Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Fluctuation-theorem and extended thermodynamics of turbulence

Fluctuation-theorem and extended thermodynamics of turbulence 1, 2,y 3,z 4,x Amilcare Porporato, Milad Hooshyar, Andrew D Bragg, and Gabriel Katul Department of Civil and Environmental Engineering and Princeton Environmental Institute, Princeton University, Princeton, NJ 08544, USA Princeton Environmental Institute and Princeton Institute for International and Regional Studies, Princeton University, Princeton, NJ 08544, USA Department of Civil and Environmental Engineering, Duke University, Durham, NC 27708, USA Department of Civil and Environmental Engineering and Nicholas School of the Environment, Duke University, Durham, NC 27708, USA Turbulent ows are out-of-equilibrium because the energy supply at large scales and its dissipation by viscosity at small scales create a net transfer of energy among all scales. Here, the energy cascade is approximated by a combined contribution of a forward drift and di usion that recover accepted phenomenological theories of turbulence. The uctuation theorem (FT) is then shown to describe the scale-wise statistics of forward and backward energy transfer and their connection to irreversibility and entropy production. The ensuing turbulence entropy may be used to formulate an extended turbulence thermodynamics. I. INTRODUCTION The statistical properties of turbulence di er from sys- tems near thermal equilibrium because the ux of en- ergy per unit mass is supplied at scales much larger than It is perhaps not coincidental that one of the most in- the scales at which energy is dissipated by the action uential experiments in the history of thermodynamics of viscosity, resulting in an energy ux (cascade) across is also a turbulence experiment. In 1849, James Prescott all scales. Such a transport is linked to multiple pro- Joule used a stirrer to show that the shaft work on a uid cesses, including vortex stretching, self-ampli cation of ends up increasing its internal energy, thereby demon- the strain-rate and viscous di usion [14, 15]. One of the strating the equivalence of heat and work. Dealing with de ning features of the turbulence cascade is that the the generation of turbulent kinetic energy (K ) and its probability of forward and backward transitions between subsequent dissipation rate () by viscosity, the Joule two energetic states at a given scale are not identical (i.e., experiment also o ers a modern link between thermody- a scale-wise `detailed balance' is not applicable [16, 17]). namics, a theory of the macroscopic e ects of microscopic uctuations, and non-equilibrium uctuations, of which The objective of this work is to illustrate that non- turbulence is a quintessential example. On the one hand, equilibrium thermodynamics, and in particular, the uc- the disparity between microscopic and macroscopic uc- tuation theorem can be extended to describe the behavior tuations appears un-reconcilable. Thermodynamic uc- of the turbulent energy cascade. While the net transfer of tuations are so small to allow a mathematical descrip- energy from large to small scales is prevalent, it is shown tion of uids as a continuum. For this reason, turbulence here that back-scatter of energy and its connection to is conveniently described by the Navier-Stokes equations time-scale irreversibility obeys the statistics predicted by assuming local thermodynamic equilibrium [1]. Turbu- the uctuation theorem [13, 17{20]. To provide a phys- lent uctuations are thus of macroscopic nature and tech- ical context, a turbulent ow conceptually analogous to nically outside the scope of traditional thermodynamics Joule's original experiment is used, where work is done [2]. On the other hand, the random-like nature of tur- on the uid system to generate K in a narrow band of bulence [3{5] invites a thermodynamic formalism to the scales, which is then dissipated as heat thereby raising problem of turbulence, including the eddy thermodynam- the internal energy. In the analysis here, constant en- ics of Richardson [6], Blackadar [7] and others [8, 9], as ergy is externally supplied at a pre-selected scale much well as the Onsager analysis of 2D turbulence [10]. Liep- larger than the Kolmogorov length scale ( ) where vis- mann [11] asserted that `turbulence can be de ned by cous e ects are signi cant. At steady state, the energy a statement of impotence reminiscent of the second law cascade develops in a manner where the energy injection of thermodynamics'. More recently, a number of stud- rate is balanced by viscous dissipation rate as in Joule's ies have argued that the uctuation theorem derived for experiment. The uctuation theorem is then applied to small systems [12, 13] can be partly applied to describe describe the forward and backward probabilities of en- macroscopic uctuations so as to explore their time re- ergy packets moving scale-wise in time through the en- versibility at multiple scales, including turbulence. ergy cascade. For analytical tractability and to illustrate connections with the uctuation theorem, simpli ed clo- sure schemes for the energy transfer rate across scales are employed. These closure schemes o er plausible expres- aporpora@princeton.edu hooshyar@princeton.edu sions for the energy cascade that are consistent with a andrew.bragg@duke.edu wide range of experiments and theories on locally homo- gaby@duke.edu geneous and isotropic turbulence. arXiv:2005.01522v1 [physics.flu-dyn] 4 May 2020 2 II. SPECTRAL ENERGY BALANCE where the coecient is to be determined depending on models for  (k) [22{25]. Substituting J and  from Eqs. (6) and (3) into the spectral energy balance in Eq. (1) For a homogeneous, isotropic turbulent ow of a New- yields tonian, incompressible viscous uid, the spectral energy balance per unit mass of uid is [3, 4, 21] @E = p 2k E(k) @E(k) @t = p(k) + #(k) (k); (1) @t @ kE(k) k @E(k) : (7) where E(k) is the turbulent kinetic energy per unit wave @k  (k)  (k) @k number k, p(k) is the production spectrum here assumed Depending on the choices made about  (k), a general to be concentrated at k = k , #(k) is the energy transfer class of non-linear di usion models for J (k) can be re- spectrum, (k) is the viscous dissipation spectrum, and covered. Here, a  (k) that is linked to E(k) is adopted, k is the wavenumber or inverse eddy-size. The normaliz- ing property E(k)dk = (1=2)K de nes the turbulent kinetic energy K . Eq. (1) makes no other assumptions 1=2 (k) = k E(k) : (8) about the velocity statistics other than homogeneity and isotropy. Because # is a scale-wise transport and cannot For the inertial subrange scales, the Kolmogorov [26] contribute to production or destruction of K , it satis es 1 scaling (hereafter referred to as K41 scaling) given by the integral constraint #dk = 0. It can be expressed 2=3 5=3 E(k) = C  k is expected to hold resulting in as the gradient of an energy ux J , 1=2 1=3 2=3 (k) = C  k (i.e. Onsager's relaxation time @J [10]), where C = 1:55 is the Kolmogorov constant. Due #(k) = ; (2) to the dissipative anomaly [27], lim  is nite, so that @k !0 3 1=4 1 in this limit  = ( =) ! 