathrm{d}\epsilon^2}\right|_{\epsilon=0}\Xi_t^\epsilon:=\chi_t,\qquad w_t,\chi_t\in\mathfrak{X}_{\operatorname{div}_{\nabla}}(\mathcal{M}),\quad\forall t\in[0,T]. \tag*{(6)}

Here, the deformation vector fields wtw_t and χt\chi_t are assumed to be semi-martingales which possess a unique decomposition into finite variation and (local) martingale parts.

Lemma 1. In terms of the deformation vector fields wtw_t and χt\chi_t, the fluctuation stochastic vector fields dv0\mathrm{d}v^{0}, dv′\mathrm{d}v^{\prime} and dv′′\mathrm{d}v^{\prime\prime} can be expressed as

∘dv0=u‾t dt,(7)\circ\mathrm{d}v^{0}=\overline{u}_t\,\mathrm{d}t, \tag*{(7)}
∘dv′=∘dwt+ad⁡wtu‾t dt,(8)\circ\mathrm{d}v^{\prime}=\circ\mathrm{d}w_t+\operatorname{ad}_{w_t}\overline{u}_t\,\mathrm{d}t, \tag*{(8)}
∘dv′′=ad⁡wtad⁡wtu‾t dt+ad⁡χt−∇wtwtu‾t dt+∘d(χt−∇wtwt)+ad⁡wt∘dwt,(9)\circ\mathrm{d}v^{\prime\prime}=\operatorname{ad}_{w_t}\operatorname{ad}_{w_t}\overline{u}_t\,\mathrm{d}t+\operatorname{ad}_{\chi_t-\nabla_{w_t}w_t}\overline{u}_t\,\mathrm{d}t+\circ\mathrm{d}\left(\chi_t-\nabla_{w_t}w_t\right)+\operatorname{ad}_{w_t}\circ\mathrm{d}w_t, \tag*{(9)}

where ∇\nabla is a fixed torsion-free connection. For convenience we will use the Levi-Civita connection induced by the metric gg.

Proof. The O(1)\mathcal{O}(1) term dv0\mathrm{d}v^{0} can be deduced immediately from the the near identity property Ξ0ϵ≡e\Xi^{\epsilon}_0\equiv e and Ad⁡eu‾t=u‾t\operatorname{Ad}_{e}\overline{u}_t=\overline{u}_t,

dv0=dgt0gt0;−1=Ad⁡Ξt0u‾t dt+∘dΞt0Ξt0;−1=u‾t dt.\mathrm{d}v^{0}=\mathrm{d}g^{0}_t g_t^{0;-1}=\operatorname{Ad}_{\Xi^{0}_t}\overline{u}_t\,\mathrm{d}t+\circ\mathrm{d}\Xi^{0}_t\Xi_t^{0;-1}=\overline{u}_t\,\mathrm{d}t.

To calculate higher order terms, we differentiate the decomposition, where the first term is a direct application of the definition of the adjoint representation (Lie chain rule),

∘dv′=ddϵ∣ϵ=0∘dvϵ=ddϵ∣ϵ=0(Ad⁡Ξtϵu‾t dt+∘dΞtϵΞtϵ;−1),ddϵ∣ϵ=0Ad⁡Ξtϵu‾t:=ad⁡wtu‾t.\circ\mathrm{d}v^{\prime}=\left.\frac{\mathrm{d}}{\mathrm{d}\epsilon}\right|_{\epsilon=0}\circ\mathrm{d}v^{\epsilon} =\left.\frac{\mathrm{d}}{\mathrm{d}\epsilon}\right|_{\epsilon=0} \left(\operatorname{Ad}_{\Xi^\epsilon_t}\overline{u}_t\,\mathrm{d}t+\circ\mathrm{d}\Xi^\epsilon_t\Xi_t^{\epsilon;-1}\right), \qquad \left.\frac{\mathrm{d}}{\mathrm{d}\epsilon}\right|_{\epsilon=0}\operatorname{Ad}_{\Xi^\epsilon_t}\overline{u}_t:=\operatorname{ad}_{w_t}\overline{u}_t.

To take the derivative of ∘dΞtϵΞtϵ;−1\circ\mathrm{d}\Xi^\epsilon_t\Xi_t^{\epsilon;-1}, we express Ξtϵ=exp⁡(Ωtϵ)\Xi^\epsilon_t=\exp(\Omega^\epsilon_t) and differentiate Ωtϵ\Omega^\epsilon_t through the chain rule on the exponential map. This procedure is known as a Magnus expansion [14]. Although we work formally, the application of this type of theory to the infinite dimensional group of volume preserving diffeomorphisms can be found in [2, 4, 5]. We state the following formulae that can be obtained through known derivatives of the exponential map at the identity [11],

∘dΞtϵΞtϵ;−1=∑k=0∞1(k+1)!ad⁡Ωtϵk∘dΩtϵ,ad⁡Ωtϵk(⋅):=ad⁡Ωtϵ(ad⁡Ωtϵk−1(⋅)),ad⁡Ωtϵ0:=I.(10)\circ\mathrm{d}\Xi^\epsilon_t\Xi_t^{\epsilon;-1} =\sum_{k=0}^{\infty}\frac{1}{(k+1)!}\operatorname{ad}_{\Omega^\epsilon_t}^{k}\circ\mathrm{d}\Omega^\epsilon_t, \qquad \operatorname{ad}_{\Omega^\epsilon_t}^{k}(\cdot):=\operatorname{ad}_{\Omega^\epsilon_t}\left(\operatorname{ad}_{\Omega^\epsilon_t}^{k-1}(\cdot)\right), \qquad \operatorname{ad}_{\Omega^\epsilon_t}^{0}:=I. \tag*{(10)}

The time derivative ∘dΩtϵ\circ\mathrm{d}\Omega^\epsilon_t is expressible as the series [3]

∘dΩtϵ=∑k=0∞Bkk!ad⁡Ωtϵk(∘dΞtϵΞtϵ;−1),(11)\circ\mathrm{d}\Omega^\epsilon_t =\sum_{k=0}^{\infty}\frac{B_k}{k!}\operatorname{ad}_{\Omega^\epsilon_t}^{k}\left(\circ\mathrm{d}\Xi^\epsilon_t\Xi_t^{\epsilon;-1}\right), \tag*{(11)}

where BkB_k denotes the Bernoulli numbers with B1=−12B_1=-\frac{1}{2}. From the initial assumption Ξt0≡e\Xi^0_t\equiv e it must be the case that Ωt0≡∘dΩt0≡∘dΞt0≡0\Omega^0_t\equiv\circ\mathrm{d}\Omega^0_t\equiv\circ\mathrm{d}\Xi^0_t\equiv0 from differentiating a constant.

Using the Magnus generator Ωtϵ\Omega^\epsilon_t, we can now compute first ϵ\epsilon-derivatives of the velocity dΞtϵΞtϵ;−1\mathrm{d}\Xi^\epsilon_t\Xi_t^{\epsilon;-1} by termwise differentiation of the power series (10),

∂∂ϵ∣ϵ=0∘dΞtϵΞtϵ;−1=∑k=0∞1(k+1)!∂∂ϵ∣ϵ=0(ad⁡Ωtϵk∘dΩtϵ)=∂∂ϵ∣ϵ=0∘dΩtϵ=:∘dδΩt.(12)\left.\frac{\partial}{\partial\epsilon}\right|_{\epsilon=0}\circ\mathrm{d}\Xi^\epsilon_t\Xi_t^{\epsilon;-1} =\sum_{k=0}^{\infty}\frac{1}{(k+1)!}\left.\frac{\partial}{\partial\epsilon}\right|_{\epsilon=0} \left(\operatorname{ad}_{\Omega^\epsilon_t}^{k}\circ\mathrm{d}\Omega^\epsilon_t\right) =\left.\frac{\partial}{\partial\epsilon}\right|_{\epsilon=0}\circ\mathrm{d}\Omega^\epsilon_t =:\circ\mathrm{d}\delta\Omega_t. \tag*{(12)}

Note that only the k=0k=0 term contributes due to the fact Ωt0=∘dΩt0=0\Omega^0_t=\circ\mathrm{d}\Omega^0_t=0. We may relate δΩt\delta\Omega_t to the first order fluctuation wtw_t using the k=0k=0 term of the Magnus series (11),

