Professional Documents
Culture Documents
Colombo Transport
Colombo Transport
Maria Colombo
Flows of vector
fields
Maria Colombo
Flows and
1 Flow of vector fields and continuity equation
continuity
equation
Smooth vs
2 Smooth vs nonsmooth theory
nonsmooth
theory The Cauchy-Lipschitz theorem for smooth vector fields
Cauchy-Lipschitz
thm
Lack of uniqueness of the flow for nonsmooth vector fields
Lack of uniqueness
Regular Lagrangian Flows and the nonsmooth theory
The nonsmooth
theory: RLF
A.e. uniqueness
of integral curves 3 A.e. uniqueness of integral curves
Ideas
Ambrosio’s
superposition
principle
4 Ideas of the proof
Interpolation Ambrosio’s superposition principle
Ill-posedness of CE
by convex integration Interpolation
Ill-posedness of CE by convex integration
Outline of the talk
Flows of vector
fields
Maria Colombo
Flows and
1 Flow of vector fields and continuity equation
continuity
equation
Smooth vs
2 Smooth vs nonsmooth theory
nonsmooth
theory The Cauchy-Lipschitz theorem for smooth vector fields
Cauchy-Lipschitz
thm
Lack of uniqueness of the flow for nonsmooth vector fields
Lack of uniqueness
Regular Lagrangian Flows and the nonsmooth theory
The nonsmooth
theory: RLF
A.e. uniqueness
of integral curves 3 A.e. uniqueness of integral curves
Ideas
Ambrosio’s
superposition
principle
4 Ideas of the proof
Interpolation Ambrosio’s superposition principle
Ill-posedness of CE
by convex integration Interpolation
Ill-posedness of CE by convex integration
Table of contents
Flows of vector
fields
Maria Colombo
Flows and
1 Flow of vector fields and continuity equation
continuity
equation
Smooth vs
2 Smooth vs nonsmooth theory
nonsmooth
theory The Cauchy-Lipschitz theorem for smooth vector fields
Cauchy-Lipschitz
thm
Lack of uniqueness of the flow for nonsmooth vector fields
Lack of uniqueness
Regular Lagrangian Flows and the nonsmooth theory
The nonsmooth
theory: RLF
A.e. uniqueness
of integral curves 3 A.e. uniqueness of integral curves
Ideas
Ambrosio’s
superposition
principle
4 Ideas of the proof
Interpolation Ambrosio’s superposition principle
Ill-posedness of CE
by convex integration Interpolation
Ill-posedness of CE by convex integration
The flow of a vector field
Flows of vector
fields
Maria Colombo
Flows and
continuity
equation
A.e. uniqueness
of integral curves It can be seen
Ideas
Ambrosio’s as a collection of trajectories X(·, x ) labelled by x ∈ Rd ;
superposition
principle
Interpolation
as a collection of diffeomorphisms X(t, ·) : Rd → Rd .
Ill-posedness of CE
by convex integration
The flow of a vector field
Flows of vector
fields
Maria Colombo
Flows and
continuity
equation
A.e. uniqueness
of integral curves It can be seen
Ideas
Ambrosio’s as a collection of trajectories X(·, x ) labelled by x ∈ Rd ;
superposition
principle
Interpolation
as a collection of diffeomorphisms X(t, ·) : Rd → Rd .
Ill-posedness of CE
by convex integration
Continuity/transport equation
Flows of vector
fields
Consider the related PDE, named continuity equation
Maria Colombo
Flows and
(
continuity ∂t µt + div (bt µt ) = 0 in (0, ∞) × Rd
equation
Smooth vs
µ0 given.
nonsmooth
theory
Cauchy-Lipschitz
thm
When bt is sufficiently smooth and µt : Rd × [0, ∞) → R is a smooth
Lack of uniqueness function, all derivatives can be computed.
The nonsmooth
theory: RLF Even if bt is only a bounded vector field and {µt }t∈[0,∞) is a
A.e. uniqueness
of integral curves
1-parameter family of finite measures, the PDE makes sense
Ideas distributionally.
Ambrosio’s
superposition
When
principle
Interpolation
divbt ≡ 0,
Ill-posedness of CE
by convex integration
the continuity equation is equivalent to the transport equation
∂t µt + b · ∇µt = 0.
Continuity/transport equation
Flows of vector
fields
Consider the related PDE, named continuity equation
Maria Colombo
Flows and
(
continuity ∂t µt + div (bt µt ) = 0 in (0, ∞) × Rd
equation
Smooth vs
µ0 given.
nonsmooth
theory
Cauchy-Lipschitz
thm
When bt is sufficiently smooth and µt : Rd × [0, ∞) → R is a smooth
Lack of uniqueness function, all derivatives can be computed.
The nonsmooth
theory: RLF Even if bt is only a bounded vector field and {µt }t∈[0,∞) is a
A.e. uniqueness
of integral curves
1-parameter family of finite measures, the PDE makes sense
Ideas distributionally.
Ambrosio’s
superposition
When
principle
Interpolation
divbt ≡ 0,
Ill-posedness of CE
by convex integration
the continuity equation is equivalent to the transport equation
∂t µt + b · ∇µt = 0.
Continuity/transport equation
Flows of vector
fields
Consider the related PDE, named continuity equation
Maria Colombo
Flows and
(
continuity ∂t µt + div (bt µt ) = 0 in (0, ∞) × Rd
equation
Smooth vs
µ0 given.
nonsmooth
theory
Cauchy-Lipschitz
thm
When bt is sufficiently smooth and µt : Rd × [0, ∞) → R is a smooth
Lack of uniqueness function, all derivatives can be computed.
The nonsmooth
theory: RLF Even if bt is only a bounded vector field and {µt }t∈[0,∞) is a
A.e. uniqueness
of integral curves
1-parameter family of finite measures, the PDE makes sense
Ideas distributionally.
Ambrosio’s
superposition
When
principle
Interpolation
divbt ≡ 0,
Ill-posedness of CE
by convex integration
the continuity equation is equivalent to the transport equation
∂t µt + b · ∇µt = 0.
Connection between continuity equation and flows
Flows of vector
fields
Solutions of the CE flow along characteristic curves of b
Maria Colombo
Ideas
Ambrosio’s
Indeed, for any test function ϕ ∈ Cc∞ (Rd ) we have
superposition
principle Z Z Z
Interpolation d d
Ill-posedness of CE ϕ dµt = ϕ(X(t, x )) dµ0 (x ) = ∇ϕ(X) · ∂t X dµ0
by convex integration dt Rd dt Rd Rd
Z Z
= ∇ϕ(X) · bt (X) dµ0 = ∇ϕ · bt dµt .
Rd Rd
Flows of vector
fields
Solutions of the CE flow along characteristic curves of b
Maria Colombo
Ideas
Ambrosio’s
Indeed, for any test function ϕ ∈ Cc∞ (Rd ) we have
superposition
principle Z Z Z
Interpolation d d
Ill-posedness of CE ϕ dµt = ϕ(X(t, x )) dµ0 (x ) = ∇ϕ(X) · ∂t X dµ0
by convex integration dt Rd dt Rd Rd
Z Z
= ∇ϕ(X) · bt (X) dµ0 = ∇ϕ · bt dµt .
Rd Rd
Flows of vector
fields
Solutions of the CE flow along characteristic curves of b
Maria Colombo
Ideas
Ambrosio’s
Indeed, for any test function ϕ ∈ Cc∞ (Rd ) we have
superposition
principle Z Z Z
Interpolation d d
Ill-posedness of CE ϕ dµt = ϕ(X(t, x )) dµ0 (x ) = ∇ϕ(X) · ∂t X dµ0
by convex integration dt Rd dt Rd Rd
Z Z
= ∇ϕ(X) · bt (X) dµ0 = ∇ϕ · bt dµt .
Rd Rd
Flows of vector
fields
Maria Colombo
Is the solution of the continuity equation starting from µ0 unique?
Flows and
continuity
equation
Smooth vs
YES if ∇b is bounded
nonsmooth
theory Given a solution νt to CE, set νet = X(t, ·)−1
# νt . An analogous
Cauchy-Lipschitz
thm computation shows that
Lack of uniqueness
The nonsmooth Z
theory: RLF d
ϕ d νet = 0,
A.e. uniqueness
of integral curves
dt Rd
Ideas
Ambrosio’s
so
superposition
principle X(t, ·)−1
# νt = ν
et = ν0 = µ0 ⇒ νt := X(t, ·)# µ0 .