0, ( ) ! 1. We may K K while the scale-wise viscous dissipation rate is given by then estimate the total time for energy to be passed from (k) = 2k E(k): (3) a nite k to an in nitely high wavenumber by Integrating Eq. (1) over k yields the energy balance of dk 3 1 2=3 1=3 (k) = p k  < 1: (9) K , k 2 1=k o dK = w ; (4) Eq. (9) implies that the steps in the energy cascade dt rapidly accelerate such that (if not interrupted by the where w = p(k)dk is the rate of work done on the action of viscosity at a nite wavenumber) the time for uid to produce turbulence and  is the dissipation rate energy to be passed to an in nitely high wavenumber of K , is nite. This nding, originally put forth by Onsager [10], foreshadows the nite time singularity in the invis- = (k)dk: (5) cid limit for such classes of  (k) models [28]. The  (k) in (8) is also singled out because it recovers the well-studied Leith's non-linear di usion approximation [29{31], The concomitant balance for internal energy U is then dU=dt =  q, where q is the heat loss to the envi- ronment. Because of uid incompressibility, tempera- @ E(k) 1 @E(k) 11=2 k E(k) = 3 2 ture uctuations resulting from dissipation have no feed- @k k k @k back on the dynamics of the turbulence, including the @E p 2k E(k) : (10) energy cascade. Finally, the entropy balance is given by @t dS=dt = q=T + , where T is the absolute temperature and  = =T is the entropy production. It is assumed When = 2, the conventional form of Leith's model be- that q is immediately delivered to a surrounding environ- comes evident [21, 29]. The latter recovers the so-called ment, acting as a thermal bath at the same temperature, warm cascade condition (i.e. a steady equipartitioned en- thus ensuring isothermal conditions. ergy spectrum, 8k : E(k) / k ) originally derived by Lee Returning to the spectral energy balance, a closure [32] under speci c conditions [30, 31, 33]. Leith's model of minimal complexity that preserves both direct energy was also derived from the so-called direct-interaction ap- cascade and an inverse cascade (or backscatter) may be proximation when a number of simpli cations are made obtained by representing the contributions to J as scale- [31]. wise drift and a di usion term linked by a timescale of For stationary conditions, and far from the production eddy relaxation  (k). A exible form for such a closure and viscous subranges, @E=@t p 2k E(k) = 0, the is proposed here as solution to the spectral budget reduces to kE(k) k @E(k) 5 3 3 J = ; (6) 2 2 E(k) = C k + C k : (11) 1 2 (k)  (k) @k 3 3=2 If p is injected at k = k , then for k > k C = C i i 1 o necessitating C = 0 to recover K41 inertial subrange scaling (also referred to as the cold cascade). For k < k , 11=3 2=3 C = 0 and C = C  k is set by the continuity 1 2 o of E(k) at k to achieve a warm cascade for = 2 [33]. The E(k)  k is also compatible with the well-known Sa man spectrum [28, 34, 35], a scaling law derived from considerations (continuity and smoothness) of how E(k) is approached as k ! 0. In the presence of viscous dissipation, the spectral bud- get equation is not analytically solvable, however, we nu- merically con rm that 5 3 3=2 3 E (k)  C k + C k f (k ) (12) o 2  K FIG. 1. The balance between energy transfer # = dJ=dk reasonably approximates the spectral energy budget, as and viscous dissipation (k) across scale based the empirical 4=3 shown in Fig 1. Here, f (k ) = exp (k ) is K K spectrum in Eq. (12) for k  = 10 . Here, = 0.33 results 5 i K the Pao correction [36] reshaping the k spectrum for in an acceptable spectral energy balance closure at steady k > 0:1 [37]. The E (k) from Eq. (12) implies that K  state. The numerical value of here di ers from the original (k) in Eq. (8) increases within the viscous sub-range Pao constant because of the choices made when deriving  (k). 3=4 when k > , which is not physically plausible. The inset shows the one-to-one correlation between #(k) and (k). The increase in  (k) is expected when E (k) decreases faster than k with increasing k. Hence, an amendment proposed by Batchelor [38] was used in the calculations p 1 where dW is the Wiener increment and b(k) = 2k . featured in Fig 1 whereby the straining rate (/  (k) ) at This equation is subjected to a unit rate of birth at k = k k is assumed to be uniform beyond scales commensurate and a state dependent killing term with rate 2k [40]. with 1= . This amendment revises the model for  (k) Since the FPE is written in the so-called transport form, as 8 the interpretation of the multiplicative term is the one of 5 1 > 2 2 C k f (k ) k < k >  K K i K Hanggi-Klimontovich [41{43]. 1 2 1 3 2 With this formal correspondence, the statistics of irre- 3 3 2 4 (k) = C  k f (k ) k  < k o  K i K K versiblity of Eq. (14) can now be analyzed [19, 43] for k > ; K K C steady-state homogeneous and isotropic turbulence with (13) energy injected at k and transported on average towards 1=2 where  = (=) is the Kolmogorov time scale. higher wavenumbers where dissipation takes place. This allows an illustration of the uctuation theorem for fully developed turbulence uctuations, linking the turbulent III. FLUCTUATION THEOREM entropy balance at k to those of forward and backward energy cascades. Fig 2a shows a numerical realization of this process in which energy packets are injected at The spectral budget in Eq. (7) can be interpreted as a k  = 10 after the termination of trajectories by dis- nonlinear Fokker-Planck equation (FPE) with scalewise i K sipation. While the energy injection at k is only related separated source/sink terms. Accordingly, the underly- ing cascade can be expressed as a stochastic process [39], statistically to the dissipation, the immediate re-injection after killing adopted here for convenience of simulation whereby trajectories represent time histories of energy packets (eddies) traveling in k-space driven by advection and visualization preserves the steady state PDF. As shown in Fig 2b, the steady state PDF of the k time and di usion, until they disappear by virtue of a killing term linked to the action of viscous dissipation. The lat- series corresponds to the empirical spectrum in Eq. (12). The injection of energy at lower k and the dissipation ter term absorbs trajectories as a state-dependent Pois- sink at higher k produce a non-zero average current and son process