Skip to main content

Full text of "Counter-flow Induced Decoupling in Super-Fluid Turbulence"

See other formats


arXiv: 1509.03566v2 [cond-mat.other] 23 Dec 2015 


Counter-flow Induced Decoupling in Super-Flnid Tnrbulence 


Dmytro Khomenko, Victor S. L’vov, Anna Pomyalov, and Itamar Procaccia 
Department of Chemical Physics, Weizmann Institute of Science, Rehovot 76100, Israel 

In mechanically driven superfluid turbulence the mean velocities of the normal- and superfluid 
components are known to coincide: Un = Us- Numerous laboratory, numerical and analytical studies 
showed that under these conditions the mutual friction between the normal- and superfluid velocity 
components couples also their fluctuations: u'^{r,t) « u'„{r,t) almost at all scales. In this paper 
we show that this is not the case in thermally driven superfluid turbulence; here the counterflow 
velocity l/ns = C/n — Us^Q- We suggest a simple analytic model for the cross correlation function 
{u'^{r,t)-u'„(r',t)) and its dependence on Pns- We demonstrate that u'^{r,t) and Ug{r,t) are 
decoupled almost in the entire range of separations \r — r'\ between the energy containing scale and 
intervortex distance. 


I. INTRODUCTION 

Much of the thinking about turbulence in quantum 
fluids like ^He at low temperature is still influenced by 
the “two fluid” model of Landau and Tisza. Within this 
model the dynamics of the superfluid ^He is described 
in terms of a viscous normal component and an inviscid 
superfluid component, each with its own density Pn{T) 
and Ps{T) and its own velocity field Un{r, t) and Us{r, t). 
Due to the quantum mechanical restriction, the circula¬ 
tion around the superfluid vortices is quantized to integer 
values of K = h/m, where h is the Plank constant and 
m is the mass of ^He atom. The quantization of circu¬ 
lation results in the appearance of characteristic “quan¬ 
tum” length scale: the mean separation between vortex 
lines, i, which is typically orders of magnitude smaller 
than the scale H of the largest (energy containing) ed¬ 
dies—*^. 

Experimental evidence^;^ indicates that superfluid tur¬ 
bulence at large scales (. is similar to classical tur¬ 
bulence if the mechanical forcing is similar. Examples 
are furnished by a towed grid— forcing or by a pres¬ 
sure drop in a channel— li. The reason for the sim¬ 
ilarity is that the interaction of normal fluid compo¬ 
nent with the quantized-vortex tangle leads to a mu¬ 
tual friction forcei^^i^ “which couples together u^{r,t) 
and Us{r,t) so strongly that they move as one fluid”—. 
This strong coupling effect was demonstrated analytically 
in Ref.— and was later confirmed by numerical simula¬ 
tions of the two-fluid model— over a wide tempera¬ 
ture range (1.44 < T < 2.157 K, corresponding to the 
ratio of densities Pn/Ps from 0.1 to 10). The simulations 
showed strong locking of normal- and superfluid veloci¬ 
ties at large scales, over one decade of the inertial range. 
In particular, it was found that even if either the nor¬ 
mal or the superfluid is forced at large scale (the dom¬ 
inant one), both fluids get locked very efficiently. Only 
detailed numerical simulations (in the framework of so- 
called shell models of turbulence) with very large inertial 
interval— showed minor decoupling of Ug and at the 
viscous edge of the inertial interval in agreement with the 
analytical result of Ref.—. 

A different situation is expected for thermally driven 


superfluid turbulence. This type of turbulence is gener¬ 
ated by a heater located at the closed end of a channel 
which is open at the other end to a superfluid helium 
bath. In this case the heat flux is carried away from the 
heater by the normal fluid alone with the mean velocity 
Un, and, by conservation of mass, a superfluid current 
with the mean velocity Ug arises in the opposite direc¬ 
tion. This gives rise to a relative (counterflow) velocity 

Uns = Un-Ug, ( 1 ) 

which is proportional to the applied heat flux. Invariably 
this counterflow excites an accompanying tangle of vortex 
lines. In counterflow experiments there is no mean mass 
flux and the mean velocities Ug and Un of the superfluid 
and the normal fluid components are related as follows: 
PnUn T PsUg = 0. 

A situation very similar to counterflow appears in su¬ 
perflows. Here superleaks (i.e. filters located at the chan¬ 
nel end with sub-micron-sized holes permeable only to 
the inviscid superfluid component) allow a net flow of the 
superfluid component in the channel. Contrary to coun¬ 
terflows, now the normal component remains stationary 
on the average: Un = 0. In both counterflows and super¬ 
flows the normal- and superfluid components are moving 
with different mean velocities and their relative velocity 
Uns ^ 0 . 

Clearly, in both cases one expects properties of the 
normal- and superfluid velocity fluctuations different 
from that in the mechanically driven “co-flow” turbu¬ 
lence, in which Un = C4 and Uns = 0. The simple reason 
for that is illustrated in Fig.[Tl in which eddies of scales 
i?i < i ?2 < Rd, are shown at three successive moments of 
time t = —T, t = 0 and t = t for co-flow (panels (a), (b) 
and (c)) and for counterflow (panels (d), (e) and (f)). 

