betaFeedback

WorksPhysics and Astronomy

Energy spectrum of two-dimensional acoustic turbulence

Abstract

We report an exact unique constant-flux power-law analytical solution of the wave kinetic equation for the turbulent energy spectrum, E(k)=C1ε acs/kE(k)=C_1 \sqrt{\varepsilon\, a c_{\rm s} }/k, of acoustic waves in 2D with almost linear dispersion law, ωk=csk[1+(ak)2]ω_k = c_{\rm s} k[1+(ak)^2], ak≪1 ak \ll 1. Here ε\varepsilon is the energy flux over scales, and C1C_1 is the universal constant which was found analytically. Our theory describes, for example, acoustic turbulence in 2D Bose-Einstein condensates (BECs). The corresponding 3D counterpart of turbulent acoustic spectrum was found over half a century ago, however, due to the singularity in 2D, no solution has been obtained until now. We show the spectrum E(k)E(k) is realizable in direct numerical simulations of forced-dissipated Gross-Pitaevskii equation in the presence of strong condensate.

Waves in nonlinear systems interact and transfer energy along scales in a cascade process with a constant flux, creating an out-of-equilibrium state known as wave turbulence. When nonlinearity is small, the Weak-Wave Turbulence (WWT) theory provides a mathematical description of the system [1, 2]. The most common applications of this theory are capillary-gravity waves [3], Alfvén waves in magnetohydrodynamics [4], Langmuir waves in plasmas [5], inertial and internal waves in rotating stratified fluids [6, 7], Kelvin waves in vortices [8], elastic plates [9], gravitational waves [10] and density waves in Bose-Einstein condensates [11].

Note, that acoustic waves in ideal compressible fluids, one of the most common examples in Nature, are non-dispersive: their frequency ωk=csk\omega_k = c_s k is linear in wave number k≡∣k∣k \equiv|k|. Accordingly, the resonance conditions

ωk=ωk1+ωk2,k=k1+k2,(1)\omega_k = \omega_{k_1} + \omega_{k_2}, \qquad k = k_1 + k_2, \tag*{(1)}

allow for interaction of the waves with parallel wave vectors only: k∥k1∥k2\boldsymbol{k}\parallel\boldsymbol{k}_1\parallel\boldsymbol{k}_2 and in the same direction [2]. Therefore in the reference frame moving with the speed of sound csc_s in the direction of kk, k1k_1 and k2k_2 all wave packets are at the rest and their overlapping time τovr\tau_{\mathrm{ovr}} goes to infinity, allowing for wave steepening and breaking effects, which requires finite shock creation time τsh\tau_{\mathrm{sh}}. In the other words, the main assumption of the WWT theory, roughly formulated as τovr≪τsh\tau_{\mathrm{ovr}} \ll\tau_{\mathrm{sh}}, fails for dispersionless acoustic waves even even at small nonlinearity.

This has caused a long-standing debate whether WWT theory is applicable for their description [12], or alternatively, if acoustic waves should be viewed as a random collection of weak shocks leading to the Kadomtsev-Petviashvilli spectrum [13]. There is a handwaving argument—yet unsupported by rigorous proof—that the theory applies to 3D acoustic turbulence because the divergence of wave packets in 3D space plays a role similar to the wave dispersion in preventing wave breaking. It was argued first in [14] and later in [15], that the WWT description is indeed possible for 3D weak acoustic turbulence. However, the respective kinetic equation for the spectrum has to be modified so that interactions of non-collinear waves are described correctly. Concerning the 2D case, it is evident that WWT theory is not applicable in its classical form. Indeed, the main equation of the theory, the wave-kinetic equation, is singular and meaningless in the 2D case. The possibility of an alternative statistical description of weak non-dispersive acoustic 2D sound was claimed in [14], but it remains an unfinished task. Fortunately, in some important physical applications, 2D sound is regularized by weak dispersion effects, and the use of the classical WWT theory becomes again possible. One of such examples, which we will use in the present Letter, is the acoustic turbulence in 2D Bose-Einstein condensates (BEC).

Recent experiments with 3D BECs have succeeded in creating wave-turbulence states where measurements can be explained using the WWT theory in the fully dispersive limit [16, 17]. On the other hand, much interest has been devoted to studying the dynamics of vortices experimentally in 2D BECs, as such states are close to hydrodynamic 2D classical turbulence [18, 19]. Unfortunately, experiments of acoustic 2D BECs still lack and no theoretical predictions are available.

In this Letter, we develop the theory of weak wave turbulence of weakly dispersive 2D sound and obtain a modified wave kinetic equation. We derive a new stationary power-law flux spectrum of Kolmogorov-Zakharov type, which corresponds to a cascade of energy from large to small spatial scales. We then determine the flux spectrum exponent and the value of the prefactor constant analytically. This prediction can not be naively guessed by dimensional arguments, unlike the existing 3D results. Our theory is then tested and validated with numerical simulations of weakly nonlinear sound in the 2D forced-dissipated Gross-Pitaevskii (GP) equation (17) which is a dynamical model for BEC.

Let us consider the classical wave-kinetic equation describing evolution of the wave action spectrum nkn_k = n(k,t)n(k,t) (with kk and tt being the wavenumber and time) driven by three-wave resonant interactions [1, 2],

∂nk∂t=St⁡k(2)\frac{\partial n_k}{\partial t}=\operatorname{St}_k \tag*{(2)}

with the wave-collision integral

St⁡k=2π∫(R12k−R2k1−Rk12)dk1dk2,(3)\operatorname{St}_k=2\pi\int\left(\mathcal{R}_{12}^{k}-\mathcal{R}_{2k}^{1}-\mathcal{R}_{k1}^{2}\right)dk_1dk_2, \tag*{(3)}
R12k=N12kδ12k(ω12k),\mathcal{R}_{12}^{k}=\mathcal{N}_{12}^{k}\delta_{12}^{k}(\omega_{12}^{k}),
N12k=∣V12k∣2[nk1nk2−nknk1−nknk2],(4)\mathcal{N}_{12}^{k}=\left|V_{12}^{k}\right|^2\left[n_{k_1}n_{k_2}-n_kn_{k_1}-n_kn_{k_2}\right], \tag*{(4)}
δ12k=δ(k−k1−k2),δ(ω12k)=δ(ωk−ωk1−ωk2),\delta_{12}^{k}=\delta(k-k_1-k_2),\quad\delta(\omega_{12}^{k})=\delta(\omega_k-\omega_{k_1}-\omega_{k_2}),

where ωk=ω(k)\omega_k=\omega(k) and V12k=V(k,k1,k2)V_{12}^{k}=V(k,k_1,k_2) is a three-wave interaction amplitude, that depends on the particular type of waves.

The central object in turbulence is the 1D energy spectrum defined as the distribution of energy in k=∣k∣k=|\mathbf{k}|, so that the energy (per unit of mass) is defined as E=∫E(k)dkE=\int E(k)dk. It is related to the waveaction spectrum as E(k)=4πωkk2nkE(k)=4\pi\omega_k k^2n_k in 3D and E(k)=2πωkknkE(k)=2\pi\omega_kkn_k in 2D.