∘dδΩt=δ∘dΩt=∂∂ϵ∣ϵ=0(∘dΞtϵΞtϵ;−1)=∘d(δΞtΞt0;−1)+∘dΞt0(δΞt−1)=∘dwt,δΩt=δΞt:=wt.(13)\circ\mathrm{d}\delta\Omega_t=\delta\circ\mathrm{d}\Omega_t =\left.\frac{\partial}{\partial\epsilon}\right|_{\epsilon=0} \left(\circ\mathrm{d}\Xi^\epsilon_t\Xi_t^{\epsilon;-1}\right) =\circ\mathrm{d}\left(\delta\Xi_t\Xi_t^{0;-1}\right)+\circ\mathrm{d}\Xi^0_t\left(\delta\Xi_t^{-1}\right) =\circ\mathrm{d}w_t, \qquad \delta\Omega_t=\delta\Xi_t:=w_t. \tag*{(13)}

Consequently, we recover ∘dv0=∘dwt+ad⁡wtu‾t dt\circ\mathrm{d}v^{0}=\circ\mathrm{d}w_t+\operatorname{ad}_{w_t}\overline{u}_t\,\mathrm{d}t.

In order to compute the second order fluctuation of dvϵ\mathrm{d}v^\epsilon we require the connection to take second derivatives. We first differentiate Ad⁡Ξtϵu‾\operatorname{Ad}_{\Xi^\epsilon_t}\overline{u} through twice application of the Lie chain rule, where the shorthand (⋅)′(\cdot)^{\prime} denotes ∂∂ϵ(⋅)\frac{\partial}{\partial\epsilon}(\cdot),

∂2∂ϵ2∣ϵ=0Ad⁡Ξtϵu‾=∂∂ϵ∣ϵ=0ad⁡Ξtϵ′Ξtϵ;−1Ad⁡Ξtϵu‾ dt=(ad⁡Ξtϵ′′Ξtϵ;−1Ad⁡Ξtϵu‾ dt+ad⁡Ξtϵ′Ξtϵ;−1ad⁡Ξtϵ′Ξtϵ;−1Ad⁡Ξtϵu‾ dt)∣ϵ=0=ad⁡χt−∇wtwtu‾t dt+ad⁡wtad⁡wtu‾t dt,\begin{aligned} \left.\frac{\partial^2}{\partial\epsilon^2}\right|_{\epsilon=0}\operatorname{Ad}_{\Xi^\epsilon_t}\overline{u} =\left.\frac{\partial}{\partial\epsilon}\right|_{\epsilon=0} \operatorname{ad}_{\Xi_t^{\epsilon\prime}\Xi_t^{\epsilon;-1}} \operatorname{Ad}_{\Xi^\epsilon_t}\overline{u}\,\mathrm{d}t \\ &=\left. \left( \operatorname{ad}_{\Xi_t^{\epsilon\prime\prime}\Xi_t^{\epsilon;-1}} \operatorname{Ad}_{\Xi^\epsilon_t}\overline{u}\,\mathrm{d}t +\operatorname{ad}_{\Xi_t^{\epsilon\prime}\Xi_t^{\epsilon;-1}} \operatorname{ad}_{\Xi_t^{\epsilon\prime}\Xi_t^{\epsilon;-1}} \operatorname{Ad}_{\Xi^\epsilon_t}\overline{u}\,\mathrm{d}t \right)\right|_{\epsilon=0} \\ &=\operatorname{ad}_{\chi_t-\nabla_{w_t}w_t}\overline{u}_t\,\mathrm{d}t +\operatorname{ad}_{w_t}\operatorname{ad}_{w_t}\overline{u}_t\,\mathrm{d}t, \end{aligned}

@page 3/9 starts_mid=0 ends_mid=1 where ξtϵ;′′=Ξtϵ;′′Ξtϵ;−1−∇Ξtϵ;′′Ξtϵ;−1Ξtϵ;′′\xi_t^{\epsilon;\prime\prime}=\Xi_t^{\epsilon;\prime\prime}\Xi_t^{\epsilon;-1}-\nabla_{\Xi_t^{\epsilon;\prime\prime}\Xi_t^{\epsilon;-1}}\Xi_t^{\epsilon;\prime\prime}. In particular, the term ∇wtwt\nabla_{w_t}w_t arises when one attempts to compute a subsequent variation of Ξtϵ;′′Ξtϵ;−1\Xi_t^{\epsilon;\prime\prime}\Xi_t^{\epsilon;-1}. With the Magnus expansion, one can write this term as δ2Ωt\delta^2\Omega_t without use of a connection, however a connection is required to relate this quantity to the fluctuations χt,wt\chi_t,w_t, see the Remark 1 below.

For the second derivative of dΞtϵΞtϵ;−1\mathrm{d}\Xi_t^\epsilon\Xi_t^{\epsilon;-1} it suffices to check the k=0,1k=0,1 terms in the series (11),

∂2∂ϵ2∣ϵ=0∘dΞtϵΞtϵ;−1=∑k=0∞1(k+1)!∂2∂ϵ2∣ϵ=0(ad⁡Ωtϵk∘dΩtϵ)=∂2∂ϵ2∣ϵ=0(∘dΩtϵ+12ad⁡Ωtϵ∘dΩtϵ)=∘dδ2Ωt+ad⁡δΩt∘dδΩt.(14)\left.\frac{\partial^2}{\partial\epsilon^2}\right|_{\epsilon=0}\circ\mathrm{d}\Xi_t^\epsilon\Xi_t^{\epsilon;-1} =\sum_{k=0}^{\infty}\frac{1}{(k+1)!}\left.\frac{\partial^2}{\partial\epsilon^2}\right|_{\epsilon=0}\left(\operatorname{ad}_{\Omega_t^\epsilon}^{k}\circ\mathrm{d}\Omega_t^\epsilon\right) =\left.\frac{\partial^2}{\partial\epsilon^2}\right|_{\epsilon=0}\left(\circ\mathrm{d}\Omega_t^\epsilon+\frac{1}{2}\operatorname{ad}_{\Omega_t^\epsilon}\circ\mathrm{d}\Omega_t^\epsilon\right) =\circ\mathrm{d}\delta^2\Omega_t+\operatorname{ad}_{\delta\Omega_t}\circ\mathrm{d}\delta\Omega_t . \tag*{(14)}

Using the relation δΩt=wt\delta\Omega_t=w_t and δ2Ωt=χt−∇wtwt\delta^2\Omega_t=\chi_t-\nabla_{w_t}w_t (proven below), we thus obtain,

∘dv′′=ad⁡wtad⁡wtu‾t dt+ad⁡χt−∇wtwtu‾t dt+∘d(χt−∇wtwt)+ad⁡wt∘dwt,\circ\mathrm{d}v^{\prime\prime}=\operatorname{ad}_{w_t}\operatorname{ad}_{w_t}\overline{u}_t\,\mathrm{d}t+\operatorname{ad}_{\chi_t-\nabla_{w_t}w_t}\overline{u}_t\,\mathrm{d}t+\circ\mathrm{d}\left(\chi_t-\nabla_{w_t}w_t\right)+\operatorname{ad}_{w_t}\circ\mathrm{d}w_t ,

as was originally claimed.