In the co-flow the quantized-vortex tangles (shown by 
blue solid lines) are swept by the superfluid component 
with the mean velocity close to Ug together with the nor¬ 
mal fluid eddies (shown by red dashed lines), which are 
swept by the normal fluid component with their mean 
velocity [/„. Since in the co-flow Ug = Un, all (normal- 
and superfluid eddies) are swept with the same velocity, 
the entire eddy configuration is moving as a whole from 
the left, in panel (a), to the right in panel (c) in the “lab¬ 
oratory” reference system, shown in all panels as a black 


2 


Co-flow 

(a) t = —T (b) t = 0 (c) t = T 







FIG. 1: Color online. Schematic view of the normal fluid eddies of scales Ri, R2 and R3 (shown by red dashed lines), swept by 
the mean normal fluid velocity Un, and of the superfluid eddies of the same scales (shown by blue solid lines) swept by the mean 
superfluid velocity Us in the co-flow with Uu = Us [panels (a), (b) and (c)] and in the counter-flow with Uns = \Un — C/s| 7^0 
[panels (d), (e) and (f)] at three consequent moment of times: t = — r [panels (a) and (d)], t = 0 [panels (b) and (e)] and t — t 
[panels (c) and (f)]. The time interval r ~ R 2 /Uns is of the order of overlapping time of the middle-scale i? 2 -eddies. 


frame. During their common motion, the mutual friction 
effectively couples the velocities and Un{r,t) = Us{r,t). 
The situation is completely different in the counter-flow, 
where the mean velocities have opposite directions and 
Dns 7 ^ 0. We have chosen for concreteness C/n > 0, there¬ 
fore the normal fluid (red dashed line) eddies are moving 
in our pictures from the left [in panel (d)] to the right [in 
panel (f)]. At the same time, Ug < 0 and superfluid (blue 
solid line) eddies are moving in the opposite direction. 

Assume that at some intermediate moment of time 
[chosen as t = 0 in panel (e)] all normal- and superfluid 
eddies of scales Ri, R2 and R3 overlap. Choose the time- 
step T, such that T ~ i?2/D„s. The largest eddies of scale 
i ?3 are almost fully overlapping during the time-step r, 
while smaller eddies of scale i?i, which were overlapping 
at t = 0, are fully separated at times t ± r. Intermedi¬ 
ate i? 2 -scale eddies are partially overlapping during the 
time-step r ~ Toi(i? 2 ). Here the “overlapping time” of 
i?-eddies To\{R) = R/U^s is the time that is required for 
eddies to be swept by the counterflow velocity U^s over 


distance of their scale R. 

This time may be small compared to the time Tcor 
required for an effective coupling of the Us{r,t) and 
Un{r,t) velocities. As we show in the last paragraph of 
Sec. Ill B[ Tcor is scale independent and may be estimated 
as Tcor ^1/where C is the vortex line density. The 
detailed analysis shows that for most eddies in the rele¬ 
vant range of scales H < R < £ the time Toi <g; Tcor and 
therefore the velocities Us{r,t) and Un{r,t) are decou¬ 
pled. This makes the energy dissipation due to mutual 
friction very effective and results in significant suppres¬ 
sion of the energy spectra of the normal- and superfluid 
turbulent velocity spectra as compared to that in the me¬ 
chanically driven turbulence, in which f7ns = 0. 

Notice that in Ref.— it was mentioned that in the 
counterflow, the coupling at all length scales must, to 
some extent, break down, because similar eddies in the 
two components are continually pulled apart, and this 
leads to dissipation at all length scales. 

The main goal of the present paper is to offer a rela- 



















3 


tively simple, physically transparent model of the cross¬ 
correlation function of the normal and superfluid veloci¬ 
ties, that accounts for non-zero value of the mean coun¬ 
terflow velocity C/ns- For simplicity we consider only the 
case of homogeneous and isotropic turbulence of an in¬ 
compressible flow of ^He. In this flow the difference 
between the counterflow and a pure superflow turbu¬ 
lence disappears due to Galilean invariance. The pa¬ 
per is organized as follows. First we overview the two- 
fluid coarse-grained Hall-Vinen-Bekarevich-Khalatnikov 
(or HVBK) model— properly generalized for the case 
of counterflow turbulence, Eqs. ([3]). Second, we sug¬ 
gest an approach that leads to a crucial simplification 
that allows us to derive analytical equations (1131) for 
the cross-correlation function of the normal- and super¬ 
fluid velocity fluctuations, £ns{k,Uns)- Third, we ana¬ 
lyze the equation for £ns{k, Uns) and show that as a rule 
£ns{k,Uns) “C £ns{k,0), See Fig. [31 Finally, in the con¬ 
cluding section, we discuss how the decoupling of veloci¬ 
ties should affect the normal- and superfluid energy spec¬ 
tra. 