In this Letter, we consider turbulence of weakly dispersive acoustic waves for which

ωk=csk[1+(ak)2].(5)\omega_k=c_sk[1+(ak)^2]. \tag*{(5)}

Here a=consta=\mathrm{const} is a dispersion length such that ak≪1ak\ll1. The interaction amplitude V12kV_{12}^{k} in case acoustic hydrodynamics [2, 11, 12, 20] is given by

V12k=V0kk1k2,V0=const.(6)V_{12}^{k}=V_0\sqrt{kk_1k_2},\qquad V_0=\mathrm{const}. \tag*{(6)}

To find the energy spectra, note that Eq. (3) conserves the total energy of the system and, therefore, it can be rewritten as the following continuity equation for E(k)E(k),

∂E(k)∂t+∂εk∂k=0,whereεk=−∫k1<kSt⁡k1ωk1 dk1.(7)\frac{\partial E(k)}{\partial t}+\frac{\partial\varepsilon_k}{\partial k}=0,\quad\text{where}\quad\varepsilon_k=-\int_{k_1<k}\operatorname{St}_{k_1}\omega_{k_1}\,dk_1. \tag*{(7)}

Dimensionally, the energy flux ε(k)∝St⁡k∝nk2\varepsilon(k)\propto\operatorname{St}_k\propto n_k^2, so that nkn_k and E(k)∝εE(k)\propto\sqrt{\varepsilon}. Assuming full self-similarity (i.e. no other dimensional parameters enter into the game), and that εk\varepsilon_k is independent of kk in an inertial range of scales, one can reconstruct E(k)E(k) from the dimensional reasoning up to a dimensionless constant CC,

E(k)=Cεωkk2.(8)E(k)=C\frac{\sqrt{\varepsilon}\omega_k}{k^2}. \tag*{(8)}

For example, for non-dispersive 3D acoustic waves with ωk=csk\omega_k=c_sk we recover the Zakharov-Sagdeev spectrum [12],

E(k)∝k−3/2,(9)E(k)\propto k^{-3/2}, \tag*{(9)}

which is a flux spectrum describing an energy cascade from large to small scales. It can also be obtained as an exact solution of Eq. (3) [12].

Recall that non-dispersive acoustic waves admit triad wave number and frequency resonances for colinear wavevectors only, i.e. k∥k1∥k2\mathbf{k}\parallel\mathbf{k}_1\parallel\mathbf{k}_2. In this case the arguments of δ12k\delta_{12}^{k} and δ(ω12k)\delta(\omega_{12}^{k}) in Eq. (3) for St⁡k\operatorname{St}_k coincide which creates a singularity. Fortunately, in 3D, this singularity is integrable and St⁡k\operatorname{St}_k remains finite. Therefore spectrum Eq. (7b) in 3D is a valid solution (see [14, 15, 20] for further discussions). In 2D, the singularity is NOT integrable and Eq. (3) for the non-dispersive case is therefore invalid. Furthermore, the energy spectrum (7b) obtained dimensionally is wrong in 2D because the extra dimensional parameter aa becomes essential (unlike in 3D).

To solve the problem of acoustic turbulence in 2D, we take into account the dispersion correction in the frequency (4). We choose a reference system with k∥x^\mathbf{k}\parallel\hat{\mathbf{x}}. Integrating in Eq. (3) R12k\mathcal{R}_{12}^{k} over k2\mathbf{k}_2 with the help of δ12k\delta_{12}^{k} and over k1yk_{1y} using δ(ω12k)\delta(\omega_{12}^{k}). We get

Δ12k=def∫k2,k1yδ(ω12k)δ12kdk1dk2=∣∂k1y(ωk−ωk1−ωk−k1)∣−1dk1x≈k1k2csk∣k1y∣,(10)\begin{aligned} \Delta_{12}^{k}\mathrel{\mathop{=}^{\mathrm{def}}}\int_{k_2,k_{1y}}\delta(\omega_{12}^{k})\delta_{12}^{k}dk_1dk_2 \\ &=\left|\partial_{k_{1y}}(\omega_k-\omega_{k_1}-\omega_{k-k_1})\right|^{-1}dk_{1x}\approx\frac{k_1k_2}{c_sk|k_{1y}|}, \tag*{(10)} \end{aligned}

where we retained the leading order in ak≪1ak\ll1 only. Analyzing resonance conditions (1) with the the dispersion law (4), for small akak we find k1y≈6ak1k2k_{1y}\approx\sqrt{6ak_1k_2} (see Supplementary Material). Substituting this into Eq. (8a), we have

Δ12k≈dk1x6csak.(11)\Delta_{12}^{k}\approx\frac{dk_{1x}}{\sqrt{6}c_sak}. \tag*{(11)}

We can use this expression to modify the dimensional analysis, which now involves an extra dimensional quantity aa. For this, it is only important to observe that Δ12k∝1/a\Delta_{12}^{k}\propto1/a. Therefore St⁡k∝nk2/a\operatorname{St}_k\propto n_k^2/a which, together with St⁡k∝ε\operatorname{St}_k\propto\varepsilon (arising from Eq. (7)), gives nk∝εan_k\propto\sqrt{\varepsilon a}. This leads to the following energy spectrum of weakly dispersive 2D acoustic turbulence,

E(k)=C1εaωkk3/2=C1εacsk,(12)E(k)=C_1\sqrt{\varepsilon a}\frac{\omega_k}{k^{3/2}}=C_1\sqrt{\varepsilon a}\frac{c_s}{k}, \tag*{(12)}

where C1C_1 is a dimensionless constant. Note that the spectrum E(k)∝1/kE(k)\propto1/k was suggested in [11] without proving that it is a solution of the kinetic equation. Below, we will prove that (9) is the unique stationary power-law flux solution of Eq. (3) and find C1C_1 analytically. To do this, we compute Δ1k2\Delta_{1}^{k_2} and Δ2k1\Delta_{2}^{k_1}, similarly to Eqs. (8a) (see the Eqs. (27) in Supplemental Material) and present the collision term in (3) in simpler form,

St⁡k=2π6csak∫dk1dk2[N12kδ(k−k1−k2)−Nk21δ(k1−k−k2)−Nk12δ(k2−k−k1)].(13)\operatorname{St}_k=\frac{2\pi}{\sqrt{6}c_sak}\int dk_1dk_2\left[\mathcal{N}_{12}^{k}\delta(k-k_1-k_2)-\mathcal{N}_{k2}^{1}\delta(k_1-k-k_2)-\mathcal{N}_{k1}^{2}\delta(k_2-k-k_1)\right]. \tag*{(13)}

Let us seek a stationary solution of Eq. (3) in form

nk=Ak−x,x=const.(14)n_k=Ak^{-x},\qquad x=\mathrm{const}. \tag*{(14)}

There is, of course, a trivial solution with x=1x=1 corresponding to the thermodynamic energy equipartition, but we will be interested in non-zero flux states only. Assuming that the integrals in Eq. (13) converge (which is to be checked a posteriori), let us apply the Kraichnan-Zakharov transform

k1=k2k~1,k2=kk~2k~1andk2=k2k~1,k1=kk~2k~1(15)k_1 = \frac{k^2}{\tilde{k}_1},\quad k_2 = \frac{k\tilde{k}_2}{\tilde{k}_1}\quad\text{and}\quad k_2 = \frac{k^2}{\tilde{k}_1},\quad k_1 = \frac{k\tilde{k}_2}{\tilde{k}_1} \tag*{(15)}

to the second and the third terms in Eq. (13) respectively. This gives (after dropping tildes on the integration variables for uniformity of notations):

St⁡k=2π6 cs a k∫dk1dk2 N12k[1−(k1k)−y−(k2k)−y]δ(k−k1−k2),(16)\operatorname{St}_k = \frac{2\pi}{\sqrt{6}\,c_s\,a\,k}\int dk_1dk_2\,\mathcal{N}_{12}^{k}\left[1-\left(\frac{k_1}{k}\right)^{-y}-\left(\frac{k_2}{k}\right)^{-y}\right]\delta(k-k_1-k_2), \tag*{(16)}

with y=5−2xy=5-2x. Thus, St⁡k=0\operatorname{St}_k=0 if y=−1y=-1, so that x=3x=3 and thus

nk∝1k3,soE(k)=2πkωknk≈2πk2csnk∝1k.(17)n_k\propto\frac{1}{k^3},\quad\text{so}\quad E(k)=2\pi k\omega_kn_k\approx2\pi k^2c_sn_k\propto\frac{1}{k}. \tag*{(17)}

This is an exact stationary solution of the Eqs. (3) and (10) coinciding with the dimensional prediction (9).