Remark 1 (Connection dependence and matrix Lie groups). The expressions (12), (14) define equations on any Lie algebra in the absence of any additional structure. As a consequence of ∘dΞt0=0\circ\mathrm{d}\Xi_t^0=0, (13) can always be used to relate the variations of Ω\Omega and Ξ\Xi. The second order term δ2Ωt\delta^2\Omega_t differs, as we must define an action of Lie algebra terms using external structures such as a connection. In order to proceed, one formally writes,

∘dδ2Ωt=δ2∘dΩt=∂2∂ϵ2∣ϵ=0(∘dΞtϵΞtϵ;−1−12ad⁡Ωtϵ∘dΞtϵΞtϵ;−1)\circ\mathrm{d}\delta^2\Omega_t=\delta^2\circ\mathrm{d}\Omega_t=\left.\frac{\partial^2}{\partial\epsilon^2}\right|_{\epsilon=0}\left(\circ\mathrm{d}\Xi_t^\epsilon\Xi_t^{\epsilon;-1}-\frac{1}{2}\operatorname{ad}_{\Omega_t^\epsilon}\circ\mathrm{d}\Xi_t^\epsilon\Xi_t^{\epsilon;-1}\right)
=∂∂ϵ∣ϵ=0(∘d[∂Ξtϵ∂ϵΞtϵ;−1]+ad⁡∂Ξtϵ∂ϵΞtϵ;−1∘dΞtϵΞtϵ;−1)−ad⁡δΩtδ(∘dΞtΞt−1)=\left.\frac{\partial}{\partial\epsilon}\right|_{\epsilon=0}\left(\circ\mathrm{d}\left[\frac{\partial\Xi_t^\epsilon}{\partial\epsilon}\Xi_t^{\epsilon;-1}\right]+\operatorname{ad}_{\frac{\partial\Xi_t^\epsilon}{\partial\epsilon}\Xi_t^{\epsilon;-1}}\circ\mathrm{d}\Xi_t^\epsilon\Xi_t^{\epsilon;-1}\right)-\operatorname{ad}_{\delta\Omega_t}\delta\left(\circ\mathrm{d}\Xi_t\Xi_t^{-1}\right)
=∘d(δ2ΞtΞt0;−1+δΞtδΞt−1)+ad⁡δΞtδ(∘dΞtΞt−1)−ad⁡δΩtδ(∘dΞtΞt−1)=∘d(χt+δΞtδΞt−1).(15)=\circ\mathrm{d}\left(\delta^2\Xi_t\Xi_t^{0;-1}+\delta\Xi_t\delta\Xi_t^{-1}\right)+\operatorname{ad}_{\delta\Xi_t}\delta\left(\circ\mathrm{d}\Xi_t\Xi_t^{-1}\right)-\operatorname{ad}_{\delta\Omega_t}\delta\left(\circ\mathrm{d}\Xi_t\Xi_t^{-1}\right)=\circ\mathrm{d}\left(\chi_t+\delta\Xi_t\delta\Xi_t^{-1}\right). \tag*{(15)}

When g\mathfrak{g} is finite dimensional, one may appeal to Ado’s Theorem to embed g⊂GL(N)\mathfrak{g}\subset GL(N) for some N∈NN\in\mathbb{N} and give meaning to δΞtδΞt−1=−δΞtΞt0;−1δΞtΞt0;−1=−δΞtδΞt=−w2\delta\Xi_t\delta\Xi_t^{-1}=-\delta\Xi_t\Xi_t^{0;-1}\delta\Xi_t\Xi_t^{0;-1}=-\delta\Xi_t\delta\Xi_t=-w^2 as a product of matrices. One can then show the second order expansion cancels significantly as,

∘dv′′=ad⁡wtad⁡wtu‾t dt+ad⁡χt−w2u‾t dt+∘d(χt−w2)+ad⁡wt∘dwt=∘dχt+ad⁡χtu‾t dt−2wt(∘dv′),\circ\mathrm{d}v^{\prime\prime}=\operatorname{ad}_{w_t}\operatorname{ad}_{w_t}\overline{u}_t\,\mathrm{d}t+\operatorname{ad}_{\chi_t-w^2}\overline{u}_t\,\mathrm{d}t+\circ\mathrm{d}\left(\chi_t-w^2\right)+\operatorname{ad}_{w_t}\circ\mathrm{d}w_t=\circ\mathrm{d}\chi_t+\operatorname{ad}_{\chi_t}\overline{u}_t\,\mathrm{d}t-2w_t\left(\circ\mathrm{d}v^{\prime}\right),

under the property that ad⁡\operatorname{ad} reduces to the commutator of GL(N)GL(N) matrices within this embedding. In the case of the diffeomorphism group, a concrete representation of (14) in terms of δΞt\delta\Xi_t requires identifying δΞtδΞt−1:=−∇wtwt\delta\Xi_t\delta\Xi_t^{-1}:=-\nabla_{w_t}w_t as a covariant derivative. In Euclidean space, we recover consistency with the ordinary chain rule where the connection is canonical,

∂∂ϵ∣ϵ=0δΞt∘Ξtϵ;−1=[D(δΞt)∘Ξtϵ;−1]⋅∂Ξtϵ;−1∂ϵ∣ϵ=0=−[D(δΞt)∘Ξtϵ;−1]⋅[D(Ξtϵ)∘Ξtϵ;−1]−1⋅(∂Ξtϵ∂ϵ∘Ξtϵ;−1)∣ϵ=0\frac{\partial}{\partial\epsilon}|_{\epsilon=0}\delta\Xi_t\circ\Xi_t^{\epsilon;-1} =[D(\delta\Xi_t)\circ\Xi_t^{\epsilon;-1}]\cdot\frac{\partial\Xi_t^{\epsilon;-1}}{\partial\epsilon}|_{\epsilon=0} =-[D(\delta\Xi_t)\circ\Xi_t^{\epsilon;-1}]\cdot[D(\Xi_t^\epsilon)\circ\Xi_t^{\epsilon;-1}]^{-1}\cdot(\frac{\partial\Xi_t^\epsilon}{\partial\epsilon}\circ\Xi_t^{\epsilon;-1})|_{\epsilon=0}
=−D(δΞt)⋅δΞt=−∇δΞtEδΞt.=-D(\delta\Xi_t)\cdot\delta\Xi_t=-\nabla_{\delta\Xi_t}^{E}\delta\Xi_t .

In the above, we have denoted by ∇E\nabla^E the Euclidean connection and DD the total (Frechêt) derivative for maps Rn↦Rn\mathbb{R}^n\mapsto\mathbb{R}^n, with the second equality was derived from the inverse function theorem. In curved geometries, the choice of ∇\nabla is non-canonical, although the Levi-Civita connection is a natural choice we take throughout.

Lemma 2. The second order fluctuation vector fields dv′′\mathrm{d}v^{\prime\prime} may be written equivalently as

∘dv′′=∘dχt+ad⁡χtu‾t dt−∇∇u‾t(wt,wt) dt+R(u‾t,wt)wt dt−2∇wt∘dv′.(16)\circ\mathrm{d}v^{\prime\prime}=\circ\mathrm{d}\chi_t+\operatorname{ad}_{\chi_t}\overline{u}_t\,\mathrm{d}t-\nabla\nabla_{\overline{u}_t}(w_t,w_t)\,\mathrm{d}t+R(\overline{u}_t,w_t)w_t\,\mathrm{d}t-2\nabla_{w_t}\circ\mathrm{d}v^{\prime}. \tag*{(16)}

where the Riemann curvature tensor R=R ijαβdxi∧dxj⊗dxα⊗∂∂xβR=R^\beta_{\ ij\alpha}\mathrm{d}x^i\wedge\mathrm{d}x^j\otimes\mathrm{d}x^\alpha\otimes\frac{\partial}{\partial x^\beta} is a tensor field of type (3,1)(3,1) and a section of Λ2(T∗M)⊗End⁡(TM)\Lambda^2(T^*M)\otimes\operatorname{End}(TM).