II. BASIC EQUATIONS OF MOTION FOR 
COUNTERFLOW TURBULENCE 

A. Two-fluid, gradually-damped HVBK equations 

As said above, the large-scale motions of superfluid 
"‘He (with characteristic scales £) are well described 
by the two-fluid model, consisting of a normal and a su¬ 
perfluid component with densities Pn(T) and Ps(T) re¬ 
spectively. Neglecting both the bulk viscosity and the 
thermal conductivity leads to the simplest model with 
two incompressible fluids, having the form of an Euler 
equation for Ug and a Navier-Stokes equation for u^, see, 
e.g. Eqs. (2.2) and (2.3) in Donnely’s textbook—. Supple¬ 
mented with quantized vortices that give rise to a mutual 
friction force Flis between the superfluid and the normal 
components, these equations are known as Hall-Vinen- 
Bekarevich-Khalatnikov (or HVBK) model— 

d u 1 

-b (ms ■ V)ms H-Vps = VgAug-Fns, (2a) 

ut Pq 

d u 1 

-b (itn • V)Mn H-Vpn = t'n Attn -b —-Fns . (2b) 

at pn Pn 

Here Pn, Ps are the pressures of the normal and the su¬ 
perfluid components: 

Pn = —[p+^\Ug - Mnp] , p, = —[p- ^\Ug - MnP] , 
pi pi 

p = Ps + Pn is the total density and is the kinematic 
viscosity of normal fluid. The mutual friction force is 
given by 

Fns = atA X [a; X («„ — Ms)] + CtUS X («„ — Ms) • 


In this equation a, a' are temperature dependent dimen¬ 
sionless mutual friction parameters and u> is tradition¬ 
ally understood as superfluid vorticity: a; = V x its and 
Cj = w/|u;|. 

Notice also that the original HVBK model does not 
take into account the important process of vortex recon¬ 
nection. In fact, vortex reconnections are responsible for 
the dissipation of the superfluid motion due to mutual 
friction. 

For temperatures above IK this the extra dissipation 
can be modeled using an effective superfluid viscosity 

v'g{T)^aK. ( 2 c) 

and, following Ref— we have added a dissipative term 
proportional to i'' to the standard HVBK model. 

The effective superfluid viscosity zy' involves a 
quantum-mechanical parameter k, proportional to the 
Plank’s constant h. This underlies the fact that the cor¬ 
responding term in Eqs. ([3]) originates from the motions 
of quantized vortex lines at quantum scales ~ £. This 
is not captured by the coarse-grained, classical HVBK 
equations. 

Bearing in mind that experimentally the counterflow 
cannot be realized for T < 1 K (due to practically zero 
normal fluid density) we cannot discuss here the deli¬ 
cate issue how to account for the superfluid dissipation 
in Eqs. ([3]) for such low temperatures. 

B. Counterflow HVBK equations 

To proceed we separate the mean velocities Un and 
Us from the turbulent velocity fluctuations, u'^{r,t) and 
M'(r,t) with zero mean. Equations (I3|) for u'^{r,t) and 
M'(r,t) may be written, as follows: 

+ V-z.'a)m'+NLK,m'} = (3a) 

(^ + I7„-V-z.„a)< + NL{<,<} = ^/'.-(Sb) 

Here the nonlinear terms NL{M',tt'} and NL{it(,,M(,} 
are quadratic in the corresponding velocities functionals. 
These terms originate from the terms m' • Vm' and from 
the Vp' terms, where the pressure fluctuations p'{r,t) 
were expressed via a quadratic velocity fluctuations func¬ 
tional, using the incompressibility condition. Eor our 
purpose we will not need to specify the nonlinear terms 
NL{it',M'} and NL{m(,, «(,}. 

Next we approximate the mutual friction fluctuation 
term In the spirit of Refj^, we write as follows: 

f^,c-a{T){K-ui)n . (3c) 

In Ref.— the characteristic superfluid vorticity Q in 
Eq. (I3cl) was understood as the root-mean-square (rms) 
vorticity: ft ~ y/ (joip). However in counterflow tur¬ 
bulence there is an additional quantum mechanism of 





4 


creating vortex lines, elucidated in pioneering works by 
Sctiwarai^: the force of mutual friction can lead to the 
stretching of the vortex lines, and this in turn can lead to 
a self-sustaining turbulence in the superfluid component 
provided that vortex lines are allowed to reconnect. This 
mechanism is leading to the creation of an additional 
peak in the superfluid energy spectrum near the inter¬ 
vortex scale £, sketched in Fig. [51 In the counterflowing 
superfluid turbulence this peak provides the main con¬ 
tribution to the rms vorticity, which cannot be described 
in the framework of the coarse-grained HVBK Eqs. (153 
and (EEl, which is valid only for scales R ^ i. There¬ 
fore n in Eq. (133 should be understood as an external 
parameter in the HVBK equations for the counterflow, 
simply estimated via the vortex line density C, which in 
its turn is proportional to the square of the counterflow 
velocity: 

VtCi K.C, Ck. ( 7 ^[/ns)^ ■ (3d) 

Here 7 ^ is a temperature dependent phenomenological 
parameter that varies from about 70 s/cm^ to about 150 
s/cm^ when T grows from 1.3 K to 1.9 K (see e.g. Eig 
9 in Ref.—) We have added here a subscript to dis¬ 
tinguish the traditional notation 7 in Eq. (I3dl) from the 
characteristic frequencies 7 s and 7 n that are used below. 

The resulting gradually damped HVBK model for tur¬ 
bulent counterflow in ^He, Eqs. (O, serves as a basis for 
our study of the correlations between normal- and super¬ 
fluid velocity correlations. We will refer to these equa¬ 
tions as the “counterflow HVBK equations”. 