To demonstrate existence and uniqueness of solution (14), let us non-dimensionalize the collision integral for arbitrary xx:

St⁡k=2πA2V026 acsI(x)k3−2x,(18)\operatorname{St}_k=\frac{2\pi A^2V_0^2}{\sqrt{6}\,a c_s}I(x)k^{3-2x}, \tag*{(18)}
I(x)=∫01q(1−q)(q−x(1−q)−x−q−x−(1−q)−x)(q−y−(1−q)−y)dq.(19)I(x)=\int_0^1 q(1-q)\left(q^{-x}(1-q)^{-x}-q^{-x}-(1-q)^{-x}\right)\left(q^{-y}-(1-q)^{-y}\right)dq. \tag*{(19)}

This integral converges if and only if 2<x<42<x<4 or x=1x=1, as we prove in the Supplemental Material. In Figure 1 we show the plot I(x)I(x) obtained numerically for the convergence range 2<x<42<x<4. As we see, x=3x=3 is the only point at which I(x)=0I(x)=0, which proves the uniqueness of the stationary power-law flux solution (9). The existence of the solution amounts to the finiteness of the constant C1C_1. We compute C1C_1 in Sec. III.C of the Supplemental Material by substituting (13) into the definition of the flux (6). The result is as follows

Figure 1. Integral (13) in the window of convergence.

ϵk=4π4A2V026 a,C1=61/4csπV0(20)\epsilon_k=\frac{4\pi^4A^2V_0^2}{\sqrt{6}\,a},\quad C_1=\frac{6^{1/4}\sqrt{c_s}}{\pi V_0} \tag*{(20)}

Notice that Eq. (3) is valid for sufficiently small nonlinearities and stochasticity of the phases [1, 2]. Let us define the interaction frequency γk\gamma_k as a frequency with which the wave packets are destroyed by the nonlinear interactions. It will also correspond to the nonlinear frequency broadening: a characteristic width of the time-Fourier spectrum at a fixed kk. Applicability of Eq. (3) requires frequency γk\gamma_k to be smaller than the characteristic frequency of interacting waves δωk\delta\omega_k. For the 3D

waves, we roughly (and definitely not rigorously) may take δωk=ωk\delta\omega_k=\omega_k. However, due to the non-integrable singularity of the non-dispersive 2D system, in 2D we should take only the dispersive part of the frequency, δωk=csa2k3\delta\omega_k=c_sa^2k^3, ignoring its linear part cskc_sk, disappearing in the reference system, co-moving with velocity csc_s in the kk-direction. On the other hand, for the wave turbulence to be considered acoustic, the dispersion must remain a small correction, i.e. ak≪1ak\ll1.

To test our theoretical predictions, we perform direct numerical simulations of the 2D GP equation for the complex wave function ψ\psi. Written in terms of the healing length ξ\xi, the speed of sound csc_s and the bulk density ρ0\rho_0, this equation reads

i∂ψ∂t=cs2ξ[−ξ2∇2+1ρ0∣ψ∣2−1]ψ+iDc∇6ψ+F(r,t),(21)i\frac{\partial\psi}{\partial t}=\frac{c_s}{\sqrt{2}\xi}\left[-\xi^2\nabla^2+\frac{1}{\rho_0}|\psi|^2-1\right]\psi+iD_c\nabla^6\psi+F(\mathbf{r},t), \tag*{(21)}

where we have also included a large-scale forcing FF and a hyper-viscous dissipation term. The healing length and the speed of sound both depend on the physical properties of the superfluid and on ρ0\rho_0, and they can be chosen arbitrarily in the dimensionless GP equation.

The GP equation is a well established model for BEC and it can be mapped to an effective compressible irrotational fluid flow via the Madelung transformation,

ψ(r,t)=ρ(r,t)exp⁡[iϕ(r,t)/2csξ],\psi(\mathbf{r},t)=\sqrt{\rho(\mathbf{r},t)}\exp\left[i\phi(\mathbf{r},t)/\sqrt{2}c_s\xi\right],

with ρ(r,t)\rho(\mathbf{r},t) and ϕ(r,t)\phi(\mathbf{r},t) being the fluid density and the velocity potential respectively. Perturbations about a still fluid with uniform density ρ(r,t)≡ρ0\rho(\mathbf{r},t)\equiv\rho_0 behave as a dispersive sound with frequency given by the Bogoliubov dispersion relation, ωk=csk1+ξ2k2/2\omega_k=c_sk\sqrt{1+\xi^2k^2/2}. In the weakly dispersive limit ξk≪1\xi k\ll1, it becomes ωk=csk(1+ξ2k2/4)\omega_k=c_sk(1+\xi^2k^2/4) i.e. the dispersion relation (4) with a=ξ/2a=\xi/2. In this limit, the three-wave interaction coefficient of Eq. (21) is of the form (5), although some ambiguities and discrepancies in the value of the coefficient V0V_0 can be found in the previous works [2, 11, 12, 20]. In Sec. I.B of the Supplemental Material, we provide the corrected derivation which leads to V0=3cs/42V_0=3\sqrt{c_s}/4\sqrt{2}. Substituting this into Eq. (20) we have the following prediction for the pre-factor constant,

C1=211/433/4π≈0.94.(22)C_1=\frac{2^{11/4}}{3^{3/4}\pi}\approx0.94. \tag*{(22)}

We simulate Eq. (21) using the standard pseudospectral code FROST [21] in a periodic domain of size LL using Nc2=10242N_c^2=1024^2 and Nc2=5122N_c^2=512^2 collocation points, denoted by Run 1 and Run 2 respectively. The nonlinear term is de-aliased twice with 2/32/3-rule following the scheme introduced in [22] in order to conserve momentum (in addition to the energy and the number of particles) in the ideal case (with F=Dc=0F=D_c=0). The Fourier transform of the forcing FF obeys the Ornstein–Uhlenbeck process dFk=−αFk dt+f0dWkdF_k=-\alpha F_k\,\mathrm{d}t+f_0\mathrm{d}W_k, where WkW_k is the Wiener process. The forcing acts only on wavenumbers such that 2π≤kL≤3×2π2\pi\le kL\le3\times2\pi. In addition, the condensate amplitude is kept constant during the evolution. We set the initial data with uniform condensate with ∣ψ0∣2=ρ0|\psi_0|^2=\rho_0, the forcing then adds the acoustic disturbances, and we evolve the system until it reaches a steady-state. We then perform averages over time. In numerics, we have set cs=1c_s=1, ρ0=1\rho_0=1 and ξ=2L/Nc\xi=2L/N_c. For forcing and dissipation we set α=1\alpha=1, f0=1.25×10−4,5×10−4f_0=1.25\times10^{-4},5\times10^{-4} and Dc=4.1×10−15,2.1×10−11D_c=4.1\times10^{-15},2.1\times10^{-11} for Run 1 and 2 respectively.