Proof. This follows from repeated use of the following identities,

ad⁡XY=−[X,Y]=−∇XY+∇YX,∇∇Z(X,Y)=∇X∇YZ−∇∇XYZ,∀X,Y,Z∈X(D),(17)\operatorname{ad}_{X}Y=-[X,Y]=-\nabla_{X}Y+\nabla_{Y}X,\qquad\nabla\nabla Z(X,Y)=\nabla_{X}\nabla_{Y}Z-\nabla_{\nabla_{X}Y}Z,\qquad\forall X,Y,Z\in\mathfrak{X}(\mathcal{D}), \tag*{(17)}

which follows directly from the definition of being a torsion-free connection and the Leibniz property. Here, we have the covariant Hessian operator ∇∇\nabla\nabla defined on tensor T∈T(M)T\in\mathcal{T}(\mathcal{M}) of type (p,q)(p,q) by ∇∇T=∇i∇jT dxi⊗dxj\nabla\nabla T=\nabla_{i}\nabla_{j}T\,dx^{i}\otimes dx^{j}. Inserting (17) into the equation for dv′′dv^{\prime\prime} thus gives,

∘dv′′=ad⁡wtad⁡wtu‾t dt+ad⁡χt−∇wtwtu‾t dt+∘d(χt−∇wtwt)+ad⁡wt∘dv′=ad⁡χt−∇wtwtu‾t dt+∘d(χt−∇wtwt)+ad⁡wt∘dv′=∘dχt+ad⁡χtu‾t dt−(∇ad⁡wtu‾twt+∇wt∘dwt+ad⁡∇wtwtu‾t dt)+ad⁡wt∘dv′=∘dχt+ad⁡χtu‾t dt+∇ad⁡wtu‾twt dt+∇wt(ad⁡wtu‾t) dt−ad⁡∇wtwtu‾t dt−2∇wt∘dv′=∘dχt+ad⁡χtu‾t dt+∇−∇wtu‾t+∇u‾twtwt dt+∇wt(−∇wtu‾t+∇u‾twt) dt+∇∇wtwtu‾t dt−∇u‾t∇wtwt dt−2∇wt∘dv′=∘dχt+ad⁡χtu‾t dt−∇∇u‾t(wt,wt) dt−∇∇wt(u‾t,wt) dt+∇∇wt(wt,u‾t) dt−2∇wt∘dv′=∘dχt+ad⁡χtu‾t dt−∇∇u‾t(wt,wt) dt+R(u‾t,wt)wt dt−2∇wt∘dv′.\begin{aligned} \circ dv^{\prime\prime}&=\operatorname{ad}_{w_t}\operatorname{ad}_{w_t}\overline{u}_t\,dt+\operatorname{ad}_{\chi_t-\nabla_{w_t}w_t}\overline{u}_t\,dt+\circ d(\chi_t-\nabla_{w_t}w_t)+\operatorname{ad}_{w_t}\circ dv^{\prime}\\ &=\operatorname{ad}_{\chi_t-\nabla_{w_t}w_t}\overline{u}_t\,dt+\circ d(\chi_t-\nabla_{w_t}w_t)+\operatorname{ad}_{w_t}\circ dv^{\prime}\\ &=\circ d\chi_t+\operatorname{ad}_{\chi_t}\overline{u}_t\,dt-\left(\nabla_{\operatorname{ad}_{w_t}\overline{u}_t}w_t+\nabla_{w_t}\circ dw_t+\operatorname{ad}_{\nabla_{w_t}w_t}\overline{u}_t\,dt\right)+\operatorname{ad}_{w_t}\circ dv^{\prime}\\ &=\circ d\chi_t+\operatorname{ad}_{\chi_t}\overline{u}_t\,dt+\nabla_{\operatorname{ad}_{w_t}\overline{u}_t}w_t\,dt+\nabla_{w_t}(\operatorname{ad}_{w_t}\overline{u}_t)\,dt-\operatorname{ad}_{\nabla_{w_t}w_t}\overline{u}_t\,dt-2\nabla_{w_t}\circ dv^{\prime}\\ &=\circ d\chi_t+\operatorname{ad}_{\chi_t}\overline{u}_t\,dt+\nabla_{-\nabla_{w_t}\overline{u}_t+\nabla_{\overline{u}_t}w_t}w_t\,dt+\nabla_{w_t}(-\nabla_{w_t}\overline{u}_t+\nabla_{\overline{u}_t}w_t)\,dt+\nabla_{\nabla_{w_t}w_t}\overline{u}_t\,dt\\ &\qquad-\nabla_{\overline{u}_t}\nabla_{w_t}w_t\,dt-2\nabla_{w_t}\circ dv^{\prime}\\ &=\circ d\chi_t+\operatorname{ad}_{\chi_t}\overline{u}_t\,dt-\nabla\nabla_{\overline{u}_t}(w_t,w_t)\,dt-\nabla\nabla_{w_t}(\overline{u}_t,w_t)\,dt+\nabla\nabla_{w_t}(w_t,\overline{u}_t)\,dt-2\nabla_{w_t}\circ dv^{\prime}\\ &=\circ d\chi_t+\operatorname{ad}_{\chi_t}\overline{u}_t\,dt-\nabla\nabla_{\overline{u}_t}(w_t,w_t)\,dt+R(\overline{u}_t,w_t)w_t\,dt-2\nabla_{w_t}\circ dv^{\prime}. \end{aligned}

In the last line, we picked up a curvature term as a consequence of the identity,

R(X,Y)Z:=−∇X∇YZ+∇Y∇XZ+∇[X,Y]Z=−∇∇Z(X,Y)+∇∇Z(Y,X).R(X,Y)Z:=-\nabla_X\nabla_YZ+\nabla_Y\nabla_XZ+\nabla_{[X,Y]}Z=-\nabla\nabla Z(X,Y)+\nabla\nabla Z(Y,X).

The curvature term R(u‾,w)wR(\overline{u},w)w vanishes in flat geometries, but otherwise may be expressed as R(u‾,w)w=R(u‾,⋅):(w⊗w)R(\overline{u},w)w=R(\overline{u},\cdot):(w\otimes w) as a contraction with a (2,1)(2,1) tensor field R(u‾,⋅)=Rijαβu‾idxj⊗dxα⊗∂∂xβR(\overline{u},\cdot)=R^{\beta}_{ij\alpha}\overline{u}^{i}dx^{j}\otimes dx^{\alpha}\otimes\frac{\partial}{\partial x^{\beta}} with a (2,0)(2,0) tensor field w⊗ww\otimes w.

With the alternative formulation of Lemma 2, the full expansion of ∘dgtϵgtϵ;−1\circ dg_t^{\epsilon}g_t^{\epsilon;-1} to O(ϵ2)O(\epsilon^{2}) is given by,

∘dgtϵgtϵ;−1=u‾t dt+ϵ(∘dwt+ad⁡wtu‾t dt)+ϵ22(∘dχt+ad⁡χtu‾t dt−∇∇u‾t(wt,wt) dt+R(u‾t,wt)wt dt−2∇wt∘dv′)+O(ϵ3).(18)\begin{aligned} \circ dg_t^{\epsilon}g_t^{\epsilon;-1} &=\overline{u}_t\,dt+\epsilon\left(\circ dw_t+\operatorname{ad}_{w_t}\overline{u}_t\,dt\right)\\ &\quad+\frac{\epsilon^{2}}{2}\left(\circ d\chi_t+\operatorname{ad}_{\chi_t}\overline{u}_t\,dt-\nabla\nabla_{\overline{u}_t}(w_t,w_t)\,dt+R(\overline{u}_t,w_t)w_t\,dt-2\nabla_{w_t}\circ dv^{\prime}\right)+O(\epsilon^{3}). \tag*{(18)} \end{aligned}

Remark 2 (Expectation of advected quantities). Advected quantities a‾t∈T(M)\overline{a}_t\in\mathcal{T}(\mathcal{M}) of the mean flow u‾t\overline{u}_t can also be perturbed in similar fashion to the perturbed stochastic velocity field ∘dgtϵgtϵ;−1\circ dg_t^{\epsilon}g_t^{\epsilon;-1} and expanded in (18) by considering the quantity atϵ:=Ξt∗ϵa‾t=gt∗ϵa0a_t^{\epsilon}:=\Xi_{t*}^{\epsilon}\overline{a}_t=g_{t*}^{\epsilon}a_0. Applying the standard Lie chain rule in the ϵ\epsilon-parameter,

ddϵ(Ξt∗ϵatϵ)=Ξt∗ϵddϵatϵ−LΞtϵgtϵ;−1atϵ,\frac{d}{d\epsilon}\left(\Xi_{t*}^{\epsilon}a_t^{\epsilon}\right)=\Xi_{t*}^{\epsilon}\frac{d}{d\epsilon}a_t^{\epsilon}-\mathcal{L}_{\Xi_t^{\epsilon}g_t^{\epsilon;-1}a_t^{\epsilon}},

where L:X(M)×T(M)→T(M)\mathcal{L}:\mathfrak{X}(\mathcal{M})\times\mathcal{T}(\mathcal{M})\to\mathcal{T}(\mathcal{M}) is the Lie derivative operator, we have the expansion

atϵ=a‾t−ϵLwta‾t+ϵ22(−Lχt−∇wtwta‾t+LwtLwta‾t)+O(ϵ3).(19)a_t^{\epsilon}=\overline{a}_t-\epsilon\mathcal{L}_{w_t}\overline{a}_t+\frac{\epsilon^{2}}{2}\left(-\mathcal{L}_{\chi_t-\nabla_{w_t}w_t}\overline{a}_t+\mathcal{L}_{w_t}\mathcal{L}_{w_t}\overline{a}_t\right)+O(\epsilon^{3}). \tag*{(19)}