Equations ([3]) allow to estimate the time Tcor required 
for the coupling of the normal and superfluid turbulent 
velocities by mutual friction. To this end we consider an 
equation for their difference, = u[^ — m', subtracting 
Eq. (153 from Eq. (i3bl) : 

9Ks , _ I r\ -^P 

o. -(-*** — ^ns’^ns ; ^ns — 

Ot Pn 

Here we dempted by ... the sweeping, viscous and non¬ 
linear terms that are irrelevant for the current discussion. 
Evidently, Tcor should be estimated as l/(ansK>C). The 
temperature dependence of Ofns, shown in Fig. [4] by a red 
line with squares, indicates that ans 1- Therefore we 
can conclude that Tcor ^ 1 /(k'C), as mentioned in Sect. 

in 

III. NORMAL - SUPERFLUID VELOCITY 
CORRELATIONS IN ^HE 

The main result of this Section is Eq. 0 for the cross¬ 
correlation function of the normal- and superfluid ve¬ 
locity turbulent fluctuations in a stationary, space ho¬ 
mogenous counterflow "^He-turbulence. This equation de¬ 
scribes how the cross-correlations depends on the coun¬ 
terflow velocity, the scale (wave-number) and the tem¬ 
perature. Its derivation requires some definitions and 



FIG. 2: Color online. The sketch of the stationary super¬ 
fluid turbulent energy spectrum in the counterflow [log-log 
coordinates, log,£a{k) vs. log(fc£)]. The spectrum £a{k) con¬ 
sists of a classical fs*(fc) and a quantum £^^{k) parts, col¬ 
ored in gray and light blue, respectively. For concreteness, 
as a large-scale classical peak we used here Lvov-Nazarenko- 
Volovik spectrum dm, found for ^He with resting normal fluid 
component, but presumably valid for counterflowing '^He in 
the fc-range with fully decoupled the normal- and superfluid 
velocities. The quantum (light blue) contribution £^ (k) has 
1/k asymptotics at large k, originated from superfluid mo¬ 
tions near the vortex cores. It is adjacent to the classical 
thermal bath part (k) oc k^ with equipartition of energy 
between degrees of freedoms. 


relationships that are common in statistical physics. We 
recall them in Appendix A. 


A. Derivation of the cross-correlation £ns{k) 

The first step in the derivation of the cross correla¬ 
tion is rewriting the counterflow HVBK Eqs. ([3]) in (fc, t)- 
representation, defined by Eq. (|Alap : 

/ d \ 

iUs ■ k -\- + llsj Vs 

-b NLfc{t)s,t!j = HgUn , (4a) 

d \ 

4“ f lAi ■ k -(- HnJ Un 

4” hlLfcl'Uni '^n} — ; (^^) 

where the mutual friction frequencies are given by 

Hs = Cr H , Hji = CTn H , rrn = OL Ps/Pn • (^^) 

The nonlinear terms NLfc{us,Us} and NLfcin 
Eqs. (|43 and (I4bp couple all fc-Fourier harmonics mak¬ 
ing their analytic solution intractable. To proceed we 
therefore simplify the equations in the spirit of the Direct 
Interaction Approximation (DIA) that was developed by 








5 


Kraichnan for classical turbulence—. This approxima¬ 
tion is equivalent to a 1-loop truncation of the Wyld dia¬ 
grammatic expansion— of the nonlinear equations with a 
1-pole approximation^^ for the Green’s function. While 
uncontrolled, this approximation served usefully in the 
study of classical turbulence, and we propose that it is 
also useful in the present context. The upshot of the 
DIA approximation is a rewriting of the nonlinear terms 
in Eqs. (ISI) and (I4bl) as a sum of two contributions—: 

NLfc{i!s,Us} = js{k)vs{k,t) - ipsik,t), (5a) 

NLfc{u„,t>„} = -fn{k)vn{k,t) - ipn{k,t) . (5b) 

The 7 s(fc) and 7 n(A:) are the charateristic frequencies and 
ips{k, t) and ipn{k, t) are the force terms. The terms pro¬ 
portional to 7 s(fc) and 7n(fc) describe the energy flux from 
fluctuations with given k to all others degrees of free¬ 
dom. In classical turbulence theory these characteristic 
frequencies are referred to as “turbulent viscosity” and 
estimated as follows: 

7n(fc) ~ y^k^£n{k) , 7s(fc) ~ . (5c) 

In turbulent systems with strong interactions these fre¬ 
quencies are the inverse turnover times of eddies of scale 
1/fc. 

The force terms in the approximation (I5al) and (HEl) 
mimic the energy influx to fluctuations with given k from 
all others degrees of freedom. In the simplest Langevin 
approach these forces are random Gaussian processes 
with zero mean and d-correlated in time: 


be treated as isotropic, in quantum turbulence there re¬ 
mains one preferred direction x of the counterflow veloc¬ 
ity t/ns- Schwarz— introduced an anisotropy index /||, 
equal to 2/3 in the case of isotropy. Numerical simula¬ 
tions (see, e.g. Ref.—) shows that /y varies between 0.74 
and 0.82, depending on the temperature and the coun¬ 
terflow velocity. Therefore the dimensionless measure of 
anisotropy 3/||/2 — 1 is below 20% in any case. According 
to our understanding, this level of anisotropy cannot af¬ 
fect significantly the results presented below. Aiming at 
simplicity and transparency of the derivation we assume 
isotropy from the very beginning, leaving a more general 
derivation (in the framework of the same formal scheme) 
for the future. For weak anisotropy all our results should 
be understood as angular averages. 

