equation
AndrédeLaire(UniversitéLille1)
JointworkwithPhilippeGravejat(Éolepolytehnique)
Shrödingerequationsandappliations
CIRMJune18th2014
Thedynamisofmagnetizationinaferromagnetimaterialisgivenbythe
LandauLifshitzequation.
Themagnetizationisadiretioneld:
~
m
(x , t) : R D × R → S
2⊂ R
3,
m~ = (m
1, m
2, m
3)
|~
m| = (m
21+ m
22+ m
23)
1/
2=
13Dferromagnetimaterials
2D:Thinmaterials.
1D:CesiumnikeltriuorideCsNiF3,theouplingbetweentheNiionsismuh
strongerthantheotherouplingsinthelattie.
∂ t
m~ = −β ~
m× ~
heff (
m~ )
| {z }
precession
~
h
eff (
m~ )
: eetivemagnetield∂ t
m~ = −β ~
m× ~
heff (
m~ )
| {z }
precession
−α~
m× (
m~ × ~
heff (~
m))
| {z }
damping
~
h
eff (
m~ )
: eetivemagnetieldα >
0:
Gilbertdampingoeientβ >
0,w.l.o.g.α
2+ β
2=
1~
heff (
m~ ) = ∆
m~ − λm
3ˆ
e3 ,∂ t
m~ = −~
m× (∆~
m− λm
3ˆ
e3),
~
m
(x , t) : R D × R → S
2, ~
m= (m
1, m
2, m
3)
~
heff (
m~ ) = ∆
m~ − λm
3ˆ
e3 ,∂ t
m~ = −~
m× (∆~
m− λm
3ˆ
e3),
~
m
(x , t) : R D × R → S
2, ~
m= (m
1, m
2, m
3)
∂ t m
1= −m
2(∆m
3− λm
3) + m
3S∆m
2∂ t m
2= −m
3∆m
1− m
1(∆m
3+ λm
3)
∂ t m
3= −m
1∆m
2+ m
2∆m
1~
heff (
m~ ) = ∆
m~ − λm
3ˆ
e3 ,∂ t
m~ = −~
m× (∆~
m− λm
3ˆ
e3),
~
m
(x , t) : R D × R → S
2, ~
m= (m
1, m
2, m
3)
Theequationishamiltonian. TheonservedHamiltonianistheLandau-Lifshitzenergy
E (~
m(t)) :=
12
Z
R N
|∇
m~ (x , t)|
2dx + λ
2
Z
R N
m
3(x , t)
2dx = E (~
m(
0)), ∀t ∈ R .
Theanisotropyparameter
λ ≥
0:λ =
0:
isotropiase,Shrödingermapequationλ >
0:
easy-planeorplanaranisotropy lim|x|→∞ m
3(x, t) =
0Inthesequel,weonsider
λ =
1andsolutionsm
withniteLandau-Lifshitzenergy.Whenthemap
m ˇ := m
1+ im
2doesnotvanish,itmaybewrittenas
ˇ m =
q
1
− m
23 exp
iϕ.
Thefuntions
v = m
3and
w = ∇ϕ
aresolutionstothehydrodynamiLandau-Lifshitz equation
∂ t v = −
div(
1− v
2)w ,
∂ t w = ∇
v (|w |
2−
1) + ∆v
1
− v
2+ v |∇v |
2(
1− v
2)
2.
(HLL)
Thelinearizedequationaroundthezerosolutionisdispersivewithdispersionrelation
ω
2= |k|
4+ |k|
2.
Dispersionveloityisatleastequaltothesoundspeed
c s =
1.Welookfortravelingwavessolutionspropagatingalongthe
x
1-axiswithspeed
c
,i.e.~
m
(x ) = ~
u(x
1− ct , x
2, . . . , x D ).
Thentheprole
~
usatises−c ∂
1~
u+ ~
u× (∆~
u− u
3ˆ
e3) =
0.