Closure via variational principle

The direct substitution of the expansion (18) into the Lagrangian (1), dropping all cubic and higher order terms, results in the following formal action,

S=E[∫0T∫M12g(u‾t,u‾t) dt+ϵ∫0T∫Mg(u‾t,∘dwt+ad⁡wtu‾t dt)+∫0T∫Mϵ22g(u‾t,∘dχt+ad⁡χtu‾t dt−∇∇u‾t(wt,wt) dt+R(u‾t,wt)wt dt−2∇wt∘dv′)+∫0T∫Mϵ22g(w˙t+ad⁡wtu‾t,w˙t+ad⁡wtu‾t) dt]:=E[I1+I2+I3+I4].(20)\begin{aligned} S=\mathbb{E}\Bigg[&\int_{0}^{T}\int_{\mathcal{M}}\frac{1}{2}g(\overline{u}_t,\overline{u}_t)\,dt+\epsilon\int_{0}^{T}\int_{\mathcal{M}}g(\overline{u}_t,\circ dw_t+\operatorname{ad}_{w_t}\overline{u}_t\,dt)\\ &+\int_{0}^{T}\int_{\mathcal{M}}\frac{\epsilon^{2}}{2}g\left(\overline{u}_t,\circ d\chi_t+\operatorname{ad}_{\chi_t}\overline{u}_t\,dt-\nabla\nabla_{\overline{u}_t}(w_t,w_t)\,dt+R(\overline{u}_t,w_t)w_t\,dt-2\nabla_{w_t}\circ dv^{\prime}\right)\\ &+\int_{0}^{T}\int_{\mathcal{M}}\frac{\epsilon^{2}}{2}g\left(\dot{w}_t+\operatorname{ad}_{w_t}\overline{u}_t,\dot{w}_t+\operatorname{ad}_{w_t}\overline{u}_t\right)\,dt\Bigg]:=\mathbb{E}[I_1+I_2+I_3+I_4]. \tag*{(20)} \end{aligned}

Here, I1I_1 is the kinetic energy of the mean velocity u‾t\overline{u}_t; I2I_2 and I3I_3 are mixed products with stochastic fluctuations wtw_t and χt\chi_t, of the mean velocity u‾t\overline{u}_t, understood as stochastic integrals. The kinetic energy of the fluctuations I4I_4 is ill-defined due to terms representing the formal product of white noise and has no classical interpretation.

In existing literature of stochastic variational principles where the action contains explicit integration against semi-martingales, e.g., [4, 7, 13, 12], terms such as I4I_4 are typically dropped. Our approach is to instead retain these terms under an average, which allow for a renormalisation of I4I_4 into a meaningful quantity. The method we employ is to interpret I4I_4 using the Wick product, which defines products of white noise through Malliavin calculus.

We very briefly discuss this construction. Since our usage is only elementary, we do not dwell substantially on definitions. For precise details, we refer the reader to [?] and [17]. One defines a white noise probability space (Ω,B,μ)(\Omega,\mathcal{B},\mu) of random distributions inside a sample space of Rn\mathbb{R}^n valued tempered distributions Ω:=S∗(R+;Rn)\Omega:= \mathcal{S}^*(\mathbb{R}_+;\mathbb{R}^n). A probability measure and Borel sigma algebra exists on this vector space via the Bochner-Milnos theorem.

One may then define L2(Ω)L^2(\Omega)-integrable random variables taking values in the Schwartz class which may be characterised by convergent Itô-Weiner chaos expansions. The elements of L2(Ω)L^2(\Omega) with a rapidly decreasing series expansion are known as the Hida test functions S(Ω)\mathcal{S}(\Omega), and elements its dual space S∗(Ω)\mathcal{S}^*(\Omega) are known as Hida distributions. Within this space, a distribution valued stochastic process W˙t\dot{W}_t may be defined. Crucial for us is that S∗(Ω)\mathcal{S}^*(\Omega) contains a closed, associative, commutative product that is defined by the formal Cauchy product of the series defined in an L2(Ω)L^2(\Omega) orthogonal basis,

S(ω,t)=∑α=(α1,…,αl)cα(t)Hα(ω),T(ω,t)=∑β=(β1,…,βm)c~β(t)Hβ(ω),S⋄T(ω,t):=∑γ=α+βcα(t)c~β(t)Hα+β(ω).S(\omega,t)=\sum_{\alpha=(\alpha_1,\ldots,\alpha_l)}c_\alpha(t)H_\alpha(\omega),\qquad T(\omega,t)=\sum_{\beta=(\beta_1,\ldots,\beta_m)}\widetilde{c}_\beta(t)H_\beta(\omega),\qquad S\mathbin{\diamond}T(\omega,t):=\sum_{\gamma=\alpha+\beta}c_\alpha(t)\widetilde{c}_\beta(t)H_{\alpha+\beta}(\omega).

The Wick product satisfies several desirable properties, firstly, E[S⋄T]=E[S]⋅E[T]\mathbb{E}[S\mathbin{\diamond}T]=\mathbb{E}[S]\cdot\mathbb{E}[T], where S,T∈S∗(Ω)S,T\in\mathcal{S}^*(\Omega) need not be independent random variables. Secondly, if either S,TS,T is deterministic (i.e. S(t)=E[S](t):=c0(t)S(t)=\mathbb{E}[S](t):=c_0(t)), then S⋄T=S⋅TS\mathbin{\diamond}T=S\cdot T reduces to the ordinary product (well defined as a chaos expansion times a constant). Finally, is the correspondence of an Itô diffusion with its S∗(Ω)\mathcal{S}^*(\Omega) valued “Wick calculus" representation

dXt=u(Xt) dt+σ(Xt)dWt⟺X˙t=u(Xt)+σ(Xt)⋄W˙t.\mathrm{d}X_t=u(X_t)\,\mathrm{d}t+\sigma(X_t)\mathrm{d}W_t\Longleftrightarrow\dot{X}_t=u(X_t)+\sigma(X_t)\mathbin{\diamond}\dot{W}_t .

which can be proven rigorously using the Skorohod integral from Malliavin calculus. Interpreting white noise as a Hida distribution value quantity, W˙∈S∗(Ω)\dot{W}\in\mathcal{S}^*(\Omega) with E[W˙t]=0\mathbb{E}[\dot{W}_t]=0, we choose to interpret the white noise products of I4I_4 in the Wick sense. With this choice the expectation distributes to cancel I4I_4. In the other integrals I1,I2,I3I_1,I_2,I_3 we have at least one deterministic factor, making the distinction between Wick and ordinary products trivial. Thus, we will proceed by making assumptions to the dynamics of wtw_t and χt\chi_t and take averages of all quantities.

Assumption 1 (Additive stochastic Taylor hypothesis)

We assume that the dynamics of the first order fluctuation vector field wtw_t is driven by Brownian motion additively such that it can be written as

∘dwt+ad⁡wtu‾t dt=ξ∘dWt,(21)\circ\mathrm{d}w_t+\operatorname{ad}_{w_t}\overline{u}_t\,\mathrm{d}t=\xi\circ\mathrm{d}W_t , \tag*{(21)}

for some ξ∈Xdivn(M)\xi\in\mathfrak{X}^{n}_{\mathrm{div}}(\mathcal{M}) that are prescribed, divergence free and temporally constant.

In this case, the Itô and Stratonovich interpretations coincide,

ξ∘dWt:=∑i=1nξi∘dWti=∑i=1nξidWti=ξdWt.\xi\circ\mathrm{d}W_t:=\sum_{i=1}^{n}\xi_i\circ\mathrm{d}W_t^i=\sum_{i=1}^{n}\xi_i\mathrm{d}W_t^i=\xi\mathrm{d}W_t .