In absence of forcing and dissipation, Eq. (21) conserves the total energy (Hamiltonian) of the system. The energy per unit of mass, written in terms of the hydrodynamic variables, consists of the kinetic, internal and quantum energies [23]:

E=1L2ρ0∫[ρ2(∇ϕ)2+c22ρ0(ρ−ρ0)2+c2ξ2(∇ρ)2]d2r.(23)E=\frac{1}{L^2\rho_0}\int\left[\frac{\rho}{2}(\nabla\phi)^2+\frac{c^2}{2\rho_0}(\rho-\rho_0)^2+c^2\xi^2(\nabla\rho)^2\right]\mathrm{d}^2\mathbf{r}. \tag*{(23)}

In the dispersiveless limit (ξ→0\xi\to0), we retrieve the standard energy for a for a compressible, isentropic, irrotational fluid [24]. The total energy spectrum is computed writing the energy as usual in quantum turbulence [23]. We also calculate the kk-space energy flux ϵ(k)\epsilon(k) directly using Eq. (21) (see the Eq. (31) of Supplemental Material for exact definitions).

In Figure 2 we show the fluxes and spectra for Run 1 and Run 2. We see that the fluxes have a pronounced plateau which indicates the presence of an inertial range (free of forcing and dissipation effects). Both runs display a stationary power-law spectrum. Remarkably, both, the power-law exponent and the pre-factor C1C_1 (calculated based on the averaged flux in the inertial range), closely agree with the theoretical predictions (9) and (16).

Figure 2. a) Energy flux. b) Energy spectra. Dashed lines correspond to the theoretical prediction (9) and (18) using the corresponding flux values in the inertial range.

In Figure 3 we show the spatio-temporal spectrum for Run1. We see that this spectrum follows closely the Bogoliubov dispersion law, which indicates that the nonlinearity is sufficiently weak. The ω\omega-width of the spectrum at each fixed kk represents the nonlinear frequency broadening; we define it as

Spatio-temporal spectrum for Run 1

Figure 3. Spatio-temporal spectrum S(ω,k)=∣ψ^(ω,k)∣2S(\omega,k)=|\hat{\psi}(\omega,k)|^2 normalized by the time-averaged spectra ∣ψ^(t,k)∣2|\hat{\psi}(t,k)|^2 of Run1 (102421024^2). The dashed and dot-dashed lines show the Bogoliubov and the pure acoustic dispersion relation.

γk=[∫0∞(ω−ωk)2∣ψ^(ω,k)∣2dω∫0∞∣ψ^(ω,k)∣2dω]1/2\gamma_k=\left[\frac{\int_0^\infty(\omega-\omega_k)^2|\hat{\psi}(\omega,k)|^2\mathrm{d}\omega}{\int_0^\infty|\hat{\psi}(\omega,k)|^2\mathrm{d}\omega}\right]^{1/2}

In Figure 4 we show the ratio γk/δωk\gamma_k/\delta\omega_k, where δωk=csa2k3\delta\omega_k=c_sa^2k^3 is the dispersive correction. Recall that the WT theory is applicable when γk/δωk≪1\gamma_k/\delta\omega_k\ll1. We see that this quantity is indeed small in the scaling range in Run1, and

Figure 4. Linear to nonlinear time ratios for Run1 (102421024^2) and Run2 (5122512^2) computed from the STS in Figure 3.

only marginally small in a rather narrow range in Run2. This indicates that the WT theory has a good predictive power even when the formal applicability condition is on the borderline of validity. The main result of our paper is the 1D energy spectrum of 2D weakly dispersive acoustic waves, Eq. (12), found as the unique stationary constant-flux solution of the kinetic equation (3) with convergent collision integral. From the physical view point such a convergence means that the main contribution to the energy balance of waves with wavenumber kk comes from their energy exchange with the “neighbouring” waves with wavenumbers k′k' of the order of kk. In the language of hydrodynamic turbulence we are dealing here with the step-by-step cascade energy transfer, local in the wavenumber space. We found the energy flux to be positive, meaning that the energy is transferred from small to large kk i.e. it is a direct energy cascade.

We tested our analytical predictions by numerical simulations of the forced-dissipated GP equation (17) in the presence of a strong condensate. The analytically found spectrum (9) was confirmed by the numerics, including both the power-law exponent and the pre-factor C1C_1 without any adjustable parameter. Such a double validation is a rare success in the theory of wave turbulence, where numerical tests were attempted by numerical simulations for various types of waves but, in most cases, only the spectrum exponent was confirmed. Wave turbulence is therefore a valid and productive approach for describing 2D superfluid BEC turbulence where interacting sound waves represent the principal mechanism of energy dissipation. Since measurements of the spectrum are experimentally accessible in BEC [16, 17], our results present verifiable predictions which could guide future experiments. The focus of the present Letter was on the weak turbulence of 2D acoustic waves, and the strong turbulence regimes would be an interesting subject for future studies.

Acknowledgments

This work was supported by the Agence Nationale de la Recherche through the project GIANTE ANR-18-CE30-0020-01, by the Simons Foundation Collaboration grant Wave Turbulence (Award ID 651471) and by NSF-BSF grant # 2020765. This work was granted access to the HPC resources of CINES, IDRIS and TGCC under the allocation 2019-A0072A11003 made by GENCI. Computations were also carried out at the Mésocentre SIGAMM hosted at the Observatoire de la Côte d’Azur.

References

Hamiltonian formulation of acoustic turbulence

In this section we review the Hamiltonian formulation of acoustic turbulence and obtain the interaction term represented by V12kV_{12}^{k} in Eq. (6) of the main text. In the acoustic limit, this term is well known to be V12k=V0kk1k2V_{12}^{k}=V_{0}\sqrt{kk_{1}k_{2}} [2, 11, 12, 20], however the constant V0V_{0} in these references takes different values. Here, we will carefully derive its value. The starting point is the action (per unit of mass) for a compressible, isentropic, irrotational fluid [24] that reads:

S=1ρ0L2∫dtd2x[−ρϕ˙−ρ2(∇ϕ)2−cs22ρ0(ρ−ρ0)2](24)\mathcal{S}=\frac{1}{\rho_{0}L^{2}}\int dt d^{2}x\left[-\rho\dot{\phi}-\frac{\rho}{2}(\nabla\phi)^{2}-\frac{c_{s}^{2}}{2\rho_{0}}(\rho-\rho_{0})^{2}\right] \tag*{(24)}

where ρ0\rho_{0} is the bulk density and csc_{s}, as it will be clear later, is the speed of sound. Note that the dimensions of the fields are [ϕ]=L2/T[\phi]=L^{2}/T and [ρ]=M/L2[\rho]=M/L^{2}.

Varying the action with respect to ρ\rho and ϕ\phi and we obtain the fluid equations,

ρ˙+∇⋅(ρ∇ϕ)=0,(25)\dot{\rho}+\nabla\cdot(\rho\nabla\phi)=0, \tag*{(25)}
ϕ˙+12∇ϕ2=−cs2ρ0(ρ−ρ0).(26)\dot{\phi}+\frac{1}{2}\nabla\phi^{2}=-\frac{c_{s}^{2}}{\rho_{0}}(\rho-\rho_{0}). \tag*{(26)}

Acoustic waves are readily obtained by linearizing the equations about ϕ=0\phi=0 and ρ=ρ0\rho=\rho_{0}, which leads to the wave equation ϕ¨=cs2∇2ϕ\ddot{\phi}=c_{s}^{2}\nabla^{2}\phi.