with a rate 2k [40]. a non-equilibrium steady state (NESS) current of energy If the steady-state solution is known (Eq. 12), the cor- towards smaller scales. The stochastic uctuating veloc- responding drift and di usion for the position in k space ity kjk is a random variable with mean current velocity of the energy packet can be formulated as a function of k. given by [43] Thus, a Langevin equation that ensures that the steady- state probability density function (PDF) abides by Eq. J (k) dE 1 2 1 1 (12) is v (k) =  2k k E  ; (15) NESS E(k) dk and a fundamental FT-type symmetry [19]. The degree dk = k 4 k  dt + b(k)dW; (14) dk of irreversibility of the NESS resulting from the cascade 4 of the path measures between forward and backward cas- cade, that is the path measures P ; P the forward and f b backward PDFs of k(t ) evaluated along stochastic tra- jectories that take place on a common interval [0; t] in steady state conditions [43, 47], is * + Z Z t 1 0 0 0 ln = E(k; t )(k; t )dkdt : (17) 0 0 Thus, S can be interpreted as a scale-wise `turbulence entropy', which together with TKE may be used to de ne an extended turbulence thermodynamics as in Richard- son's [6], providing a measure of the number of turbulent states at wave number k to be linked to the correspond- ing portion of TKE. Neglecting momentarily the viscous subrange, TKE can be linked to the area under the spec- trum given in Eq. (11), namely FIG. 2. (a) A numerical realization of the stochastic process 1= given by Eq. (14). The trajectories are terminated following a K =2 Edk = state-dependent Poisson process with rate 2k and initiated 0 (18) at k  = 10 . (b) The steady-state PDF from the numeri- i K 1=2 2 2 2 3C  k Re Re ; cal simulation and the approximate spectrum in Eq. (12) are i 5=3 compared. For reference, the red lines show k (K41 or cold cascade) and k (warm cascade or Sa man spectrum). where Re = l u= is a Reynolds number formed from a 1=3 characteristic length l = 1=k and velocity u = (=k ) . i i i 3=4 This de nition ensures that l = = Re consistent i K towards dissipation may be given by the rate of energy with expectations for many turbulent ows [14]. The transfer to smaller scales. According to the formalism integrated entropy S = S (k)dk can also be obtained I t of stochastic thermodynamics [44, 45], the non-zero cur- as rent velocity may be associated with a positive `turbulent entropy' production rate  [43], 1= 2 2 v (k) dE NESS  S = ln Edk = 2 1 I (k) = 2  2E E  : b(k) dk (19) 11 5 (16) 3=4 2 3=4 k k Re + ln C  k Re : i i o i 2=3 5=3 3 3 In the inertial sub-range where E(k) = C  k 3=2 (i.e. f (k )  1), J = (11=3)C  (a constant), K o p p One objection that can be raised here is that S is 1=3 5=3 1=3 2=3 v = C  k and  = (121=9) C  k . NESS o o not an entropy in the sense of Clausius, but an extended These scaling laws have also been con rmed by simu- entropy related to macroscopic turbulent uctuations. lations (not shown) of the stochastic process in Eq. (14), However, similar to classical thermodynamics, we may Eqs. (15) and (16) using the model spectrum and relax- consider the TKE to correspond to an internal kinetic en- ation time scale in Eqs. (12) and (13). ergy of turbulence, so that dS =dK = 1=T , analogous to I I an e ective temperature for the turbulent system. When such a de nition is combined with Eq 18, the extended IV. AN EXTENDED TURBULENCE turbulence thermodynamics can proceed as follows: THERMODYNAMICS The term (k)  0 can be shown to be the source dK dS T = = term in the balance equation for the turbulent entropy I dRe dRe S = ln E [46]. In this view, the asymmetry in the 5=4 1=2 (20) turbulence cascade that not only transports energy for- Re Re 18C  k : ward to higher wave numbers (the forward turbulence o i 3=4 8 + 3 ln C  k Re cascade) but also backward (the so-called back scatter) o i contributing energy to larger eddies, is also linked to a turbulent entropy production. Speci cally, if one consid- An integrated positive turbulence-entropy production ers the (random) path of a turbulent energy packet over for the integrated turbulence entropy, S , can be readily a period of time [0; t], the ensemble average of the ratio obtained assuming that molecular dissipation primarily 5 V. CONCLUSIONS The uctuation theorem has been used to analytically link the shape of the energy spectrum with the imbalance between forward and backward probabilities of energy packets moving scale-wise in time across the energy cas- cade. The di erence between these two aforementioned probabilities is the main cause why the `detailed balance' or `microscopic reversibility' (i.e. at equilibrium, each el- ementary process is in equilibrium with its reverse pro- cess) is not applicable to turbulence, and why turbulent uctuations are presumed to be far from equilibrium. The previous result unfolded a connection between the turbulent entropy production rate measuring the e ec- tive spreading of energy-packet trajectories in the cas- cade, the thermodynamic entropy production, and the Reynolds number Re for an externally prescribed injec- tion scale 1=k (often dictated by boundary conditions or geometry). As rst pointed out by Landau [48] and sub- stantiated in later studies [49], the nite Re is an indica- tor of the number of degrees of freedom of the turbulence 9=4 cascade, N  (l = )  Re . d i K An additional foresight from this analysis is that the shape of the spectrum at low k. It is shown here that the Sa man spectrum is linked to v = 0 and  = 0 NESS FIG. 3. The relation between entropy S (a) and turbulent (no scale-wise entropy production and the detailed bal- temperature T (b) with TKE K . For these results, it was ance is satis ed as expected for warm cascades). From 7 2 assumed that k = 1 m and  = 8:95  10 m =s. J in Eq. (6), the condition for v = J=E 6= 0 assum- ing E(k) / k can now be derived for the large scales (k < k ). With > 0, the condition dJ=dk > 0 im- posed by the energy balance necessitates > (=2 for acts in the neighborhood of k = 1= . For k < k where K i the Sa man spectrum and the associated Leith's model) E  C k and f (k )  1, v  0 and   0. As 2  K NESS for k=k < 1. It also follows that J < 0 (or v < 0) i NESS a result, when < , a state where the current towards larger scales is caused by the dominance of the backscatter over the forward drift. From the perspective of = 2 (i.e. Leith's model), the Sa man ( = 2) spectrum results in 1= dJ=dk = 0 whereas the Batchelor [50] spectrum ( = 4) = E dk = yields dJ=dk > 0 (i.e. forward drift still dominates over (21) backscattering). However, the Karman spectrum [35, 51] 3=2 3 4 3 C  k Re ln Re: o i often used in reshaping the inertial subrange spectrum at production scales in boundary-layer turbulence yields a non-monotonic dJ=dk