Multiplying Eqs. (153 and (l6bl) by v*, and v*, respec¬ 
tively and averaging, we get equations for the velocity 
correlations Ann, Ess and the cross-correlation de¬ 
fined by Eqs. (I A4I) : 


A 

- d 
d 


+ Ts 


+ Tr 


Ess — f 2 sR.e[£’ns] + R.e[$ss] , 


Enn — fdnRe[Ens] + Re[<I>nn] , 


idt 


-I- ik ■ Un 


Er,s = 


^sEj^n GnE/s! 
■ + ‘I’ns • 


(7a) 

(7b) 

(7c) 


These equations involve the presently unknown simulta¬ 
neous cross-correlations of the velocities and the forces, 
defined similarly to Eqs. (IA2I) : 


{cps{k,t)-cp*{k\t')) = {2TTf6{k-k')6{t-t')(pl^{k), 

{(pn{k,t)-ip*^{k',t')) = {2TTfS{k-k')S{t-t')ipl^{k), 

{(Psik,t) ■ ipl{k',t')) = 0 . (5d) 


Here the Delta functions 5{k — k') originate from the 
space homogeneity. An important difference from the 
traditional Langevin approach is that our turbulent sys¬ 
tem is not in the thermodynamic equilibrium and there¬ 
fore the correlation amplitudes and are not de¬ 
termined by fluctuation-dissipation theorems. We will 
show below that these amplitudes may be expressed via 
the energy spectra Esik) and 6n{k). 

With these approximations the counterflow HVBK 
Eqs. (HJ become linear in Vs and vp. 


■_5 

.m 

■_5 

-Ft 


ik • Us -(- Tg 


-(- ik • Un -(- Tn 


Vs{k,t) = nsVn{k,t) + Ps{k,t), (6a) 
t)n(fc, t) = DnUs(fc, t) + Pn{k, t) , (6b) 


Tn — 7n + , Tg — 7 s -|- Og -|- v'Ji^ . (6c) 


Glearly, counterflow turbulence in a channel is 
anisotropic due to the existence of two preferred direc¬ 
tions: the stream-wise direction x and the wall-normal 
direction y. Even far away from the wall, in the chan¬ 
nel core, where classical hydrodynamic turbulence can 


{Pn{k,t) ■ v*{k',t)) = (27r)^$nn(fe) d(A; - fe'), (8a) 

{ps{k,t) ■ v*{k',t)) = (27r)^$ss(fe) (5(fe - fc')! (8b) 

{pnik,t) ■ v*{k',t)) = {2Trf^nsik)S{k~-k'), (8c) 
{psik,t) ■ v^{k',t)) = (27r)^$sn(fe) ^(fc - fc') ■ (8d) 

To find these correlations, we rewrite Eqs. ([5]) in Fourier 
(fc, u;)-representation: 


[i{k ■ Us — oj) 

-h rs]vs{k,uj) 

(9a) 


= nsVn{k,Ul) + Ps{k,Uj) , 


[i{k ■ Un — oj) 

+ rn]vn{k,uj) 

(9b) 


= nnVs{k,Uj) + (pn{k,Uj) , 


where psik,uj) 

and ipn{k,uj) are the 

{k,uj)- 

representation of the force terms Psik-,t) or 

Pn{k,t)- 


The solution of the linear Eqs. ([S]) reads: 

Vs = -[(i(fc-U'n-w)+rn)<^s + Gs^n]/A,(10a) 
Vn = -[(i(fc-I7s-a;)-hrs)^n + Gn^s]/A, (10b) 
A = {uj — k ■ Un + iE'n){t^ — k ■ Us + iTs) (10c) 

where for brevity we suppressed the arguments {k,uj) in 
all functions. 
















6 


Multiplying the two Eqs. (fTOl) by <^n and (ps, respec¬ 
tively and averaging, we get equations for the (cross)- 
correlations <l>ns(fc,w) and d>sn(fc,a;) which give after in¬ 
tegration over u! the simultaneous cross-correlation func¬ 
tions: 


$s„(fc) = 


$ns(fc) = 


^s/n f du 

27r J A*(fe,w) 
2'k J A*{k,uj) 


(lla) 

(llb) 


To compute the above integrals we found the solutions 
of the equations A(fc, w) = 0 with respect to w: 


w 



- ik ■ {Un + Us) -|- Tg -|- Tn 
Tn -|- ik ■ c/ns) + dOglln ■ 


( 12 ) 


Using these solutions, after relatively simple analysis, 
we find that both roots have positive imaginary parts: 
lm[a;+] > 0 and lm[a;_] > 0. Therefore, the integral in 
Eqs. dm vanishes. Now Eq. in the stationary case 
gives: 


E^ns(fc) 

A 

B 


A 


B + ik ■ Uns 
f^sE'„n(fc) -|- flnEss{k) , 


= Tn 


(13a) 

(13b) 

(13c) 


Averaging Eq. (I13a|) with respect to all orientations of k 
we get: 


(^ns(fe)),„gi 3 = ^ arctan (^) ■ (13d) 

Using Eqs. (IA5I) this can be finally rewritten as follows: 


£ns{k) 

D{0 

m 


£^Hk)DiO, 

fls£n{k) + Unfs(fc) 

rg(fc) + rn(fc) 
^ arctan[C(A:)], 
kUns 

rg(fc) + r„(fc) • 


(14a) 

(14b) 

(14c) 

(14d) 


Here fns^ (k) is the cross-correlation function for zero 
counterflow velocity which was previously found in 
Ref.—. The dimensionless “decoupling function” D{() 
of the dimensionless “decoupling parameter” C{k), de¬ 
scribes the decoupling of the normal- and superfluid ve¬ 
locity fluctuations, caused by the counterflow velocity. 