Note that the action (20) is not written in a Hamiltonian way. Making the following change of variables ρ=ρ0A2\rho=\rho_{0}A^{2} and p=2Aϕp=2A\phi, after substituting in (20), we obtain

S=∫dtd2xL212(Ap˙−A˙p)−∫dtHwithH=∫d2xL2[18(∇p−p∇AA)2+cs22(A2−1)2](27)\mathcal{S}=\int dt\frac{d^{2}x}{L^{2}}\frac{1}{2}(A\dot{p}-\dot{A}p)-\int dtH\quad\text{with}\quad H=\int\frac{d^{2}x}{L^{2}}\left[\frac{1}{8}\left(\nabla p-\frac{p\nabla A}{A}\right)^{2}+\frac{c_{s}^{2}}{2}(A^{2}-1)^{2}\right] \tag*{(27)}

where the equation of motion are now given by

p˙=δHδA,A˙=−δHδp.(28)\dot{p}=\frac{\delta H}{\delta A},\qquad\dot{A}=-\frac{\delta H}{\delta p}. \tag*{(28)}

We remark that the units of the new fields are [A]=1[A]=1 and [p]=L2/T[p]=L^{2}/T. The Hamiltonian per unit of mass has units [H]=L2/T2[H]=L^{2}/T^{2} as usual in hydrodynamics.

Acoustic waves

Waves are obtained by making A→1+A~A\to1+\widetilde{A} and p→p~p\to\widetilde{p}. Dropping tildes and keeping the terms up to the cubic order, we rewrite the Hamiltonian as H=H2+H3H=H_{2}+H_{3}, where the second and third order terms are

H2=∫d2xL2[18(∇p)2+2cs2A2],H3=∫d2xL2[2cs2A3−14p(∇p)⋅(∇A)].(29)H_{2}=\int\frac{d^{2}x}{L^{2}}\left[\frac{1}{8}(\nabla p)^{2}+2c_{s}^{2}A^{2}\right],\qquad H_{3}=\int\frac{d^{2}x}{L^{2}}\left[2c_{s}^{2}A^{3}-\frac{1}{4}p(\nabla p)\cdot(\nabla A)\right]. \tag*{(29)}

We now assume that the fields are periodic and write them as p(x)=∑kpkeik⋅xp(\mathbf{x})=\sum_{k}p_{k}e^{i\mathbf{k}\cdot\mathbf{x}} and A(x)=∑kAkeik⋅xA(\mathbf{x})=\sum_{k}A_{k}e^{i\mathbf{k}\cdot\mathbf{x}}. The Hamiltonian and the action become:

S=∫dt∑k12(Akp˙k∗−A˙kpk∗)−∫dt(H2+H3),(30)\mathcal{S}=\int dt\sum_{k}\frac{1}{2}(A_{k}\dot{p}_{k}^{*}-\dot{A}_{k}p_{k}^{*})-\int dt(H_{2}+H_{3}), \tag*{(30)}
H2=∑k18k2∣pk∣2+2cs2∣Ak∣2,(31)H_{2}=\sum_{k}\frac{1}{8}k^{2}|p_{k}|^{2}+2c_{s}^{2}|A_{k}|^{2}, \tag*{(31)}
H3=∑1,2,32cs2A1A2A3δ1,2,3+14p1p2A3k2⋅k3δ1,2,3,(32)H_{3}=\sum_{1,2,3}2c_{s}^{2}A_{1}A_{2}A_{3}\delta_{1,2,3}+\frac{1}{4}p_{1}p_{2}A_{3}\mathbf{k}_{2}\cdot\mathbf{k}_{3}\delta_{1,2,3}, \tag*{(32)}

where δ1,2,3\delta_{1,2,3} is 1 if k1+k2+k3=0\mathbf{k}_{1}+\mathbf{k}_{2}+\mathbf{k}_{3}=0, and 0 otherwise.

In order to write the Hamiltonian and the action in the canonical form, we perform the following change of variables

pk=i12(αk2)14(ak−a−k∗),(33)p_k = i\frac{1}{\sqrt{2}}\left(\frac{\alpha}{k^2}\right)^{\frac{1}{4}}\left(a_k-a_{-k}^*\right), \tag*{(33)}
Ak=12(k2α)14(ak+a−k∗),(34)A_k = \frac{1}{\sqrt{2}}\left(\frac{k^2}{\alpha}\right)^{\frac{1}{4}}\left(a_k+a_{-k}^*\right), \tag*{(34)}

where α=16cs2\alpha=16c_s^2. The value of this coefficient is set in order to kill the off-diagonal terms in H2H_2. At the leading order, the action becomes

S2=∫dti2∑k(a˙kak∗−aka˙k∗)−∫dtH2,withH2=∑kcsk∣ak∣2=∑kωk∣ak∣2.(35)S_2=\int dt\frac{i}{2}\sum_{\mathbf{k}}\left(\dot{a}_ka_k^*-a_k\dot{a}_k^*\right)-\int dtH_2,\quad\text{with}\quad H_2=\sum_{\mathbf{k}}c_sk\lvert a_k\rvert^2=\sum_{\mathbf{k}}\omega_k\lvert a_k\rvert^2. \tag*{(35)}

Then,

δS2δak∗=0⟹ia˙k=∂H∂ak∗=ωkak.(36)\frac{\delta S_2}{\delta a_k^*}=0\Longrightarrow i\dot{a}_k=\frac{\partial H}{\partial a_k^*}=\omega_ka_k. \tag*{(36)}

H3 terms

The cubic part of the Hamiltonian requires some tedious work. Keeping only resonant terms we obtain the following contributions

∑A1A2A3δ1,2,3=123/2∑1,2,3k1k2k3α34(a1+a−1∗)(a2+a−2∗)(a3+a−3∗)=323/2∑1,2,3k1k2k3α34(a1a2∗a3∗+a1∗a2a3)δ2,31,(37)\sum A_1A_2A_3\delta_{1,2,3}=\frac{1}{2^{3/2}}\sum_{1,2,3}\frac{\sqrt{k_1k_2k_3}}{\alpha^{\frac{3}{4}}}(a_1+a_{-1}^*)(a_2+a_{-2}^*)(a_3+a_{-3}^*)=\frac{3}{2^{3/2}}\sum_{1,2,3}\frac{\sqrt{k_1k_2k_3}}{\alpha^{\frac{3}{4}}}(a_1a_2^*a_3^*+a_1^*a_2a_3)\delta_{2,3}^1, \tag*{(37)}

where δ2,31=δ−1,2,3\delta_{2,3}^1=\delta_{-1,2,3}. The second term requires more manipulations