In particular, taking the expectation of (21) implies that E[wt]=:w‾t\mathbb{E}[w_t]=:\overline{w}_t is advected by the mean flow u‾t\overline{u}_t

∂tw‾t+ad⁡w‾tu‾t=0.\partial_t\overline{w}_t+\operatorname{ad}_{\overline{w}_t}\overline{u}_t=0 .

which is equivalent for wtw_t to satisfy Taylor’s frozen turbulence hypothesis in expectation. From the additive stochastic Taylor hypothesis (21), via direct calculation we find the dynamics of the fluctuation correlation tensor Ft:=E[wt⊗wt]∈X(M)⊗X(M)F_t:=\mathbb{E}[w_t\otimes w_t]\in\mathfrak{X}(\mathcal{M})\otimes\mathfrak{X}(\mathcal{M}) as

∂tFt+Lu‾tFt=(ξ⊗ξ),(22)\partial_tF_t+\mathcal{L}_{\overline{u}_t}F_t=(\xi\otimes\xi), \tag*{(22)}

in geometric notation. In terms of index notations, it can be written as

∂tFtij+u‾tk∇kFtij−Ftkj∇ku‾ti−Ftik∇ku‾tj=ξiξj.\partial_tF_t^{ij}+\overline{u}_t^k\nabla_kF_t^{ij}-F_t^{kj}\nabla_k\overline{u}_t^i-F_t^{ik}\nabla_k\overline{u}_t^j=\xi^i\xi^j .

Under the additive stochastic Taylor Hypothesis, the velocity decomposition up to O(ϵ2)O(\epsilon^2) becomes

∘dgtϵ gtϵ,−1=uˉt dt+ϵξ⋄dWt+ϵ22(∘dχt+ad⁡χtuˉt dt−∇∇uˉt(wt,wt) dt+R(uˉt,wt)wt dt−2∇wtξ⋄dWt),\mathop{\circ\mathrm{d}g}_{t}^{\epsilon}\,g_{t}^{\epsilon,-1}=\bar{u}_{t}\,\mathrm{d}t+\epsilon\xi\diamond\mathrm{d}W_{t}+\frac{\epsilon^{2}}{2}\left(\mathop{\circ\mathrm{d}\chi}_{t}+\operatorname{ad}_{\chi_{t}}\bar{u}_{t}\,\mathrm{d}t-\nabla\nabla\bar{u}_{t}(w_{t},w_{t})\,\mathrm{d}t+R(\bar{u}_{t},w_{t})w_{t}\,\mathrm{d}t-2\nabla_{w_{t}}\xi\diamond\mathrm{d}W_{t}\right),

which can be expressed in Wick calculus and distributional white noise W˙t\dot{W}_{t} as

vtϵ=∂tgtϵgtϵ,−1=uˉt+ϵξ⋄W˙t+ϵ22(∂tχt+ad⁡χtuˉt−∇∇uˉt(wt,wt)+R(uˉt,wt)wt−2∇wtξ⋄W˙t)−ϵ32(∇ξi∇wtξi+∇wtξiξi+∇wt∇ξiξi) dt.(23)v_{t}^{\epsilon}=\partial_{t}g_{t}^{\epsilon}g_{t}^{\epsilon,-1}=\bar{u}_{t}+\epsilon\xi\diamond\dot{W}_{t}+\frac{\epsilon^{2}}{2}\left(\partial_{t}\chi_{t}+\operatorname{ad}_{\chi_{t}}\bar{u}_{t}-\nabla\nabla\bar{u}_{t}(w_{t},w_{t})+R(\bar{u}_{t},w_{t})w_{t}-2\nabla_{w_{t}}\xi\diamond\dot{W}_{t}\right) -\frac{\epsilon^{3}}{2}\left(\nabla_{\xi_{i}}\nabla_{w_{t}}\xi_{i}+\nabla_{w_{t}}\xi_{i}\xi_{i}+\nabla_{w_{t}}\nabla_{\xi_{i}}\xi_{i}\right)\,\mathrm{d}t . \tag*{(23)}

Here, the Stratonovich to Itô correction for the additive stochastic Taylor hypothesis results in higher-order terms that are dropped in the expansion by ϵ\epsilon. Noting that there is no canonical choice when applying the expectation, one may apply it to the velocity decomposition expressed on ∂tgtϵ∈TgtϵSDiff⁡(M)\partial_{t}g_{t}^{\epsilon}\in T_{g_{t}^{\epsilon}}\operatorname{SDiff}(M) or on vtϵ=∂tgtϵ,−1=TeSDiff⁡(M)≃X(M)v_{t}^{\epsilon}=\partial_{t}g_{t}^{\epsilon,-1}=T_{e}\operatorname{SDiff}(M)\simeq\mathfrak{X}(M), which corresponds to the Lagrangian and Eulerian expressions of the fluid velocity respectively. In taking the expectation of the Euler-Poincaré variational principle, we make the modelling choice to apply the expectation directly to the vector field vtϵv_{t}^{\epsilon} at fixed Eulerian coordinates. Thus, we obtain the expected mean flow expansion in terms of ϵ\epsilon,

vˉtϵ:=E[vtϵ]=uˉt+ϵ22(∂tχˉt+ad⁡χˉtuˉt−∇∇uˉt:Ft+R(uˉt,⋅):Ft)+O(ϵ3).(24)\bar{v}_{t}^{\epsilon}:=\mathbb{E}[v_{t}^{\epsilon}]=\bar{u}_{t}+\frac{\epsilon^{2}}{2}\left(\partial_{t}\bar{\chi}_{t}+\operatorname{ad}_{\bar{\chi}_{t}}\bar{u}_{t}-\nabla\nabla\bar{u}_{t}:F_{t}+R(\bar{u}_{t},\cdot):F_{t}\right)+O(\epsilon^{3}) . \tag*{(24)}

Assumption 2 (Expectation orthogonality hypothesis) We assume that the material derivative of the expectation of the second order fluctuation vector field χt\chi_{t} is orthogonal to the mean flow uˉt\bar{u}_{t} under metric pairing. That is

∫Mg(uˉt,∂tχˉt+∇uˉtχˉt)=0,(25)\int_{M}g(\bar{u}_{t},\partial_{t}\bar{\chi}_{t}+\nabla_{\bar{u}_{t}}\bar{\chi}_{t})=0 , \tag*{(25)}

where χˉt:=E[χt]\bar{\chi}_{t}:=\mathbb{E}[\chi_{t}]. Noting that uˉt\bar{u}_{t} is assumed to be deterministic and E[⋅]\mathbb{E}[\cdot] commutes with spatial derivatives and integrals, the condition (25) can be rewritten as

∫Mg(uˉt,∘dχt+∇uˉtχt dt)=dMt,\int_{M}g(\bar{u}_{t},\mathop{\circ\mathrm{d}\chi}_{t}+\nabla_{\bar{u}_{t}}\chi_{t}\,\mathrm{d}t)=\mathrm{d}M_{t},

where Mt∈RM_{t}\in\mathbb{R} is a martingale.

Under the assumptions [1] and [2], substituting the white noise velocity expansion (23) into the white noise Lagrangian (20) we have

