ABSENCE OF ANOMALOUS DISSIPATION FOR

VORTEX SHEETS
arXiv:2504.18523v1 [math.AP] 25 Apr 2025

TAREK M. ELGINDI1 , MILTON C. LOPES FILHO2 ,

AND HELENA J. NUSSENZVEIG LOPES2

Abstract. A family of solutions of the incompressible Navier-

Stokes equations is said to present anomalous dissipation if energy
dissipation due to viscosity does not vanish in the limit of small
viscosity. In this article we present a proof of absence of anoma-
lous dissipation for 2D flows on the torus, with an arbitrary non-
negative measure plus an integrable function as initial vorticity
and square-integrable initial velocity. Our result applies to flows
with forcing and provides an explicit estimate for the dissipation at
small viscosity. The proof relies on a new refinement of a classical
inequality due to J. Nash.

Compiled on: April 28, 2025

Contents
1. Introduction 1
2. Real analysis preliminaries 3
3. A refinement of Nash’s inequality 6
4. An extension to bounded Radon measures 9
5. No anomalous dissipation 12
6. No anomalous dissipation for vortex sheets 17
7. Comments and conclusions 21
Acknowledgments 22
References 23

1. Introduction
This article concerns families of solutions of the incompressible Navier-
Stokes equations on the torus T2 = [−π, π]2 , given by:
1
2 ELGINDI, LOPES FILHO, AND NUSSENZVEIG LOPES


 ∂t uν + (uν · ∇)uν = −∇pν + ν∆uν + F ν , in T2 × (0, T )
div uν = 0, in T2 × [0, T )
 uν (·, 0) = uν , in T2 ,
0
(1.1)
ν ν
where ν > 0 is the viscosity, u the flow velocity, p is the pressure and
F ν = F ν (·, t) is an external force. If ν = 0, system (1.1) is called the
Euler system. Energy methods yield the identity:
1d
ˆ ˆ ˆ
|u | dx = −ν |ω | dx + F ν · uν dx,
ν 2 ν 2
(1.2)
2 dt
where ω ν = curl uν is the vorticity of the flow. The term
ˆ ˆ
ν
ζ ≡ν |ω ν |2 dx dt (1.3)

is called the energy dissipation term, or simply the dissipation term, in

this context.
For initially smooth flows, it is easy to see that ζ ν vanishes as ν → 0,
leading to solutions of the Euler system which satisfy an exact energy
balance. However, this is not necessarily the case if the flow is initially
nonsmooth. A sequence of solutions {uν } for which limν→0 ζ ν > 0 is
said to exhibit anomalous dissipation. The issue of presence or absence
of anomalous dissipation in the vanishing viscosity limit is central to
the study of turbulence, see for example [9] and references therein.
In [11], F. Jin and co-authors proved that, under suitable assump-
tions on the forcings, strong convergence of the vanishing viscosity
approximations is equivalent to the inviscid limit satisfying energy bal-
ance. Part of their proof involves showing that strong convergence im-
plies absence of anomalous dissipation; the converse of which, however,
is not necessarily true. Additionally, it is proved in [11] that strong
convergence of the vanishing viscosity approximation follows if initial
vorticity is in a rearrangement-invariant space compactly embedded in
H −1, see also [15]. The conceptual gap between initial vorticities in
these spaces and initial vorticities whose singular parts are nonnega-
tive bounded Radon measures, called vortex sheet initial data is that
the L2 -compactness of velocity approximations is lost. Consequently, a
compensated compactness argument is called for to show that the weak
limit of the vanishing viscosity approximation is a weak solution of the
inviscid equations, see [4, 16, 21]. In view of [11, Theorem 2.7] these
weak solutions can, indeed, have an energy defect. Still, it makes sense
to ask whether the viscous approximations exhibit anomalous dissipa-
tion. This question was answered in the negative by L. De Rosa and
 ABSENCE OF ANOMALOUS DISSIPATION 3

J. Park in the nice recent work [18] and, independently, in the present
work.
Although the results obtained here are similar to those obtained in
[18], there are a few differences. First, our work applies to flows with
forcing, and it allows for more general ways in which initial data may
be approximated. Also, our proof is quite different, involving direct
quantitative estimates, and, as a result, provides a rate of vanishing of
dissipation with respect to viscosity. The key ingredient of our work
is a new refinement of Nash’s inequality, given in Proposition 3.2. In
contrast, De Rosa and Park use the classical Nash inequality (3.1) to
obtain a key estimate, and a mollification which makes clever use of
parabolic scaling.
The remainder of this paper is organized as follows. In Section 2
we prove a convolution inequality and use it to prove an interpolation
inequality. In Section 3 we derive a refinement of an inequality due to
J. Nash, which, in turn, is a limiting case of the Gagliardo-Niremberg
inequalities. In Section 4 we present an adaptation, to bounded mea-
sures, of the inequality in the previous section, replacing the uniform
integrability of L1 functions with an estimate of the total mass of a
measure in small disks. It is only in the final sections that we turn to
solutions of the incompressible flow equations. In Section 5 we prove
absence of anomalous dissipation, assuming suuitable uniform bounds
on the Navier-Stokes solutions, as viscosity vanishes. In Section 6 we
apply the result Section 5 to flows with initial vorticities whose singular
part is a nonnegative measure and we obtain a rate for the vanishing
of the dissipation. Finally, in Section 7, we remark on an extension of
our results, describe an example which shows the necessity of some of
the conditions we use to prove absence of anomalous dissipation and
we present a few concluding remarks.

2. Real analysis preliminaries

In this section we will develop some real analysis preliminaries which
will serve as a basis for our work.
Let f and g be smooth functions on Rd and consider the convolution
ˆ
f ∗ g = (f ∗ g)(x) ≡ f (y)g(x − y) dy.

By Young’s inequality, we have that

kf ∗ gkL2 ≤ kf kL1 kgkL2 . (2.1)

4 ELGINDI, LOPES FILHO, AND NUSSENZVEIG LOPES

In what follows we will show that we can refine (2.1) if g is sup-

ported in a ball B∗ , replacing kf kL1 in (2.1) by something smaller (in
particular, related to the integral of |f | on balls of the same size as B∗ ).
Lemma 2.1. Let f ∈ L1 (Rd ) and g ∈ L2 (Rd ) and assume that supp g
is contained in a ball B∗ . Then,
!
r
kf ∗ gkL2 ≤ kf kL1 sup kf kL1 (B) kgkL2
|B|=|B∗ |

Proof. Let us begin by noting that

!
kf ∗ gkL∞ ≤ sup kf kL1 (B) kgkL∞ . (2.2)
B=|B∗ |

From Young’s inequality we have, additionally,

kf ∗ gkL1 ≤ kf kL1 kgkL1 . (2.3)
We now claim that
!1/2
1/2
kf ∗ gkL2 ≤ kf kL1 sup kf kL1 (B) kgkL2 . (2.4)
|B|=|B∗ |

To see this let LpB∗ denote the Banach space of Lp functions that
vanish identically outside of B∗ .
Fix f ∈ L1 (Rd ) and consider the linear operator I given by
I : LpB∗ (Rd ) → Lp (Rd )
.
g 7→ f ∗ g
Then, from (2.2), we have
kIkL∞
B∗ →L
∞ ≤ sup kf kL1 (B) .
|B|=|B∗ |

Also, (2.3) gives

kIkL1B →L1 ≤ kf kL1 .
∗

We use the Riesz-Thorin theorem, see [8], to conclude that

!1/2
1/2
kIkL2B →L2 ≤ kf kL1 sup kf kL1 (B) .
∗
|B|=|B∗ |

This establishes the claim and, thus, concludes the proof.