Taking
~
u×
andusingthat~
u
× (~
u× ∆~
u) = ~
u(~
u· ∆~
u) − ∆~
u(~
u· ~
u) = −~
u|∇~
u|
2− ∆~
u,
weobtain
− ∆ ~
u= |∇~
u|
2~
u+ u
23~
u− u
3ˆ
e3+ c~
u× ∂
1~
u.
(TWc
)Welookfortravelingwavessolutionspropagatingalongthe
x
1-axiswithspeed
c
,i.e.~
m
(x ) = ~
u(x
1− ct , x
2, . . . , x D ).
Thentheprole
~
usatises−c ∂
1~
u+ ~
u× (∆~
u− u
3ˆ
e3) =
0.
Taking
~
u×
andusingthat~
u
× (~
u× ∆~
u) = ~
u(~
u· ∆~
u) − ∆~
u(~
u· ~
u) = −~
u|∇~
u|
2− ∆~
u,
weobtain
− ∆ ~
u= |∇~
u|
2~
u+ u
23~
u− u
3ˆ
e3+ c~
u× ∂
1~
u.
(TWc
)Trivialsolutions: onstantsin
S
1× {
0}
.Itisenoughtoonsider
c ≥
0.propertiesfor
D =
2,
3. Theyderivednumeriallyabranhoftravelingwavesforsubsonispeeds
|c | <
1.LinandWei(2010)provedtheexisteneofsmallspeedtravellingwaveswitha
vortex-antivortexpairwhen
D =
2.d.L.(2014): thenon-existeneofnon-onstantsubsonismallenergytravelling
waves for
D =
2,
3,
4.d.L.(2014): alsosmoothnessandtheiralgebraideayatinnityfor
D ≥
2.Somenumerial resultsfor
D =
2 (PapaniolaouSpathis)0 15 30 45 60
15 30
p E
c → 0 c → 1
E = p
c≈ 0 . 78
Somenumerial resultsfor
D =
2 (PapaniolaouSpathis) Funtionu
3for
c =
0.
5.Somenumerial resultsfor
D =
2 (PapaniolaouSpathis) Figure(a): speedc =
0.
2. Figure(b): speedc =
0.
95.Let
~
ubeasolutionof (TWc
). Usingthestereographivariableψ = u
1+ iu
21
+ u
3,
wehavethat
ψ
satises∆ψ +
1− |ψ|
21
+ |ψ|
2ψ + ic∂
1ψ =
2ψ ¯
1
+ |ψ|
2(∇ψ · ∇ψ),
whihseemslikeaperturbedequationforthetravelingwavesforaNonlinearShrödinger
equation(GrossPitaevskiiequation).
Let
~
ubeasolutionof (TWc
). Usingthestereographivariableψ = u
1+ iu
21
+ u
3,
wehavethat
ψ
satises∆ψ +
1− |ψ|
21
+ |ψ|
2ψ + ic∂
1ψ =
2ψ ¯
1
+ |ψ|
2(∇ψ · ∇ψ)
thatseemslikeaperturbedequationforthetravelingwavesforanonlinearShrödinger
equation(GrossPitaevskiiequation):
∆Ψ + (
1− |Ψ|
2)Ψ − ic ∂
1Ψ =
0.
(GP)Solitonsfor
D =
1Indimensionone,thetravelingwavesolutionsarethefollowingexpliitdarksolitons.
Lemma(The one-dimensionalase)
Let
D =
1andc ≥
0. Assumethat~
uisanontrivialniteenergysolutionof (TWc
).Then0
≤ c <
1and(uptoinvarianes)thesolutionisgivenbyu
1= c
seh( p
1
− c
2x ), u
2=
tanh( p
1
− c
2x), u
3= p
1
− c
2seh( p
1
− c
2x).
Multisolitonsfor
D =
1TheLandau-Lifshitzequationisintegrablebymeansoftheinversesatteringmethod. In
partiular,thereareexpliitformulaeformulti-solitons. Theybehavelikeasumof
orderedsolitonsas
t → ∞
.(Loading...)