S=E[∫0T∫M12g(uˉt,uˉt) dt+∫0T∫Mϵg(uˉt,ξ)⋄W˙t dt+∫0T∫Mϵ22g(ξ,ξ)⋄W˙t⋄W˙t dt+∫0T∫Mϵ22g(uˉt,∂tχt+ad⁡χtuˉt−∇∇uˉt(wt,wt)+R(uˉt,wt)wt−2∇wtξ⋄W˙t) dt]=∫0T∫M12g(uˉt,uˉt)+ϵ22g(uˉt,∂tχˉt+ad⁡χˉtuˉt−∇∇uˉt:Ft+R(uˉt,⋅):Ft) dt=∫0T∫M12g(uˉt,uˉt)+ϵ22g(uˉt,−∇χˉtuˉt−∇∇uˉt:Ft+R(uˉt,⋅):Ft) dt.\begin{aligned} S=\mathbb{E}\left[\int_{0}^{T}\int_{M}\frac{1}{2}g(\bar{u}_{t},\bar{u}_{t})\,\mathrm{d}t+\int_{0}^{T}\int_{M}\epsilon g(\bar{u}_{t},\xi)\diamond\dot{W}_{t}\,\mathrm{d}t+\int_{0}^{T}\int_{M}\frac{\epsilon^{2}}{2}g(\xi,\xi)\diamond\dot{W}_{t}\diamond\dot{W}_{t}\,\mathrm{d}t+\int_{0}^{T}\int_{M}\frac{\epsilon^{2}}{2}g\left(\bar{u}_{t},\partial_{t}\chi_{t}+\operatorname{ad}_{\chi_{t}}\bar{u}_{t}-\nabla\nabla\bar{u}_{t}(w_{t},w_{t})+R(\bar{u}_{t},w_{t})w_{t}-2\nabla_{w_{t}}\xi\diamond\dot{W}_{t}\right)\,\mathrm{d}t\right] \\ &=\int_{0}^{T}\int_{M}\frac{1}{2}g(\bar{u}_{t},\bar{u}_{t})+\frac{\epsilon^{2}}{2}g\left(\bar{u}_{t},\partial_{t}\bar{\chi}_{t}+\operatorname{ad}_{\bar{\chi}_{t}}\bar{u}_{t}-\nabla\nabla\bar{u}_{t}:F_{t}+R(\bar{u}_{t},\cdot):F_{t}\right)\,\mathrm{d}t \\ &=\int_{0}^{T}\int_{M}\frac{1}{2}g(\bar{u}_{t},\bar{u}_{t})+\frac{\epsilon^{2}}{2}g\left(\bar{u}_{t},-\nabla_{\bar{\chi}_{t}}\bar{u}_{t}-\nabla\nabla\bar{u}_{t}:F_{t}+R(\bar{u}_{t},\cdot):F_{t}\right)\,\mathrm{d}t . \end{aligned}

Here, we have used the property that E[b(Xt)⋄W˙t]=0\mathbb{E}[b(X_{t})\diamond\dot{W}_{t}]=0 which is analogous to the martingale property to Itô integrals in the second equality and the orthogonality condition (25) in the third equality. Noting that χˉt+div⁡∇Ft=E[χt+div⁡(wt⊗wt)]=E[χt+∇wtwt]∈Xdiv⁡(M)\bar{\chi}_{t}+\operatorname{div}\nabla F_{t}=\mathbb{E}[\chi_{t}+\operatorname{div}(w_{t}\otimes w_{t})]=\mathbb{E}[\chi_{t}+\nabla_{w_{t}}w_{t}]\in\mathfrak{X}_{\operatorname{div}}(M), through integration by parts and appropriate boundary conditions we have

S:=∫0T∫M12g(uˉt,uˉt)+ϵ22g(∇uˉt,∇uˉt⋅Ft)+ϵ22gijRklniuˉtjuˉtlFtkn.(26)S:=\int_{0}^{T}\int_{M}\frac{1}{2}g(\bar{u}_{t},\bar{u}_{t})+\frac{\epsilon^{2}}{2}g(\nabla\bar{u}_{t},\nabla\bar{u}_{t}\cdot F_{t})+\frac{\epsilon^{2}}{2}g_{ij}R^{i}_{kln}\bar{u}^{j}_{t}\bar{u}^{l}_{t}F^{kn}_{t}. \tag*{(26)}

Using the Euler-Poincaré constrained variations for uˉt\bar{u}_t, δuˉt=∂tuˉt−ad⁡ηtuˉt\delta\bar{u}_t = \partial_t\bar{u}_t - \operatorname{ad}_{\eta_t}\bar{u}_t for arbitrary ηt∈Xdiv⁡∇(M)\eta_t \in\mathfrak{X}_{\operatorname{div}\nabla}(\mathcal{M}) vanishing at the boundaries, Hamilton’s principle δS=0\delta S = 0 yields the following Euler-Poincaré equation

(∂t+Luˉt)δℓδuˉt=−dp,whereδℓδuˉt=uˉt♭−ϵ2div⁡∇(∇uˉt♭⋅Ft)+ϵ2R(uˉt,⋅):Ft.(27)\left(\partial_t+\mathcal{L}_{\bar{u}_t}\right)\frac{\delta\ell}{\delta\bar{u}_t}=-\mathrm{d}p,\qquad\text{where}\qquad\frac{\delta\ell}{\delta\bar{u}_t}=\bar{u}_t^{\flat}-\epsilon^2\operatorname{div}\nolimits_{\nabla}\left(\nabla\bar{u}_t^{\flat}\cdot F_t\right)+\epsilon^2R(\bar{u}_t,\cdot):F_t . \tag*{(27)}

Here, d:Λk(M)→Λk+1(M)\mathrm{d}:\Lambda^k(\mathcal{M})\to\Lambda^{k+1}(\mathcal{M}) is the exterior derivative, div⁡∇(∇uˉt♭⋅Ft):=∇j(∇kuˉt♭Ftjk)\operatorname{div}\nolimits_{\nabla}\left(\nabla\bar{u}_t^{\flat}\cdot F_t\right):=\nabla_j\left(\nabla_k\bar{u}_t^{\flat}F_t^{jk}\right), and the pressure contribution arises from the equivalence class of the dual space to the space divergence free vector fields. The Lie derivative Luˉ\mathcal{L}_{\bar{u}} acting on m∈Λ1(M)m\in\Lambda^1(\mathcal{M}) can be expressed using covariant derivatives as

Luˉm=∇uˉm+(∇uˉ)Tm.\mathcal{L}_{\bar{u}}m=\nabla_{\bar{u}}m+(\nabla\bar{u})^{T}m .

Together with the evolution of FF derived from the stochastic Taylor hypothesis [1], we have the closed system of equations

∂tδℓδuˉt+∇uˉtδℓδuˉt+(∇uˉt)Tδℓδuˉt=−dp,∂tFt+∇Ft⋅uˉt−∇uˉt⋅Ft−(∇uˉt)T⋅Ft=(ξ⊗ξ),div⁡∇(uˉt)=div⁡∇(ξi)=0.(28)\begin{aligned} \partial_t\frac{\delta\ell}{\delta\bar{u}_t}+\nabla_{\bar{u}_t}\frac{\delta\ell}{\delta\bar{u}_t}+(\nabla\bar{u}_t)^{T}\frac{\delta\ell}{\delta\bar{u}_t}&=-\mathrm{d}p,\\ \partial_tF_t+\nabla_{F_t}\cdot\bar{u}_t-\nabla_{\bar{u}_t}\cdot F_t-(\nabla\bar{u}_t)^{T}\cdot F_t&=(\xi\otimes\xi),\\ \operatorname{div}\nolimits_{\nabla}(\bar{u}_t)=\operatorname{div}\nolimits_{\nabla}(\xi_i)&=0 . \tag*{(28)} \end{aligned}

which are the forced anisotropic Lagrangian averaged Euler equations. Several simplification can be made by specialising to flat geometries such that curvature RR vanishes. In this case, we have the following proposition.

Proposition 1. Assuming the correlation tensor FtF_t has the dynamics

∂tFt+LuˉtFt=Gt,(29)\partial_tF_t+\mathcal{L}_{\bar{u}_t}F_t=G_t , \tag*{(29)}

for arbitrary, time dependent Gt∈X(M)⊗X(M)G_t\in\mathfrak{X}(\mathcal{M})\otimes\mathfrak{X}(\mathcal{M}) which can have functional dependence on uˉt\bar{u}_t, wtw_t and FtF_t. Denoting by C(q)=div⁡∇(∇q⋅Ft)C(q)=\operatorname{div}\nolimits_{\nabla}(\nabla q\cdot F_t) for all arbitrary vector qt∈X(M)q_t\in\mathfrak{X}(\mathcal{M}), we have the commutation relation

[∂t+∇uˉt,C]qt=div⁡∇(∇qtGt).(30)[\partial_t+\nabla_{\bar{u}_t},C]q_t=\operatorname{div}\nolimits_{\nabla}(\nabla q_tG_t). \tag*{(30)}

Proof. In the case ∂tF~t+LuˉtF~t=0\partial_t\widetilde{F}_t+\mathcal{L}_{\bar{u}_t}\widetilde{F}_t=0, it was shown in e.g. [18, 15], that [∂t+uˉt⋅∇,C~]qt=0\left[\partial_t+\bar{u}_t\cdot\nabla,\widetilde{C}\right]q_t=0 for C~u=div⁡∇E(∇u⋅F~t)\widetilde{C}u=\operatorname{div}\nolimits_{\nabla^E}(\nabla u\cdot\widetilde{F}_t) for Euclidean connection ∇E\nabla^E. Since the contribution from GtG_t enters only in term ∂t(C(qt))=∂tdiv⁡∇(∇qt⋅Ft)\partial_t(C(q_t))=\partial_t\operatorname{div}\nolimits_{\nabla}(\nabla q_t\cdot F_t), we have the required result through a similar direct computation and switching of covariant derivatives, which always holds under the assumption that R≡0R\equiv0.