in the rising limb of E(k) as k ! k . In the limit of high Re the above relationships approach i 2 3 3=4 to K / Re ,  / Re , S / Re , and T / These considerations have also led to a new perspec- I I I 5=4+ Re , where and are deviations due to the log- tive on the turbulence-thermodynamics formalism link- 1 2 arithmic terms. At Re  1, the spectrum within the iner- ing the emergence of turbulent modes to store disorderly tial subrange vanishes with production scales being com- kinetic energy to key macroscopic quanti es such as the mensurate to the Kolmogorov microscale ( k  1). For Reynolds number and the turbulence temperature. It K i this case,   0 (detailed balance holds) although will be of interest to compute S and T based on an en- I I I the cascade still generates small K , S , and T . The key ergy spectrum including intermittency corrections, to as- I I quantities in this turbulent thermodynamics are plotted sess how intermittency might play a role in the proposed in Fig. 3. The fundamental equation, S = S (K ) shows extended thermodynamics. One might also conjecture I I a downward concavity that ensures entropy production the existence of an extended global turbulence pressure by `combining' turbulent ows of di erent TKE, while to link ow con gurations to kinetic energy and entering the dependence of T on K shows a turbulent TKE ca- as a natural variable in a Gibbs turbulent free energy to pacity (dT =dK ) that is not constant but decays with the provide a uni ed criterion for turbulent transition and Reynolds number. development. It is hoped that future investigations will 6 contribute further elements to picture of turbulence as 1338694 and the Carbon Mitigation Initiative at Prince- a nonequilibrium phase transition, of which several ele- ton University. M.H acknowledges support from the ments are beginning to emerge [52]. Princeton Institute for International and Regional Stud- ies and the Princeton Environmental Institute. G.K. ac- knowledges support from the NSF grants AGS-1644382 ACKNOWLEDGMENTS and IOS-1754893. A.P. acknowledges support from the US National Sci- ence Foundation (NSF) grants EAR-1331846 and EAR- [1] R. L. Panton, Incompressible ow (John Wiley & Sons, [26] A. N. Kolmogorov, Dokl. Akad. Nauk SSSR 30, 299 2006). (1941). [2] K. R. Sreenivasan, Reviews of Modern Physics 71, S383 [27] G. Falkovich, K. Gaw edzki, and M. Vergassola, Rev. (1999). Mod. Phys. 73, 913 (2001). [3] G. Batchelor, \The theory of homogeneous turbulence," [28] P. A. Davidson, Turbulence: an introduction for scien- (Cambridge Univ Press, London, 1953) p. 197. tists and engineers (Oxford university press, 2015) p. 656 [4] A. Monin and A. Yaglom, Statistical uid mechanics: pp. Mechanics of turbulence (MIT Press, Cambridge, Mass, [29] C. Leith, Physics of Fluids (1958-1988) 11, 1612 (1968). 1975) p. 874. [30] D. Besnard, F. Harlow, R. Rauenzahn, and C. Zemach, [5] J. Hinze, in Turbulence: An introduction to its mecha- Theoretical and computational uid dynamics 8, 1 nism and theory (McGraw Hill, New York, 1959) p. 586. (1996). [6] L. F. Richardson, Proc. R. Soc. Lond. A 97, 354 (1920). [31] T. Clark, R. Rubinstein, and J. Weinstock, Journal of [7] A. K. Blackadar, Journal of Meteorology 12, 165 (1955). Turbulence , N35 (2009). [8] V. Nevzglyadov, in Soviet Physics Doklady, Vol. 5 (1961) [32] T. Lee, Quarterly of Applied Mathematics 10, 69 (1952). p. 1172. [33] C. Connaughton and S. Nazarenko, Physical review let- [9] N. Reinke, A. Fuchs, D. Nickelsen, and J. Peinke, Journal ters 92, 044501 (2004). of Fluid Mechanics 848, 117 (2018). [34] P. Sa man, Journal of Fluid Mechanics 27, 581 (1967). [10] G. L. Eyink and K. R. Sreenivasan, Reviews of modern [35] J. Meyers and C. Meneveau, Physics of Fluids 20, 065109 physics 78, 87 (2006). (2008). [11] H. W. Liepmann, American Scientist 67, 221 (1979). [36] Y.-H. Pao, Physics of Fluids (1958-1988) 8, 1063 (1965). [12] D. J. Evans and D. J. Searles, Advances in Physics 51, [37] A. G. Lamorgese, D. A. Caughey, and S. B. Pope, 1529 (2002). Physics of Fluids (1994-present) 17, 015106 (2005). [13] G. Gallavotti and E. G. D. Cohen, Journal of Statistical [38] G. Batchelor, Journal of Fluid Mechanics 5, 113 (1959). Physics 80, 931 (1995). [39] C. W. Gardiner, Handbook of stochastic methods for [14] H. Tennekes, J. L. Lumley, J. Lumley, et al., A rst physics, chemistry and the natural sciences (Springer- course in turbulence (MIT press, 1972) p. 300 pp. Verlag, 1994). [15] M. Carbone and A. D. Bragg, Journal of Fluid Mechanics [40] E. Daly and A. Porporato, Physical Review E 75, 011119 883 (2020). (2007). [16] H. Xu, A. Pumir, G. Falkovich, E. Bodenschatz, [41] P. Hanggi, Helvetica Physica Acta 51, 183 (1978). M. Shats, H. Xia, N. Francois, and G. Bo etta, Pro- [42] Y. L. Klimontovich, Physica A: Statistical Mechanics and ceedings of the National Academy of Sciences 111, 7558 its Applications 163, 515 (1990). (2014). [43] A. Porporato, P. Kramer, M. Cassiani, E. Daly, and [17] A. Fuchs, S. Queir os, P. Lind, A. Girard, F. Bouchet, J. Mattingly, Physical Review E 84, 041142 (2011). M. W achter, and J. Peinke, Physical Review Fluids 5, [44] U. Seifert, Reports on progress in physics 75, 126001 034602 (2020). (2012). [18] D. J. Evans, E. G. D. Cohen, and G. P. Morriss, Physical [45] C. Jarzynski, Annu. Rev. Condens. Matter Phys. 2, 329 review letters 71, 2401 (1993). (2011). [19] A. Porporato, J. R. Rigby, and E. Daly, Physical Review [46] U. Seifert, Physical review letters 95, 040602 (2005). Letters 98, 094101 (2007). [47] C. Maes and K. Neto cny,  Journal of statistical physics [20] E. Zorzetto, A. Bragg, and G. Katul, Physical Review 110, 269 (2003). Fluids 3, 094604 (2018). [48] L. Landau and E. Lifshitz, Fluid mechanics (Addison- [21] S. Panchev, Random Functions and Turbulence (Perga- Wesley Pub., 1959) p. 532 pp. mon Press, 1971) p. 444. [49] P. Constantin, C. Foias, O. P. Manley, and R. Temam, [22] L. Onsager, Il Nuovo Cimento (1943-1954) 6, 279 (1949). Journal of Fluid Mechanics 150, 427 (1985). [23] S. Corrsin, Physics of Fluids 7, 1156 (1964). [50] G. K. Batchelor and I. Proudman, Philosophical Trans- [24] L. Danaila and R. A. Antonia, Physics of Fluids 21, actions of the Royal Society of London. Series A, Math- 111702 (2009). ematical and Physical Sciences 248, 369 (1956). [25] D. Poggi, G. Katul, and B. Vidakovic, Boundary-layer [51] S. B. Pope, Turbulent ows (Cambridge University Press, meteorology 139, 83 (2011). Cambridge, MA, 2001) p. 773 pp. 7 [52] N. Goldenfeld and H.-Y. Shih, Journal of Statistical Physics 167, 575 (2017). http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Physics arXiv (Cornell University)

Fluctuation-theorem and extended thermodynamics of turbulence

Download PDF
7 pages

Loading next page...