Multisolitonsfor
D =
1TheLandau-Lifshitzequationisintegrablebymeansoftheinversesatteringmethod. In
partiular,thereareexpliitformulaeformulti-solitons. Theybehavelikeasumof
orderedsolitonsas
t → ∞
.(Loading...)
For
c 6=
0,thesolitonu ˇ =
1− u
32 12
exp
i ϕ
isrepresentedinthehydrodynami frameworkbythepairv c = v c , w c
= u
3, ∂ x ϕ ,
where
v c (x) = (
1− c
2)
12osh
((
1− c
2)
12x ,
and
w c (x) = c v c (x )
1
− v c (x )
2.
For
c 6=
0,thesolitonu ˇ =
1− u
32 12
exp
i ϕ
isrepresentedinthehydrodynami frameworkbythepairv c = v c , w c
= u
3, ∂ x ϕ ,
where
v c (x) = (
1− c
2)
12osh
((
1− c
2)
12x ,
and
w c (x) = c v c (x )
1
− v c (x )
2.
Ahainofsolitonsisdenedasaperturbationofasumofsolitons
S c , a , s (x ) = X N
j=
1s j v c j (x − a j ) = X N
j=
1s j v c j (x − a j ), s j w c j (x − a j ) ,
with
a = (a
1, . . . , a N ) ∈ R N
,c = (c
1, . . . , c N ) ∈ (−
1,
1) N
ands = (s
1, . . . , s N ) ∈ {±
1} N
.Theenergyspaeisdenedas
E ( R ) =
m : R → S
2,
s.
t. m ′ ∈ L
2( R )
andm
3∈ L
2( R ) .
Theorem(Zhou-Guo'84,Sulem-Sulem-Bardos'86)
Let
m
0∈ E ( R )
. Thereexistsaglobalsolutionm ∈ L ∞ ( R , E( R ))
to∂ t m + m × (∂ xx m − m
3e
3) =
0,
(LL)withinitialdatum
m
0.Remark: nouniqueness
Theorem(Existeneanduniquenessofsmoothsolutions)
Let
k ≥
1andm
0∈ E ( R )
,with[m
0] ′ ∈ H k ( R )
. Thereexistsauniqueglobalsolutionm : R × R → S
2to (LL)withinitialdatumm
0,suhthatInthehydrodynamiframework, theLandau-Lifshitzenergyisequalto
E (v , w ) =
12
Z
R
(∂ x v )
21
− v
2+ (
1− v
2)w
2+ v
2.
Thenon-vanishingspaeisdenedas
N V( R ) =
v = (v , w ) ∈ H
1( R ) × L
2( R ),
s.
t.
maxx∈R |v(x)| <
1.
Anotheronservedquantityisthemomentum:
P(v , w ) = Z
R
vw.
Theorem1(d.L-Gravejat, 2014)
Let
v
0= (v
0, w
0) ∈ N V( R )
.(i )
ThereexistsanumberT
max>
0andauniquesolutionv ∈ C
0([
0, T
max), N V( R ))
to(HLL)withinitialdatum
v
0 suhthatthereexistsmoothsolutionsv n ∈ C ∞ ( R × [
0, T ])
to(HLL)satisfying
v n → v
inC
0([
0, T ], N V( R )),
asn → ∞,
foranyT < T
max.
For
c 6=
0,thesolitonv c
isaminimizerofthevariationalproblemE
min(p c ) =
infE(v ), v : R → C
s.
t. P(v ) = p c .
Thespeed
c
isrelatedtothemomentump c
throughtheformulap c =
2artan(
1− c)
12c
.
TheEuler-Lagrangeequationisy
E ′ (v c ) = cP ′ (v c ).
