Stability in Sobolev and related inequalities
Jean Dolbeault
http://www.ceremade.dauphine.fr/∼dolbeaul Ceremade, Université Paris-Dauphine
June 15, 2018
CIMI – EFI workshop onStability of functional inequalities and applications(13-15/6/2018)
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Outline
The stability issue in critical Sobolev and related inequalities Sobolev and Hardy-Littlewood-Sobolev inequalities Joint work with G. Jankowiak
Subcritical interpolation inequalities
BOn the Euclidean space: joint work with G. Toscani BOn the sphere: joint work with M.J. Esteban and M. Loss Reverse HLS inequality
BA quick introduction to a new family of inequalities for mean-field diffusion equations
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
A question by H. Brezis and E. Lieb
(Brezis, Lieb (1985))Is there a natural way to bound
Sdk∇uk2L2(Rd)− kuk2L2∗
(Rd)
from below in terms of the“distance” off from the set of optimal [Aubin-Talenti] functionswhend≥3 ?
(Bianchi, Egnell 1990)There is a positive constantαsuch that Sdk∇uk2L2(Rd)− kuk2L2∗(Rd)≥α inf
ϕ∈Mk∇u− ∇ϕk2L2(Rd)
(Cianchi, Fusco, Maggi, Pratelli 2009)(also a version for k∇ukpLp(Rd)) There are constantsαandκsuch that
Sdk∇uk2L2(Rd)≥(1 +κ λ(u)α)kuk2L2∗
(Rd)
where λ(u) = infϕ∈M
ku−ϕk2∗ L2∗
(Rd)
kuk2∗
L2∗ (Rd)
: kuk2L∗2∗
(Rd)=kϕk2L∗2∗
(Rd)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
Sobolev and Hardy-Littlewood-Sobolev inequalities
BStability in a weaker norm but with explicit constants BFrom duality to improved estimates based on Yamabe’s flow
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
Sobolev and HLS
As it has been noticed by E. Lieb, Sobolev’s inequality inRd,d≥3, kuk2L2∗(Rd)≤Sdk∇uk2L2(Rd) ∀u∈D1,2(Rd) (1) and the Hardy-Littlewood-Sobolev inequality
Sdkvk2
L2
d d+2(Rd)
≥ Z
Rd
v(−∆)−1v dx ∀v∈Ld+22d(Rd) (2) aredual of each other. HereSd is the Aubin-Talenti constant and 2∗= d−22d
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
Improved Sobolev inequality by duality
Theorem
(JD, G. Jankowiak)Assume thatd≥3 and letq= d+2d−2. There exists a positive constantC≤1 such that
Sdkwqk2
Ld+22d(Rd)
− Z
Rd
wq(−∆)−1wqdx
≤CSdkwk
8 d−2
L2∗(Rd)
hk∇wk2L2(Rd)−Sdkwk2L2∗
(Rd)
i
for anyw∈D1,2(Rd)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
Proof: the completion of a square
Integrations by parts show that Z
Rd
|∇(−∆)−1v|2dx= Z
Rd
v(−∆)−1v dx and, ifv=uq withq=d+2d−2,
Z
Rd
∇u· ∇(−∆)−1v dx= Z
Rd
u v dx= Z
Rd
u2∗dx
Hence the expansion of the square 0≤
Z
Rd
Sdkuk
4 d−2
L2∗(Rd)∇u− ∇(−∆)−1v
2
dx
shows that 0≤Sdkuk
8 d−2
L2∗(Rd)
h
Sdk∇uk2L2(Rd)− kuk2L2∗
(Rd)
i
−h
Sdkuqk2
L
2d d+2(Rd)
− Z
Rd
uq(−∆)−1uqdxi
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
Using a nonlinear flow to relate Sobolev and HLS
Consider thefast diffusionequation
∂v
∂t = ∆vm t >0, x∈Rd (3) If we defineH(t) :=Hd[v(t,·)], with
Hd[v] :=
Z
Rd
v(−∆)−1v dx−Sdkvk2
L