Notice that in future comparisons of the experimental 
or numerical data with Eqs. m one needs to bear in 
mind that the counterflow velocity affects not only the 
decoupling function D{^), but also the energy spectra 
£’n(fc) and £s{k) in Eq. (Il4bl) for £^\k). 



FIG. 3: The decoupling function £ns/£ns^ = D{k/kx) vs the 
dimensionless wave number k/kx- 


Considering the limits of small and large values of the 
decoupling parameter C we get from Eq. (I14ap : 


£nsik) = 
£ns{k) = 


1 - 


C(fc)^ 


: for C(fc) < 1, (15a) 


2ak) 


£^\k), 


for ({k) ^ 1 . (15b) 


We choose the crossover value ~ 2 such that 
D{Cx) = 1/2. Below we show that with good accu¬ 
racy C(fc) oc k. Therefore, we can consider D[({k)] as 
a function of k and present in Fig. [3] the decoupling ratio 
due to counterflow velocity £ns{k)/£^\k) = D[({k)] , as 
a function of k/kx- Our estimate below shows that the 
crossover wave number kx (for which £ns{k) = £^\k)/2) 
is independent of the counterflow velocity and typically is 
in the relevant interval of scales, between -k/H and 'k/I. 


B. Typical value of the decoupling parameter C,{k) 

To clarify what are the typical values of (/(fc) in realis¬ 
tic conditions and how ((k) depends on the temperature 
and the counterflow velocity we note— that the main 
contributions to Tn and Tg, Eq. (l6cl) . come from and 
Ug, given by Eq. (IS^ : 

Tn “t“ Tg ~ Uji Ug = CTns H ; Oins — CXCXn = . ( 16 ) 

Pn 

Indeed, for scales fc/ <C 1 the viscous terms Vs,nk^ ^ 
7s,n(fc) and may be safely neglected, while for scales near 
the intervortex distance they are of the same order of 
magnitude. Moreover, for k£ ~ 1, Vs,nk^ ~ 7s,n(fc) ~ 
Ug^ii) if one estimates Ug^n ~ Ud in a classical manner via 
the root- mean square of the vorticity, see e.g. Refs.— 

Uci ^ However, as we explained above, in the 

counterflow there is an additional quantum mechanism 




























7 



FIG. 4: Color online. Temperature dependence of the mu¬ 
tual friction parameters for ^He, Ref.—: a for the superfluid 
Eq. (l3aD . blue line with circles; a^iT) — apsjpn in the normal 
fluid Eq. ((30, green line with triangles; and Ons = a + a-n = 
apjpvL in Eqs. (1161) . red line with squares. 


of the random vortex tangle excitation with scales if the 
order of 1. This mechanism provides the leading contri¬ 
bution to ris.n and, consequently, the leading contribu¬ 
tion to Tg^n, as written in Eq. (1161) . 

The temperature dependence of ans{T) = B{T)/2, 
where B{T) is the coefficient in the Vinen equation, tabu¬ 
lated in Ref.—, is shown in Fig. (K together with a{T) and 
a-a{T). The opposite temperature dependence of a{T) 
and a-o_{T) results in a weak temperature dependence of 
the parameter a^siT) in Eg. (ITSl) : it varies between 0.7 
and 0.5 in the relevant for counterflow experiments tem¬ 
perature range 1.4 -F 1.9 K . 

Now Eqs. (l3dl) and (flBl) together with Eq. (I14dl) give: 


Q{k) 


k 

C^ns ^ U ns 


(17a) 


Clearly, (^{k) oc k and it reaches its maximal value Cmax at 
the highest k value which is permissible in our approach, 
i.. k ~ fcmax — this is at the edge of the applicability. 
With £ ~ ~ this gives a simple estimate 

of Cmax, independent of Un&'- 


Cmax=i-50, io^T^lAK. (17b) 

Ckns K Ic 


Here for the numerical estimate we used Ons — 0.6, 
7 ^ ~ 100 s/cm^ and k ~ 10“^cm^/s. An important 
conclusion is that for large k the normal- and superfluid 
velocities are practically fully decoupled: for k ^ fc^ax 
C(fc) ~ 50 and the ratio Sns/^’n? is about 0.03 according 
to Eq. (|15bl) . 

An even more important conclusion is that according 
to Eq. (|17bll the range of wave numbers fcmax > k > 
kx, where fns/^ns^ < 1/2, extends over more than one 


decade: 

Cmax 


Cx 


2a„s K 7^ 


25, forr«1.4K. 


(17c) 

Equation (I17all allows us to estimate also the minimal 
value Cmin, which is attained at fcmin — tt/H: 


Cmin 


fins 

cm 

for T Ri 1.4 K, C/„s = 1 — 


(17d) 

H = 1 cm . 


This means that the value kx^ for which C{kx) = 2, 
is few times larger than /cmin — t^/H. Therefore for 
large scales (between H and i?x — T^/kx) we expect 
significant coupling of the normal- and superfluid 
velocities: for ( = 0.5 Eg. (I15bl) give Sns/^ns^ — 0.9. 
The value of Cmin is inversely proportional to Uns 
and for Uns > 1 cm/s become even smaller than 0.5. 
Accordingly, for Uns > 1 cm/s the interval between H 
and Rx become larger and the coupling between the 
normal and superfluid velocities at the largest scale H is 
even stronger: the ratio Sns/^ns^ > 0.9. 