 
/lp/arxiv-cornell-university/fluctuation-theorem-and-extended-thermodynamics-of-turbulence-ZI0IRrQXnR

References (82)

ISSN
1364-5021
eISSN
ARCH-3341
DOI
10.1098/rspa.2020.0468
Publisher site
See Article on Publisher Site

Abstract

1, 2,y 3,z 4,x Amilcare Porporato, Milad Hooshyar, Andrew D Bragg, and Gabriel Katul Department of Civil and Environmental Engineering and Princeton Environmental Institute, Princeton University, Princeton, NJ 08544, USA Princeton Environmental Institute and Princeton Institute for International and Regional Studies, Princeton University, Princeton, NJ 08544, USA Department of Civil and Environmental Engineering, Duke University, Durham, NC 27708, USA Department of Civil and Environmental Engineering and Nicholas School of the Environment, Duke University, Durham, NC 27708, USA Turbulent ows are out-of-equilibrium because the energy supply at large scales and its dissipation by viscosity at small scales create a net transfer of energy among all scales. Here, the energy cascade is approximated by a combined contribution of a forward drift and di usion that recover accepted phenomenological theories of turbulence. The uctuation theorem (FT) is then shown to describe the scale-wise statistics of forward and backward energy transfer and their connection to irreversibility and entropy production. The ensuing turbulence entropy may be used to formulate an extended turbulence thermodynamics. I. INTRODUCTION The statistical properties of turbulence di er from sys- tems near thermal equilibrium because the ux of en- ergy per unit mass is supplied at scales much larger than It is perhaps not coincidental that one of the most in- the scales at which energy is dissipated by the action uential experiments in the history of thermodynamics of viscosity, resulting in an energy ux (cascade) across is also a turbulence experiment. In 1849, James Prescott all scales. Such a transport is linked to multiple pro- Joule used a stirrer to show that the shaft work on a uid cesses, including vortex stretching, self-ampli cation of ends up increasing its internal energy, thereby demon- the strain-rate and viscous di usion [14, 15]. One of the strating the equivalence of heat and work. Dealing with de ning features of the turbulence cascade is that the the generation of turbulent kinetic energy (K ) and its probability of forward and backward transitions between subsequent dissipation rate () by viscosity, the Joule two energetic states at a given scale are not identical (i.e., experiment also o ers a modern link between thermody- a scale-wise `detailed balance' is not applicable [16, 17]). namics, a theory of the macroscopic e ects of microscopic uctuations, and non-equilibrium uctuations, of which The objective of this work is to illustrate that non- turbulence is a quintessential example. On the one hand, equilibrium thermodynamics, and in particular, the uc- the disparity between microscopic and macroscopic uc- tuation theorem can be extended to describe the behavior tuations appears un-reconcilable. Thermodynamic uc- of the turbulent energy cascade. While the net transfer of tuations are so small to allow a mathematical descrip- energy from large to small scales is prevalent, it is shown tion of uids as a continuum. For this reason, turbulence here that back-scatter of energy and its connection to is conveniently described by the Navier-Stokes equations time-scale irreversibility obeys the statistics predicted by assuming local thermodynamic equilibrium [1]. Turbu- the uctuation theorem [13, 17{20]. To provide a phys- lent uctuations are thus of macroscopic nature and tech- ical context, a turbulent ow conceptually analogous to nically outside the scope of traditional thermodynamics Joule's original experiment is used, where work is done [2]. On the other hand, the random-like nature of tur- on the uid system to generate K in a narrow band of bulence [3{5] invites a thermodynamic formalism to the scales, which is then dissipated as heat thereby raising problem of turbulence, including the eddy thermodynam- the internal energy. In the analysis here, constant en- ics of Richardson [6], Blackadar [7] and others [8, 9], as ergy is externally supplied at a pre-selected scale much well as the Onsager analysis of 2D turbulence [10]. Liep- larger than the Kolmogorov length scale ( ) where vis- mann [11] asserted that `turbulence can be de ned by cous e ects are signi cant. At steady state, the energy a statement of impotence reminiscent of the second law cascade develops in a manner where the energy injection of thermodynamics'. More recently, a number of stud- rate is balanced by viscous dissipation rate as in Joule's ies have argued that the uctuation theorem derived for experiment. The uctuation theorem is then applied to small systems [12, 13] can be partly applied to describe describe the forward and backward probabilities of en- macroscopic uctuations so as to explore their time re- ergy packets moving scale-wise in time through the en- versibility at multiple scales, including turbulence. ergy cascade. For analytical tractability and to illustrate connections with the uctuation theorem, simpli ed clo- sure schemes for the energy transfer rate across scales are employed. These closure schemes o er plausible expres- aporpora@princeton.edu hooshyar@princeton.edu sions for the energy cascade that are consistent with a andrew.bragg@duke.edu wide range of experiments and theories on locally homo- gaby@duke.edu geneous and isotropic turbulence. arXiv:2005.01522v1 [physics.flu-dyn] 4 May 2020 2 II. SPECTRAL ENERGY BALANCE where the coecient is to be determined depending on models for  (k) [22{25]. Substituting J and  from Eqs. (6) and (3) into the spectral energy balance in Eq. (1) For a homogeneous, isotropic turbulent ow of a New- yields tonian, incompressible viscous uid, the spectral energy balance per unit mass of uid is [3, 4, 21] @E = p 2k E(k) @E(k) @t = p(k) + #(k) (k); (1) @t @ kE(k) k @E(k) : (7) where E(k) is the turbulent kinetic energy per unit wave @k  (k)  (k) @k number k, p(k) is the production spectrum here assumed Depending on the choices made about  (k), a general to be concentrated at k = k , #(k) is the energy transfer class of non-linear di usion models for J (k) can be re- spectrum, (k) is the viscous dissipation spectrum, and covered. Here, a  (k) that is linked to E(k) is adopted, k is the wavenumber or inverse eddy-size. The normaliz- ing property E(k)dk = (1=2)K de nes the turbulent kinetic energy K . Eq. (1) makes no other assumptions 1=2 (k) = k E(k) : (8) about the velocity statistics other than homogeneity and isotropy. Because # is a scale-wise transport and cannot For the inertial subrange scales, the Kolmogorov [26] contribute to production or destruction of K , it satis es 1 scaling (hereafter referred to as K41 scaling) given by the integral constraint #dk = 0. It can be expressed 2=3 5=3 E(k) = C  k is expected to hold resulting in as the gradient of an energy ux J , 1=2 1=3 2=3 (k) = C  k (i.e. Onsager's relaxation time @J [10]), where C = 1:55 is the Kolmogorov constant. Due #(k) = ; (2) to the dissipative anomaly [27], lim  is nite, so that @k !0 3 1=4 1 in this limit  = ( =) ! 