2d d+2(Rd)
then we observe that 1
2H0 =− Z
Rd
vm+1dx+Sd
Z
Rd
vd+22d dx 2dZ
Rd
∇vm· ∇vd−2d+2dx wherev=v(t,·) is a solution of (3). With the choicem= d−2d+2, we find thatm+ 1 = d+22d
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
A preliminary observation
Proposition
(JD)Assume thatd≥3 andm= d−2d+2. Ifv is a solution of (3)with nonnegative initial datum inL2d/(d+2)(Rd), then
1 2
d dt
Z
Rd
v(−∆)−1v dx−Sdkvk2
Ld+22d(Rd)
= Z
Rd
vm+1dx 2dh
Sdk∇uk2L2(Rd)− kuk2L2∗(Rd)
i≥0
The HLS inequality amounts toH≤0 and appears as a consequence of Sobolev, that isH0 ≥0 if we show that lim supt>0H(t) = 0
Notice thatu=vmis an optimal function for (1) ifvis optimal for (2)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
Improved Sobolev inequality
By integrating along the flow defined by (3), we can actually obtain optimal integral remainder terms which improve on the usual Sobolev inequality (1), but only whend≥5 for integrability reasons
Theorem
(JD)Assume thatd≥5 and letq= d+2d−2. There exists a positive constantC≤ 1 +d2
1−e−d/2
Sd such that
Sdkwqk2
Ld+22d(Rd)
− Z
Rd
wq(−∆)−1wqdx
≤CkwkLd−228∗(Rd)
hk∇wk2L2(Rd)−Sdkwk2L2∗(Rd)
i
for anyw∈D1,2(Rd)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
Solutions with separation of variables
Consider the solution of ∂v∂t = ∆vm vanishing att=T: vT(t, x) =c(T−t)α(F(x))d+2d−2 whereF is the Aubin-Talenti solution of
−∆F =d(d−2)F(d+2)/(d−2)
Letkvk∗:= supx∈Rd(1 +|x|2)d+2|v(x)|
Lemma
(M. del Pino, M. Saez),(J. L. Vázquez, J. R. Esteban, A. Rodriguez) For any solutionv with initial datumv0∈L2d/(d+2)(Rd),v0>0, there existsT >0,λ >0 andx0∈Rd such that
t→Tlim−
(T−t)−1−m1 kv(t,·)/v(t,·)−1k∗= 0
withv(t, x) =λ(d+2)/2vT(t,(x−x0)/λ)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
Improved inequality: proof (1/2)
The functionJ(t) :=R
Rdv(t, x)m+1dxsatisfies J0=−(m+ 1)k∇vmk2L2(Rd)≤ −m+ 1
Sd
J1−2d
Ifd≥5, then we also have J00= 2m(m+ 1)
Z
Rd
vm−1(∆vm)2dx≥0 Notice that
J0
J ≤ −m+ 1
Sd J−2d ≤ −κ with κ T = 2d d+ 2
T Sd
Z
Rd
vm+10 dx −2d
≤ d 2
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
Improved inequality: proof (2/2)
By theCauchy-Schwarz inequality, we have J02
(m+ 1)2 =k∇vmk4L2(Rd)= Z
Rd
v(m−1)/2∆vm·v(m+1)/2dx 2
≤ Z
Rd
vm−1(∆vm)2dx Z
Rd
vm+1dx=CstJ00J so thatQ(t) :=k∇vm(t,·)k2L2(Rd)
R
Rdvm+1(t, x)dx−(d−2)/d is monotone decreasing, and
H0 = 2J(SdQ−1), H00= J0
J H0+ 2J SdQ0≤ J0 J H0≤0 H00≤ −κH0 with κ= 2d
d+ 2 1 Sd
Z
Rd
v0m+1dx −2/d
By writing that−H(0) =H(T)−H(0)≤H0(0) (1−e−κ T)/κand using the estimateκ T ≤d/2, the proof is completed
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
d = 2: Onofri’s and log HLS inequalities
H2[v] :=
Z
R2
(v−µ) (−∆)−1(v−µ)dx− 1 4π
Z
R2
vlog v
µ
dx Withµ(x) := π1(1 +|x|2)−2. Assume thatv is a positive solution of
∂v
∂t = ∆ log (v/µ) t >0, x∈R2 Proposition
Ifv=µ eu/2 is a solution with nonnegative initial datumv0 in L1(R2) such thatR