IV. SUMMARY AND DISCUSSION 

We demonstrated that the cross-correlation function 
between normal- and superfluid velocity fluctuations 
£ns{k) in a turbulent counterflow of ^He is strongly af¬ 
fected by the relative velocity Uns- As described by 
Eqs. (fT4l) and illustrated in Fig.[0 this effect is governed 
by a dimensionless decoupling parameter ({k) oc k/Uns, 
given by Eq. (I17al) . This parameter increases with k and 
when k ~ kmux — 7r/f it reaches its maximum Cmax ^ 1 , 
as estimated in Eq. (I17bl) . Accordingly, the normal- and 
superfluid velocity fluctuations of small scales (i.e. for 
large wave numbers) are almost fully decoupled: the cor¬ 
relation £ns is much smaller than its value ~ 1 for 
Uns = 0. On the contrary, at large scales the energy con¬ 
taining fluctuations oi R ^ H are almost fully coupled: 

— £ns ^ ^’ns^. The crossover scale Rx, for which 
£ns = 5 is a few times smaller than H. Therefore 
the large scale fluctuations, ioi H > R > Rx may be 
qualitatively considered as coupled: Sns ^ On 

the other hand, in the large interval of small scales, for 
Rx ^ R ^ ^ the normal- and superfluid velocities may 
be considered as effectively decoupled: £ns ^ 

The coupling or decoupling of normal- and superfluid 
velocities crucially affects the energy dissipation due to 
the mutual friction. Correspondingly it also affects the 
energy spectra £s{k,t) and £n{k,t). To see this let us 
consider the evolution equations for these objects, which 
may be obtained multiplying Eq. (l3aD and (|3bl) in {k,t)- 
representation by Vs{k,t) and Vn{k,t) respectively, and 























averaging with respect to the turbulent statistics and di¬ 
rections of k : 


d 

2dt 

d 

2 dt 


£s{k,t) -I- AfCs 


(18a) 


= ns[£ns{k,t) - £s{k,t)] , 

+ k‘^Vn £n{k,t) + AfCn (18b) 

= nn[£ns{k,t) - £n{k,t)] . 


Here AfCs,n are nonlinear terms. For k ^ kx, due to 
the decoupling £ns{k) -C £s{k). Therefore it may be ne¬ 
glected on the RHS of Eq. (I18ap . which becomes —£ls£s- 
This is similar to the equation for £s for superfluid turbu¬ 
lence in ^He, where mutual friction drastically suppresses 
the energy spectrum £s{k) instead of the classical 

Kolmogorov spectrum £{k) oc k~^/^ one finds the spec¬ 
trum discussed by Lvov, Nazarenko and Volovik—: 


£s{k) oc 


1 


r 1 



(19) 


that terminates at some critical value fc*. This means 
that, provided that there exists a full decoupling of 
the velocities, the situation in counter-flowing superfluid 
component of “^He becomes similar to that in ^He tur¬ 
bulence with a normal fluid component at rest. Thus 
one expects that the spectrum (fTOl) describes the energy 
distribution between scales for fc ^ /cx. 

For k < kx due to the partial velocity correlations 
the energy dissipation is much weaker than for k > kx, 
although it cannot be neglected as in co-flowing “^He, with 
classical Kolmogorov-1941 (K41) energy spectrum. Thus 
we can expect only moderate suppression of the energy 
spectrum as compared to the K41 case, as was recently 
observed in Ref.—. 

A more detailed analysis of the energy spectra £s{k) 
and £’n(fc) in the counter-flowing ^He that accounts for 
the decoupling of the normal and superfluid turbulent 
velocity fluctuations and the resulting energy dissipation 
due to the mutual friction is beyond the scope of this 
paper. 


Acknowledgments 

We acknowledge L. Skrbek, S. Babuin and E. Varga 
for numerous and useful discussions of similarities and 
differences between co- and counter-flowing turbulence 
of superfluid ^He that inspired current research. VSL 
and AP acknowledge kind hospitality in Prague univer¬ 
sity and support of EuHIT project “V-Front” that make 
their visit possible. 

Appendix A: Some Definitions and Known 
Relationships 

To find the cross-correlation («(, • m() we need to re¬ 
call some definitions and relationships required for our 


derivation, which are well-known in statistical physics. 
The first is the set of Fourier transforms in the following 
normalization: 



/ (27r)3 ’ 

(Ala) 

Vn,sik,t) = 


(Alb) 


drdt Mn^s(r, t) exp[i(a;t — k ■ r)] 

. (Ale) 


The same normalization will be used for other objects of 
interest. 

Next we define the simultaneous correlations and cross¬ 
correlations in fc-representation, (proportional to S{k — 
k') due to homogeneity): 

(n„(fc,t).<(fc',t)) = (27r)3F;„„(fe)(5(fc-fe'),(A2a) 
(u,(fc,t)-<(fc',t)) = i27rfE,,ik)dik-k'),iA2h) 
{v^{k,t)-v:ik',t)) = (27r)3F;ns(fe)5(fe-fc') ■(A2 c) 

We also need to define cross-correlations (?;„ • ^s) 

(fc, u;)-representation: 

{Vn{k,ui) -v^ik^u)) (A3a) 

= (27r)^£’ns(fe, w) S{k — k') S(uj — uj') . 