∑p1p2A3k2⋅k3δ1,2,3=−123/2∑1,2,3α14k1k2k3k3(k2⋅k3)(a1−a−1∗)(a2−a−2∗)(a3+a−3∗)δ1,2,3(38)\sum p_1p_2A_3\mathbf{k}_2\cdot\mathbf{k}_3\delta_{1,2,3}=-\frac{1}{2^{3/2}}\sum_{1,2,3}\frac{\alpha^{\frac{1}{4}}}{\sqrt{k_1k_2k_3}}k_3(\mathbf{k}_2\cdot\mathbf{k}_3)(a_1-a_{-1}^*)(a_2-a_{-2}^*)(a_3+a_{-3}^*)\delta_{1,2,3} \tag*{(38)}
=−12×23/2∑1,2,3α14k1k2k3k3(k1⋅k3+k2⋅k3)(a1−a−1∗)(a2−a−2∗)(a3+a−3∗)δ1,2,3(39)=-\frac{1}{2\times2^{3/2}}\sum_{1,2,3}\frac{\alpha^{\frac{1}{4}}}{\sqrt{k_1k_2k_3}}k_3(\mathbf{k}_1\cdot\mathbf{k}_3+\mathbf{k}_2\cdot\mathbf{k}_3)(a_1-a_{-1}^*)(a_2-a_{-2}^*)(a_3+a_{-3}^*)\delta_{1,2,3} \tag*{(39)}
=12×23/2∑1,2,3α14k1k2k3k33(a1−a−1∗)(a2−a−2∗)(a3+a−3∗)δ1,2,3,(40)=\frac{1}{2\times2^{3/2}}\sum_{1,2,3}\frac{\alpha^{\frac{1}{4}}}{\sqrt{k_1k_2k_3}}k_3^3(a_1-a_{-1}^*)(a_2-a_{-2}^*)(a_3+a_{-3}^*)\delta_{1,2,3}, \tag*{(40)}

where form the second to third line we used the resonant condition k1+k2=−k3\mathbf{k}_1+\mathbf{k}_2=-\mathbf{k}_3. Again, keeping only resonant terms, changing summation variables and using symmetries, we can replace inside the sum

k33(a1−a−1∗)(a2−a−2∗)(a3+a−3∗)δ1,2,3→(a1a2∗a3∗+a1∗a2a3)(k13−2k33)δ2,31→(a1a2∗a3∗+a1∗a2a3)(k13−k23−k33)δ2,31(41)k_3^3(a_1-a_{-1}^*)(a_2-a_{-2}^*)(a_3+a_{-3}^*)\delta_{1,2,3}\rightarrow(a_1a_2^*a_3^*+a_1^*a_2a_3)(k_1^3-2k_3^3)\delta_{2,3}^1\rightarrow(a_1a_2^*a_3^*+a_1^*a_2a_3)(k_1^3-k_2^3-k_3^3)\delta_{2,3}^1 \tag*{(41)}

Finally, using the resonant condition k1=k2+k3\mathbf{k}_1=\mathbf{k}_2+\mathbf{k}_3, we have k13−k23−k33=(k2+k3)3−k23−k33=3(k2+k3)k2k3=3k1k2k3k_1^3-k_2^3-k_3^3=(k_2+k_3)^3-k_2^3-k_3^3=3(k_2+k_3)k_2k_3=3k_1k_2k_3. Gathering all the terms

H3=123/2∑(a1a2∗a3∗+c.c)δ2,31k1k2k3[2cs23α3/4+143α1/42]=∑V2,31(a1a2∗a3∗+c.c)δ2,31,(42)H_3=\frac{1}{2^{3/2}}\sum(a_1a_2^*a_3^*+c.c)\delta_{2,3}^1\sqrt{k_1k_2k_3}\left[2c_s^2\frac{3}{\alpha^{3/4}}+\frac{1}{4}\frac{3\alpha^{1/4}}{2}\right]=\sum V_{2,3}^1(a_1a_2^*a_3^*+c.c)\delta_{2,3}^1, \tag*{(42)}

where V2,31=V0k1k2k3V_{2,3}^1=V_0\sqrt{k_1k_2k_3}, with

V0=123/2[2cs23α3/4+143α1/42]=342cs,(43)V_0=\frac{1}{2^{3/2}}\left[2c_s^2\frac{3}{\alpha^{3/4}}+\frac{1}{4}\frac{3\alpha^{1/4}}{2}\right]=\frac{3}{4\sqrt{2}}\sqrt{c_s}, \tag*{(43)}

the formula used to obtained Eq. (22) from Eq. (20) in the main text.

Wave vector triad

Figure S1. Wave vector triad. We choose a coordinate system such that kˉ\bar{k} is aligned with the xx-axis. θ1k\theta_{1k} is the angle between kˉ\bar{k} and kˉ1\bar{k}_1, k1yk_{1y} is the yy-component of kˉ1\bar{k}_1.

Analysis of the collision term

Analysis of R12kR^k_{12} term

According to Eqs. (3) of the main text, the first term in the collision integral (R12kR^k_{12}) contains

Δ12k≡∫k2,k1yδ(ω12k)δ12k dk1dk2≈k1k2 dk1xcsk∣k1y∣.(44)\Delta^k_{12} \equiv\int_{k_2,k_{1y}} \delta(\omega^k_{12})\delta^k_{12}\,dk_1dk_2 \approx\frac{k_1k_2\,dk_{1x}}{c_s k\lvert k_{1y}\rvert}. \tag*{(44)}

From Figure 1 we can see that to leading order

k1y=k1sin⁡(θ1k)≈k1θ1k.(45)k_{1y}=k_1\sin(\theta_{1k})\approx k_1\theta_{1k}. \tag*{(45)}

To find θ1k\theta_{1k} consider the wave number resonance condition,

k2=k−k1⇒k22=k12+k2−2k1⋅k=k12+k2−2k1kcos⁡θ1k≈(k−k1)2+kk1θ1k2⇒k2≈(k−k1)+kk12k2θ1k2.(46)\begin{aligned} k_2&=k-k_1 \Rightarrow k_2^2=k_1^2+k^2-2\boldsymbol{k}_1\cdot\boldsymbol{k}=k_1^2+k^2-2k_1k\cos\theta_{1k}\approx(k-k_1)^2+kk_1\theta_{1k}^2\\ &\Rightarrow k_2\approx(k-k_1)+\frac{kk_1}{2k_2}\theta_{1k}^2. \tag*{(46)} \end{aligned}

Next, consider the frequency resonance condition,

k1+k2−k=−a2(k13+k23−k3),(47)k_1+k_2-k=-a^2(k_1^3+k_2^3-k^3), \tag*{(47)}

and substitute k2k_2 from Eq. (46) to the LHS and k2=(k−k1)k_2=(k-k_1) to the RHS of this equation. This gives kk1θ1k22k2=3a2kk1k2\frac{kk_1\theta_{1k}^2}{2k_2}=3a^2kk_1k_2 or θ1k2=6a2k22\theta_{1k}^2=6a^2k_2^2 and

k1y=6ak1k2.(48)k_{1y}=\sqrt{6}a k_1k_2. \tag*{(48)}

Together with Eq. (44), this finally gives

Δ12k≈dk1x6 csak=δ(k−k1−k2) dk1dk26 csak,(49)\Delta^k_{12}\approx\frac{dk_{1x}}{\sqrt{6}\,c_sak}=\frac{\delta(k-k_1-k_2)\,dk_1dk_2}{\sqrt{6}\,c_sak}, \tag*{(49)}

also shown in Eq. (8.b) in the main text. Here, we have replaced dk1xdk_{1x} by dk1dk_1 and inserted δ(k−k1−k2)dk2=1\delta(k-k_1-k_2)dk_2=1 to stress that k=k1+k2k=k_1+k_2 in the used approximation.

Contribution to R1k2kR^k_{1k_2} and R2k1kR^k_{2k_1}

Similar derivations using the wave number and frequency resonance conditions lead to

Δ2k1≈dk1x6 csak=δ(k1−k−k2) dk1dk26 csak\Delta^1_{2k}\approx\frac{dk_{1x}}{\sqrt{6}\,c_sak}=\frac{\delta(k_1-k-k_2)\,dk_1dk_2}{\sqrt{6}\,c_sak}

and

Δ1k2≈δ(k2−k−k1) dk1dk26 csak.(50)\Delta^2_{1k}\approx\frac{\delta(k_2-k-k_1)\,dk_1dk_2}{\sqrt{6}\,c_sak}. \tag*{(50)}

Together, Eqs. (49) and (50) lead to the collision integral (10) of the main text.