R2v0dx= 1,v0logv0∈L1(R2)andv0 logµ∈L1(R2), then
d
dtH2[v(t,·)] = 1 16π
Z
R2
|∇u|2dx− Z
R2
eu2 −1 u dµ
≥ 161πR
R2|∇u|2dx+R
R2u dµ−log R
R2eudµ
≥0
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
Another improvement
Jd[v] :=
Z
Rd
vd+22d dx and Hd[v] :=
Z
Rd
v(−∆)−1v dx−Sdkvk2
Ld+22d(Rd)
Theorem (J.D., G. Jankowiak)
Assume thatd≥3. Then we have 0≤Hd[v] +SdJd[v]1+2dϕ
Jd[v]d2−1 h
Sdk∇uk2L2(Rd)− kuk2L2∗
(Rd)
i
∀u∈D, v=ud+2d−2 whereϕ(x) :=√
C2+ 2Cx−C for anyx≥0 Proof: H(t) =−Y(J(t))∀t∈[0, T),κ0:= H
0 0
J0 and consider the differential inequality
Y0
CSds1+2d+Y
≤ d+ 2
2d Cκ0S2ds1+4d, Y(0) = 0, Y(J0) =−H0
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Duality Yamabe flow
... but C = 1 is not optimal
Theorem
(JD, G. Jankowiak)In the inequality
Sdkwqk2
Ld+22d(Rd)
− Z
Rd
wq(−∆)−1wqdx
≤CdSdkwk
8 d−2
L2∗(Rd)
hk∇wk2L2(Rd)−Sdkwk2L2∗(Rd)
i
we have
d
d+ 4 ≤Cd<1
based on a (painful) linearization Extensions:
fractional Laplacian operator (Jankowiak, Nguyen) Moser-Trudinger-Onofri inequality
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
Subcritical interpolation inequalities
BEuclidean space: fast diffusion, entropies and improved asymptotic expansions
Based on papers with A. Blanchet, M. Bonforte, G. Grillo, J.L. Vázquez and papers with G. Toscani
BSphere: explicit remainder terms based on nonlinear diffusions Joint work with MJ. Esteban and M. Loss
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
Higher order matching asymptotics
(J.D., G. Toscani)For somem∈(mc,1) with mc:= (d−2)/d, we consider onRd the fast diffusion equation
∂u
∂τ +∇ · u∇um−1
= 0
The strategy is easy to understand using a time-dependent rescaling and the relative entropy formalism. Define the functionvsuch that
u(τ, y+x0) =R−dv(t, x), R=R(τ), t= 12 logR , x= y R Thenvhas to be a solution of
∂v
∂t +∇ ·h v
σd2(m−mc)∇vm−1−2xi
= 0 t >0, x∈Rd with (as long as we make no assumption onR)
2σ−d2(m−mc)=R1−d(1−m)dR dτ
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
Refined relative entropy
Consider the family of the Barenblatt profiles Bσ(x) :=σ−d2 CM+σ1|x|2m−11
∀x∈Rd (4) Note thatσis a function oft: as long as dσdt 6= 0, the Barenblatt profile Bσ is nota solution but we may still consider the relative entropy
Fσ[v] := 1 m−1
Z
Rd
vm−Bσm−m Bσm−1(v−Bσ) dx Let us briefly sketch the strategy of our method before giving all details
The time derivative of this relative entropy is d
dtFσ(t)[v(t,·)] = dσ dt
d dσFσ[v]
|σ=σ(t)
| {z } choose it = 0
⇐⇒ MinimizeFσ[v] w.r.t.σ ⇐⇒R
Rd|x|2Bσdx=R
Rd|x|2v dx
+ m
m−1 Z
Rd
vm−1−Bσ(t)m−1∂v
∂tdx
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
The entropy / entropy production estimate
According to the definition ofBσ, we know that 2x=σd2(m−mc)∇Bm−1σ
Using the new change of variables, we know that d
dtFσ(t)[v(t,·)] =−m σ(t)d2(m−mc) 1−m
Z
Rd
v ∇h
vm−1−Bσ(t)m−1i
2
dx