0, ( ) ! 1. We may K K while the scale-wise viscous dissipation rate is given by then estimate the total time for energy to be passed from (k) = 2k E(k): (3) a nite k to an in nitely high wavenumber by Integrating Eq. (1) over k yields the energy balance of dk 3 1 2=3 1=3 (k) = p k  < 1: (9) K , k 2 1=k o dK = w ; (4) Eq. (9) implies that the steps in the energy cascade dt rapidly accelerate such that (if not interrupted by the where w = p(k)dk is the rate of work done on the action of viscosity at a nite wavenumber) the time for uid to produce turbulence and  is the dissipation rate energy to be passed to an in nitely high wavenumber of K , is nite. This nding, originally put forth by Onsager [10], foreshadows the nite time singularity in the invis- = (k)dk: (5) cid limit for such classes of  (k) models [28]. The  (k) in (8) is also singled out because it recovers the well-studied Leith's non-linear di usion approximation [29{31], The concomitant balance for internal energy U is then dU=dt =  q, where q is the heat loss to the envi- ronment. Because of uid incompressibility, tempera- @ E(k) 1 @E(k) 11=2 k E(k) = 3 2 ture uctuations resulting from dissipation have no feed- @k k k @k back on the dynamics of the turbulence, including the @E p 2k E(k) : (10) energy cascade. Finally, the entropy balance is given by @t dS=dt = q=T + , where T is the absolute temperature and  = =T is the entropy production. It is assumed When = 2, the conventional form of Leith's model be- that q is immediately delivered to a surrounding environ- comes evident [21, 29]. The latter recovers the so-called ment, acting as a thermal bath at the same temperature, warm cascade condition (i.e. a steady equipartitioned en- thus ensuring isothermal conditions. ergy spectrum, 8k : E(k) / k ) originally derived by Lee Returning to the spectral energy balance, a closure [32] under speci c conditions [30, 31, 33]. Leith's model of minimal complexity that preserves both direct energy was also derived from the so-called direct-interaction ap- cascade and an inverse cascade (or backscatter) may be proximation when a number of simpli cations are made obtained by representing the contributions to J as scale- [31]. wise drift and a di usion term linked by a timescale of For stationary conditions, and far from the production eddy relaxation  (k). A exible form for such a closure and viscous subranges, @E=@t p 2k E(k) = 0, the is proposed here as solution to the spectral budget reduces to kE(k) k @E(k) 5 3 3 J = ; (6) 2 2 E(k) = C k + C k : (11) 1 2 (k)  (k) @k 3 3=2 If p is injected at k = k , then for k > k C = C i i 1 o necessitating C = 0 to recover K41 inertial subrange scaling (also referred to as the cold cascade). For k < k , 11=3 2=3 C = 0 and C = C  k is set by the continuity 1 2 o of E(k) at k to achieve a warm cascade for = 2 [33]. The E(k)  k is also compatible with the well-known Sa man spectrum [28, 34, 35], a scaling law derived from considerations (continuity and smoothness) of how E(k) is approached as k ! 0. In the presence of viscous dissipation, the spectral bud- get equation is not analytically solvable, however, we nu- merically con rm that 5 3 3=2 3 E (k)  C k + C k f (k ) (12) o 2  K FIG. 1. The balance between energy transfer # = dJ=dk reasonably approximates the spectral energy budget, as and viscous dissipation (k) across scale based the empirical 4=3 shown in Fig 1. Here, f (k ) = exp (k ) is K K spectrum in Eq. (12) for k  = 10 . Here, = 0.33 results 5 i K the Pao correction [36] reshaping the k spectrum for in an acceptable spectral energy balance closure at steady k > 0:1 [37]. The E (k) from Eq. (12) implies that K  state. The numerical value of here di ers from the original (k) in Eq. (8) increases within the viscous sub-range Pao constant because of the choices made when deriving  (k). 3=4 when k > , which is not physically plausible. The inset shows the one-to-one correlation between #(k) and (k). The increase in  (k) is expected when E (k) decreases faster than k with increasing k. Hence, an amendment proposed by Batchelor [38] was used in the calculations p 1 where dW is the Wiener increment and b(k) = 2k . featured in Fig 1 whereby the straining rate (/  (k) ) at This equation is subjected to a unit rate of birth at k = k k is assumed to be uniform beyond scales commensurate and a state dependent killing term with rate 2k [40]. with 1= . This amendment revises the model for  (k) Since the FPE is written in the so-called transport form, as 8 the interpretation of the multiplicative term is the one of 5 1 > 2 2 C k f (k ) k < k >  K K i K Hanggi-Klimontovich [41{43]. 1 2 1 3 2 With this formal correspondence, the statistics of irre- 3 3 2 4 (k) = C  k f (k ) k  < k o  K i K K versiblity of Eq. (14) can now be analyzed [19, 43] for k > ; K K C steady-state homogeneous and isotropic turbulence with (13) energy injected at k and transported on average towards 1=2 where  = (=) is the Kolmogorov time scale. higher wavenumbers where dissipation takes place. This allows an illustration of the uctuation theorem for fully developed turbulence uctuations, linking the turbulent III. FLUCTUATION THEOREM entropy balance at k to those of forward and backward energy cascades. Fig 2a shows a numerical realization of this process in which energy packets are injected at The spectral budget in Eq. (7) can be interpreted as a k  = 10 after the termination of trajectories by dis- nonlinear Fokker-Planck equation (FPE) with scalewise i K sipation. While the energy injection at k is only related separated source/sink terms. Accordingly, the underly- ing cascade can be expressed as a stochastic process [39], statistically to the dissipation, the immediate re-injection after killing adopted here for convenience of simulation whereby trajectories represent time histories of energy packets (eddies) traveling in k-space driven by advection and visualization preserves the steady state PDF. As shown in Fig 2b, the steady state PDF of the k time and di usion, until they disappear by virtue of a killing term linked to the action of viscous dissipation. The lat- series corresponds to the empirical spectrum in Eq. (12). The injection of energy at lower k and the dissipation ter term absorbs trajectories as a state-dependent