E
E = p
2
Theorem2(d.L.-Gravejat 2014)
Let
c
0∈ (−
1,
1) N
,s
0∈ {±
1} N
,andv
0= (v
0, w
0) ∈ N V( R )
. Assumethatc
10< · · · <
0< · · · < c N
0.
Thereexistfournumbers
α ∗ >
0,ν ∗ >
0 ,A ∗ >
0andL ∗ >
0 ,suhthat,ifv
0− S c
0,a
0,s
0H
1×L
2= α
0< α ∗ ,
forpositions
a
0∈ R N
suhthat mina k+
0 1− a
0k ,
1≤ k ≤ N −
1= L
0> L ∗ ,
thenthereexistauniquesolution
v ∈ C
0( R + , N V( R ))
to (HLL)withinitialdatumv
0,aswellas
N
funtionsa k ∈ C
1( R + , R )
,witha k (
0) = a
0k
,suhthata ′ k (t) − c k
0≤ A ∗ α
0+ e −ν ∗ L
0,
Corollary3(d.L-Gravejat2014)
Let
c
0∈ (−
1,
0) ∪ (
0,
1)
,a
0∈ R
,s
0∈ {±
1}
,andv
0∈ N V( R )
. Thereexisttwonumbersα ∗ >
0andA ∗ >
0suhthat,ifv
0− s
0v c
0(· − a
0)
H
1×L
2< α ∗ ,
thenthereexistauniquesolution
v ∈ C
0( R + , N V( R ))
to (HLL)withinitialdatumv
0,aswellasafuntion
a ∈ C
1( R + , R )
,witha(
0) = a
0,suhthata ′ (t) − c
0≤ A ∗ α ∗ ,
and
v(·, t) − s
0v c
0(· − a(t))
H
1×L
2≤ A ∗ α ∗ ,
forany
t ∈ R +
.Itfollowsfromtheexisteneofsmoothsolutionsto(LL).
Lemma
Let
k ≥
4andsetN V k ( R ) =
v = (v , w ) ∈ H k+
1( R ) × H k ( R ),
s.
t.
maxx∈R |v(x)| <
1.
Givenapair
v
0∈ N V k ( R )
,thereexistsanumberT
max>
0andauniquesolutionv ∈ L ∞ ([
0, T
max), N V k ( R ))
to(HLL)withinitialdatumv
0. ThemaximaltimeT
maxis
haraterizedbytheondition
lim
t→T
max maxx∈R |v (x , t)| =
1,
ifT
max< +∞.
Givenasmoothsolution
v = (v , w )
to(HLL),wedenethemapΨ =
12
∂ x v
(
1− v
2)
12+ i (
1− v
2)
12w
exp
i θ,
wherethe phase
θ
isgivenbyθ(x , t) = −
Z x
−∞
v (y , t) w(y , t) dy.
Wealsoset
F (v , Ψ)(x ,t) = Z x
−∞
v(y, t) Ψ(y, t) dy.
Themap
Ψ
issolutiontothenonlinearShrödingerequationi ∂ t Ψ + ∂ xx Ψ+
2|Ψ|
2Ψ +
12
v
2Ψ
−
ReΨ(
1−
2F (v , Ψ))
(
1−
2F (v , Ψ)) =
0,
whilethefuntion
v
satisesthesystemowmaporrespondingtothissystem.
Lemma
Let
T ∗ >
0. Giventwosmoothsolutions(v
1, Ψ
1)
and(v
2, Ψ
2)
totheprevioussystemwithinitialdatum
(v j
0, Ψ
0j )
,thereexistanumberτ >
0 ,dependingonlyonk v j
0k L
2 andkΨ
0j k L
2,andauniversalonstantK
suhthatv
1− v
2C
0([
0,T],L
2) +
Ψ
1− Ψ
2C
0([
0,T],L
2)
≤ K v
10− v
20L
2+ Ψ
01− Ψ
02L
2,
forany
T ∈ [
0,
min{τ, T ∗ }]
.Theorem1derivesfromexpressingthisontinuitypropertyintermsofthepair
v = (v , w )
.ThestrategyissimilartotheonedevelopedbyMartel-Merle-Tsai'02,'05,forthe
Korteweg-deVriesandthenonlinearShrödingerequations(seealso
Béthuel-Gravejat-Smets'12).