This object is related to the simultaneous {v^ ■ v*) cross¬ 
correlation (IA2c|) via the frequency integral: 


{vn{k,t) ■ v*{k',t)) 


dujEns{k,uj) . (A3b) 


Here and below “tilde” marks the objects defined in 
(fc, u;)-representation. 

It is known also that the fc-integration of the correla¬ 
tions (|A2I) produces their one-point second moment: 


dk 

(^ 
' dk 
(27r)= 
■ dk 
(27r)3 


Enn{k) 

= {\Un{r,t)\^) , 

(A4a) 

^Ess(fc) 

= (|ws(r,t)p) , 

(A4b) 

■Ens{k) 

= {Un{r,t) ■ Us{r,t)) . 

(A4c) 


In the isotropic case, each of the three correlations E,,Xk) 
is independent of the direction of k: E,,,{k) = E, JJi) and 
J ... dk = Att J .. .k^ dk. This allows the introduction of 
the one-dimensional energy spectra £s, £n and the cross¬ 
correlation £ns as follows: 


f„(fc) = ^E^4k), 

£nsik) = ^Ens{k) . 


£s{k) 


2tt'^ 


Ess{k) , 


(A5) 




















9 


^ R. J. Donnelly, Quantized Vortices in Hellium II (Cam¬ 
bridge 3 University Press, Cambridge, 1991) 

^ Quantized Vortex Dynamics and Superfluid Turbulence, 
edited by C.F. Barenglii, R.J. Donnelly and W.F. Vinen, 
Lecture Notes in Physics 571 (Springer-Verlag, Berlin, 
2001) 

® L. Skrbek, K.R. Sreenivasan Phys Fluids 24 011301 (2012). 
^ C. F. Barenghi, V. S. L’vov, and P.-E. Roche, Proc Natl 
Acad Sci USA 111 , 4683 (2014). 

® M. R. Smith, R. J. Donnelly, N. Goldenfeld, and W. F. 

Vinen, Phys. Rev. Lett. 71, 2583 (1993). 

® P.L. Walstrom, J. C. Weisend, J.R. Maddocks and S. W. 

Van Sciver, Criogenics 28, 101 (1988). 

^ S. Babuin E. Varga L. Skrbek, J Low Temp Phys 175 
324 (2014). 

® H. E. Hall and W. F. Vinen, Proc. Roy. Soc. A 238, 204 
(1956). 

® K.W. Schwarz and J. R. Rozen, Phys. Rev. Lett., 66, 1896 
(1991). 

V.S. L’vov, S.V. Nazarenko and L. Skrbek, J. Low Tem¬ 
perature Physics, 145, 125 (2006). 

P-E Roche, C.F. Barenghi, E. Leveque Europhys Lett 87 
54006 (2009). 

Tchoufag J, Sagaut P Phys Fluids 22 125103 (2010). 

L. Boue, V S. L’vov., Y. Nagar., S. V. Nazarenko, A. 
Pomyalov., and 1. Procaccia. Phys. Rev. B 91, 144501 
(2015). 

L. Boue, V.S. Lvov, A. Pomyalov, 1. Procaccia Phys Rev 
Lett 110 014502 (2013). 


W. F. Vinen, J. Low Temp. Phys., 175 305 (2014). 

I.L. Bekarevich, I.M. Khalatnikov,Sov. Phys. JETP 13 (3), 
643-646 (1961). 

W. F. Vinen and J. J. Niemela, J. Low Temp. Phys. 128, 
167 (2002) 

V. S. L’vov, S. V. Nazarenko, G. E. Volovik, JETP Letters 
80, 479 (2004). 

K. W. Schwarz, Phys. Rev. B 38, 2398 (1988). 

L. Kondaurova, V. S. L’vov, A. Pomyalov and 1. Procac¬ 
cia,Phys. Rev. B 90 094501 (2014) 

R. H. Kraichnan, J. Fluid Mech. 5,497 (1959). 

H. W. Wyld, Ann. Phys. (N. Y.) 14, 143 (1961). 

V.S. L’vov, Yu. L’vov, A.G. Newell and V.E. Zakharov. 
Phys. Rev. E. 56, 390 (1997). 

V.S. L’vov and 1. Procaccia, Phys. Rev. E, 52, 3840 (1995) 
and 52, 3858 (1995). 

L. Kondaurova, V. S. L’vov, A. Pomyalov and 1. Procaccia, 
Phys. Rev. B, 89, 014502 (2014). 

L. Boue, V S. L’vov., Y. Nagar., S. V. Nazarenko, A. 
Pomyalov., and 1. Procaccia. Phys. Rev. B 91, 144501 
(2015). 

R. J. Donnelly and C. F. Barenghi,J. Phys. Ghem. Ref. 
Data, 27, No. 6, 1217(1998) 

L. Boue, V.S. L’vov, A. Pomyalov, 1. Procaccia, Phys. Rev. 
B 85, 104502 (2012) 

A. Marakov, J. Gao, W. Guo, S. W. Van Sciver,G. G. Ihas, 
D. N. McKinsey, and W. F. Vinen, Phys. Rev B 91 094503 
(2015). 


15 

16 

17 

18 

19 

20 

21 

22 

23 

24 

25 

26 

27 

28 

29