Pois- sink at higher k produce a non-zero average current and son process with a rate 2k [40]. a non-equilibrium steady state (NESS) current of energy If the steady-state solution is known (Eq. 12), the cor- towards smaller scales. The stochastic uctuating veloc- responding drift and di usion for the position in k space ity kjk is a random variable with mean current velocity of the energy packet can be formulated as a function of k. given by [43] Thus, a Langevin equation that ensures that the steady- state probability density function (PDF) abides by Eq. J (k) dE 1 2 1 1 (12) is v (k) =  2k k E  ; (15) NESS E(k) dk and a fundamental FT-type symmetry [19]. The degree dk = k 4 k  dt + b(k)dW; (14) dk of irreversibility of the NESS resulting from the cascade 4 of the path measures between forward and backward cas- cade, that is the path measures P ; P the forward and f b backward PDFs of k(t ) evaluated along stochastic tra- jectories that take place on a common interval [0; t] in steady state conditions [43, 47], is * + Z Z t 1 0 0 0 ln = E(k; t )(k; t )dkdt : (17) 0 0 Thus, S can be interpreted as a scale-wise `turbulence entropy', which together with TKE may be used to de ne an extended turbulence thermodynamics as in Richard- son's [6], providing a measure of the number of turbulent states at wave number k to be linked to the correspond- ing portion of TKE. Neglecting momentarily the viscous subrange, TKE can be linked to the area under the spec- trum given in Eq. (11), namely FIG. 2. (a) A numerical realization of the stochastic process 1= given by Eq. (14). The trajectories are terminated following a K =2 Edk = state-dependent Poisson process with rate 2k and initiated 0 (18) at k  = 10 . (b) The steady-state PDF from the numeri- i K 1=2 2 2 2 3C  k Re Re ; cal simulation and the approximate spectrum in Eq. (12) are i 5=3 compared. For reference, the red lines show k (K41 or cold cascade) and k (warm cascade or Sa man spectrum). where Re = l u= is a Reynolds number formed from a 1=3 characteristic length l = 1=k and velocity u = (=k ) . i i i 3=4 This de nition ensures that l = = Re consistent i K towards dissipation may be given by the rate of energy with expectations for many turbulent ows [14]. The transfer to smaller scales. According to the formalism integrated entropy S = S (k)dk can also be obtained I t of stochastic thermodynamics [44, 45], the non-zero cur- as rent velocity may be associated with a positive `turbulent entropy' production rate  [43], 1= 2 2 v (k) dE NESS  S = ln Edk = 2 1 I (k) = 2  2E E  : b(k) dk (19) 11 5 (16) 3=4 2 3=4 k k Re + ln C  k Re : i i o i 2=3 5=3 3 3 In the inertial sub-range where E(k) = C  k 3=2 (i.e. f (k )  1), J = (11=3)C  (a constant), K o p p One objection that can be raised here is that S is 1=3 5=3 1=3 2=3 v = C  k and  = (121=9) C  k . NESS o o not an entropy in the sense of Clausius, but an extended These scaling laws have also been con rmed by simu- entropy related to macroscopic turbulent uctuations. lations (not shown) of the stochastic process in Eq. (14), However, similar to classical thermodynamics, we may Eqs. (15) and (16) using the model spectrum and relax- consider the TKE to correspond to an internal kinetic en- ation time scale in Eqs. (12) and (13). ergy of turbulence, so that dS =dK = 1=T , analogous to I I an e ective temperature for the turbulent system. When such a de nition is combined with Eq 18, the extended IV. AN EXTENDED TURBULENCE turbulence thermodynamics can proceed as follows: THERMODYNAMICS The term (k)  0 can be shown to be the source dK dS T = = term in the balance equation for the turbulent entropy I dRe dRe S = ln E [46]. In this view, the asymmetry in the 5=4 1=2 (20) turbulence cascade that not only transports energy for- Re Re 18C  k : ward to higher wave numbers (the forward turbulence o i 3=4 8 + 3 ln C  k Re cascade) but also backward (the so-called back scatter) o i contributing energy to larger eddies, is also linked to a turbulent entropy production. Speci cally, if one consid- An integrated positive turbulence-entropy production ers the (random) path of a turbulent energy packet over for the integrated turbulence entropy, S , can be readily a period of time [0; t], the ensemble average of the ratio obtained assuming that molecular dissipation primarily 5 V. CONCLUSIONS The uctuation theorem has been used to analytically link the shape of the energy spectrum with the imbalance between forward and backward probabilities of energy packets moving scale-wise in time across the energy cas- cade. The di erence between these two aforementioned probabilities is the main cause why the `detailed balance' or `microscopic reversibility' (i.e. at equilibrium, each el- ementary process is in equilibrium with its reverse pro- cess) is not applicable to turbulence, and why turbulent uctuations are presumed to be far from equilibrium. The previous result unfolded a connection between the turbulent entropy production rate measuring the e ec- tive spreading of energy-packet trajectories in the cas- cade, the thermodynamic entropy production, and the Reynolds number Re for an externally prescribed injec- tion scale 1=k (often dictated by boundary conditions or geometry). As rst pointed out by Landau [48] and sub- stantiated in later studies [49], the nite Re is an indica- tor of the number of degrees of freedom of the turbulence 9=4 cascade, N  (l = )  Re . d i K An additional foresight from this analysis is that the shape of the spectrum at low k. It is shown here that the Sa man spectrum is linked to v = 0 and  = 0 NESS FIG. 3. The relation between entropy S (a) and turbulent (no scale-wise entropy production and the detailed bal- temperature T (b) with TKE K . For these results, it was ance is satis ed as expected for warm cascades). From 7 2 assumed that k = 1 m and  = 8:95  10 m =s. J in Eq. (6), the condition for v = J=E 6= 0 assum- ing E(k) / k can now be derived for the large scales (k < k ). With > 0, the condition dJ=dk > 0 im- posed by the energy balance necessitates > (=2 for acts in the neighborhood of k = 1= . For k < k where K i the Sa man spectrum and the associated Leith's model) E  C k and f (k )  1, v  0 and   0. As 2  K NESS for k=k < 1. It also follows that J < 0 (or v < 0) i NESS a result, when < , a state where the current towards larger scales is caused by the dominance of the backscatter over the forward drift. From the perspective of = 2 (i.e. Leith's model), the Sa man ( = 2) spectrum results in 1= dJ=dk = 0 whereas the Batchelor [50] spectrum ( = 4) = E dk = yields dJ=dk > 0 (i.e. forward drift still dominates over (21) backscattering). However, the Karman spectrum [35, 51] 3=2 3 4 3 C  k Re ln Re: o i often used in reshaping the inertial subrange spectrum at production scales in boundary-layer turbulence yields a non-monotonic dJ=dk in the rising limb of E(k) as k ! k . In the limit of high Re the above relationships approach i 2 3 3=4 to K / Re ,  / Re , S / Re , and T / These considerations have also led to a new perspec- I I I 5=4+ Re , where and are deviations due to the log- tive on the turbulence-thermodynamics formalism link- 1 2 arithmic terms. At Re  1, the spectrum within the iner- ing the emergence of turbulent modes to store disorderly tial subrange vanishes with production scales being com- kinetic energy to key macroscopic quanti es such as the mensurate to the Kolmogorov microscale ( k  1). For Reynolds number and the turbulence temperature. It K i this case,   0 (detailed balance holds) although will be of interest to compute S and T based on an en- I I I the cascade still generates small K , S , and T . The key ergy spectrum including intermittency corrections, to as- I I quantities in this turbulent thermodynamics are plotted sess how intermittency might play a role in the proposed in Fig. 3. The fundamental equation, S = S (K ) shows extended thermodynamics. One might also conjecture I I a downward concavity that ensures entropy production the existence of an extended global turbulence pressure by `combining' turbulent ows of di erent TKE, while to link ow con gurations to kinetic energy and entering the dependence of T on K shows a turbulent TKE ca- as a natural variable in a Gibbs turbulent free energy to pacity (dT =dK ) that is not constant but decays with the provide a uni ed criterion for turbulent transition and Reynolds number. development. It is hoped that future investigations will 6 contribute further elements to picture of turbulence as 1338694 and the Carbon Mitigation Initiative at Prince- a nonequilibrium phase transition, of which several ele- ton University. M.H acknowledges support from the ments are beginning to emerge [52]. Princeton Institute for International and Regional Stud- ies and the Princeton Environmental Institute. G.K. ac- knowledges support from the NSF grants AGS-1644382 ACKNOWLEDGMENTS and IOS-1754893. A.P. acknowledges support from the US National Sci- ence Foundation (NSF) grants EAR-1331846 and EAR- [1] R. L. Panton, Incompressible ow (John Wiley & Sons, [26] A. N. Kolmogorov, Dokl. Akad. Nauk SSSR 30, 299 2006). (1941). [2] K. R. Sreenivasan, Reviews of Modern Physics 71, S383 [27] G. Falkovich, K. Gaw edzki, and M. Vergassola, Rev. (1999). Mod. Phys. 73, 913 (2001). [3] G. Batchelor, \The theory of homogeneous turbulence," [28] P. A. Davidson, Turbulence: an introduction for scien- (Cambridge Univ Press, London, 1953) p. 197. tists and engineers (Oxford university press, 2015) p. 656 [4] A. Monin and A. Yaglom, Statistical uid mechanics: pp. Mechanics of turbulence (MIT Press, Cambridge, Mass, [29] C. Leith, Physics of Fluids (1958-1988) 11, 1612 (1968). 1975) p. 874. [30] D. Besnard, F. Harlow, R. Rauenzahn, and C. Zemach, [5] J. Hinze, in Turbulence: An introduction to its mecha- Theoretical and computational uid dynamics 8, 1 nism and theory (McGraw Hill, New York, 1959) p. 586. (1996). [6] L. F. Richardson, Proc. R. Soc. Lond. A 97, 354 (1920). [31] T. Clark, R. Rubinstein, and J. Weinstock, Journal of [7] A. K. Blackadar, Journal of Meteorology 12, 165 (1955). Turbulence , N35 (2009). [8] V. Nevzglyadov, in Soviet Physics Doklady, Vol. 5 (1961) [32] T. Lee, Quarterly of Applied Mathematics 10, 69 (1952). p. 1172. [33] C. Connaughton and S. Nazarenko, Physical review let- [9] N. Reinke, A. Fuchs, D. Nickelsen, and J. Peinke, Journal ters 92, 044501 (2004). of Fluid Mechanics 848, 117 (2018). [34] P. Sa man, Journal of Fluid Mechanics 27, 581 (1967). [10] G. L. Eyink and K. R. Sreenivasan, Reviews of modern [35] J. Meyers and C. Meneveau, Physics of Fluids 20, 065109 physics 78, 87 (2006). (2008). [11] H. W. Liepmann, American Scientist 67, 221 (1979). [36] Y.-H. Pao, Physics of Fluids (1958-1988) 8, 1063 (1965). [12] D. J. Evans and D. J. Searles, Advances in Physics 51, [37] A. G. Lamorgese, D. A. Caughey, and S. B. Pope, 1529 (2002). Physics of Fluids (1994-present) 17, 015106 (2005). [13] G. Gallavotti and E. G. D. Cohen, Journal of Statistical [38] G. Batchelor, Journal of Fluid Mechanics 5, 113 (1959). Physics 80, 931 (1995). [39] C. W. Gardiner, Handbook of stochastic methods for [14] H. Tennekes, J. L. Lumley, J. Lumley, et al., A rst physics, chemistry and the natural sciences (Springer- course in turbulence (MIT press, 1972) p. 300 pp. Verlag, 1994). [15] M. Carbone and A. D. Bragg, Journal of Fluid Mechanics [40] E. Daly and A. Porporato, Physical Review E 75, 011119 883 (2020). (2007). [16] H. Xu, A. Pumir, G. Falkovich, E. Bodenschatz, [41] P. Hanggi, Helvetica Physica Acta 51, 183 (1978). M. Shats, H. Xia, N. Francois, and G. Bo etta, Pro- [42] Y. L. Klimontovich, Physica A: Statistical Mechanics and ceedings of the National Academy of Sciences 111, 7558 its Applications 163, 515 (1990). (2014). [43] A. Porporato, P. Kramer, M. Cassiani, E. Daly, and [17] A. Fuchs, S. Queir os, P. Lind, A. Girard, F. Bouchet, J. Mattingly, Physical Review E 84, 041142 (2011). M. W achter, and J. Peinke, Physical Review Fluids 5, [44] U. Seifert, Reports on progress in physics 75, 126001 034602 (2020). (2012). [18] D. J. Evans, E. G. D. Cohen, and G. P. Morriss, Physical [45] C. Jarzynski, Annu. Rev. Condens. Matter Phys. 2, 329 review letters 71, 2401 (1993). (2011). [19] A. Porporato, J. R. Rigby, and E. Daly, Physical Review [46] U. Seifert, Physical review letters 95, 040602 (2005). Letters 98, 094101 (2007). [47] C. Maes and K. Neto cny,  Journal of statistical physics [20] E. Zorzetto, A. Bragg, and G. Katul, Physical Review 110, 269 (2003). Fluids 3, 094604 (2018). [48] L. Landau and E. Lifshitz, Fluid mechanics (Addison- [21] S. Panchev, Random Functions and Turbulence (Perga- Wesley Pub., 1959) p. 532 pp. mon Press, 1971) p. 444. [49] P. Constantin, C. Foias, O. P. Manley, and R. Temam, [22] L. Onsager, Il Nuovo Cimento (1943-1954) 6, 279 (1949). Journal of Fluid Mechanics 150, 427 (1985). [23] S. Corrsin, Physics of Fluids 7, 1156 (1964). [50] G. K. Batchelor and I. Proudman, Philosophical Trans- [24] L. Danaila and R. A. Antonia, Physics of Fluids 21, actions of the Royal Society of London. Series A, Math- 111702 (2009). ematical and Physical Sciences 248, 369 (1956). [25] D. Poggi, G. Katul, and B. Vidakovic, Boundary-layer [51] S. B. Pope, Turbulent ows (Cambridge University Press, meteorology 139, 83 (2011). Cambridge, MA, 2001) p. 773 pp. 7 [52] N. Goldenfeld and H.-Y. Shih, Journal of Statistical Physics 167, 575 (2017).

Journal

Physics – arXiv (Cornell University)

Published: May 4, 2020

There are no references for this article.