Itresultsfromthefollowingquantiationoftheminimizingnatureofasoliton.
Lemma
Let
c ∈ (−
1,
0) ∪ (
0,
1)
,andsetH c = E ′′ (v c ) − cP ′′ (v c )
. ThereexistsanumberΛ c >
0suhthat
h H c (ε), εi L
2≥ Λ c kεk
2H
1×L
2,
foranypair
ε ∈ H
1( R ) × L
2( R )
suhthath∂ x v c , εi L
2= hP ′ (v c ), εi L
2=
0.
When
ε = v − v c
satisesthetwoorthogonalityonditions,wehaveE (v) − cP(v)= E (v c ) − cP(v c ) +
12
hH c (ε), εi L
2+ O kεk
3H
1×L
2Λ
Inordertoguaranteethetwoorthogonalityonditions,weintroduemodulation
parameters. Given
α >
0andL >
0 ,wesetV(α, L) = n
v ∈ N V( R ),
s.
t.
infa k+
1
>a k +L kv − S c ∗ , a , s ∗ k H
1×L
2< α o .
Lemma
Thereexistnumbers
α
1>
0,ν
1>
0andL
1>
0suhthat,givena solutionv ∈ C
0([
0, T ], H
1( R ) × L
2( R ))
to(HLL),withv(·, t) ∈ V(α, L),
forα < α
1andL > L
1,thereexisttwofuntions
a ∈ C
1([
0, T ], R N )
andc ∈ C
1([
0, T ], (−
1,
1) N )
suhthatthemap
ε(·, t) = v(·, t) − S c(t),a(t),s ∗ ,
satisestheorthogonalityonditions
∂ x v c k (t),a k (t) , ε(·, t)
L
2=
P ′ (v c k (t),a k (t) ), ε(·, t)
L
2=
0,
forany1
≤ k ≤ N
andanyt ∈ [
0, T ]
. Moreover,wehavekε(·, t)k H
1×L
2+
X N
k=
|c k (t) − c k
0| = O(α),
For
ν
smallenough,weintroduethefuntionsφ
1=
1,φ N+
1=
0,andφ j (x) =
12
1
+
tanhν
x − a j−
1(t) + a j (t)
2
,
for2
≤ j ≤ N
,andwesetF(v) = E (v) −
X N
j=
1c j
0P j (v),
whereP j (v) = Z
R
φ j − φ j+
1vw.
Lemma
Thereexistnumbers
Λ >
0 ,α
2>
0,ν
2>
0andL
2>
0suhthat,givena solutionv ∈ C
0([
0, T ], H
1( R ) × L
2( R ))
to(HLL),withv(·, t) ∈ V(α, L),
for
α < α
2 andL > L
2,wehavefort ∈ [
0, T ]
:Theonservationlawforthemomentumwritesas
∂ t vw
=−
12
∂ x
v
2+ w
2 1−
3v
2+
3− v
2(
1− v
2)
2(∂ x v )
2−
12
∂ xxx
ln 1− v
2.
Lemma
Thereexistnumbers
α
3>
0,ν
3>
0andL
3>
0suhthat,givena solutionv ∈ C
0([
0, T ], H
1( R ) × L
2( R ))
to(HLL),withv(·, t) ∈ V(α, L),
for
α < α
3 andL > L
3,wehave
F ′ (t) ≤ O
exp(−ν
3(L + t) ,
forany
t ∈ [
0, T ]
.Asymptotistabilityofthesolitons(YakineBahri,PhDThesis).
Construtionandorbitalstabilityofsolitonsinhigherdimension.
Asymptotistabilityinhigherdimension?