Interpolation
Ill-posedness of CE
by convex integration
NO if b is less regular
As soon as uniqueness for the ODE fails.
Regularity of b matters
Flows of vector
fields
Maria Colombo
Is the solution of the continuity equation starting from µ0 unique?
Flows and
continuity
equation
Smooth vs
YES if ∇b is bounded
nonsmooth
theory Given a solution νt to CE, set νet = X(t, ·)−1
# νt . An analogous
Cauchy-Lipschitz
thm computation shows that
Lack of uniqueness
The nonsmooth Z
theory: RLF d
ϕ d νet = 0,
A.e. uniqueness
of integral curves
dt Rd
Ideas
Ambrosio’s
so
superposition
principle X(t, ·)−1
# νt = ν
et = ν0 = µ0 ⇒ νt := X(t, ·)# µ0 .
Interpolation
Ill-posedness of CE
by convex integration
NO if b is less regular
As soon as uniqueness for the ODE fails.
Regularity of b matters
Flows of vector
fields
Maria Colombo
Is the solution of the continuity equation starting from µ0 unique?
Flows and
continuity
equation
Smooth vs
YES if ∇b is bounded
nonsmooth
theory Given a solution νt to CE, set νet = X(t, ·)−1
# νt . An analogous
Cauchy-Lipschitz
thm computation shows that
Lack of uniqueness
The nonsmooth Z
theory: RLF d
ϕ d νet = 0,
A.e. uniqueness
of integral curves
dt Rd
Ideas
Ambrosio’s
so
superposition
principle X(t, ·)−1
# νt = ν
et = ν0 = µ0 ⇒ νt := X(t, ·)# µ0 .
Interpolation
Ill-posedness of CE
by convex integration
NO if b is less regular
As soon as uniqueness for the ODE fails.
Table of contents
Flows of vector
fields
Maria Colombo
Flows and
1 Flow of vector fields and continuity equation
continuity
equation
Smooth vs
2 Smooth vs nonsmooth theory
nonsmooth
theory The Cauchy-Lipschitz theorem for smooth vector fields
Cauchy-Lipschitz
thm
Lack of uniqueness of the flow for nonsmooth vector fields
Lack of uniqueness
Regular Lagrangian Flows and the nonsmooth theory
The nonsmooth
theory: RLF
A.e. uniqueness
of integral curves 3 A.e. uniqueness of integral curves
Ideas
Ambrosio’s
superposition
principle
4 Ideas of the proof
Interpolation Ambrosio’s superposition principle
Ill-posedness of CE
by convex integration Interpolation
Ill-posedness of CE by convex integration
Theory for smooth vector fields
Flows of vector
fields
Maria Colombo
Flows and
continuity
equation
Smooth vs
Cauchy-Lipschitz Theorem
nonsmooth
theory Let bt a vector field with ∇bt locally bounded. Then for every x ∈ Rd
there exists a unique maximal solution X(·, x ) : [0, TX (x )) → Rd of
Cauchy-Lipschitz
thm
Lack of uniqueness
The nonsmooth
the ODE.
theory: RLF
A.e. uniqueness
of integral curves
For every x ∈ Rd such that TX (x ) < ∞ the trajectory X(·, x ) blows
Ideas
up properly
Ambrosio’s
superposition
lim |X(t, x )| = ∞.
principle t→TX (x )
Interpolation
Ill-posedness of CE
by convex integration
Theory for smooth vector fields
Flows of vector
fields
Maria Colombo
Flows and
continuity
equation
Smooth vs
Cauchy-Lipschitz Theorem
nonsmooth
theory Let bt a vector field with ∇bt locally bounded. Then for every x ∈ Rd
there exists a unique maximal solution X(·, x ) : [0, TX (x )) → Rd of
Cauchy-Lipschitz
thm
Lack of uniqueness
The nonsmooth
the ODE.
theory: RLF
A.e. uniqueness
of integral curves
For every x ∈ Rd such that TX (x ) < ∞ the trajectory X(·, x ) blows
Ideas
up properly
Ambrosio’s
superposition
lim |X(t, x )| = ∞.
principle t→TX (x )
Interpolation
Ill-posedness of CE
by convex integration
Nonsmooth theory: lack of uniqueness
Flows of vector
fields One-dimensional autonomous vector field with lack of uniqueness
Maria Colombo p
Flows and
b(x ) = 2 |x |, x ∈R
continuity
equation Given x0 = −c 2 < 0, the 1-parameter family of curves that stop at the
Smooth vs
nonsmooth origin for an arbitrary time T ≥ 0, solve the ODE.
theory
Cauchy-Lipschitz
thm
x x
Lack of uniqueness
The nonsmooth
theory: RLF
(t − c − T )2
A.e. uniqueness
of integral curves
Ideas
Ambrosio’s
superposition
principle 0 c c+T t
Interpolation
Ill-posedness of CE
by convex integration
−(t − c)2
x0 = −c2
Nonsmooth theory: lack of uniqueness
Flows of vector
fields
Maria Colombo
Flows and
x
continuity x
equation
Smooth vs
nonsmooth
theory
Cauchy-Lipschitz t t
thm
Lack of uniqueness
The nonsmooth
theory: RLF
A.e. uniqueness
of integral curves
Ideas
Ambrosio’s
superposition
principle
Interpolation
Ill-posedness of CE
by convex integration
Between all the possible integral curves, a “better selection” could be
made by the ones that do not stop in 0. In other words, we wish to
select a collection of integral curves that “do not concentrate”.
Selection of a flow
Flows of vector
fields
Maria Colombo
Regular lagrangian flows
Flows and Given a vector field b : (0, T ) × Rd → Rd , the map
continuity
equation X : [0, ∞) × Rd → Rd is a regular Lagrangian flow of b if:
Smooth vs
nonsmooth
(i) for L d -a.e. x ∈ Rd , X(·, x ) solves the ODE ẋ (t) = bt (x (t))
theory
Cauchy-Lipschitz
starting from x ;
(ii) X(t, ·)# L d ≤ C L d for every t ∈ [0, T ] and for some C > 0.
thm
Lack of uniqueness
The nonsmooth
theory: RLF
A.e. uniqueness
of integral curves Theorem ([Di Perna-Lions ’89)
Ideas
Ambrosio’s
, [Ambrosio ’04]] Let us assume that |∇bt | ∈ L1loc (Rd ),
superposition
principle div bt ∈ L∞ (Rd ) and
Interpolation
Ill-posedness of CE
by convex integration |bt (x )|
∈ L1 (Rd ) + L∞ (Rd ).
1 + |x |
Flows of vector
fields
Maria Colombo
Regular lagrangian flows
Flows and Given a vector field b : (0, T ) × Rd → Rd , the map
continuity
equation X : [0, ∞) × Rd → Rd is a regular Lagrangian flow of b if:
Smooth vs
nonsmooth
(i) for L d -a.e. x ∈ Rd , X(·, x ) solves the ODE ẋ (t) = bt (x (t))
theory
Cauchy-Lipschitz
starting from x ;
(ii) X(t, ·)# L d ≤ C L d for every t ∈ [0, T ] and for some C > 0.
thm
Lack of uniqueness
The nonsmooth
theory: RLF
A.e. uniqueness
of integral curves Theorem ([Di Perna-Lions ’89)
Ideas
Ambrosio’s
, [Ambrosio ’04]] Let us assume that |∇bt | ∈ L1loc (Rd ),
superposition
principle div bt ∈ L∞ (Rd ) and
Interpolation
Ill-posedness of CE
by convex integration |bt (x )|
∈ L1 (Rd ) + L∞ (Rd ).
1 + |x |
Flows of vector
fields
Maria Colombo
The regularity assumption |∇bt | ∈ L1loc (Rd ) can be replaced by
Flows and
continuity
equation
Assumption
Smooth vs
nonsmooth
For every compactly supported µ0 ∈ L∞ (Rd ) there exists a unique
theory
Cauchy-Lipschitz
bounded, compactly supported solution of the CE starting from µ0 .
thm
Lack of uniqueness
The nonsmooth
This is satisfied by several classes of vector fields:
theory: RLF
A.e. uniqueness
when locally ∇bt is a matrix-valued finite measure (namely, bt is
of integral curves
a function of bounded variation BVloc (Rd ; Rd ) function),
Ideas
Ambrosio’s
[Ambrosio 04];
superposition
principle by singular integrals of L1 functions, for instance convolutions of
Interpolation
Ill-posedness of CE the form h ∗ |xx|d with h ∈ L1 (Rd ), [Bouschut, Crippa 13] and of
by convex integration
measures, with some additional structure [Bohun, Bouschut,
Crippa 13].
Remarks on the DiPerna-Lions theorem
Flows of vector
fields
Maria Colombo
The regularity assumption |∇bt | ∈ L1loc (Rd ) can be replaced by
Flows and
continuity
equation
Assumption
Smooth vs
nonsmooth
For every compactly supported µ0 ∈ L∞ (Rd ) there exists a unique
theory
Cauchy-Lipschitz
bounded, compactly supported solution of the CE starting from µ0 .
thm
Lack of uniqueness
The nonsmooth
This is satisfied by several classes of vector fields:
theory: RLF
A.e. uniqueness
when locally ∇bt is a matrix-valued finite measure (namely, bt is
of integral curves
a function of bounded variation BVloc (Rd ; Rd ) function),
Ideas
Ambrosio’s
[Ambrosio 04];
superposition
principle by singular integrals of L1 functions, for instance convolutions of
Interpolation
Ill-posedness of CE the form h ∗ |xx|d with h ∈ L1 (Rd ), [Bouschut, Crippa 13] and of
by convex integration
measures, with some additional structure [Bohun, Bouschut,
Crippa 13].
Remarks on the DiPerna-Lions theorem
Flows of vector
fields
Maria Colombo
Flows and
continuity a different approach to this result was proposed by [Crippa, De
equation
Smooth vs
Lellis, 08]. To show the uniqueness of the flow, they consider a
nonsmooth
theory
functional of the type
Cauchy-Lipschitz
div bt ∈ BMO(Rd )
Ambrosio’s
superposition
principle
Interpolation
Ill-posedness of CE
by convex integration
[Mucha, 2010], [C., Crippa, Spirito 2016].
Idea of the proof - existence and uniqueness
Flows of vector
fields
Maria Colombo
The real difficulty is with uniqueness.
Well posedness of the CE ⇒ uniqueness of the flow. (in the
Flows and
continuity class of bounded, compactly supported solutions). Assume by
equation
contradiction that there exists a set A ⊆ Rd such that two flows
Smooth vs
nonsmooth X(·, x ) and Y (·, x ) start at every x ∈ A. Taking a subset, we can
theory
Cauchy-Lipschitz
assume that the two flows are disjoint at a later time t0 .
thm
Lack of uniqueness
The nonsmooth
theory: RLF
A.e. uniqueness
X(·, x)
of integral curves
Ideas
Ambrosio’s
superposition Y(·, x)
principle
Interpolation
Ill-posedness of CE
by convex integration t0
Flows of vector
fields
Maria Colombo
The real difficulty is with uniqueness.
Well posedness of the CE ⇒ uniqueness of the flow. (in the
Flows and
continuity class of bounded, compactly supported solutions). Assume by
equation
contradiction that there exists a set A ⊆ Rd such that two flows
Smooth vs
nonsmooth X(·, x ) and Y (·, x ) start at every x ∈ A. Taking a subset, we can
theory
Cauchy-Lipschitz
assume that the two flows are disjoint at a later time t0 .
thm
Lack of uniqueness
The nonsmooth
theory: RLF
A.e. uniqueness
X(·, x)
of integral curves
Ideas
Ambrosio’s
superposition Y(·, x)
principle
Interpolation
Ill-posedness of CE
by convex integration t0
Flows of vector
fields
Maria Colombo
The real difficulty is with uniqueness.
Well posedness of the CE ⇒ uniqueness of the flow. (in the
Flows and
continuity class of bounded, compactly supported solutions). Assume by
equation
contradiction that there exists a set A ⊆ Rd such that two flows
Smooth vs
nonsmooth X(·, x ) and Y (·, x ) start at every x ∈ A. Taking a subset, we can
theory
Cauchy-Lipschitz
assume that the two flows are disjoint at a later time t0 .
thm
Lack of uniqueness
The nonsmooth
theory: RLF
A.e. uniqueness
X(·, x)
of integral curves
Ideas
Ambrosio’s
superposition Y(·, x)
principle
Interpolation
Ill-posedness of CE
by convex integration t0
Flows of vector
fields
Maria Colombo
Flows and
continuity
equation Proof of the well posedness of the CE.
Smooth vs
nonsmooth
theory Proposition ([DiPerna-Lions, ’89], [Ambrosio ’04])
Cauchy-Lipschitz
thm
Lack of uniqueness
for any divergence free b ∈ L1t W 1,1 x ,loc , u0 ∈ L∞
c there exists a unique
The nonsmooth
theory: RLF
solution u ∈ L∞ ∞
t L x ,c to
A.e. uniqueness
of integral curves ∂t ut + div (bt ut ) = 0
Ideas
Ambrosio’s
superposition The statement holds more in general when the integrability of ∇b is
principle
Interpolation coupled with the integrability of u (DiPerna-Lions range).
Ill-posedness of CE
by convex integration
Idea of the proof - well posedness
Flows of vector
fields
Maria Colombo
Flows and
continuity
equation Proof of the well posedness of the CE.
Smooth vs
nonsmooth
theory Proposition ([DiPerna-Lions, ’89], [Ambrosio ’04])
Cauchy-Lipschitz
thm
Lack of uniqueness
for any divergence free b ∈ L1t W 1,1 x ,loc , u0 ∈ L∞
c there exists a unique
The nonsmooth
theory: RLF
solution u ∈ L∞ ∞
t L x ,c to
A.e. uniqueness
of integral curves ∂t ut + div (bt ut ) = 0
Ideas
Ambrosio’s
superposition The statement holds more in general when the integrability of ∇b is
principle
Interpolation coupled with the integrability of u (DiPerna-Lions range).
Ill-posedness of CE
by convex integration
Idea of the proof - well posedness
Flows of vector
fields
Maria Colombo
Flows and
Proof of the well posedness of the CE.
continuity
equation
Proposition ([DiPerna-Lions, ’89], [Ambrosio ’04])
Smooth vs
nonsmooth
theory Let r , p ∈ [1, ∞] satisfy
Cauchy-Lipschitz 1 1
thm
+ ≤ 1.
Lack of uniqueness
p r
The nonsmooth
theory: RLF
A.e. uniqueness
Then, for any divergence free b ∈ L1t L1t Wx1,r , u0 ∈ Lpc there exists a
of integral curves
unique solution u ∈ L∞ p
t L x to
Ideas
Ambrosio’s
superposition
principle
∂t ut + div (bt ut ) = 0
Interpolation
Ill-posedness of CE
by convex integration The statement holds more in general when the integrability of ∇b is
coupled with the integrability of u (DiPerna-Lions range).
Idea of the proof - well posedness
Flows of vector
fields
Maria Colombo By linearity, we show that any bounded, compactly supported solution
Flows and u(t, x ) of the CE
continuity
equation ∂t ut + div (bt ut ) = 0
Smooth vs
nonsmooth u0 (0, ·) = 0 =⇒ u(t, ·) ≡ 0 for any t > 0.
theory
Cauchy-Lipschitz
thm Formally, multiply the equation by u and integrate
Lack of uniqueness
u(t, x )2
The nonsmooth Z Z Z
theory: RLF d
A.e. uniqueness
dx = u∂t u dx = − u div (ub) dx
of integral curves dt Rd 2 Rd Rd
u2
Z Z
Ideas
Ambrosio’s = u∇u · b dx = div b dx
Rd 2
superposition
principle Rd
u2
Interpolation
Z
Ill-posedness of CE
by convex integration ≤C dx
Rd 2
Flows of vector
fields
Maria Colombo By linearity, we show that any bounded, compactly supported solution
Flows and u(t, x ) of the CE
continuity
equation ∂t ut + div (bt ut ) = 0
Smooth vs
nonsmooth u0 (0, ·) = 0 =⇒ u(t, ·) ≡ 0 for any t > 0.
theory
Cauchy-Lipschitz
thm Formally, multiply the equation by u and integrate
Lack of uniqueness
u(t, x )2
The nonsmooth Z Z Z
theory: RLF d
A.e. uniqueness
dx = u∂t u dx = − u div (ub) dx
of integral curves dt Rd 2 Rd Rd
u2
Z Z
Ideas
Ambrosio’s = u∇u · b dx = div b dx
Rd 2
superposition
principle Rd
u2
Interpolation
Z
Ill-posedness of CE
by convex integration ≤C dx
Rd 2
Flows of vector
fields
Maria Colombo By linearity, we show that any bounded, compactly supported solution
Flows and u(t, x ) of the CE
continuity
equation ∂t ut + div (bt ut ) = 0
Smooth vs
nonsmooth u0 (0, ·) = 0 =⇒ u(t, ·) ≡ 0 for any t > 0.
theory
Cauchy-Lipschitz
thm Formally, multiply the equation by u and integrate
Lack of uniqueness
u(t, x )2
The nonsmooth Z Z Z
theory: RLF d
A.e. uniqueness
dx = u∂t u dx = − u div (ub) dx
of integral curves dt Rd 2 Rd Rd
u2
Z Z
Ideas
Ambrosio’s = u∇u · b dx = div b dx
Rd 2
superposition
principle Rd
u2
Interpolation
Z
Ill-posedness of CE
by convex integration ≤C dx
Rd 2
Flows of vector
fields
Maria Colombo We repeat the computation after convolving the equation with a
Flows and
smooth kernel ρε . The function u ε = u ∗ ρε solves
continuity
equation
Smooth vs
∂t utε + div (bt utε ) = div (bt utε ) − div (bt ut )ε = C ε (ut , bt ).
nonsmooth
theory
Cauchy-Lipschitz
To handle the commutator, we employ the lemma
thm
Lack of uniqueness
The nonsmooth
Lemma
theory: RLF
A.e. uniqueness If u ∈ L∞ d 1 d
c (R ) and |∇b| ∈ Lloc (R ), then
of integral curves
Ideas
Ambrosio’s
lim C ε (u, b) = 0 in L1 (Rd ).
superposition
ε→0
principle
Interpolation
Ill-posedness of CE Uniqueness ⇒ Existence. Consider a smooth approximation bε of
by convex integration
the vector field b by convolution and consider the approximating flows
Xε . They converge (in a suitable weak sense) to a limit collection of
curves, which is the limit flow by uniqueness.
Idea of the proof - well posedness
Flows of vector
fields
Maria Colombo We repeat the computation after convolving the equation with a
Flows and
smooth kernel ρε . The function u ε = u ∗ ρε solves
continuity
equation
Smooth vs
∂t utε + div (bt utε ) = div (bt utε ) − div (bt ut )ε = C ε (ut , bt ).
nonsmooth
theory
Cauchy-Lipschitz
To handle the commutator, we employ the lemma
thm
Lack of uniqueness
The nonsmooth
Lemma
theory: RLF
A.e. uniqueness If u ∈ L∞ d 1 d
c (R ) and |∇b| ∈ Lloc (R ), then
of integral curves
Ideas
Ambrosio’s
lim C ε (u, b) = 0 in L1 (Rd ).
superposition
ε→0
principle
Interpolation
Ill-posedness of CE Uniqueness ⇒ Existence. Consider a smooth approximation bε of
by convex integration
the vector field b by convolution and consider the approximating flows
Xε . They converge (in a suitable weak sense) to a limit collection of
curves, which is the limit flow by uniqueness.
Idea of the proof - well posedness
Flows of vector
fields
Maria Colombo We repeat the computation after convolving the equation with a
Flows and
smooth kernel ρε . The function u ε = u ∗ ρε solves
continuity
equation
Smooth vs
∂t utε + div (bt utε ) = div (bt utε ) − div (bt ut )ε = C ε (ut , bt ).
nonsmooth
theory
Cauchy-Lipschitz
To handle the commutator, we employ the lemma
thm
Lack of uniqueness
The nonsmooth
Lemma
theory: RLF
A.e. uniqueness If u ∈ L∞ d 1 d
c (R ) and |∇b| ∈ Lloc (R ), then
of integral curves
Ideas
Ambrosio’s
lim C ε (u, b) = 0 in L1 (Rd ).
superposition
ε→0
principle
Interpolation
Ill-posedness of CE Uniqueness ⇒ Existence. Consider a smooth approximation bε of
by convex integration
the vector field b by convolution and consider the approximating flows
Xε . They converge (in a suitable weak sense) to a limit collection of
curves, which is the limit flow by uniqueness.
Table of contents
Flows of vector
fields
Maria Colombo
Flows and
1 Flow of vector fields and continuity equation
continuity
equation
Smooth vs
2 Smooth vs nonsmooth theory
nonsmooth
theory The Cauchy-Lipschitz theorem for smooth vector fields
Cauchy-Lipschitz
thm
Lack of uniqueness of the flow for nonsmooth vector fields
Lack of uniqueness
Regular Lagrangian Flows and the nonsmooth theory
The nonsmooth
theory: RLF
A.e. uniqueness
of integral curves 3 A.e. uniqueness of integral curves
Ideas
Ambrosio’s
superposition
principle
4 Ideas of the proof
Interpolation Ambrosio’s superposition principle
Ill-posedness of CE
by convex integration Interpolation
Ill-posedness of CE by convex integration
Two questions
Flows of vector
fields
Maria Colombo
Question: a.e. uniqueness of integral curves
Flows and
continuity
equation Does any divergence free b ∈ L1t Wx1,p admit a unique integral curve
Smooth vs (namely, γ ∈ W 1,1 (0, T ) solution of the ODE γ̇(t) = u(t, γ)) for a.e.
nonsmooth
theory initial datum x ∈ Rd ?
Cauchy-Lipschitz
thm
Lack of uniqueness Open since the pioneering works of DiPerna-Lions and Ambrosio.
The nonsmooth
theory: RLF
A.e. uniqueness
(Related) question: well posedness of the CE
of integral curves
Ill-posedness of CE
by convex integration
Flows of vector
fields
Maria Colombo
Question: a.e. uniqueness of integral curves
Flows and
continuity
equation Does any divergence free b ∈ L1t Wx1,p admit a unique integral curve
Smooth vs (namely, γ ∈ W 1,1 (0, T ) solution of the ODE γ̇(t) = u(t, γ)) for a.e.
nonsmooth
theory initial datum x ∈ Rd ?
Cauchy-Lipschitz
thm
Lack of uniqueness Open since the pioneering works of DiPerna-Lions and Ambrosio.
The nonsmooth
theory: RLF
A.e. uniqueness
(Related) question: well posedness of the CE
of integral curves
Ill-posedness of CE
by convex integration
Flows of vector
fields
Maria Colombo
Question: a.e. uniqueness of integral curves
Flows and
continuity
equation Does any divergence free b ∈ L1t Wx1,p admit a unique integral curve
Smooth vs (namely, γ ∈ W 1,1 (0, T ) solution of the ODE γ̇(t) = u(t, γ)) for a.e.
nonsmooth
theory initial datum x ∈ Rd ?
Cauchy-Lipschitz
thm
Lack of uniqueness Open since the pioneering works of DiPerna-Lions and Ambrosio.
The nonsmooth
theory: RLF
A.e. uniqueness
(Related) question: well posedness of the CE
of integral curves
Ill-posedness of CE
by convex integration
Flows of vector
fields
Maria Colombo
Flows and
For any b ∈ Lip then
continuity
equation
|b(x ) − b(y )| ≤ C |x − y | ∀x , y ,
Smooth vs
nonsmooth
theory
Cauchy-Lipschitz
thm
Lack of uniqueness
The nonsmooth
Let X(·, x ) be the RLF, γx an integral curve from x ∈ Rd .
theory: RLF
A.e. uniqueness d
of integral curves
|X(t, x ) − γx (t)| ≤ |b(X(t, x )) − b(γx (t))|
Ideas dt
Ambrosio’s
superposition ≤ C |X(t, x ) − γx (t)|
principle
Interpolation
Ill-posedness of CE
by convex integration By Gronwall inequality, if b ∈ Lip we have everywhere uniqueness.
If b ∈ W 1,p , p > d, we have a.e. uniqueness [Caravenna, Crippa -
Jabin].
The case p > d: a.e. uniqueness
Flows of vector
fields
Maria Colombo
Flows and
For any b ∈ Lip then
continuity
equation
|b(x ) − b(y )| ≤ C |x − y | ∀x , y ,
Smooth vs
nonsmooth
theory
Cauchy-Lipschitz
thm
Lack of uniqueness
The nonsmooth
Let X(·, x ) be the RLF, γx an integral curve from x ∈ Rd .
theory: RLF
A.e. uniqueness d
of integral curves
|X(t, x ) − γx (t)| ≤ |b(X(t, x )) − b(γx (t))|
Ideas dt
Ambrosio’s
superposition ≤ C |X(t, x ) − γx (t)|
principle
Interpolation
Ill-posedness of CE
by convex integration By Gronwall inequality, if b ∈ Lip we have everywhere uniqueness.
If b ∈ W 1,p , p > d, we have a.e. uniqueness [Caravenna, Crippa -
Jabin].
The case p > d: a.e. uniqueness
Flows of vector
fields
Smooth vs
nonsmooth
|b(x ) − b(y )| ≤ (g(x ) + g(y ))|x − y | ∀x , y p > 1,
theory
Cauchy-Lipschitz
thm
Lack of uniqueness
A.e. uniqueness
of integral curves
d
Ideas |X(t, x ) − γx (t)| ≤ |b(X(t, x )) − b(γx (t))|
Ambrosio’s dt
superposition
principle ≤ C |X(t, x ) − γx (t)|
Interpolation
Ill-posedness of CE
by convex integration
Flows of vector
fields
Maria Colombo
Lusin-Lipschitz inequality
Flows and For any b ∈ W 1,p then there exists g ∈ Lp such that
continuity
equation
Ideas d
Ambrosio’s |X(t, x ) − γx (t)| ≤ |b(X(t, x )) − b(γx (t))|
superposition
principle dt
Interpolation
Ill-posedness of CE
≤ C |X(t, x ) − γx (t)|
by convex integration
Flows of vector
fields
(Asymmetric) Lusin-Lipschitz inequality
Maria Colombo
Flows and
For any b ∈ W 1,p then there exists g ∈ Lp such that
continuity
equation
|b(x ) − b(y )| ≤ (g(x ) + g(y ))|x − y | ∀x , y p > 1,
Smooth vs
nonsmooth
theory
Cauchy-Lipschitz
thm
Lack of uniqueness
|b(x ) − b(y )| ≤ g(x )|x − y | ∀x p > d.
The nonsmooth
theory: RLF
A.e. uniqueness
Let X(·, x ) be the RLF, γx an integral curve from x ∈ Rd . We use
of integral curves the asymmetric Lusin inequality
Ideas
Ambrosio’s d
superposition
principle |X(t, x ) − γx (t)| ≤ |b(X(t, x )) − b(γx (t))|
Interpolation
dt
Ill-posedness of CE
by convex integration
≤ g(X(t, x ))|X(t, x ) − γx (t)|
Flows of vector
fields
(Asymmetric) Lusin-Lipschitz inequality
Maria Colombo
Flows and
For any b ∈ W 1,p then there exists g ∈ Lp such that
continuity
equation
|b(x ) − b(y )| ≤ (g(x ) + g(y ))|x − y | ∀x , y p > 1,
Smooth vs
nonsmooth
theory
Cauchy-Lipschitz
thm
Lack of uniqueness
|b(x ) − b(y )| ≤ g(x )|x − y | ∀x p > d.
The nonsmooth
theory: RLF
A.e. uniqueness
Let X(·, x ) be the RLF, γx an integral curve from x ∈ Rd . We use
of integral curves the asymmetric Lusin inequality
Ideas
Ambrosio’s d
superposition
principle |X(t, x ) − γx (t)| ≤ |b(X(t, x )) − b(γx (t))|
Interpolation
dt
Ill-posedness of CE
by convex integration
≤ g(X(t, x ))|X(t, x ) − γx (t)|
Flows of vector
fields Key observation: for a.e. x it holds
Maria Colombo
Z T
Flows and
continuity g(X(t, x )) dt < ∞.
equation 0
Smooth vs
nonsmooth
theory
Indeed, integrating in x and by incompressibility
Cauchy-Lipschitz
thm Z T Z T
Lack of uniqueness
The nonsmooth kg(X(t, ·))kLp dt ≤ C kgkLp dt ≤ CT k∇b(t, ·)kLp < ∞.
theory: RLF
0 0
A.e. uniqueness
of integral curves
Does the Lusin-Lipschitz inequality imply uniqueness for p < d?
Ideas
Ambrosio’s
[Crippa, De Lellis] used it to infer uniqueness of the RLF.
superposition
principle
Interpolation d
Ill-posedness of CE |X(t, x ) − γx (t)| ≤ |u(X(t, x )) − u(γx (t))|
by convex integration
dt
≤ g(X(t, x )) + g(γx (t)) |X(t, x ) − γx (t)|
RT
For a.e. x it holds 0
g(X(t, x )) + g(Y (t, x )) dt < ∞.
The case p < d: uniqueness of RLF
Flows of vector
fields Key observation: for a.e. x it holds
Maria Colombo
Z T
Flows and
continuity g(X(t, x )) dt < ∞.
equation 0
Smooth vs
nonsmooth
theory
Indeed, integrating in x and by incompressibility
Cauchy-Lipschitz
thm Z T Z T
Lack of uniqueness
The nonsmooth kg(X(t, ·))kLp dt ≤ C kgkLp dt ≤ CT k∇b(t, ·)kLp < ∞.
theory: RLF
0 0
A.e. uniqueness
of integral curves
Does the Lusin-Lipschitz inequality imply uniqueness for p < d?
Ideas
Ambrosio’s
[Crippa, De Lellis] used it to infer uniqueness of the RLF.
superposition
principle
Interpolation d
Ill-posedness of CE |X(t, x ) − γx (t)| ≤ |u(X(t, x )) − u(γx (t))|
by convex integration
dt
≤ g(X(t, x )) + g(γx (t)) |X(t, x ) − γx (t)|
RT
For a.e. x it holds 0
g(X(t, x )) + g(Y (t, x )) dt < ∞.
The case p < d: uniqueness of RLF
Flows of vector
fields Key observation: for a.e. x it holds
Maria Colombo
Z T
Flows and
continuity g(X(t, x )) dt < ∞.
equation 0
Smooth vs
nonsmooth
theory
Indeed, integrating in x and by incompressibility
Cauchy-Lipschitz
thm Z T Z T
Lack of uniqueness
The nonsmooth kg(X(t, ·))kLp dt ≤ C kgkLp dt ≤ CT k∇b(t, ·)kLp < ∞.
theory: RLF
0 0
A.e. uniqueness
of integral curves
Does the Lusin-Lipschitz inequality imply uniqueness for p < d?
Ideas
Ambrosio’s
[Crippa, De Lellis] used it to infer uniqueness of the RLF.
superposition
principle
Interpolation d
Ill-posedness of CE |X(t, x ) − Y (t, x )| ≤ |u(X(t, x )) − u(Y (t, x ))|
by convex integration
dt
≤ g(X(t, x )) + g(Y (t, x )) |X(t, x ) − Y (t, x )|
RT
For a.e. x it holds 0
g(X(t, x )) + g(Y (t, x )) dt < ∞.
The case p < d: uniqueness of RLF
Flows of vector
fields Key observation: for a.e. x it holds
Maria Colombo
Z T
Flows and
continuity g(X(t, x )) dt < ∞.
equation 0
Smooth vs
nonsmooth
theory
Indeed, integrating in x and by incompressibility
Cauchy-Lipschitz
thm Z T Z T
Lack of uniqueness
The nonsmooth kg(X(t, ·))kLp dt ≤ C kgkLp dt ≤ CT k∇b(t, ·)kLp < ∞.
theory: RLF
0 0
A.e. uniqueness
of integral curves
Does the Lusin-Lipschitz inequality imply uniqueness for p < d?
Ideas
Ambrosio’s
[Crippa, De Lellis] used it to infer uniqueness of the RLF.
superposition
principle
Interpolation d
Ill-posedness of CE |X(t, x ) − Y (t, x )| ≤ |u(X(t, x )) − u(Y (t, x ))|
by convex integration
dt
≤ g(X(t, x )) + g(Y (t, x )) |X(t, x ) − Y (t, x )|
RT
For a.e. x it holds 0
g(X(t, x )) + g(Y (t, x )) dt < ∞.
Main result
Flows of vector
fields
Maria Colombo
Flows and
continuity
equation
If p < d then the a.e. uniqueness for trajectories does not hold.
Smooth vs
nonsmooth
theory Theorem ([B.-Colombo-DeLellis, ’20])
Cauchy-Lipschitz
thm
Lack of uniqueness For every d ≥ 2, p < d and s < ∞ there exist a divergence free
The nonsmooth
theory: RLF
velocity field b ∈ Ct (Wx1,p ∩ Lsx ) and a set A ⊂ Td such that
A.e. uniqueness
of integral curves
L d (A) > 0;
Ideas for any x ∈ A there are at least two integral curves of u starting
Ambrosio’s
superposition at x .
principle
Interpolation
Ill-posedness of CE
by convex integration
The case p = d
Flows of vector
fields
A.e. uniqueness
What about the critical case p = d?
of integral curves
If ∇b ∈ L1t Ld,1
x , the a.e. uniqueness for integral curves holds.
Ideas
Ambrosio’s Recall that
superposition
principle
Interpolation
Z ∞ q dx λ 1/q
Ill-posedness of CE kf kLr ,q := λL d ({|f | ≥ λ})1/r
by convex integration
0 λ
Flows of vector
fields
A.e. uniqueness
What about the critical case p = d?
of integral curves
If ∇b ∈ L1t Ld,1
x , the a.e. uniqueness for integral curves holds.
Ideas
Ambrosio’s Recall that
superposition
principle
Interpolation
Z ∞ q dx λ 1/q
Ill-posedness of CE kf kLr ,q := λL d ({|f | ≥ λ})1/r
by convex integration
0 λ
Flows of vector
fields
A.e. uniqueness
What about the critical case p = d?
of integral curves
If ∇b ∈ L1t Ld,1
x , the a.e. uniqueness for integral curves holds.
Ideas
Ambrosio’s Recall that
superposition
principle
Interpolation
Z ∞ q dx λ 1/q
Ill-posedness of CE kf kLr ,q := λL d ({|f | ≥ λ})1/r
by convex integration
0 λ
Flows of vector
fields
Maria Colombo
Flows and
1 Flow of vector fields and continuity equation
continuity
equation
Smooth vs
2 Smooth vs nonsmooth theory
nonsmooth
theory The Cauchy-Lipschitz theorem for smooth vector fields
Cauchy-Lipschitz
thm
Lack of uniqueness of the flow for nonsmooth vector fields
Lack of uniqueness
Regular Lagrangian Flows and the nonsmooth theory
The nonsmooth
theory: RLF
A.e. uniqueness
of integral curves 3 A.e. uniqueness of integral curves
Ideas
Ambrosio’s
superposition
principle
4 Ideas of the proof
Interpolation Ambrosio’s superposition principle
Ill-posedness of CE
by convex integration Interpolation
Ill-posedness of CE by convex integration
Lagrangian uniqueness vs Eulerian uniqueness
Flows of vector
fields
A measure valued solution µ ∈ L∞ t (M+ ) to (CE) with velocity b is a
Maria Colombo superposition solution if for µ0 -a.e. x ∈ Td there exists
Flows and
ηx ∈ P(C ([0, T ], Td )) such that
continuity
equation ηx is concentrated on integral curves of b starting at x ;
Smooth vs
nonsmooth
we have the representation formula µ = (et )# (µ0 ⊗ ηx ),
theory
Cauchy-Lipschitz
Z Z Z
thm
Lack of uniqueness
φ dµt = φ(γ(t)) dηx (γ) dµ0 (x ).
The nonsmooth
theory: RLF
Ideas
Ambrosio’s
Theorem ( [Ambrosio ’04] )
superposition
principle
Interpolation
Let b : [0, T ] × Td → Rd , µ ∈ L∞
t (M+ ) solution of CE with
Ill-posedness of CE
by convex integration Z T Z
|b(t, x )| dx µt (x ) dxt < ∞.
0
Flows of vector
fields
A measure valued solution µ ∈ L∞ t (M+ ) to (CE) with velocity b is a
Maria Colombo superposition solution if for µ0 -a.e. x ∈ Td there exists
Flows and
ηx ∈ P(C ([0, T ], Td )) such that
continuity
equation ηx is concentrated on integral curves of b starting at x ;
Smooth vs
nonsmooth
we have the representation formula µ = (et )# (µ0 ⊗ ηx ),
theory
Cauchy-Lipschitz
Z Z Z
thm
Lack of uniqueness
φ dµt = φ(γ(t)) dηx (γ) dµ0 (x ).
The nonsmooth
theory: RLF
Ideas
Ambrosio’s
Theorem ( [Ambrosio ’04] )
superposition
principle
Interpolation
Let b : [0, T ] × Td → Rd , µ ∈ L∞
t (M+ ) solution of CE with
Ill-posedness of CE
by convex integration Z T Z
|b(t, x )| dx µt (x ) dxt < ∞.
0
Flows of vector
fields
A measure valued solution µ ∈ L∞ t (M+ ) to (CE) with velocity b is a
Maria Colombo superposition solution if for µ0 -a.e. x ∈ Td there exists
Flows and
ηx ∈ P(C ([0, T ], Td )) such that
continuity
equation ηx is concentrated on integral curves of b starting at x ;
Smooth vs
nonsmooth
we have the representation formula µ = (et )# (µ0 ⊗ ηx ),
theory
Cauchy-Lipschitz
Z Z Z
thm
Lack of uniqueness
φ dµt = φ(γ(t)) dηx (γ) dµ0 (x ).
The nonsmooth
theory: RLF
Ideas
Ambrosio’s
Theorem ( [Ambrosio ’04] )
superposition
principle
Interpolation
Let b : [0, T ] × Td → Rd , µ ∈ L∞
t (M+ ) solution of CE with
Ill-posedness of CE
by convex integration Z T Z
|b(t, x )| dx µt (x ) dxt < ∞.
0
Flows of vector
fields A.e. uniqueness of integral curves implies uniqueness of positive
Maria Colombo solutions to (CE).
Flows and
continuity
equation Proposition
Smooth vs
nonsmooth Let b ∈ L1t Wx1,1 divergence free whose integral curves are unique a.e.
theory
Cauchy-Lipschitz and X its RLF.
Then positive solutions µ ∈ L∞
thm
1
Lack of uniqueness t Lx to (CE) are unique and have the
The nonsmooth
theory: RLF representation
A.e. uniqueness
of integral curves
µt = (Xt )# µ0 for any t ∈ [0, T ].
Ideas
Ambrosio’s
superposition
principle
Interpolation
Indeed, by the superposition principle and a.e. uniqueness of integral
Ill-posedness of CE
by convex integration
curves, ηx ∈ P(C ([0, T ], Td )) must satisfy ηx := δX (·,t) . Hence
Z Z Z Z
φ dx µt = φ(γ(t)) d(γ) dµ0 (x ) = φ(X (t, x )) dµ0 (x ).
Lagrangian uniqueness vs Eulerian uniqueness
Flows of vector
fields A.e. uniqueness of integral curves implies uniqueness of positive
Maria Colombo solutions to (CE).
Flows and
continuity
equation Proposition
Smooth vs
nonsmooth Let b ∈ L1t Wx1,1 divergence free whose integral curves are unique a.e.
theory
Cauchy-Lipschitz and X its RLF.
Then positive solutions µ ∈ L∞
thm
1
Lack of uniqueness t Lx to (CE) are unique and have the
The nonsmooth
theory: RLF representation
A.e. uniqueness
of integral curves
µt = (Xt )# µ0 for any t ∈ [0, T ].
Ideas
Ambrosio’s
superposition
principle
Interpolation
Indeed, by the superposition principle and a.e. uniqueness of integral
Ill-posedness of CE
by convex integration
curves, ηx ∈ P(C ([0, T ], Td )) must satisfy ηx := δX (·,t) . Hence
Z Z Z Z
φ dx µt = φ(γ(t)) d(γ) dµ0 (x ) = φ(X (t, x )) dµ0 (x ).
Lagrangian uniqueness vs Eulerian uniqueness
Flows of vector
fields A.e. uniqueness of integral curves implies uniqueness of positive
Maria Colombo solutions to (CE).
Flows and
continuity
equation Proposition
Smooth vs
nonsmooth Let b ∈ L1t Wx1,1 divergence free whose integral curves are unique a.e.
theory
Cauchy-Lipschitz and X its RLF.
Then positive solutions µ ∈ L∞
thm
1
Lack of uniqueness t Lx to (CE) are unique and have the
The nonsmooth
theory: RLF representation
A.e. uniqueness
of integral curves
µt = (Xt )# µ0 for any t ∈ [0, T ].
Ideas
Ambrosio’s
superposition
principle
Interpolation
Indeed, by the superposition principle and a.e. uniqueness of integral
Ill-posedness of CE
by convex integration
curves, ηx ∈ P(C ([0, T ], Td )) must satisfy ηx := δX (·,t) . Hence
Z Z Z Z
φ dx µt = φ(γ(t)) d(γ) dµ0 (x ) = φ(X (t, x )) dµ0 (x ).
Lagrangian uniqueness vs Eulerian uniqueness
Flows of vector
fields A.e. uniqueness of integral curves implies uniqueness of positive
Maria Colombo solutions to (CE).
Flows and
continuity
equation Proposition
Smooth vs
nonsmooth Let b ∈ L1t Wx1,1 divergence free whose integral curves are unique a.e.
theory
Cauchy-Lipschitz and X its RLF.
Then positive solutions µ ∈ L∞
thm
1
Lack of uniqueness t Lx to (CE) are unique and have the
The nonsmooth
theory: RLF representation
A.e. uniqueness
of integral curves
µt = (Xt )# µ0 for any t ∈ [0, T ].
Ideas
Ambrosio’s
superposition
principle
Interpolation
Indeed, by the superposition principle and a.e. uniqueness of integral
Ill-posedness of CE
by convex integration
curves, ηx ∈ P(C ([0, T ], Td )) must satisfy ηx := δX (·,t) . Hence
Z Z Z Z
φ dx µt = φ(γ(t)) dηx (γ) dµ0 (x ) = φ(X (t, x )) dµ0 (x ).
Lagrangian uniqueness vs Eulerian uniqueness
Flows of vector
fields A.e. uniqueness of integral curves implies uniqueness of positive
Maria Colombo solutions to (CE).
Flows and
continuity
equation Proposition
Smooth vs
nonsmooth Let b ∈ L1t Wx1,1 divergence free whose integral curves are unique a.e.
theory
Cauchy-Lipschitz and X its RLF.
Then positive solutions µ ∈ L∞
thm
1
Lack of uniqueness t Lx to (CE) are unique and have the
The nonsmooth
theory: RLF representation
A.e. uniqueness
of integral curves
µt = (Xt )# µ0 for any t ∈ [0, T ].
Ideas
Ambrosio’s
superposition
principle
Interpolation
Indeed, by the superposition principle and a.e. uniqueness of integral
Ill-posedness of CE
by convex integration
curves, ηx ∈ P(C ([0, T ], Td )) must satisfy ηx := δX (·,t) . Hence
Z Z Z Z
φ dx µt = φ(γ(t)) dδX (·,t) (γ) dµ0 (x ) = φ(X (t, x )) dµ0 (x ).
Sharp summability, p > n
Flows of vector
fields
Maria Colombo
Flows and
continuity
(Related) question: well posedness of the CE
equation
Smooth vs
Let u ∈ L1t Wx1,r divergence free. Is the CE well-posed in the class of
nonsmooth
theory
positive solutions u ∈ L∞ p
t Lx under the minimal summability
Cauchy-Lipschitz
thm
requirement r + p ∗ < 1, namely p1 + 1r < 1 + d1 ?
1 1
Lack of uniqueness
The nonsmooth
theory: RLF From Ambrosio’s superposition principle and the a.e. uniqueness of
A.e. uniqueness
of integral curves
integral curves for p > d we infer that
Ideas
Ambrosio’s Corollary (Caravenna-Crippa 2018)
superposition
Flows of vector
fields
Maria Colombo
Flows and
continuity
(Related) question: well posedness of the CE
equation
Smooth vs
Let u ∈ L1t Wx1,r divergence free. Is the CE well-posed in the class of
nonsmooth
theory
positive solutions u ∈ L∞ p
t Lx under the minimal summability
Cauchy-Lipschitz
thm
requirement r + p ∗ < 1, namely p1 + 1r < 1 + d1 ?
1 1
Lack of uniqueness
The nonsmooth
theory: RLF From Ambrosio’s superposition principle and the a.e. uniqueness of
A.e. uniqueness
of integral curves
integral curves for p > d we infer that
Ideas
Ambrosio’s Corollary (Caravenna-Crippa 2018)
superposition
Flows of vector
fields
Is there a (p-dependent) family of inequalities which interpolates
Maria Colombo
between
Flows and
continuity
equation |b(x ) − b(y )| ≤ (g(x ) + g(y ))|x − y | p<n
Smooth vs
nonsmooth
theory |b(x ) − b(y )| ≤ g(x )|x − y | p>n
Cauchy-Lipschitz
thm
Lack of uniqueness
?
The nonsmooth
theory: RLF
Theorem (Brué-Colombo-De Lellis (2020))
A.e. uniqueness
of integral curves
Let b ∈ W 1,p , 1 < p < d, α ∈ [0, dp ). Then there exists g ∈ Lp such
Ideas
Ambrosio’s
that
superposition
principle
Interpolation |u(x ) − u(y )| ≤ (g(x ) + g(x )α g(y )1−α )|x − y | ∀x , y .
Ill-posedness of CE
by convex integration
Remark
The range of α is optimal.
Interpolating I
Flows of vector
fields
Is there a (p-dependent) family of inequalities which interpolates
Maria Colombo
between
Flows and
continuity
equation |b(x ) − b(y )| ≤ (g(x ) + g(y ))|x − y | p<n
Smooth vs
nonsmooth
theory |b(x ) − b(y )| ≤ g(x )|x − y | p>n
Cauchy-Lipschitz
thm
Lack of uniqueness
?
The nonsmooth
theory: RLF
Theorem (Brué-Colombo-De Lellis (2020))
A.e. uniqueness
of integral curves
Let b ∈ W 1,p , 1 < p < d, α ∈ [0, dp ). Then there exists g ∈ Lp such
Ideas
Ambrosio’s
that
superposition
principle
Interpolation |u(x ) − u(y )| ≤ (g(x ) + g(x )α g(y )1−α )|x − y | ∀x , y .
Ill-posedness of CE
by convex integration
Remark
The range of α is optimal.
Interpolating I
Flows of vector
fields
Is there a (p-dependent) family of inequalities which interpolates
Maria Colombo
between
Flows and
continuity
equation |b(x ) − b(y )| ≤ (g(x ) + g(y ))|x − y | p<n
Smooth vs
nonsmooth
theory |b(x ) − b(y )| ≤ g(x )|x − y | p>n
Cauchy-Lipschitz
thm
Lack of uniqueness
?
The nonsmooth
theory: RLF
Theorem (Brué-Colombo-De Lellis (2020))
A.e. uniqueness
of integral curves
Let b ∈ W 1,p , 1 < p < d, α ∈ [0, dp ). Then there exists g ∈ Lp such
Ideas
Ambrosio’s
that
superposition
principle
Interpolation |u(x ) − u(y )| ≤ (g(x ) + g(x )α g(y )1−α )|x − y | ∀x , y .
Ill-posedness of CE
by convex integration
Remark
The range of α is optimal.
Interpolating II
Flows of vector
fields
Maria Colombo
Flows and
continuity
equation Corollary
Smooth vs
nonsmooth Let u ∈ W 1,p , p < d. Positive solutions u ∈ L∞ r
t Lx of the CE are well
theory
Cauchy-Lipschitz posed in the range of exponent
thm
Lack of uniqueness
The nonsmooth 1 1 1 p−1
theory: RLF + <1+
A.e. uniqueness
r p d −1 p
of integral curves
Ideas This range but it is strictly contained in the range for which the
equations make sense p1 + 1r < 1 + d1 . What happens in between?
Ambrosio’s
superposition
principle
Interpolation
Ill-posedness of CE
by convex integration
Partial result by [Cheskidov, Luo ’20].
Interpolating II
Flows of vector
fields
Maria Colombo
Flows and
continuity
equation Corollary
Smooth vs
nonsmooth Let u ∈ W 1,p , p < d. Positive solutions u ∈ L∞ r
t Lx of the CE are well
theory
Cauchy-Lipschitz posed in the range of exponent
thm
Lack of uniqueness
The nonsmooth 1 1 1 p−1
theory: RLF + <1+
A.e. uniqueness
r p d −1 p
of integral curves
Ideas This range but it is strictly contained in the range for which the
equations make sense p1 + 1r < 1 + d1 . What happens in between?
Ambrosio’s
superposition
principle
Interpolation
Ill-posedness of CE
by convex integration
Partial result by [Cheskidov, Luo ’20].
Interpolating II
Flows of vector
fields
Maria Colombo
Flows and
continuity
equation
Corollary
Smooth vs
nonsmooth
Let u ∈ W 1,p , p < d. Positive solutions u ∈ L∞ r
t Lx of the CE are well
theory posed in the range of exponent
Cauchy-Lipschitz
thm
Lack of uniqueness 1 1 1 p−1
The nonsmooth + <1+
theory: RLF
r p d −1 p
A.e. uniqueness
of integral curves
Ideas
This range strictly contains the DiPerna-Lions range p1 + 1r ≤ 1 but it
Ambrosio’s
superposition
is strictly contained in the range for which the equations make sense
1 1 1
p + r < 1 + d . What happens in between? Partial result by
principle
Interpolation
Ill-posedness of CE
by convex integration
[Cheskidov, Luo ’20].
Nonuniqueness by convex integration
Flows of vector
fields
Maria Colombo
If we produce an example of nonuniqueness of positive solutions of
Flows and the continuity equations in some range of exponents we have disproved
continuity
equation the a.e. uniqueness of integral curves.
Smooth vs
nonsmooth
theory
Theorem ([B.-Colombo-DeLellis, ’20] )
Cauchy-Lipschitz
1 1
thm
Let d ≥ 2, p ∈ (1, ∞), p + p0 = 1, r ∈ [1, ∞] be such that
Lack of uniqueness
The nonsmooth
theory: RLF
1 1 1
A.e. uniqueness + >1+ .
of integral curves p r d
Ideas
Ambrosio’s
superposition Then there exist
principle 0
Interpolation a divergence-free vector field b ∈ Ct (Wx1,r ∩ Lpx ),
Ill-posedness of CE
by convex integration
a positive, nonconstant u ∈ Ct Lpx with u(0, ·) = 1,
which solve CE.
Nonuniqueness by convex integration
Flows of vector
fields
Maria Colombo
If we produce an example of nonuniqueness of positive solutions of
Flows and the continuity equations in some range of exponents we have disproved
continuity
equation the a.e. uniqueness of integral curves.
Smooth vs
nonsmooth
theory
Theorem ([B.-Colombo-DeLellis, ’20] )
Cauchy-Lipschitz
1 1
thm
Let d ≥ 2, p ∈ (1, ∞), p + p0 = 1, r ∈ [1, ∞] be such that
Lack of uniqueness
The nonsmooth
theory: RLF
1 1 1
A.e. uniqueness + >1+ .
of integral curves p r d
Ideas
Ambrosio’s
superposition Then there exist
principle 0
Interpolation a divergence-free vector field b ∈ Ct (Wx1,r ∩ Lpx ),
Ill-posedness of CE
by convex integration
a positive, nonconstant u ∈ Ct Lpx with u(0, ·) = 1,
which solve CE.
Nonuniqueness by convex integration
Flows of vector
fields
Maria Colombo
If we produce an example of nonuniqueness of positive solutions of
Flows and the continuity equations in some range of exponents we have disproved
continuity
equation the a.e. uniqueness of integral curves.
Smooth vs
nonsmooth
theory
Theorem ([B.-Colombo-DeLellis, ’20] )
Cauchy-Lipschitz
1 1
thm
Let d ≥ 2, p ∈ (1, ∞), p + p0 = 1, r ∈ [1, ∞] be such that
Lack of uniqueness
The nonsmooth
theory: RLF
1 1 1
A.e. uniqueness + >1+ .
of integral curves p r d
Ideas
Ambrosio’s
superposition Then there exist
principle 0
Interpolation a divergence-free vector field b ∈ Ct (Wx1,r ∩ Lpx ),
Ill-posedness of CE
by convex integration
a positive, nonconstant u ∈ Ct Lpx with u(0, ·) = 1,
which solve CE.
Remarks
Flows of vector
fields
Maria Colombo The main theorem follows: any velocity field obtained in the
Flows and
previous theorem does not have the a.e. uniqueness for integral
continuity
equation
curves. Indeed
Smooth vs Since div b = 0, the function ū ≡ 1 solves CE.
nonsmooth
theory
The u constructed in this theorem is a second distinct solution!
Cauchy-Lipschitz
thm
As seen before, a.e. uniqueness of integral curves implies
Lack of uniqueness uniqueness of positive solutions to (CE).
The nonsmooth
theory: RLF
The construction is based on convex integration scheme, as in the
A.e. uniqueness
of integral curves groundbreaking works [DeLellis-Székelyhidi, ’09-’13], [Isett ’16] for
Ideas the Euler equation and [Buckmaster-Vicol ’17] for Navier-Stokes.
Ambrosio’s
superposition
principle
The first ill-posedness result for (CE) with Sobolev velocity field
Interpolation has been proven in [Modena-Székelyhidi, ’18], [Modena-Sattig,
Ill-posedness of CE
by convex integration ’19].
Main novelties: positive solutions, a simpler convex integration
scheme in any dimension.
Remarks
Flows of vector
fields
Maria Colombo The main theorem follows: any velocity field obtained in the
Flows and
previous theorem does not have the a.e. uniqueness for integral
continuity
equation
curves. Indeed
Smooth vs Since div b = 0, the function ū ≡ 1 solves CE.
nonsmooth
theory
The u constructed in this theorem is a second distinct solution!
Cauchy-Lipschitz
thm
As seen before, a.e. uniqueness of integral curves implies
Lack of uniqueness uniqueness of positive solutions to (CE).
The nonsmooth
theory: RLF
The construction is based on convex integration scheme, as in the
A.e. uniqueness
of integral curves groundbreaking works [DeLellis-Székelyhidi, ’09-’13], [Isett ’16] for
Ideas the Euler equation and [Buckmaster-Vicol ’17] for Navier-Stokes.
Ambrosio’s
superposition
principle
The first ill-posedness result for (CE) with Sobolev velocity field
Interpolation has been proven in [Modena-Székelyhidi, ’18], [Modena-Sattig,
Ill-posedness of CE
by convex integration ’19].
Main novelties: positive solutions, a simpler convex integration
scheme in any dimension.
Remarks
Flows of vector
fields
Maria Colombo The main theorem follows: any velocity field obtained in the
Flows and
previous theorem does not have the a.e. uniqueness for integral
continuity
equation
curves. Indeed
Smooth vs Since div b = 0, the function ū ≡ 1 solves CE.
nonsmooth
theory
The u constructed in this theorem is a second distinct solution!
Cauchy-Lipschitz
thm
As seen before, a.e. uniqueness of integral curves implies
Lack of uniqueness uniqueness of positive solutions to (CE).
The nonsmooth
theory: RLF
The construction is based on convex integration scheme, as in the
A.e. uniqueness
of integral curves groundbreaking works [DeLellis-Székelyhidi, ’09-’13], [Isett ’16] for
Ideas the Euler equation and [Buckmaster-Vicol ’17] for Navier-Stokes.
Ambrosio’s
superposition
principle
The first ill-posedness result for (CE) with Sobolev velocity field
Interpolation has been proven in [Modena-Székelyhidi, ’18], [Modena-Sattig,
Ill-posedness of CE
by convex integration ’19].
Main novelties: positive solutions, a simpler convex integration
scheme in any dimension.
Remarks
Flows of vector
fields
Maria Colombo The main theorem follows: any velocity field obtained in the
Flows and
previous theorem does not have the a.e. uniqueness for integral
continuity
equation
curves. Indeed
Smooth vs Since div b = 0, the function ū ≡ 1 solves CE.
nonsmooth
theory
The u constructed in this theorem is a second distinct solution!
Cauchy-Lipschitz
thm
As seen before, a.e. uniqueness of integral curves implies
Lack of uniqueness uniqueness of positive solutions to (CE).
The nonsmooth
theory: RLF
The construction is based on convex integration scheme, as in the
A.e. uniqueness
of integral curves groundbreaking works [DeLellis-Székelyhidi, ’09-’13], [Isett ’16] for
Ideas the Euler equation and [Buckmaster-Vicol ’17] for Navier-Stokes.
Ambrosio’s
superposition
principle
The first ill-posedness result for (CE) with Sobolev velocity field
Interpolation has been proven in [Modena-Székelyhidi, ’18], [Modena-Sattig,
Ill-posedness of CE
by convex integration ’19].
Main novelties: positive solutions, a simpler convex integration
scheme in any dimension.
The end
Flows of vector
fields
Maria Colombo
Flows and
continuity
equation
Smooth vs
nonsmooth
theory
Cauchy-Lipschitz
thm
Lack of uniqueness
Thank you for your attention!
The nonsmooth
theory: RLF
A.e. uniqueness
of integral curves
Ideas
Ambrosio’s
superposition
principle
Interpolation
Ill-posedness of CE
by convex integration