Letw:=v/Bσand observe that the relative entropy can be written as Fσ[v] = m
1−m Z
Rd
h
w−1− 1
m wm−1i Bσmdx
(Repeating) define therelative Fisher informationby Iσ[v] :=
Z
Rd
1 m−1∇
(wm−1−1)Bσm−1
2
Bσw dx
so that d
dtFσ(t)[v(t,·)] =−m(1−m)σ(t)Iσ(t)[v(t,·)] ∀t >0 When linearizing, one more mode is killed andσ(t)scales out
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
Improved rates of convergence
Theorem (J.D., G. Toscani)
Letm∈(me1,1),d≥2,v0∈L1+(Rd)such that v0m,|y|2v0∈L1(Rd) E[v(t,·)]≤C e−2γ(m)t ∀t≥0
where
γ(m)=
((d−2)m−(d−4))2
4 (1−m) ifm∈(me1,me2] 4 (d+ 2)m−4d ifm∈[me2, m2]
4 ifm∈[m2,1)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
Spectral gaps and best constants
mc=d−2d 0
m1=d−1d
m2=d+1d+2
! m2:=d+4d+6
m 1
2 4
Case 1 Case 2 Case 3
γ(m)
(d= 5)
! m1:=d+2d
(A. Blanchet, M. Bonforte, J.D., G. Grillo, J.-L. Vázquez)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
Best matching Barenblatt profiles
(Repeating) Consider thefast diffusion equation
∂u
∂t +∇ ·h u
σd2(m−mc)∇um−1−2xi
= 0 t >0, x∈Rd with a nonlocal, time-dependent diffusion coefficient
σ(t)= 1 KM
Z
Rd
|x|2u(x, t)dx , KM :=
Z
Rd
|x|2B1(x)dx where
Bλ(x) :=λ−d2 CM+λ1|x|2m−11
∀x∈Rd and define the relative entropy
Fλ[u] := 1 m−1
Z
Rd
um−Bλm−m Bm−1λ (u−Bλ) dx
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
Three ingredients for global improvements
1 infλ>0Fλ[u(x, t)] =Fσ(t)[u(x, t)] so that d
dtFσ(t)[u(x, t)] =−Jσ(t)[u(·, t)]
where the relative Fisher information is Jλ[u] :=λd2(m−mc) m
1−m Z
Rd
u
∇um−1− ∇Bλm−1
2dx
2 In the Bakry-Emery method, there isan additional (good) term 4
"
1 + 2Cm,d
Fσ(t)[u(·, t)]
Mγσ0d2(1−m)
# d
dt Fσ(t)[u(·, t)]
≥ d
dt Jσ(t)[u(·, t)]
3 TheCsiszár-Kullback inequalityis also improved Fσ[u]≥ m
8R
RdBm1 dxCM2 ku−Bσk2L1(Rd)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
improved decay for the relative entropy
0.0 0.2 0.4 0.6 0.8 1.0
0.2 0.4 0.6 0.8 1.0
t!→f(t)e4t f(0)
t (a)
(b) (c)
Figure:Upper bounds on the decay of the relative entropy:
t7→f(t)e4t/f(0) (a): estimate given by the entropy-entropy production method
(b): exact solution of a simplified equation
(c): numerical solution (found by a shooting method)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
A Csiszár-Kullback(-Pinsker) inequality
Letm∈(me1,1) withme1= d+2d and consider the relative entropy Fσ[u] := 1
m−1 Z
Rd
um−Bσm−m Bm−1σ (u−Bσ) dx
Theorem
Letd≥1,m∈(me1,1)and assume thatuis a nonnegative function in L1(Rd)such that um andx7→ |x|2uare both integrable onRd. If kukL1(Rd)=M andR
Rd|x|2u dx=R
Rd|x|2Bσdx, then Fσ[u]
σd2(1−m) ≥ m 8R
RdB1mdx
CMku−BσkL1(Rd)+1 σ Z
Rd
|x|2|u−Bσ|dx 2
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
An improved Gagliardo-Nirenberg inequality: setting
The inequality
kfkL2p(Rd)≤CGNp,dk∇fkθL2(Rd)kfk1−θLp+1(
Rd)
withθ=θ(p) :=p−1p d+2−pd(d−2), 1< p≤ d−2d ifd≥3 and 1< p <∞ ifd= 2, can be rewritten, in a non-scale invariant form, as
Z
Rd
|∇f|2dx+ Z
Rd
|f|p+1dx≥Kp,d Z
Rd
|f|2pdx γ
withγ=γ(p, d) := d+2−pd−p(d−4)(d−2). Optimal function are given by fM,y,σ(x) = 1
σd2
CM+|x−y|2 σ
−p−11
∀x∈Rd whereCM is determined byR
RdfM,y,σ2p dx=M Md:=
fM,y,σ : (M, y, σ)∈Md:= (0,∞)×Rd×(0,∞)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
An improved Gagliardo-Nirenberg inequality
Relative entropyfunctional R(p)[f] := inf
g∈M(p)d
Z
Rd
hg1−p |f|2p−g2p
−p+12p |f|p+1−gp+1i dx
Theorem
Letd≥2,p >1 and assume thatp < d/(d−2) ifd≥3. If R
Rd|x|2|f|2pdx R
Rd|f|2pdxγ =d(p−1)σ∗M
γ−1
∗
d+2−p(d−2) , σ∗(p) :=
4d+2−p(p−1)2(d−2)(p+1)
d−p(d−4)4p
for anyf ∈Lp+1∩D1,2(Rd), then we have Z
Rd
|∇f|2dx+
Z
Rd
|f|p+1dx−Kp,d
Z
Rd
|f|2pdx γ
≥Cp,d
R(p)[f]2 R
Rd|f|2pdxγ By Csiszár-Kullback: control of
|f|2p−g2p
4 L1(Rd)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
Best matching Barenblatt profiles are delayed
Letube such that
v(τ, x) = µd R(D τ)du
1
2logR(D τ), µ x R(D τ)
withτ7→R(τ) given as the solution to 1
R dR dτ =
µ2 KM
Z
Rd
|x|2v(τ, x)dx
−d2(m−mc)
, R(0) = 1 Then
1 R
dR dτ =
R2(τ)σ
1
2logR(D τ)
−d2(m−mc)
that isR(τ) =R0(τ)≤R0(τ) where R1 dRdτ0 = R20(τ)σ(0)−d2(m−mc)
and asymptotically asτ → ∞,R(τ) =R0(τ−δ) for somedelayδ >0
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
The interpolation inequalities on S
dOn thed-dimensional sphere, let us consider the interpolation inequality
k∇uk2L2(Sd)+ d
p−2kuk2L2(Sd)≥ d
p−2kuk2Lp(Sd) ∀u∈H1(Sd, dµ) where the measuredµis the uniform probability measure on Sd⊂Rd+1 corresponding to the measure induced by the Lebesgue measure onRd+1, and the exposantp≥1,p6= 2, is such that
p≤2∗:= 2d d−2
ifd≥3. We adopt the convention that 2∗=∞ifd= 1 ord= 2. The casep= 2 corresponds to the logarithmic Sobolev inequality
k∇uk2L2(Sd)≥d 2
Z
Sd
|u|2 log |u|2 kuk2L2(
Sd)
!
dµ ∀u∈H1(Sd, dµ)\ {0}
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
The Bakry-Emery method
Entropy functional
Ep[ρ] := p−21 h R
Sdρ2p dµ− R
Sdρ dµ2pi
if p6= 2 E2[ρ] :=R
Sdρlog
ρ kρkL1 (Sd)
dµ Fisher informationfunctional
Ip[ρ] :=R
Sd|∇ρ1p|2dµ
Bakry-Emery (carré du champ) method: use the heat flow
∂ρ
∂t = ∆ρ
and compute dtdEp[ρ] =−Ip[ρ] and dtdIp[ρ]≤ −dIp[ρ] to get d
dt(Ip[ρ]−dEp[ρ])≤0 =⇒ Ip[ρ]≥dEp[ρ]
withρ=|u|p, ifp≤2#:=(d−1)2d2+12
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
The evolution under the fast diffusion flow
To overcome the limitationp≤2#, one can consider a nonlinear diffusion of fast diffusion / porous medium type
∂ρ
∂t = ∆ρm. (5)
(Demange),(JD, Esteban, Kowalczyk, Loss): for anyp∈[1,2∗] Kp[ρ] := d
dt
Ip[ρ]−dEp[ρ]
≤0
1.0 1.5 2.5 3.0
0.0 0.5 1.5 2.0
(p, m)admissible region,d= 5
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
Improved interpolation inequalities in the sphere
Let
λ? := inf
v∈H1+(Sd, dµ) R
Sdv dµ= 1 R
Sdx|v|pdµ= 0 R
Sd(∆v)2dµ R
Sd|∇v|2ν dµ > d
and consider the inequality Z
Sd
|∇f|2ν dµ+ λ
p−2kfk22≥ λ p−2kfk2p
∀f ∈H1(Sd, dµ) s.t.
Z
Sd
x|f|pdµ= 0 Proposition
For anyp∈(2,2#), the inequality holds with λ≥d+ (d−1)2
d(d+ 2)(2#−p) (λ?−d)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
p = 2: the logarithmic Sobolev case
λ?=d+ 2 (d+ 2) 2 (d+ 3) +p
2 (d+ 3) (2d+ 3) Proposition
Letd≥2. For anyu∈H1(Sd, dµ)\ {0}such that R
Sdx|u|2dµ= 0, we have
Z
Sd
|∇u|2dµ≥ δ 2
Z
Sd
|u|2log |u|2
kuk22
dµ withδ:=d+2d 4d−1
2 (d+3)+√
2 (d+3) (2d+3)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
Stability under antipodal symmetry
With the additional restriction ofantipodal symmetry, that is u(−x) =u(x) ∀x∈Sd
Theorem
Ifp∈(1,2)∪(2,2∗), we have Z
Sd
|∇u|2dµ≥ d p−2
1 + (d2−4) (2∗−p) d(d+ 2) +p−1
kuk2Lp(Sd)− kuk2L2(Sd)
for anyu∈H1(Sd, dµ)with antipodal symmetry. The limit casep= 2 corresponds to the improved logarithmic Sobolev inequality
Z
Sd
|∇u|2dµ≥ d 2
(d+ 3)2 (d+ 1)2
Z
Sd
|u|2 log |u|2 kuk2L2(
Sd)
! dµ
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Fast diffusion equations and best matching onRd Improved interpolation inequalities on the sphere
The optimal constant in the antipodal framework
æ
æ æ
æ æ
1.5 2.0 2.5 3.0
6 7 8 9 10 11 12
Numerical computation of the optimal constant when d= 5and 1≤p≤10/3≈3.33. The limiting value of the constant is numerically
found to be equal toλ?= 21−2/pd≈6.59754 with d= 5andp= 10/3
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Basic properties Relaxation Free energy
Reverse Hardy-Littewood-Sobolev inequality
Joint work with J. A. Carrillo, M. G. Delgadino, R. Frank, F. Hoffmann
BA family of inequalities
BExistence of minimizers and relaxation
BNo concentration and regularity of measure valued minimizers BFree Energy
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Basic properties Relaxation Free energy
The reverse HLS inequality
For anyλ >0 and any measurable function ρ≥0 onRN, let Iλ[ρ] :=
Z Z
RN×RN
|x−y|λρ(x)ρ(y)dx dy N ≥1, 0< q <1, α:= 2N−q(2N+λ)
N(1−q) Convention: ρ∈Lp(RN) ifR
RN|ρ(x)|pdxfor anyp >0 Theorem
Iλ[ρ] ≥CN,λ,q
Z
RN
ρ(x)dx αZ
RN
ρ(x)qdx
(2−α)/q
(6) holds for anyρ∈L1+∩Lq(RN)with CN,λ,q>0 if and only if
q > N/(N+λ)
If eitherN= 1,2or if N ≥3andq≥min
1−2/N ,2N/(2N+λ) , then there is a radial nonnegative optimizerρ∈L1∩Lq(RN)
J. Dolbeault Sobolev and related inequalities
Critical Sobolev and HLS inequalities Subcritical interpolation inequalities Reverse Hardy-Littewood-Sobolev inequality
Basic properties Relaxation Free energy
0 2 4 6 8 10 12
0.0 0.2 0.4 0.6 0.8 1.0
N = 4, region of the parameters(λ, q) for whichCN,λ,q>0
J. Dolbeault Sobolev and related inequalities