Corollary 1. When the curvature RR vanishes, the Euler-Poincaré equation can (27) be expressed equivalently as

(1−ϵ2C)(∂t+∇uˉt)uˉt−ϵ2C(u)i∇ui=−∇p~−div⁡∇(∇uˉt(ξ⊗ξ)),(31)(1-\epsilon^2C)(\partial_t+\nabla_{\bar{u}_t})\bar{u}_t-\epsilon^2C(u)^i\nabla u_i=-\nabla\widetilde{p}-\operatorname{div}\nolimits_{\nabla}\left(\nabla\bar{u}_t(\xi\otimes\xi)\right), \tag*{(31)}

where p~=p+12∇(∣uˉt∣2)\widetilde{p}=p+\frac{1}{2}\nabla(|\bar{u}_t|^2) is the modified pressure.

Conclusion

In this work, we considered an averaged Euler-Poincaré variational principle to construct sub-grid scale ideal fluid models. In this construction, the total fluid flow was assumed stochastic and averaged through the expectation in analogy with the GLM theory. The first step of this construction, made in section 2, is the formulation of the mean velocity, defined by the factorisation of the total flow map into the fluctuation and mean maps. Then, the Magnus expansion of the stochastic flow map is considered so that the expectation operator is applied to vector fields and their actions. Our averages then occurred only at the tangent space of the identity, bypassing the difficulty in defining the expectation of diffeomorphism group-valued quantities. In section 3, we generalised the KE Lagrangian functional such that it is well-defined when the input velocity contains white noise terms using tools from Malliavin calculus. After making dynamical assumptions on the fluctuations vector fields in assumptions 1 and 2, we explicitly computed the averaged Lagrangian and derived a stochastically averaged version of the anisotropic Lagrangian averaged Euler’s equation through an Euler-Poincaré variational principle. Following this work, several open problems present themselves to be addressed in the future. First, the closure dynamics of wtw_t and χt\chi_t can be modified to include additional physics. For example, one can assume transport type stochastic perturbation for the dynamics of wtw_t, dwt+ad⁡wtut+ad⁡wtξ∘dWt=0d w_t + \operatorname{ad}_{w_t} u_t + \operatorname{ad}_{w_t} \xi\circ dW_t = 0 as to introduce dissipation into the dynamics of the correlation tensor FtF_t. Second, the driving stochastic processes can be generalised from Brownian motions to geometric rough paths. For example, fractional Brownian motion or lifts of Brownian motion that incorporate a noise-induced drift. Such drifts are natural in the context of multiscale limits [6]. Lastly, the same modelling procedure can be generalised from incompressible Euler’s fluid using SDiff⁡(M)\operatorname{SDiff}(M) to semi-direct products such that sub-grid scale models for a large class of fluid flows can be derived.

Acknowledgments. We wish to thank D. Holm, O. Street and J. Woodfield for several thoughtful suggestions during the course of this work. TD is supported during the present work by European Research Council (ERC) Synergy grant Stochastic Transport in Upper Ocean Dynamics (STUOD) – DLV-856408. RH is grateful for the support by Office of Naval Research (ONR) grant award N00014-22-1-2082, Stochastic Parameterization of Ocean Turbulence for Observational Networks.

Disclosure of Interests. The authors have no competing interests to declare that are relevant to the content of this article.

References

  1. [1]Andrews, D.G., Mcintyre, M.E.: An exact theory of nonlinear waves on a Lagrangian-mean flow. Journal of Fluid Mechanics 89(4), 609–646 (1978). https://doi.org/10.1017/S0022112078002773
  2. [2]Arnaudon, M., Chen, X., Cruzeiro, A.B.: Stochastic Euler-Poincaré reduction. Journal of Mathematical Physics 55 (2014). https://doi.org/10.1063/1.4893357
  3. [3]Blanes, S., Casas, F., Oteo, J., Ros, J.: The magnus expansion and some of its applications. Physics Reports 470(5), 151–238 (2009). https://doi.org/10.1016/j.physrep.2008.11.001
  4. [4]Chen, X., Cruzeiro, A.B., Ratiu, T.S.: Stochastic Variational Principles for Dissipative Equations with Advected Quantities. Journal of Nonlinear Science 33, 5 (2 2023). https://doi.org/10.1007/s00332-022-09846-1
  5. [5]Cipriano, F., Cruzeiro, A.B.: Navier-Stokes equations and diffusions on the group of homeomorphisms of the torus. Communications in Mathematical Physics 275, 255–269 (10 2007). https://doi.org/10.1007/s00220-007-0306-3
  6. [6]Cotter, C.J., Gottwald, G.A., Holm, D.D.: Stochastic partial differential fluid equations as a diffusive limit of deterministic Lagrangian multi-time dynamics. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 473(2205), 20170388 (2017). https://doi.org/10.1098/rspa.2017.0388
  7. [7]Diamantakis, T., Holm, D.D., Pavliotis, G.A.: Variational Principles on Geometric Rough Paths and the Lévy Area Correction. SIAM Journal on Applied Dynamical Systems 22, 1182–1218 (6 2023). https://doi.org/10.1137/22M1522164
  8. [8]Diamantakis, T., Hu, R.: Variational closures for composite homogenised fluid flows (2024), https://arxiv.org/abs/2409.10408
  9. [9]Foias, C., Holm, D.D., Titi, E.S.: The navier-stokes-alpha model of fluid turbulence. Physica D: Nonlinear Phenomena 152–153, 505–519 (2001). https://doi.org/10.1016/S0167-2789(01)00191-9
  10. [10]Gjaja, I., Holm, D.D.: Self-consistent hamiltonian dynamics of wave mean-flow interaction for a rotating stratified incompressible fluid. Physica D: Nonlinear Phenomena 98(2), 343–378 (1996). https://doi.org/10.1016/0167-2789(96)00104-2
  11. [11]Hall, B.C.: Lie Groups, Lie Algebras, and Representations. Springer New York (2003). https://doi.org/10.1007/978-0-387-21554-9
  12. [12]Holm, D.D., Hu, R.: Stochastic effects of waves on currents in the ocean mixed layer. Journal of Mathematical Physics 62 (2021). https://doi.org/10.1063/5.0045010
  13. [13]Holm, D.D.: Variational principles for stochastic fluid dynamics. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 471(2176), 20140963 (2015)pubmed.ncbi.nlm.nih.gov/27547083
  14. [14]Magnus, W.: On the exponential solution of differential equations for a linear operator. Communications on Pure and Applied Mathematics 7(4), 649–673 (1954). https://doi.org/10.1002/cpa.3160070404
  15. [15]Marsden, J.E., Shkoller, S.: The anisotropic lagrangian averaged euler and navier-stokes equations. Archive for Rational Mechanics and Analysis 166, 27–46 (1 2003). https://doi.org/10.1007/s00205-002-0207-8
  16. [16]Mémin, E.: Fluid flow dynamics under location uncertainty. Geophysical & Astrophysical Fluid Dynamics 108, 119–146 (3 2014). https://doi.org/10.1080/03091929.2013.836190
  17. [17]Nunno, G.D., Øksendal, B., Proske, F.: Malliavin Calculus for Lévy Processes with Applications to Finance. Springer Berlin Heidelberg, Berlin, Heidelberg (2009). https://doi.org/10.1007/978-3-540-78572-9
  18. [18]Shkoller, S.: The Lagrangian Averaged Euler (LAE-α) Equations with Free-Slip or Mixed Boundary Conditions, pp. 113–165. Springer-Verlag (2002). https://doi.org/10.1007/0-387-21791-6_5

Paper details

vlnc:W_fxmgbvzu
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents