Abstract
We describe ideal incompressible hydrodynamics on the hyperbolic plane which is an infinite surface of constant negative curvature. We derive equations of motion, general symmetries and conservation laws, and then consider turbulence with the energy density linearly increasing with time due to action of small-scale forcing. In a flat space, such energy growth is due to an inverse cascade, which builds a constant part of the velocity autocorrelation function proportional to time and expanding in scales, while the moments of the velocity difference saturate during a time depending on the distance. For the curved space, we analyze the long-time long-distance scaling limit, that lives in a degenerate conical geometry, and find that the energy-containing mode linearly growing with time is not constant in space. The shape of the velocity correlation function indicates that the energy builds up in vortical rings of arbitrary diameter but of width comparable to the curvature radius of the hyperbolic plane. The energy current across scales does not increase linearly with the scale, as in a flat space, but reaches a maximum around the curvature radius. That means that the energy flux through scales decreases at larger scales so that the energy is transferred in a non-cascade way, that is the inverse cascade spills over to all larger scales where the energy pumped into the system is cumulated in the rings. The time-saturated part of the spectral density of velocity fluctuations contains a finite energy per unit area, unlike in the flat space where the time-saturated spectrum behaves as \(\,k^{-5/3}\).
Similar content being viewed by others
Notes
More general volume forms could be also used [3].
For non-compact \(\,M\,\) we shall assume that \(\,\Phi \,\) do not move points outside a compact set and that \(\,v\,\) have compact supports, although weaker assumptions about the behavior at infinity may be more natural and/or more physiqcal.
Field theorists will notice a similarity of the following calculation to those in nonlinear sigma models.
References
Arneodo, A., et al.: Structure functions in turbulence, in various flow configurations, at Reynolds number between 30 and 5000, using extended self-similarity. Europhys. Lett. 34, 411–416 (1996)
Arnold, V.I.: Sur la géométrie différentielle des groupes de Lie de dimension infinie et ses applications à l’hydrodynamique des fluides parfaits. Ann. Inst. Fourier (Grenoble) 16, 319–361 (1966)
Arnold, V.I., Khesin, B.: Topological Methods in Hydrodynamics, vol. 125. Springer, New York (1997)
Batchelor, G.K.: Computation of the energy spectrum in homogeneous two-dimensional turbulence. Phys. Fluids Suppl. 12, 233–239 (1969)
Bernard, D., Boffetta, G., Celani, A., Falkovich, G.: Conformal invariance in two-dimensional turbulence. Nat. Phys. 2, 124–128 (2006)
Bernard, D., Boffetta, G., Celani, A., Falkovich, G.: Inverse turbulent cascades and conformally invariant curves. Phys. Rev. Lett. 98, 024501 (2007)
Boffetta, A.B., Vergassola, M.: Inverse cascade in two-dimensional turbulence: deviations from Gaussianity. Phys. Rev. E 61, R29R32 (2000)
Boffetta, G., Musacchio, S.: Evidence for the double cascade scenario in two dimensional turbulence. Phys. Rev. E 82, 016307 (2010)
G. Carron: \(L^2\) harmonics forms on non-compact manifolds, arXiv:0704.3194 [math.DG]
Chan, C.H., Czubak, M.: Non-uniqueness of the Leray-Hopf solutions in the hyperbolic setting. Dyn PDE 10, 43–77 (2013)
Ebin, D.G., Marsden, J.: Groups of diffeomorphisms and the motion of an incompressible fluid. Ann. Math. 92, 102–163 (1970)
Falkovich, G.: Fluid Mechanics, a Short Course for Physicists. Cambridge University Press, Cambridge (2011)
Falkovich, G., Gawedzki, K., Vergassola, M.: Particles and fields in fluid turbulence. Rev. Mod. Phys. 73, 913–975 (2001)
Frisch, U.: Turbulence: the Legacy of A. N. Kolmogorov. Cambridge University Press, Cambridge (1995)
Frisch U., Villone B.: Cauchy’s almost forgotten Lagrangian formulation of the Euler equation for 3D incompressible flow, arXiv:1402.4957 [math.HO]
Gradshteyn, I.S., Ryzhik, I.M.: Table of Integrals, Series, and Products, 7th edn. Academic Press, Amsterdam (2007)
Hilbert, D.: Über Flächen von constanter Gaußscher Krümmung. Trans. Am. Math. Soc. 2, 8799 (1901)
Kellay, H., Goldburg, W.I.: Two-dimensional turbulence: a review of some recent experiments. Rep. Prog. Phys. 65, 845–894 (2002)
Khesin, B., Misiolek, G.: The Euler and Navier–Stokes equations on the hyperbolic plane. Proc. Nat. Acad. Sci. 109, 18324–18326 (2012)
Kraichnan, R.H.: Inertial ranges in two-dimensional turbulence. Phys. Fluids 10, 1417–1423 (1967)
Kuksin, S., Shirikyan, A.: Mathematics of Two-dimensional Turbulence. Cambridge University Press, Cambridge (2012)
Paret, J., Tabeling, P.: Intermittency in the 2D inverse cascade of energy: experimental observations. Phys. Fluids 10, 3126–3136 (1998)
Vilenkin, N.J.: Special functions and the theory of group representations. Am. Math. Soc. Trans. 2, 207–316 (1968)
Vishik, M.J., Fursikov, A.V.: Mathematical Problems of Statistical Hydrodynamics. Kluwer Academic Publishers, Dordrecht (1988)
Vucelja, M., Falkovich, G., Turitsyn, K.S.: Fractal iso-contours of passive scalar in two-dimensional smooth random flows. J. Stat. Phys. 147, 424–435 (2012)
Xia, H., Shats, M., Falkovich, G.: Spectrally condensed turbulence in thin layers. Phys. Fluids 21, 125101 (2009)
Acknowledgments
The work of G. F. was supported by the grants of BSF and Minerva Foundation. The work of K. G. was a part of the STOSYMAP project ANR-11-BS01-015-02. Both authors are grateful to their home institutions for the support of mutual visits. We thank R. Chetrite and A. B. Zamolodchikov for useful discussions and U. Frisch for additional references and for pointing out a mistake in the original version of the paper.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix 1: Some Notions of Riemannian Geometry
Let \(M\) be a \(d\)-dimensional oriented manifold equipped with a Riemannian metric \(\,g_{ij}{\mathrm {d}}x^i{\mathrm {d}}x^j\). Such a metric may be used to lower the indices of the contravariant tensor fields, e.g. \(v_i=g_{ij}v^j\). One has \(\,v^j=g^{ij}v_i\), where \(\,(g^{ij})\,\) is the matrix inverse to \(\,(g_{ij})\). The metric together with the orientation induces the volume form \(\,\chi =\sqrt{g}\,{\mathrm {d}}x^1\wedge \cdots \wedge {\mathrm {d}}x^d \equiv \sqrt{g}\,{\mathrm {d}}x\), where \(\,\sqrt{g}\,\) is the shorthand notation for \(\,\sqrt{\mathrm{det}(g_{ij})}\). A Riemannian metric induces the Levi-Civita connection acting on tensor fields by
where \(\,\Gamma ^k_{ij}\,\) are given by the Christoffel symbols
For the Levi-Civita connection, the covariant derivatives of the tensors \(g_{ij}\) and \(g^{ij}\) vanish so that \(\nabla _i\) commutes with the raising and lowering of indices. The Riemann curvature tensor is defined by the formula
and the Ricci tensor by
The scalar curvature is \( S\,=\,g^{jl}({\mathrm {Ric}})_{jl}\). The totally antisymmetric tensors may be identified with the differential forms by the correspondence
The exterior derivative in terms of this identification is given by the formula
The Hodge star on \(r\)-forms is defined by
It satisfies \(\,*^2=(-1)^{(d-r)r}\). The \(L^2\) scalar product of \(r\)-forms is given by
The (formal) adjoint of the exterior derivative \(d\) mapping \(r\)-forms to \((r+1)\)-forms is the operator
mapping \((r+1)\)-forms to \(r\)-forms. In terms of the tensor components,
The (negative) Laplace–Beltrami operator is
In the action on functions, it gives
The Laplace–Beltrami operator act also on vector fields \(\,v\,\) by the formula
where \(\,v^{\flat }\,\) is the 1-form represented by \((v_i)\) associated to the vector field represented by \(\,(v^i)\). We also need the Lie derivative of tensors w.r.t. vector fields \(\,u\,\) given by the formulae
in the action on the vector fields and 1-forms, respectively, and by the Leibniz rule on higher tensors. In particular, the vector fields \(\,u\,\) are called Killing vectors if \(\,\mathcal{L}_u\,\) annihilate the metric tensor \(\,g\):
We also consider divergenceless vector fields \(\,v\,\) defined by the condition \(\,\mathcal{L}_v\chi =0\). In coordinates, the condition is equivalent to the conditions \(\,\partial _k(v^k\sqrt{g})=0\,\) or \(\,\nabla _kv^k=\sqrt{g}^{-1} \partial _k(v^k\sqrt{g})=0\). Yet another form of this condition is the requirement that \(\,{\mathrm {d}}^\dagger v^\flat =0\). Locally, every 1-form satisfying such condition can be written via \((d-2)\)-form \(\,\psi \,\):
Appendix 2: Variation Principle for Euler Equation
It will be convenient to perform the extremization over all diffeomorphisms imposing the volume-preservation constraint by adding to the action a term
with a Lagrange multiplier \(\,\lambda (t,x)\). For the variation of \(\,S_g\,\) at volume-preserving \(\,\Phi \), we obtainFootnote 3:
where \(\,u^j(t,\Phi (t,x))=\delta \Phi ^j(t,x)\,\) and \(\,\nabla _k\,\) is the covariant derivative with respect to the Levi-Civita connection. On the other hand, the variation of \(\,S_{Lm}\,\) is given by
where we have introduced the pressure by the formula
Integrating by parts in the last term, we finally obtain
Equating to zero the variation \(\,\delta S(\Phi ,\lambda )\,\) for \(\,S(\Phi ,\lambda )=S_g(\Phi )+S_{Lm}(\Phi ,\lambda )\), we obtain the (generalized) Euler equation (2.2) together with the volume-preserving condition on \(\,\Phi (t)\,\) which, in terms of the velocity is the incompressibility condition (2.3). On noncompact manifolds, these should be accompanied by the decay conditions at infinity that assure that the integrations by parts above may be performed and eliminate the solutions \(\,v^i\,=\,g^{ij}\partial _jf\,\) with \(\,\Delta f=g^{ij}\nabla _i\partial _jf=0\,\) and arbitrary \(t\)-dependence (solving also the unforced Navier–Stokes equations on Einstein manifolds, see [10, 19]).
Appendix 3: Velocity Covariance in Terms of Stream Functions
\({{\mathbb H}}_{_R}\,\) is a contractible space. On the other hand, due to incompressibility of velocity, the equal-time 2-point function \(\,\langle (*v^\flat )(\mathcal{X}_1) \otimes (*v^\flat (\mathcal{X}_2)\rangle \) is a closed 1-form in its dependence on each of the two points. We may then obtain a stream-function correlator \(\,\langle \psi (\mathcal{X}_1)\,\psi (\mathcal{X}_2)\rangle \,\) satisfying Eq. (6.26) by integrating \(\,\langle (*v^\flat )(\mathcal{X}_1) \otimes (*v^\flat (\mathcal{X}_2)\rangle \,\) in each variable from a fixed point of \(\,{{\mathbb H}}_{_R}\,\) to \(\,\mathcal{X}_1\,\) and \(\,\mathcal{X}_2\), respectively. Function \(\,\langle \psi (\mathcal{X}_1)\,\psi (\mathcal{X}_2)\rangle \,\) obtained this way is symmetric and of positive type, but it depends on the choice of the initial point of integration. As a result, the \(\,SL(2,{{\mathbb R}})\)-covariance of the velocity 2-point function does not imply the \(\,SL(2,{{\mathbb R}})\,\) invariance of \(\,\langle \psi (\mathcal{X}_1)\, \psi (\mathcal{X}_2)\rangle \). However, for \(\,\gamma \in SL(2,{{\mathbb R}})\), the function
is annihilated by the product of exterior derivatives \(\,d(1)d(2)\,\) in \(\,\mathcal{X}_1\,\) and \(\,\mathcal{X}_2\). It has then to be of the form \(\,f^1_\gamma (\mathcal{X}_1)+f^2_\gamma (\mathcal{X}_2)\,\) where, by symmetry, we may take \(\,f^1_\gamma (\mathcal{X})=f^2_\gamma (\mathcal{X})=\frac{1}{2}\,f_\gamma (\mathcal{X},\mathcal{X})\). Definition (16.1) implies the cocycle condition
Taking, in particular, \(\,\mathcal{X}=\mathcal{X}_0\,\) corresponding to \(\,w=i\,\) and \(\,\gamma _2\in SO(2,{{\mathbb R}})\), which is the stabilizer subgroup of \(\,\mathcal{X}_0\), we obtain
Taking also \(\,\gamma _1\in SO(2,{{\mathbb R}})\), we infer that \(\,f^1_\gamma (\mathcal{X}_0)\,\) is an additive function of \(\,\gamma \in SO(2,{{\mathbb R}})\), so it must be equal to zero when restricted to such \(\gamma \). Consequently, for general \(\,\gamma _1\), one has \(\,f^1_{\gamma _1\gamma _2}(\mathcal{X}_0)\,=\,f^1_{\gamma _1}(\mathcal{X}_0)\), so that \(\,f^1_\gamma (\mathcal{X}_0)\,\) defines a function \(\,g(\mathcal{X})\,\) on \(\,{{\mathbb H}}_{_R}\cong SL(2,{{\mathbb R}})/SO(2,{{\mathbb R}})\,\) (vanishing at \(\,\mathcal{X}=\mathcal{X}_0\)) and it determines \(\,f^1_\gamma (\mathcal{X})\,\) by the formula
following from the cocycle condition (16.2). This, in fact, provides a general solution of this condition if we consider arbitrary functions \(\,g(\mathcal{X})\,\) on \(\,{{\mathbb H}}_{_R}\). Note that the symmetric function \(\,\Psi (\mathcal{X}_1,\mathcal{X}_2)= \big \langle \,\psi (\mathcal{X}_1)\,\psi (\mathcal{X}_2)\big \rangle -g(\mathcal{X}_1)-g(\mathcal{X}_2)\,\) satisfies the relation \(\,\Psi (\gamma \mathcal{X}_1,\gamma \mathcal{X}_2)=\Psi (\mathcal{X}_1,\mathcal{X}_2)\), hence it depends only on the hyperbolic distance \(\,\delta _{12}\). It defines the same velocity correlators if used in formula (6.26) instead of \(\,\big \langle \psi (\mathcal{X}_1)\, \psi (\mathcal{X}_2)\big \rangle \). If \(\,\Psi '\,\) is another such functions, then
for some function \(\,g^1(\mathcal{X})\). The \(SL(2,{{\mathbb R}})\)-covariance of \(\,\Psi \,\) and \(\,\Psi '\,\) implies then that
but for each \(\,\gamma \,\) there is \(\,\mathcal{X}_1\,\) such that \(\,\gamma \mathcal{X}_1=\mathcal{X}_1\,\) and we infer that \(\,c_\gamma =0\,\) so that function \(\,g^1\,\) is constant.
Appendix 4: Forcing Covariance on \(\,{{\mathbb H}}_{_{R/\lambda }}\)
Equation (7.2) implies the formulae:
The asymptotic behavior (12.5), (12.6) and (12.7) follow from these expressions in a straightforward way.
Appendix 5: Rescaled Velocity 2-Point Functions on \(\,{{\mathbb H}}_{_{R}}\)
Assuming the form (12.27) of the 2-point correlator of the stream function on \({{\mathbb H}}_{_{R}}\), the rescaled version of the velocity 2-point functions given by Eq. (6.26) may be written in the form:
where \(\,x\,\) is given by the right hand side of Eq. (12.28) with \(\,R\,\) replaced by \(\,R/\lambda \). The above expressions should be compared to Eqs. (12.16), (12.17) and (12.18) for the velocity correlators on \(\,{{\mathbb H}}_{_0}\).
Appendix 6: Harmonic Analysis on \(\,{{\mathbb H}}_{_R}\,\) in the \(\,R\rightarrow 0\,\) limit
To study the harmonic analysis on \(\,{{\mathbb H}}_{_R}\,\) in the limit \(\,R\rightarrow 0\), we rewrite it using the coordinates \(\,(r,\varphi )\). The harmonic analysis in space \(\,L^2({{\mathbb H}}_{_R})\,\) is realized by the formulae that combine the Gel’fand-Graev transformation \(\,\psi \mapsto \psi ^{GG}\,\) from functions on the hyperboloid \(\,{{\mathbb H}}_{_R}\,\) to functions on the cone \(\,{{\mathbb H}}_{_0}\,\)
with the Fourier transform in the angular direction and Mellin transform in the radial one [23]. In the coordinates \(\,(r,\varphi )\,\) it is given by the relations:
which reproduce Eqs. (6.9), (6.15) and (6.16) upon setting \(\,\sigma =Rk\).
To study the \(\,R\rightarrow 0\,\) limit, we use the asymptotic expansion from [16] holding for large positive \(\,x\):
Substituting this to Eq. (19.2) and using the relation \(\,\sigma \tanh (\pi \sigma ) = {|\Gamma (\frac{1}{2}+i\sigma )|^2} /{|\Gamma }\) \({(i\sigma )|^2} \), we obtain
where we have set
Hence, the limiting decomposition at \(\,R=0\,\) of functions on the cone \(\,{{\mathbb H}}_{_0}\,\) is given by the Mellin transform in \(\,r\,\) combined with the Fourier transform in the angular variable \(\,\varphi \):
Action (11.21) of \(\,\mathrm{Diff}S^1\,\) on \(\,{{\mathbb H}}_{_0}\,\) induces the unitary action of that infinite-dimensional group in the space \(\,L^2({{\mathbb H}}_{_0})\,\) of functions on \(\,{{\mathbb H}}_{_0}\,\) square-integrable with respect to the volume measure \(\,\chi _0\), defined by the formula
Note that this action commutes with the Laplacian \(\,\Delta =\partial _r r^2\partial _r\). The functions \(\,\mathrm{e}^{im\varphi }\,r^{-\frac{1}{2}+i\sigma }\,\) are eigenvectors of \(\,\Delta \,\) corresponding to the eigenvalue \(-(\frac{1}{4}+\sigma ^2)\). The decomposition (19.8) realizes the decomposition of \(\,L^2({{\mathbb H}}_{_0})\,\) into the spectral eigenspaces of \(\,\Delta \,\) and, at the same time, the decomposition of the unitary representation of \(\,\mathrm{Diff}S^1\,\) in \(\,L^2({{\mathbb H}}_{_0})\,\) into the irreducible components:
Unlike for \(\,R>0\), where only \(\,\sigma >0\,\) appeared in the decomposition (6.8), now \(\,\sigma >0\,\) and \(\,\sigma <0\,\) give rise to inequivalent representations.
Appendix 7: Real Analyticity of the Spectrum
Spectrum \(\,\mathcal{E}(k)\,\) is related to the stream function correlator \(\,\Psi \,\) by the Fourier transform on \(\,{{\mathbb H}}_{_R}\,\) (6.25) and Eq. (6.31) which imply the inverse transform relation
where the integral has to be interpreted in the appropriate sense (by analytic continuation) since, for large \(\,x\), \(\,|\mathcal{P}^{-\frac{1}{2}+i\sigma }_{00}(x)|=\mathcal{O}(x^{-\frac{1}{2}})\), see Eqs. (19.5). Let us split the integration on the right hand side of (20.1) into the one from 1 to 2 and the rest. The first integral gives an entire function of \(\,k\), as is easily seen from Eq. (6.11), and it contributes to \(\,\mathcal{E}(k)\,\) a term \(\,O(k^2)\,\) or less for small \(\,k\). For the other integral, we shall use the expression
for \(\,\zeta \equiv x+\sqrt{x^2-1}\). Expanding the hypergeometric functions into a power series in \(\,\zeta ^{-2}\), we see that only the contribution
of the leading term into the integral needs a special treatment. Using the asymptotic expansion for \(\,\Psi _{stat}\,\) rewritten in terms of variable \(\,\zeta \), we see that the integral (20.3) contributes few simple poles and a function analytic in \(\,k\,\) in a neighborhood of the real axis. The pole at zero appears only if the power \(\,x^{-\frac{1}{2}}\,\) occurs in the large \(\,x\,\) expansion of \(\,\Psi _{stat}\). The other terms of the expansion of the hypergeometric functions contribute absolutely convergent integrals so that
is again analytic in \(\,k\,\) in a neighborhood of the real axis. The prefactors \(\,\pi ^{-\frac{1}{2}}\Gamma (\mp iRk)/\) \(\Gamma (\frac{1}{2}\mp iRk)\,\) in Eq. (20.2) are meromorphic in \(\,k\,\) in such a region with a simple pole at \(\,k=0\). Equation (20.1) implies than that the contribution of \(\,\Psi _{stat}\,\) is analytic around \(\,k\ge 0\). As for the contribution to \(\,\Psi _0\), a presence of the \(\,x^{-\frac{2}{2}}\,\) term in its asymptotic expansion could leave us with a pole at \(\,k=0\,\) of up to the second order (due to up to two derivatives over \(\,k\,\) used to produced the logarithmic terms of \(\,\Psi _0\)), but such pole would be incompatible with the finiteness of the energy density at finite times. Hence the contribution of \(\,\Psi _0\,\) to the spectrum should also be analytic around \(\,k\ge 0\).
Rights and permissions
About this article
Cite this article
Falkovich, G., Gawȩdzki, K. Turbulence on Hyperbolic Plane: The Fate of Inverse Cascade. J Stat Phys 156, 10–54 (2014). https://doi.org/10.1007/s10955-014-0995-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10955-014-0995-x