 ABSENCE OF ANOMALOUS DISSIPATION 5

Lemma 2.2. There ˆ exists a universal constant C > 0 such that, if

f ∈ H 1 (T2 ) with f = 0, 0 < α < 1/2, 0 < ε < 1 and N ∈ R+ , then:
T2
  
1
ˆ
kf k2L2 2
≤C N kf kL1 sup |f (y)| dy + kf kL1 α
z∈T2 {|z−y|<εα } Nε

1
+ 2 k∇f k2L2 . (2.5)
N
Proof. We begin with the Plancherel identity:
X
kf k2L2 = |fb(k)|2 , (2.6)
k∈Z2

where fb(k) is the k-th Fourier coefficient of f .

Consider the ℓ∞ norm of z = (z1 , z2 ) on T2 , |z|∞ = max{|z1 |, |z2|}.
Using (2.6), we have
X X
kf k2L2 = |fb(k)|2 + |fb(k)|2
|k|∞ <N |k|≥N
X 1 X X 1
≤ |fb(k)|2 + 2 C|k|2 |fb(k)|2 ≤ |fb(k)|2 + 2 k∇f k2L2 .
N N
|k|∞ <N |k|≥N |k|∞ <N
(2.7)
We introduce the projection of f onto the first N ≡ ⌈N⌉ Fourier
modes, with respect to the ℓ∞ -norm on T2 :
 ∨
PN f = fb1|k|∞ <N . (2.8)
P
Therefore |k|∞ <N |fb(k)|2 ≤ kPN f k2L2 .
Note that, since we are working with the ℓ∞ -norm on T2 , it follows
that ˆ
PN f (x) = f (x − y)DN (y) dy, (2.9)
T2
where DN is the
P square Dirichlet kernel, DN (z1 , z2 ) = dN (z1 )dN (z2 ),
with dN (z) = −N ≤k≤N eikz , the 1-dimensional Dirichlet kernel. Hence
sin[(2N + 1)(z1 /2)] sin[(2N + 1)(z2 /2)]
DN (z1 , z2 ) = , (2.10)
sin(z1 /2) sin(z2 /2)
see [19].
With this notation we now write, for ε < 1:
PN f = f ∗ (DN 1|z|∞<εα ) + f ∗ (DN 1|z|∞ ≥εα ) ≡ PNI f + PNO f. (2.11)
6 ELGINDI, LOPES FILHO, AND NUSSENZVEIG LOPES

We estimate each of these terms separately. For PNI we will use

the refined estimate for convolutions given in Lemma 2.1 with g =
DN 1|z|∞<εα . We deduce that
 ˆ 
I 2
kPN f kL2 ≤ kf kL1(T2 ) sup |f (y)| dy kDN k2L2 (T2 ) . (2.12)
z∈T2 {|z−y|∞ <εα }

It is easy to see that

kDN k2L2 = kdN k4L2 = 4πN 2 . (2.13)
Using (2.13) in (2.12) yields
ˆ
kPNI f k2L2 2
≤ CN kf kL1 (T2 ) sup |f (y)| dy. (2.14)
z∈T2 {|z−y|∞ <εα }

Next we estimate PNO f . We begin with the elementary observation

that, if 0 < ρ < π, then
ˆ π
π2
d2N (x) dx ≤ . (2.15)
ρ ρ
Using (2.15) and the formula for DN given in (2.10) we find
ˆ π
2 2 1
kDN 1|z|∞ ≥εα kL2 (T2 ) ≤ 8kdN kL2 (T1 ) d2N (x) dx ≤ CN α . (2.16)
εα ε
By Young’s inequality we deduce
1
kPNO f k2L2 ≤ Ckf k2L1 N α . (2.17)
ε
From (2.14) and (2.17) it follows that
 
1
ˆ
2 2
kPN f kL2 ≤ CN kf kL1 sup |f (y)| dy + kf kL1 .
z∈T2 {|z−y|∞ <εα } Nεα
(2.18)
Finally, putting together (2.18) and (2.7) yields the desired result.


3. A refinement of Nash’s inequality

Our main results depend on a refinement of an inequality due to J.
Nash which, in dimension 2, corresponds to
 2
2
kf kL2 (R2 ) ≤ kf k2L1 (R2 ) k∇f k2L2 (R2 ) , (3.1)
for all f ∈ H 1 (R2 ) ∩ L1 (R2 ), see [1, 17]. This is the content of the
present Section.
Fix K > 0. Let F be a family of functions in L1 such that
(A) kf kL1 ≤ K for all f ∈ F, and
 ABSENCE OF ANOMALOUS DISSIPATION 7

(B)
ˆ
lim sup sup |f (y)| dy = 0.
r→0+ f ∈F z∈T2 {|z−y|<r}

Let
 
1 r
ˆ
η(r) = max sup sup |f (y)| dy, , for 0 ≤ r ≤ π.
K f ∈F z∈T2 {|z−y|<r} π
(3.2)
It is immediate that, if 0 ≤ r ≤ π and f ∈ F, then:

0 ≤ η ≤ 1; r ≤ πη(r); (3.3)
ˆ
sup |f (y)| dy ≤ K η(r). (3.4)
z∈T2 {|z−y|<r}

Proposition 3.1. Let K > 0 and consider F ⊂ H 1 (T2 ) a family of

mean-free functions satisfying A and B. Then there exists a universal
constant C > 0 such that
r  
−1/4
kf k2L2 ≤ C (K 2 + 1) k∇f kL2 η k∇f kL2 (3.5)

for all f ∈ F such that k∇f kL2 > 1.

Proof. Let f ∈ F and √ assume k∇f kL2 > 1. Let 0 < ε < 1 and use
α = 1/4 and N = 1/ ε in Lemma 2.2. Then (2.5) becomes
!
2
kf kL kf k
ˆ
1
L1 1/4
kf k2L2 ≤C sup |f (y)| dy + ε + εk∇f k2L2 .
ε z∈T2 {|z−y|<ε1/4} ε
(3.6)

We wish to use ε∗ in (3.6), where

r  
∗ 1 −1/4
ε ≡ η k∇f kL2 ,
k∇f kL2

with η the function that was introduced in (3.2). We note in passing

−1/4
that the hypothesis k∇f kL2 > 1 implies that k∇f kL2 < 1, thus in the
domain of η; since 0 ≤ η ≤ 1 (see (3.3)) we conclude that 0 < ε∗ < 1.
Let ε = ε∗ . Then, for the first term on the right-hand-side of (3.6),
we have
8 ELGINDI, LOPES FILHO, AND NUSSENZVEIG LOPES

kf kL1 1

ˆ
sup |f (y)| dy ≤ ∗ K 2 η (ε∗ )1/4
ε∗ z∈T2
{|z−y|<(ε∗ )1/4 } ε
! r 
K 2 k∇f kL2 1 
2 −1/4
≤r  η 1/4
= K k∇f kL2 η k∇f kL2 ,
−1/4 k∇f kL2
η k∇f kL2
(3.7)
where we used (3.4) in the first estimate and (3.3) in the second one.
Next, we estimate the second term in (3.6):
r  
kf k2L1 ∗ 1/4 K 2 ∗ 1/4 2 −1/4
(ε ) ≤ ∗ πη((ε ) ) ≤ πK k∇f kL2 η k∇f kL2 , (3.8)
ε∗ ε
where we first used (3.3), followed by the previous estimate, (3.7).
Finally, for the last term of (3.6) we have:
r  
∗ 2 −1/4
ε k∇f kL2 = k∇f kL2 η k∇f kL2 . (3.9)

Substituting (3.7), (3.8) and (3.9) in (3.6) yields the desired result.

Let us consider an extension of the function η, defined in (3.2), to
all of [0, +∞), given by

η(r) if 0 ≤ r ≤ π,
η = η(r) = (3.10)
1 if r > π.
We will use η to prove our next result.
Proposition 3.2. Let K > 0 and let F ⊂ H 1 (T2 ) be a family of mean-
free functions such that A and B hold. Then there exists Υ ∈ C1 ,
convex, increasing and superquadratic, such that


Υ kf k2L2 ≤ k∇f k2L2 ,
for all f ∈ F.
Proof. Let f ∈ F. We have already established that, if k∇f kL2 > 1,
then r  
−1/4
kf k2L2 ≤ C k∇f kL2 η k∇f kL2 ,

for some constant C > 0, where C depends on K and additional uni-

versal constants. Using the Poincaré inequality and the definition of η,
it is not difficult to deduce that
 ABSENCE OF ANOMALOUS DISSIPATION 9

r  
−1/4
kf k2L2 ≤ Ck∇f kL2 η πk∇f kL2 (3.11)
for all f ∈ F. We note that the function η is continuous and non-
decreasing.
Now let ˆ x q
Φ(x) = C η (πy −1/4 ) dy.
0
It is not difficult to see that Φ is concave, increasing, and sublinear.
Furthermore, since η is non-decreasing, we have that
q
Φ(x) ≥ Cx η (πx−1/4 ).
Consequently, we have from (3.11) that
 
kf k2L2 ≤ Φ k∇f kL2 .

Since Φ is C1 , increasing, concave and sublinear it follows that it has

an inverse Φ−1 which is C1 , increasing, convex and superlinear. Hence,
Υ = Υ(x) = (Φ−1 )2 (x) (3.12)
belongs to C1 , is increasing, convex and superquadratic. Clearly, the
function Υ is such that Υ(kf k2L2 ) ≤ k∇f k2L2 for all f ∈ F, as desired.
This concludes the proof.


4. An extension to bounded Radon measures

The purpose of this section is to obtain a special version of Proposi-
tion 3.2 valid for families of mean-free functions in H 1 (T2 ) which can
be written as the sum of a nonnegative Radon measure in H −1 and
an Lp (T2 ) function, for some p > 1. We assume that this family is
uniformly bounded in H −1 + Lp .
Let µ be a nonnegative measure in H −1 (T2 ). We begin by recalling
a basic estimate on the mass of µ in small discs, namely
ˆ
dµ(y) ≤ CkµkH −1 | log ρ|−1/2 . (4.1)
{|z−y|<ρ}

This estimate was established, in this form, in [20], where the author
used it to study limits of singular solutions of the 2D incompressible
Euler equations. It is also at the heart of the proof of existence of weak
solutions to the 2D incompressible Euler equations with vortex sheet
initial data of a distinguished sign, see [4].
We begin with a proposition which is analogous to Proposition 3.1.
10 ELGINDI, LOPES FILHO, AND NUSSENZVEIG LOPES
ˆ
1 2
Proposition 4.1. Let f ∈ H (T ) such that f = 0. Assume that
T2
f = µ + w, with µ ∈ BM ∩ H −1 , µ ≥ 0, and w ∈ Lp (T2 ), for some
p > 1. Then there exists a universal constant C > 0 such that,
k∇f kL2
kf k2L2 ≤ C [kf kL1 (kµkH −1 + kwkLp + kf kL1 ) + 1] p ,
4
log k∇f kL2
(4.2)
whenever k∇f kL2 > e2 .
ˆ
Proof. We begin by estimating |f | dz. We have, using (4.1):
|z−y|<ρ
ˆ ˆ ˆ
|f | dz ≤ µ dz + |w| dz
|z−y|<ρ |z−y|<ρ |z−y|<ρ

≤ kµkH −1 | log ρ|−1/2 + CkwkLp ρ2p/p−1 .

√
Let 0 < ε < 1 and choose N = 1/ ε in Lemma 2.2. Then (2.5)
becomes
"
1

kf k2L2 ≤C kf kL1 kµkH −1 | log ε|−1/2 + kwkLp ε2αp/(p−1) + kf kL1 ε−α+1/2
ε
#
+ εk∇f k2L2 . (4.3)

Recall we chose α < 1/2, so the power of ε in the third term is

positive. Clearly, the power of ε in the second term is also positive.
Therefore, if 0 < ε < 1, then the logarithmic term dominates the
algebraic powers of ε.
In summary, it follows from (4.3) and that, for all 0 < ε < 1,
 
2 1 −1/2 2
kf kL2 ≤ C kf kL1 (kµkH −1 | + kwkLp + kf kL1 ) | log ε| + εk∇f kL2 .
ε
(4.4)
Assume that
k∇f kL2 > e2 (4.5)
and choose
1
ε∗ = p . (4.6)
k∇f kL2 log k∇f k2L2
4

Clearly 0 < ε∗ < 1.

It is easy to see that, using ε = ε∗ in (4.4), we obtain the desired
result.

 ABSENCE OF ANOMALOUS DISSIPATION 11

Proposition 4.1 allows us to adapt Proposition 3.2 to bounded Radon

measures. The main difference between our next result and that of
Proposition 3.2 is that, since we have a precise estimate on the behavior
of our functions in small disks, this leads to more precise control on
the relative behavior of k∇f k2L2 and kf k2L2 .
Proposition 4.2. Let F ⊂ H 1 (T2 ) be a family of mean-free functions.
Assume that, if f ∈ F then f = µ + w, with µ ∈ BM ∩ H −1 , µ ≥ 0,
and w ∈ Lp (T2 ), for some p > 1. Let K > 0 and assume further that,
for all f ∈ F, there exists a decomposition f = µ + w as above such
that kµkBM + kµkH −1 + kwkLp ≤ K.
Then there exists

Φ ∈ C1 , concave, increasing, sublinear, Φ(x) =
−1/4
O x| log x| for large x, such Υ = (Φ−1 )2 satisfies


Υ kf k2L2 ≤ k∇f k2L2 , for all f ∈ F.
Proof. It follows from (4.2) that, whenever k∇f kL2 > e2 , we have
k∇f kL2
kf k2L2 ≤ C p , (4.7)
4
log k∇f kL2
where C = C(K).
We introduce
x
Φ1 = Φ1 (x) ≡ C √ , for x > e2 .
4
log x
It follows from an elementary calculation that, if x > e2 , then Φ′1 > 0
and Φ′′1 < 0, so that Φ1 is increasing and concave in this interval. In
addition, limx→+∞ Φ′1 (x) = 0.
We extend Φ1 to [0, e2 ] so that the extension, still called Φ1 , belongs
to C1 and is concave and increasing on all of [0, +∞).
Let, also,
Φ2 = Φ2 (x) = CP x2 ,
with CP the constant from the Poincaré inequality for mean-free func-
tions.
Choose L > 0 such that Φ2 (e2 ) = LΦ1 (e2 ) and let ϕ(x) ≡ LΦ1 (x) −
Φ2 (x). Then ϕ(0) = 0 = ϕ(e2 ) and ϕ is concave. It follows that
ϕ(x) ≥ 0 for x ∈ [0, e2 ], i.e. Φ2 (x) ≤ LΦ1 (x) on [0, e2 ].
Let Φ = Φ(x) = (L + 1)Φ1 (x). Clearly
Φ ∈ C1 , Φ is concave,
increasing, sublinear, Φ(x) = O x| log x|−1/4 for large x, as desired.
Next, in view of (4.7), the definition of Φ, and using the Poincaré
inequality for mean-free functions together with the inequality Φ2 ≤ Φ
on [0, e2 ], we obtain
kf k2L2 ≤ Φ(k∇f kL2 ), for all f ∈ F.
12 ELGINDI, LOPES FILHO, AND NUSSENZVEIG LOPES

It is immediate that the inverse function, Φ−1 , belongs to C1 and

that it is increasing, convex and superlinear.
Set
Υ = Υ(x) = (Φ−1 )2 (x). (4.8)
1
Then Υ is C , convex, increasing, superquadratic, and satisfies the
desired estimate.
This concludes the proof.


5. No anomalous dissipation
In this section we finally turn our attention to the incompressible
flow equations, namely, the Navier-Stokes system (1.1) and the Euler
system below:

 ∂t u + (u · ∇)u = −∇p + F, in T2 × (0, T )
div u = 0, in T2 × [0, T ) (5.1)
 2
u(0, ·) = u0 , in T .
We will apply Proposition 3.2 to show that, under certain assump-
tions, there is no anomalous dissipation for a family of weak solutions of
the 2D incompressible Navier-Stokes equations (1.1) whose weak limit
is a weak solution of the Euler equations.
Let us begin by recalling the definition of a weak solution of (5.1).
Definition 5.1. Fix T > 0, let u0 ∈ L2 (T2 ) be a divergence-free vector
field. Let F ∈ L1 (0, T ; L2 (T2 )) and assume that div F (·, t) = 0 a.e.
t ∈ (0, T ). A vector field u ∈ L∞ (0, T ; L2(T2 )) is said to be a weak
solution of (5.1), if the following conditions are satisfied:
(1) for all divergence-free test vector fields Φ ∈ Cc∞ (T2 × [0, T )), we
have
ˆ Tˆ ˆ
{u · ∂t Φ + u · DΦu} dx dt + u0 (x) · Φ(x, 0) dx
0 T2 T2
ˆ Tˆ
= F · Φ dx dt; (5.2)
0 T2
(2) div u(·, t) = 0 holds in the sense of distributions, a.e. t ∈ [0, T ].
The problem of existence of weak solutions with vortex sheet initial
data, i.e. u0 ∈ L2 , div u0 = 0, curl u0 = ω0 ∈ BM, was first addressed
in pioneering work by R. DiPerna and A. Majda, see [5–7], and, in
this generality, it remains an open problem. If ω0 ∈ L1 ∩ H −1 then the
existence of a weak solution was established in [21]. If the singular part
of ω0 belongs to BM ∩ H −1 and has a distinguished sign then existence
 ABSENCE OF ANOMALOUS DISSIPATION 13

of a weak solution with this initial data is due to J.-M. Delort, see [4].
Uniqueness of weak solutions has not been proved for such irregular
initial data and it is unlikely to hold. It is particularly worthy of note
that nonuniqueness was recently proved in the setting ω ∈ L∞ p
t Lx , for
some p > 1 but close to 1, see [2].
It is natural to concentrate on those weak solutions which arise as
vanishing viscosity limits of solutions of the Navier-Stokes equations.
These are called physically realizable weak solutions, originally intro-
duced in [3, Definition 3] for unforced flows, see also [12, Definition
2.2], and in [14, Definition 2.5] for flows with forcing.
Anomalous dissipation refers to the vanishing viscosity approxima-
tion of a physically realizable weak solution. Our first theorem applies
to such a family of approximations, called physical realizations of a
given physically realizable weak solution.
Theorem 5.2. Fix T > 0. Let {uν0 }ν>0 be a family of divergence-free
vector fields in L2 (T2 ). Let uν ∈ L∞ (0, T ; L2(T2 )) ∩ L2 (0, T ; H 1(T2 ))
be the solution to 2D Navier-Stokes with viscosity ν, with initial data
uν0 , and forcing F ν ∈ L2 (0, T ; L2 (T2 )). Set ω ν ≡ curl uν . Assume the
following conditions:

H(a) uν0 → u0 strongly in L2 (T2 );

H(b) F ν ⇀ F weakly in L2 (0, T ; L2 (T2 ));

H(c) ω ν is bounded in L∞ (0, T ; L1 (T2 )).

If additionally
ˆ
lim sup sup sup |ω ν | dx = 0, (5.3)
r→0+ ν>0 t∈(0,T ) z∈T2 {|x−z|<r}

then, passing to subsequences as needed, uν converges, weak-∗ L∞ (0, T ; L2 (T2 )),

to a (physically realizable) weak solution of 2D Euler with initial data
u0 and
ˆ T
lim sup ν kω ν (·, t)k2L2 dt = 0.
ν→0+ 0

Proof. We begin by noting that, from hypotheses H(a), H(b) and

H(c), we have:
kuν0 kL2 + kω ν kL∞ ν
1 + kF kL2 L2 ≤ K,
t Lx t x

for some K > 0 and for all ν > 0.

14 ELGINDI, LOPES FILHO, AND NUSSENZVEIG LOPES

Recall the energy balance identity for uν :

ˆ t ˆ t
1 ν 2 1 ν 2 ν 2
ku (t)kL2 = ku0 kL2 − ν kω (τ )kL2 dτ + hF ν , uν iL2 dτ, (5.4)
2 2 0 0

Therefore, from (5.4), we find

kuν (·, t)kL2 ≤ C, for t ∈ [0, T ], (5.5)
for some C = C(K, T ) > 0, see also [11, Lemma 3.1].
It follows from (5.4) and (5.5) that the dissipation term is bounded:
ˆ t
ν kω ν (·, τ )k2L2 dτ ≤ C, (5.6)
0

for some C = C(K, T ) > 0 and for all t ∈ [0, T ], see also [11, (3.6)].
The evolution of ω ν is governed by the vorticity formulation of the
Navier-Stokes equations (1.1) on the torus, given by

 ∂t ω ν + (uν · ∇)ω ν = ν∆ω ν + curl F ν , in T2 × (0, T )
div uν = 0, curl uν = ω ν , in T2 × [0, T ) (5.7)
 ν ν
ω (0, ·) = ω0 , in T2 .
Let r > 0. Energy methods for the vorticity equation (5.7) provide
the following estimate:
ˆ t
1 t ν
ˆ
ν 2 ν 2 ν 2
kω (·, t)kL2 ≤ kω (·, r)kL2 −ν k∇ω (·, τ )kL2 dτ + kF (·, τ )k2L2 dτ,
r ν r
(5.8)
for any t ≥ r, see for example [11, (3.14)]. It follows from (5.8) together
with (5.6) that
C
kω ν (·, t)k2L2 ≤ , (5.9)
tν
for all t > 0, see also [11, Lemma 3.7].
Define: ˆ t
ζ ν = ζ ν (t) = ν kω ν (τ )k2L2 dτ. (5.10)
0
Let
F = {ω ν (·, t), 0 < t < T, ν > 0}.
Note that, in view of (5.9), F ⊂ H 1 (T2 ) and, furthermore, since
ω ν = curl uν , it follows that ω ν is mean-free. By hypothesis H(c) and
assumption (5.3) this family satisfies conditions A and B. Thus we can
use Proposition 3.2 to obtain a convex, increasing and superquadratic
function Υ such that
Υ(kω ν (·, t)k2L2 ) ≤ k∇ω ν (·, t)k2L2 . (5.11)
 ABSENCE OF ANOMALOUS DISSIPATION 15

Fix δ > 0 and introduce

ˆ t
ζδν (t) =ν kω ν (·, τ )k2L2 dτ.
δ
From (5.8) and (5.9) we obtain
ˆ t
ν 2 C 2
νkω (·, t)kL2 ≤ − ν k∇ω ν (·, τ )k2L2 dτ + K. (5.12)
δ δ
Now, since the left-hand-side of (5.12) is nonnegative, it follows that
ˆ t
2 C
ν k∇ω ν (·, τ )k2L2 dτ ≤ K + .
δ δ
Using (5.11) we find
ˆ t
2 C
ν Υ(kω ν (·, τ )k2L2 ) dτ ≤ K + .
δ δ
Recall Υ is convex. Then Jensen’s inequality yields
 ˆ t 
2 1 ν 2 C
ν (t − δ)Υ kω (·, τ )kL2 dτ ≤ K + .
t−δ δ δ
Therefore we have
 ν 
2 ζδ (t) C
ν (t − δ)Υ ≤K+ , (5.13)
ν(t − δ) δ
which we re-write as
 
ζδν K + C/δ
Υ ≤ 2 . (5.14)
ν(t − δ) ν (t − δ)
For convenience let M = K + C/δ.
Recall that, by construction, Υ = (Φ−1 )2 , where Φ is a concave,
increasing, sublinear function. Therefore (5.14) yields
  √
−1 ζδν M
Φ ≤ √ .
ν(t − δ) ν t−δ
Thus, since Φ is increasing, we obtain
√ !
ζδν (t) M
≤Φ √ .
ν(t − δ) ν t−δ
Hence √ !
M
ζδν (t)
≤ ν(t − δ) Φ √
ν t−δ
√ √ 
√ √ Φ M /ν t − δ
= M t−δ √ √ .
M /ν t − δ
16 ELGINDI, LOPES FILHO, AND NUSSENZVEIG LOPES
√
M
Let xν = √ . We have shown that
ν T −δ
√ √ Φ (xν )
ζδν (T ) ≤ M T − δ . (5.15)
xν
Since xν → +∞ as ν → 0+ , it follows from the sublinearity of Φ
that the right-hand-side above vanishes as ν → 0+ .
We have established, therefore, that, for every δ > 0,
lim sup ζδν (T ) = 0. (5.16)
ν→0+
To conclude the proof it remains to examine
ˆ δ
ν kω ν (·, τ )k2L2 dτ.
0
Recall that, from (5.5), u is bounded in L∞ (0, T ; L2 (T2 )). Passing
ν

to subsequences as needed it follows that uν converges weakly-∗ in

L∞ (0, T ; L2(T2 )). That its weak-∗ limit u is a weak solution of the
2D Euler equations follows by a simple adaptation of the proof of
[4, Théorème 2.1.2], see also [20, Theorem 3.3] and [10, Proposition
2.1], using the key assumption (5.3).
Rewrite the energy balance equation (5.4), this time from 0 to δ, as:
ˆ δ ˆ δ
ν 2 1 ν 2 1 ν 2
ν kω (·, τ )kL2 dτ = ku0 kL2 − ku (δ)kL2 + hF ν , uν iL2 dτ.
0 2 2 0
We wish to take the lim supν→0+ on both sides of this equation. To this
end first recall that, since uν ⇀ u weak-∗ in L∞ (0, T ; L2(T2 )) we have,
by weak lower semicontinuity, that
ku(δ)k2L2 ≤ lim inf
+
kuν (δ)k2L2 .
ν→0
In addition,
ˆ δ √
hF ν , uν iL2 dτ ≤ kF ν kL2t L2x δkuν kL∞ 2.
t Lx
0

Therefore, recalling hypothesis H(a), that uν0 → u0 strongly in L2 (T2 ),

we obtain
ˆ δ
1
√
lim sup ν kω ν (τ )k2L2 dτ ≤ ku0k2L2 − ku(δ)k2L2 + C δ. (5.17)
ν→0+ 0 2
Now, it was established in [11, Lemma 3.5] that, under the current
hypotheses, the inviscid solution is right-continuous in L2 at t = 0,
that is,
lim+ ku(t) − u0 k2L2 = 0,
t→0
 ABSENCE OF ANOMALOUS DISSIPATION 17

Hence the right-hand-side of (5.17) is vanishingly small as δ → 0+ .

Putting together (5.16) and (5.17) and recalling the definition of ζ ν ,
(5.10), we obtain
ˆ δ
ν ν
lim sup ζ (T ) ≤ lim sup ζδ (T ) + lim sup ν kω ν (τ )k2L2 dτ
ν→0+ ν→0+ ν→0+ 0
1
√
≤ ku0 k2L2 − ku(δ)k2L2 + C δ.
2
Above δ > 0 is arbitrary and the left-hand-side above vanishes as
δ → 0+ . This establishes that
lim sup ζ ν (T ) = 0.
ν→0+

This concludes the proof.



6. No anomalous dissipation for vortex sheets

In this section we will use Theorem 5.2 to show there is no anomalous
dissipation for physical realizations of weak solutions with vorticities
of vortex sheet regularity. We note that, while Theorem 5.2 requires
assumptions on the viscous solutions, namely H(c) and (5.3), for the
results in the present section we place assumptions only on the data.
Fix T > 0 and let uν ∈ L∞ (0, T ; L2 (T2 )) ∩ L2 (0, T ; H 1(T2 )) be the
solution to the 2D Navier-Stokes equations (1.1) with divergence-free
initial data uν0 ∈ L2 (T2 ) and forcing F ν ∈ L2 (0, T ; L2 (T2 )). Assume
hypotheses H(a) and H(b) hold true. Let ω0ν = curl uν0 .
Theorem 6.1. Let ω ν ≡ curl uν . Suppose that
H1(a) {ω0ν }ν>0 is weakly compact in L1 (T2 );

H1(b) for almost every τ ∈ (0, T ), {curl F ν (·, τ )}ν>0 is weakly compact
in L1 (T2 ) and
ˆ T
sup k curl F ν (·, τ )kL1 dτ < ∞.
0 ν>0

Then we have:
ˆ T
lim sup ν kω ν (τ )k2L2 dτ = 0.
ν→0+ 0

Remark 6.2. It is always possible to find solutions of Navier-Stokes

uν with forcing F ν satisfying conditions H(a), H(b) as well as H1(a)
and H1(b) by, for instance, choosing uν0 , ω0ν , F ν to be mollifications of
given divergence-free u0 ∈ L2 , ω0 ∈ L1 and appropriate F , respectively.
18 ELGINDI, LOPES FILHO, AND NUSSENZVEIG LOPES

Indeed, this is a consequence of an easy adaptation of the proof of [21,

Theorem1].
Proof. We wish to use Theorem 5.2 so we need only check that hy-
potheses H(c) and (5.3) hold true.
We begin by noting that, from the assumptions H(a), H(b), H1(a),
H1(b), we have:
kuν0 kL2 + kF ν kL2t L2x + kω0ν kL1 + k curl F ν kL1t L1x ≤ K,
for some K > 0 and for all ν > 0.
Standard energy estimates give
sup kω ν (·, t)kL1 ≤ kω0ν kL1 + k curl F ν kL1t L1x ≤ K.
t∈(0,T )

Thus hypothesis H(c) is verified.

We now turn to (5.3).
For f ∈ L1 (T2 ) let us recall the definition of the rearrangement
invariant maximal function Ms (f ):
ˆ
Ms (f ) ≡ sup |f (x)| dx.
E⊂T2 ,|E|=s E

The following estimate was established in [11, Proposition 4.5]:

ˆ T
ν ν
Ms (ω (t)) ≤ Ms (ω0 ) + Ms (curl F ν (τ )) dτ. (6.1)
0

We use (6.1) to propagate the uniform integrability of ω0ν and curl F ν (τ ),

which follow from H1(a) and H1(b). We conclude that Ms (ω ν (t)) → 0
as s → 0, uniformly in t ∈ [0, T ] and ν > 0, and, consequently, (5.3)
holds true.
The proof is complete.

Remark 6.3. We note that, in fact, our analysis yields a rate of van-
ishing, with respect to viscosity, of the dissipation, albeit not very ex-
plicit, given by (5.15).
We will see, in our next result, that, in the case of physically real-
izable solutions for which the singular part of the initial vorticity is a
bounded, nonnegative, Radon measure in H −1 , we obtain a very explicit
rate.
Next we apply Theorem 5.2 to show absence of anomalous dissipation
for physically realizations uν such that uν0 is compact in L2 and ω0ν is
bounded in BM+ + Lp , for some p ≥ 1. We first address the case p > 1.
 ABSENCE OF ANOMALOUS DISSIPATION 19

Theorem 6.4. Fix p > 1. Let ω ν ≡ curl uν . Assume, in addition to

H(a) and H(b), that
H2(a) ω0ν = µν0 + w0ν , with {µν0 }ν>0 bounded in BM(T2 ), µν0 ≥ 0, and
{w0ν }ν>0 bounded in Lp (T2 );

H2(b) {curl F ν }ν>0 bounded in L1 (0, T ; Lp(T2 )).

Then ˆ T
lim sup ν kω ν (τ )k2L2 dτ = 0.
ν→0+ 0

Remark 6.5. As in Remark 6.2 we note that there always exists a

family of Navier-Stokes solutions uν satisfying conditions H(a), H(b),
H2(a) and H2(b). Again, it is enough to take uν0 and F ν to be mol-
lifications (in space) of u0 ∈ L2 (T2 ), ω0 = µ0 + w0 , with µ0 ∈ BM+ ,
w0 ∈ Lp , and appropriate F . That such initial data and forcing leads
to a physically realizable solution is an easy adaptation of [16], see also
[4] and [20].
Proof. We start by noticing that the hypotheses imply
kuν0 kL2 + kF ν kL2t L2x + kµν0 kBM + kw0ν kLp + k curl F ν kL1t Lpx ≤ K,
for some K > 0 and for all ν > 0. We also have, increasing the constant
K without renaming, that
kw0ν kL1 + k curl F ν kL1t L1x ≤ K.
As in Theorem 5.2, (5.5), (5.6), (5.9), we note that
ˆ t
ν C
ku (·, t)kL2 ≤ C, ν kω ν (·, τ )k2L2 dτ ≤ C, and kω ν (·, t)k2L2 ≤ ,
0 tν
for some C = C(K, T ) > 0 and for all t > 0.
We now derive additional vorticity estimates. We write the vorticity
as
ω ν = µν + w ν ,
where µν satisfies the equation
∂t µν + uν · ∇µν = ν∆µν , (6.2)
with µν (·, 0) = µν0 . Consequently w ν is a solution of
∂t w ν + uν · ∇w ν = ν∆w ν + curl F ν , (6.3)
and w ν (·, 0) = w0ν .
Standard estimates and the parabolic maximum principle now give
kµν (·, t)kL1 ≤ kµν0 kBM ,
sign[µν (·, t)] = sign[µν0 ],
20 ELGINDI, LOPES FILHO, AND NUSSENZVEIG LOPES

and
kw ν (·, t)kLp ≤ kw0ν kLp + k curl F ν kL1t Lpx .
We also have
kw ν (·, t)kL1 ≤ kw0ν kL1 + k curl F ν kL1t L1x .
Furthermore, since {uν } is bounded in L∞ (0, T ; L2 (T2 )) it follows
that {ω ν (·, t)} is bounded in H −1 (T2 ). Recall that Lq (T2 ) is continu-
ously embedded into H −1 (T2 ) for all q > 1. Thus w ν (·, t) is bounded
in H −1 (T2 ). Therefore µν (·, t) is also bounded in H −1 (T2 ). We have,
hence, the following collection of estimates:
(1) kω ν (·, t)kL1 ≤ kµν (·, t)kL1 + kw ν (·, t)kL1 ≤ K,
(2) kw ν (·, t)kLp ≤ K,
(3) kµν (·, t)kH −1 ≤ kω ν (·, t)kH −1 + kw ν (·, t)kH −1 ≤ K.
These bounds are enough to ensure that hypotheses H(c) and (5.3)
are valid. Moreover, we can use the convex, increasing, superquadratic
function Υ obtained in Proposition 4.2, in the corresponding part of
the proof of Theorem 5.2.
This concludes the proof.

Remark 6.6. We argued, in the proof of Theorem 6.4, that we can
use the function Υ obtained in Proposition 4.2. Recall that Υ is the
square of the inverse of a function Φ which, near infinity, is a multiple
of x| log x|−1/4 . Therefore, under the hypotheses of Theorem 6.4, we
find that, for small ν > 0, (5.15) becomes
ζδν (T ) . | log ν|−1/4 , as ν → 0+ . (6.4)
In our final result we consider the case of vorticities in BM+ + L1 .
Corollary 6.7. Let ω ν ≡ curl uν . Assume, in addition to H(a) and
H(b), that
H3(a) ω0ν = µν0 + w0ν , with {µν0 }ν>0 bounded in BM(T2 ), µν0 ≥ 0, and
{w0ν }ν>0 bounded in L1 (T2 ), where ω0ν = curl uν0 ;

H3(b) {curl F ν }ν>0 bounded in L1 (0, T ; L1(T2 )).

Then
ˆ T
lim sup ν kω ν (τ )k2L2 dτ = 0.
ν→0+ 0

The proof is a standard combination of ideas from the proofs of

Theorem 6.1 and Theorem 6.4, so we omit it.
 ABSENCE OF ANOMALOUS DISSIPATION 21

7. Comments and conclusions

Hypothesis (5.3) means that no Dirac deltas appear in the vanishing
viscosity limit of vorticity. This is a condition which has appeared many
times in the literature and, in particular, it is central to J.-M. Delort’s
existence result for vortex sheet evolution with distinguished sign, see
[4] and see also [20]. A weaker, time-averaged, version of this hypothesis
was used to extend Delort’s result to the case of nonnegative mirror-
symmetric vortex sheets, see [13]. This raises the natural question
as to whether no anomalous dissipation can be established under this
time-averaged version of (5.3) as well. In fact, this issue was already
addressed by De Rosa and Park in [18] with an affirmative answer.
We observe, below, that our version can also be carried out under this
weaker hypothesis and we obtain rates as well. More precisely, let
0 < t1 < t2 < T . Consider a family F ⊂ L∞ (t1 , t2 ; L1 (T2 )) and assume
the following generalizations of A and B are valid:
(A’) kf kL∞ 1 ≤ K for all f ∈ F, and
t Lx

(B’)
ˆ t2 ˆ
lim sup sup |f (y, t)| dy dt = 0.
r→0+ f ∈F t1 z∈T2 {|z−y|<r}

If, additionally, F ⊂ L2 (t1 , t2 ; H 1 (T2 )) then it is possible to prove gen-

eralizations of Proposition 3.1, redefining appropriately the function η,
and Proposition 3.2, so that kf kL2(T2 ) and k∇f kL2 (T2 ) are substituted
by kf kL2 ((t1 ,t2 )×T2 ) and k∇f kL2 ((t1 ,t2 )×T2 ) , respectively. The function Υ
in the statement of Proposition 3.2 can be chosen independent of t1 ,
t2 .
Moreover, if (B’) is substituted by
(B”)
ˆ t2 ˆ
sup sup |f (y, t)| dy dt . | log r|−1/2 .
f ∈F t1 z∈T2 {|z−y|<r}

then we can generalize Propositions 4.1, and 4.2. The function Υ in

the statement of Proposition 4.2 can be chosen independent of t1 , t2
and follows the same construction as before.
Therefore, if we assume the weaker hypothesis (B’) for the family
F = {ω ν }ν>0 then the conclusion of Theorem 5.2 remains valid. If
(B′′ ) holds for the singular part of vorticity, while the continuous part
is bounded in Lp for some p > 1, then we can obtain the same rate of
dissipation as in (6.4).
22 ELGINDI, LOPES FILHO, AND NUSSENZVEIG LOPES

The proof of the results claimed above is a straightforward adapta-

tion of the work presented here.
Next we wish to discuss a particular illustration of anomalous dissi-
pation. For simplicity, we will present the example in the full plane.
Let ϕ ∈ Cc∞ (0, +∞) and assume that
ˆ +∞ ˆ +∞
1
sϕ(s) ds = 0 and s|ϕ(s)| ds = .
0 0 2π
Let ω0 = ω0 (x) = ϕ(|x|), x ∈ R2 . For each ν > 0 set
ν ν 1 x
ω0 = ω0 (x) = 2 ω0 .
ν ν
ˆ
Then ω0ν (x) dx = 0 and kω0ν kL1 (R2 ) = 1. Note that, if ω ν is the
R2
solution of the heat equation ∂t ω ν = ν∆ω ν with initial data ω0ν then
ω ν is also the solution of the ˆ (unforced) Navier-Stokes equations with
⊥
x
velocity uν (x, t) = ω ν (y, t) dy since, in this case, the
2π|x|2 B(0;|x|)
nonlinear term is curl-free. It is straightforward to show that:
(1) kω ν (·, t)kL1 ≤ 1
ˆ T
(2) ν kω ν (·, s)k2L2 ds ≥ C > 0, for all ν > 0.
0
Referring back to Theorem 5.2 we note that (H(b)) and (H(c)) hold
true, but
e 0 , weakly in L1 ,
|uν |2 ⇀ Cδ0 and |ω ν | ⇀ Cδ
0 0

so that both (H(a)), strong convergence of the initial data, and (5.3),
the no-Diracs condition, are violated. It would be interesting to have
examples that show that each one of these conditions are needed to
prove absence of anomalous dissipation, specially the no-Diracs condi-
tion.
Finally, let us add a couple of final remarks and open problems. First,
absence of anomalous dissipation does not imply strong convergence of
the approximating sequence. This remains an important open problem,
equivalent to absence of inviscid dissipation for the limiting flow, as it
was proved in [11]. Second, the rates of vanishing for the dissipation
which we obtained here were not shown to be optimal.

Acknowledgments
TME acknowledges support from a Simons Fellowship and the NSF
DMS-2043024. MCLF was partially supported by CNPq, through
grant # 304990/2022-1, and by FAPERJ, through grant # E-26/201.209/2021.
 ABSENCE OF ANOMALOUS DISSIPATION 23

HJNL acknowledges the support of CNPq, through grant # 305309/2022-

6, and that of FAPERJ, through grant # E-26/201.027/2022. This
work was carried out while the authors were visiting AIM and the De-
partment of Mathematics at Princeton University. They thank both
for their gracious hospitality.

References
[1] E. Bouin, J. Dolbeault, and C. Schmeiser, A variational proof of Nash’s in-
equality, Atti Accad. Naz. Lincei, Cl. Sci. Fis. Mat. Nat., IX. Ser., Rend. Lincei,
Mat. Appl. 31 (2020), no. 1, 211–223 (English). ↑6
[2] E. Bruè, M. Colombo, and A. Kumar, Flexibility of two-dimensional euler flows
with integrable vorticity, 2024. ↑13
[3] A Cheskidov, M Lopes Filho, H. Nussenzveig Lopes, and R Shvydkoy, Energy
conservation in two-dimensional incompressible ideal fluids, Communications
in Mathematical Physics 348 (2016), no. 1, 129–143. ↑13
[4] J.-M. Delort, Existence of vortex sheets in dimension two, J. Am. Math. Soc.
4 (1991), no. 3, 553–586 (French). ↑2, 9, 13, 16, 19, 21
[5] R. J. DiPerna and A. Majda, Reduced Hausdorff dimension and concentration-
cancellation for two- dimensional incompressible flow, J. Am. Math. Soc. 1
(1988), no. 1, 59–95 (English). ↑12
[6] R. J. Diperna and A. J. Majda, Concentrations in regularizations for 2 − D
incompressible flow, Commun. Pure Appl. Math. 40 (1987), no. 3, 301–345
(English). ↑12
[7] R. J. DiPerna and A. J. Majda, Oscillations and concentrations in weak solu-
tions of the incompressible fluid equations, Commun. Math. Phys. 108 (1987),
667–689 (English). ↑12
[8] G. B. Folland, Real analysis. Modern techniques and their applications., 2nd
ed., Pure Appl. Math., Wiley-Intersci. Ser. Texts Monogr. Tracts, New York,
NY: Wiley, 1999 (English). ↑4
[9] U. Frisch, Turbulence: The legacy of A. N. Kolmogorov, Cambridge University
Press, 1995. ↑2
[10] D. Iftimie, M. C. Lopes Filho, and H. J. Nussenzveig Lopes, Weak vortic-
ity formulation of the incompressible 2D Euler equations in bounded domains,
Comm. Partial Differential Equations 45 (2020), no. 2, 109–145. MR4045802
↑16
[11] F. Jin, S. Lanthaler, M. C. L. Filho, and H. J. N. Lopes, Sharp conditions for
energy balance in two-dimensional incompressible ideal flow with external force
(2024). ↑2, 14, 16, 18, 22
[12] S. Lanthaler, S. Mishra, and C. Parés-Pulido, On the conservation of energy
in two-dimensional incompressible flows, Nonlinearity 34 (2021), no. 2, 1084,
available at 2001.06195. ↑13
[13] M. C. Lopes Filho, H. J. Nussenzveig Lopes, and Z. Xin, Existence of vortex
sheets with reflection symmetry in two space dimensions, Arch. Ration. Mech.
Anal. 158 (2001), no. 3, 235–257. MR1842346 ↑21
[14] M. Lopes Filho and H. Nussenzveig Lopes, Energy balance for forced two-
dimensional incompressible ideal fluid flow, Philosophical Transactions of the
Royal Society A 380 (2022), no. 2219, 20210095. ↑13
24 ELGINDI, LOPES FILHO, AND NUSSENZVEIG LOPES

[15] M. C. Lopes Filho, H. J. Nussenzveig Lopes, and E. Tadmor, Approximate

solutions of the incompressible Euler equations with no concentrations, Ann.
Inst. H. Poincaré C Anal. Non Linéaire 17 (2000), no. 3, 371–412. MR1771138
↑2
[16] A. J. Majda, Remarks on weak solutions for vortex sheets with a distinguished
sign, Indiana Univ. Math. J. 42 (1993), no. 3, 921–939 (English). ↑2, 19
[17] J. F. Nash, Continuity of solutions of parabolic and elliptic equations, Am. J.
Math. 80 (1958), 931–954 (English). ↑6
[18] L. D. Rosa and J. Park, No anomalous dissipation in two-dimensional incom-
pressible fluids (2024), available at 2403.04668. ↑3, 21
[19] W. Rudin, Principles of mathematical analysis. 3rd ed, 1976 (English). ↑5
[20] S. Schochet, The weak vorticity formulation of the 2-D Euler equations and
concentration- cancellation, Commun. Partial Differ. Equations 20 (1995),
no. 5-6, 1077–1104 (English). ↑9, 16, 19, 21
[21] I. Vecchi and S. Wu, On L1 -vorticity for 2-D incompressible flow, Manuscr.
Math. 78 (1993), no. 4, 403–412 (English). ↑2, 12, 18

1
Department of Mathematics, Duke University, Durham, NC 27708-
0320 – USA
Email address: [email protected]
2
Instituto de Matemática, Universidade Federal do Rio de Janeiro,
Cidade Universitária – Ilha do Fundão, Caixa Postal 68530, 21941-909
Rio de Janeiro, RJ – BRAZIL
Email address: [email protected]
Email address: [email protected]