Proof of the interaction locality

Convergence of the integral in Eq. (13) of the main text is referred to as the interaction locality property. First, we note that this integral is trivially convergent for x=1x=1 because the integrand is identically equal to zero. This exponent corresponds to the thermodynamic energy equipartition state, i.e. a trivial zero-flux equilibrium which we will not be interested in. Thus, below we will consider the cases with x≠1x\ne1.

Infrared locality

Consider first the infra-red (IR) locality, i.e. convergence of the integral in Eq. (13) of the main text, in the region k1≪kk_{1}\ll k. We take into account that for the acoustic turbulence V12k∝kk1k2V_{12}^{k}\propto\sqrt{k k_{1} k_{2}} and integrate over k2k_{2} with the help of the δ\delta-functions. Then the leading term is

St⁡k∝∫0k(Nk1,k−k1k−Nk,k1k+k1) dk1∝∫0kk1nk1[(nk−k1−nk)−(nk−nk+k1)] dk1=∫0kk1nk1[nk+k1+nk−k1−2nk] dk1∝d2nkdk2∫0kk1nk1k12 dk1∝∫0kk13nk1 dk1.(51)\begin{aligned} \operatorname{St}_{k}\propto\int_{0}^{k}\left(\mathcal{N}_{k_{1},k-k_{1}}^{k}-\mathcal{N}_{k,k_{1}}^{k+k_{1}}\right)\,dk_{1}\propto\int_{0}^{k}k_{1}n_{k_{1}}\left[(n_{k-k_{1}}-n_{k})-(n_{k}-n_{k+k_{1}})\right]\,dk_{1} \\ &=\int_{0}^{k}k_{1}n_{k_{1}}\left[n_{k+k_{1}}+n_{k-k_{1}}-2n_{k}\right]\,dk_{1}\propto\frac{d^{2}n_{k}}{dk^{2}}\int_{0}^{k}k_{1}n_{k_{1}}k_{1}^{2}\,dk_{1}\propto\int_{0}^{k}k_{1}^{3}n_{k_{1}}\,dk_{1}. \tag*{(51)} \end{aligned}

We see that this integral converges for any nk∝k−xn_{k}\propto k^{-x} with x<4x<4 including x=3x=3.

Ultraviolet locality

Consider now the ultraviolet (UV) locality, i.e. convergence of integral in Eq. (12) of the main text, in the region k1≫kk_{1}\gg k. Now the leading term is

St⁡k∝∫k∞(Nk1,k−k1k−Nk,k1k+k1) dk1∝∫k∞k12[nknk1−k−nk1(nk+nk1−k)+nknk1−nk+k1(nk+nk1)] dk1≈nk∫k∞k12[nk1−k−nk1+k] dk1≈−2k∫k∞k12dnk1dk1 dk1∝∫k∞k12 dk1k1x+1.(52)\begin{aligned} \operatorname{St}_{k}\propto\int_{k}^{\infty}\left(\mathcal{N}_{k_{1},k-k_{1}}^{k}-\mathcal{N}_{k,k_{1}}^{k+k_{1}}\right)\,dk_{1}\propto\int_{k}^{\infty}k_{1}^{2}\left[n_{k}n_{k_{1}-k}-n_{k_{1}}\left(n_{k}+n_{k_{1}-k}\right)+n_{k}n_{k_{1}}-n_{k+k_{1}}\left(n_{k}+n_{k_{1}}\right)\right]\,dk_{1} \\ &\approx n_{k}\int_{k}^{\infty}k_{1}^{2}\left[n_{k_{1}-k}-n_{k_{1}+k}\right]\,dk_{1}\approx-2k\int_{k}^{\infty}k_{1}^{2}\frac{dn_{k_{1}}}{dk_{1}}\,dk_{1}\propto\int_{k}^{\infty}\frac{k_{1}^{2}\,dk_{1}}{k_{1}^{x+1}}. \tag*{(52)} \end{aligned}

We see that in the UV-region this integral converges for any x>2x>2, including x=3x=3.

The overall conclusion is that the collision integral converges for 2<x<42<x<4 and actual scaling exponent x=3x=3 is exactly in the middle of the locality window. This phenomenon is called counterbalanced locality of the collision integral, which quite common property of the kinetic equations.

Energy spectrum, energy flux, and constant C1C_{1}

Energy spectrum

‡

The total energy (Eq. (23) of the main text) can be rewritten as

E=1L2ρ0∫[ρ2(∇ϕ)2+cs22ρ0(ρ−ρ0)2+cs2ξ2(∇ρ)2] d2r=1L2ρ0∫[cs2ξ2∣∇ψ∣2+cs22ρ0(ρ−ρ0)2] d2r.(53)E=\frac{1}{L^{2}\rho_{0}}\int\left[\frac{\rho}{2}(\nabla\phi)^{2}+\frac{c_{s}^{2}}{2\rho_{0}}(\rho-\rho_{0})^{2}+c_{s}^{2}\xi^{2}(\nabla\rho)^{2}\right]\,d^{2}r=\frac{1}{L^{2}\rho_{0}}\int\left[c_{s}^{2}\xi^{2}|\nabla\psi|^{2}+\frac{c_{s}^{2}}{2\rho_{0}}(\rho-\rho_{0})^{2}\right]\,d^{2}r. \tag*{(53)}

The energy spectrum is then computed taking into account that the total energy is the sum of two quadratic quantities and using the definition of the cross spectrum of two fields ff and gg that is defined in terms of their Fourier transform f^\hat{f} and g^\hat{g} as

Ef,g(k)=1Δk∑k−Δk/2<∣k∣<k+Δk/2f^kg^k∗E_{f,g}(k)=\frac{1}{\Delta_{k}}\sum_{k-\Delta_{k}/2<|\mathbf{k}|<k+\Delta_{k}/2}\hat{f}_{k}\hat{g}_{k}^{*}

for some small Δk\Delta_{k}. Note that by the Parseval theorem ∫f(x)g∗(x) dx=L2∑kEf,g(k)Δk≈∫Ef,g(k) dk\int f(x)g^{*}(x)\,dx=L^{2}\sum_{k}E_{f,g}(k)\Delta_{k}\approx\int E_{f,g}(k)\,dk. The total energy spectrum is then computed as E(k)=Ekin(k)+Eint(k)E(k)=E_{\mathrm{kin}}(k)+E_{\mathrm{int}}(k), where ρ0Ekin(k)=cs2ξ2E∇ψ,∇ψ(k)\rho_{0}E_{\mathrm{kin}}(k)=c_{s}^{2}\xi^{2}E_{\nabla\psi,\nabla\psi}(k) and 2ρ02Eint(k)=cs2Eρ−ρ0,ρ−ρ0(k)2\rho_{0}^{2}E_{\mathrm{int}}(k)=c_{s}^{2}E_{\rho-\rho_{0},\rho-\rho_{0}}(k).

Energy flux

The energy flux can be computed as usual in hydrodynamics [?], but adapting it to GP dynamics as

ε(k)=−∑p=0k∂E(p)∂t∣GP=−∑p=0k∂Ekin(p)∂t∣GP+∂Eint(p)∂t∣GP,(54)\varepsilon(k)=-\sum_{p=0}^{k}\left.\frac{\partial E(p)}{\partial t}\right|_{\mathrm{GP}}=-\sum_{p=0}^{k}\left.\frac{\partial E_{\mathrm{kin}}(p)}{\partial t}\right|_{\mathrm{GP}}+\left.\frac{\partial E_{\mathrm{int}}(p)}{\partial t}\right|_{\mathrm{GP}}, \tag*{(54)}

where the label GP means that time derivatives are computed using the GP equation (without forcing and dissipation). Namely, we have

∂Ekin(p)∂t∣GP=2cs2ξ2ρ0E∇ψ,∇ψ(k)(55)\left.\frac{\partial E_{\mathrm{kin}}(p)}{\partial t}\right|_{\mathrm{GP}}=\frac{2c_s^2\xi^2}{\rho_0}E_{\nabla\psi,\nabla\psi}(k) \tag*{(55)}
∂Eint(p)∂t∣GP=cs2ρ0Eρ−ρ0,ρ(k)(56)\left.\frac{\partial E_{\mathrm{int}}(p)}{\partial t}\right|_{\mathrm{GP}}=\frac{c_s^2}{\rho_0}E_{\rho-\rho_0,\rho}(k) \tag*{(56)}

where dψ=−ics2ξ[−ξ2∇2+1ρ0∣ψ∣2−1]ψd\psi=-i\frac{c_s}{\sqrt{2}\xi}\left[-\xi^2\nabla^2+\frac{1}{\rho_0}|\psi|^2-1\right]\psi and dρ=ψdψ∗+dψψ∗d\rho=\psi d\psi^*+d\psi\psi^*. Note that lim⁡k→∞ε(k)=0\lim_{k\to\infty}\varepsilon(k)=0 because of the energy conservation of the GP equation.

Dimensionless prefactor C1C_1

To compute C1C_1, we substitute Eq.(15) into the definition of the flux (6) (both in the main text), substitute dk1=2πk1dk1d\mathbf{k}_1=2\pi k_1dk_1 (since the integrand is a function of the k1=∣k1∣k_1=|\mathbf{k}_1| and the polar angle is immediately integrated out) and integrate with respect to k1k_1. This leads to

εk=−2π2A2V026aI(x)(3−x)k6−2x.(57)\varepsilon_k=-\frac{2\pi^2A^2V_0^2}{\sqrt{6}a}\frac{I(x)}{(3-x)}k^{6-2x}. \tag*{(57)}

For actual value x=3x=3 this equation has an uncertainties zero divided by zero, which can be resolved according the the L’Hopital rule:

lim⁡x→3I(x)(3−x)=dI(x)dx∣x=3=∫0112log⁡q1q1−1 dq1=2π2.\lim_{x\to3}\frac{I(x)}{(3-x)}=\left.\frac{dI(x)}{dx}\right|_{x=3}=\int_0^1\frac{12\log q_1}{q_1-1}\,dq_1=2\pi^2.

Now Eq. (53) with x=3x=3 give for the energy flux:

εk=4π4A2V026a.(58)\varepsilon_k=\frac{4\pi^4A^2V_0^2}{\sqrt{6}a}. \tag*{(58)}

Thus, Eq. (54), together with Eqs.(9) and (11) of the main text finally give:

C1=61/4csπV0.(59)C_1=\frac{6^{1/4}\sqrt{c_s}}{\pi V_0}. \tag*{(59)}

for the pre-factor C1C_1 in Eq. (9) of the main text.

References

  1. [1]V. Zakharov, V. L’vov, and G. Falkovich, Kolmogorov Spectra of Turbulence, wave turbulence (Springer, 2012).
  2. [2]S. Nazarenko, Wave Turbulence (Springer Berlin Heidelberg, series: Lecture Notes in Physics, 2011).
  3. [3]E. Falcon and N. Mordant, Annual Review of Fluid Mechanics 54, annurev (2022), ISSN 0066-4189, 1545-4479.
  4. [4]S. Galtier, S. V. Nazarenko, A. C. Newell, and A. Pouquet, Journal of Plasma Physics 63, 447 (2000), ISSN 0022-3778, 1469-7807.
  5. [5]V. E. Zakharov, Soviet Physics—JETP 35, 908-914 (1972).
  6. [6]P. Caillol and V. Zeitlin, Dynamics of Atmospheres and Oceans 32, 81 (2000), ISSN 03770265.
  7. [7]S. Galtier, Physical Review E 68, 015301 (2003), ISSN 1063-651X, 1095-3787.
  8. [8]V. S. L’vov and S. Nazarenko, Low Temperature Physics 36, 785 (2010), ISSN 1063-777X, 1090-6517.
  9. [9]G. Düring, C. Josserand, and S. Rica, Physical review letters 97, 025503 (2006).
  10. [10]S. Galtier and S. V. Nazarenko, Physical review letters 119, 221101 (2017).
  11. [11]S. Dyachenko, A. C. Newell, A. Pushkarev, and V. E. Zakharov, Physica D: Nonlinear Phenomena 57, 96 (1992).
  12. [12]V. E. Zakharov and R. Z. Sagdeev, Dokl. Akad. Nauk SSSR 192, 297 (1970).
  13. [13]B. B. Kadomtsev and V. I. Petviashvili, Doklady Akademii Nauk SSSR 208, 794 (1973).
  14. [14]A. C. Newell and P. J. Aucoin, Journal of Fluid Mechanics 49, 593-609 (1971).
  15. [15]V. S. L’vov, Y. L’vov, A. C. Newell, and V. Zakharov, Phys. Rev. E 56, 390 (1997).
  16. [16]N. Navon, A. L. Gaunt, R. P. Smith, and Z. Hadzibabic, Nature 539, 72 (2016), ISSN 0028-0836, 1476-4687.
  17. [17]N. Navon, C. Eigen, J. Zhang, R. Lopes, A. L. Gaunt, K. Fujimoto, M. Tsubota, R. P. Smith, and Z. Hadzibabic, Science 366, 382 (2019), ISSN 0036-8075, 1095-9203.
  18. [18]S. P. Johnstone, A. J. Groszek, P. T. Starkey, C. J. Billington, T. P. Simula, and K. Helmerson, Science 364, 1267 (2019), ISSN 0036-8075, 1095-9203.
  19. [19]G. Gauthier, M. T. Reeves, X. Yu, A. S. Bradley, M. A. Baker, T. A. Bell, H. Rubinsztein-Dunlop, M. J. Davis, and T. W. Neely, Science 364, 1264 (2019), ISSN 0036-8075, 1095-9203.
  20. [20]V. E. Zakharov and S. V. Nazarenko, Physica D: Nonlinear Phenomena 201, 203 (2005).
  21. [21]N. P. Müller, J. I. Polanco, and G. Krstulovic, Phys. Rev. X 11, 011053 (2021), URL https://link.aps.org/doi/10.1103/PhysRevX.11.011053.DOI
  22. [22]G. Krstulovic and M. Brachet, Phys. Rev. E 83, 066311 (2011), URL https://link.aps.org/doi/10.1103/PhysRevE.83.066311.DOI
  23. [23]C. Nore, M. Abid, and M. E. Brachet, Phys. Rev. Lett. 78, 3896 (1997).
  24. [24]L. D. Landau and E. M. Lifshitz, Fluid Mechanics, Second Edition: Volume 6 (Course of Theoretical Physics), Course of theoretical physics / by L. D. Landau and E. M. Lifshitz, Vol. 6 (Butterworth-Heinemann, 1987), 2nd ed., ISBN 0750627670.

Paper details

vlnc:W_c3ctem3m
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