EPFL, Seminar of Probability and Stochastic Processes
Regularity structures and renormalisation
of FitzHugh-Nagumo SPDEs
in three space dimensions
Nils Berglund
MAPMO, Université d’Orléans
11 April 2016
with Christian Kuehn (TU Munich)
Nils Berglund
nils.berglund@univ-orleans.fr
http://www.univ-orleans.fr/mapmo/membres/berglund/
Plan
1. Motivation and main result
2. Introduction to regularity structures
3. Extension to FitzHugh–Nagumo
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
0/18
1. FitzHugh–Nagumo SDE
dut = [ut − ut3 + vt ] dt + σ dWt
dvt = ε[a − ut − bvt ] dt
.
ut : membrane potential of neuron
.
vt : gating variable (proportion of open ion channels)
−ut
v
ε = 0.1
b=0
a = √13 + 0.02
σ = 0.03
u
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
t
11 April 2016
1/18
FitzHugh–Nagumo SPDE
∂t u = ∆u + u − u 3 + v + ξ
∂t v = a1 u + a2 v
u = u(t, x) ∈ R, v = v (t, x) ∈ R (or Rn ), (t, x) ∈ D = R+ × Td , d = 2, 3
. ξ(t, x) Gaussian space-time white noise: E ξ(t, x)ξ(s, y ) = δ(t − s)δ(x − y )
.
ξ: distribution defined by hξ, ϕi = Wϕ , {Wh }h∈L2 (D) , E[Wh Wh0 ] = hh, h0 i
(Link to simulation)
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
2/18
Main result
Mollified noise: ξ ε = %ε ∗ ξ 1
where %ε (t, x) = εd+2
% εt2 , xε with % compactly supported, integral 1
Theorem [NB & C. Kuehn, Elec J Probab 21 (18):1-48 (2016)]
There exists a choice of renormalisation constant C (ε), limε→0 C (ε) = ∞,
such that
∂t u ε = ∆u ε + [1 + C (ε)]u ε − (u ε )3 + v ε + ξ ε
∂t v ε = a1 u ε + a2 v ε
admits a sequence of local solutions (u ε , v ε ), converging in probability to a
limit (u, v ) as ε → 0.
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
3/18
Main result
Mollified noise: ξ ε = %ε ∗ ξ 1
where %ε (t, x) = εd+2
% εt2 , xε with % compactly supported, integral 1
Theorem [NB & C. Kuehn, Elec J Probab 21 (18):1-48 (2016)]
There exists a choice of renormalisation constant C (ε), limε→0 C (ε) = ∞,
such that
∂t u ε = ∆u ε + [1 + C (ε)]u ε − (u ε )3 + v ε + ξ ε
∂t v ε = a1 u ε + a2 v ε
admits a sequence of local solutions (u ε , v ε ), converging in probability to a
limit (u, v ) as ε → 0.
Local solution means up to a random possible explosion time
Initial conditions should be in appropriate Hölder spaces
. C (ε) log(ε−1 ) for d = 2 and C (ε) ε−1 for d = 3
. Similar results for more general cubic nonlinearity and v ∈ Rn
.
.
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
3/18
2. Mild solutions of SPDE
∂t u = ∆u + F (u) + ξ
u(0, x) = u0 (x)
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
4/18
2. Mild solutions of SPDE
∂t u = ∆u + F (u) + ξ
u(0, x) = u0 (x)
Construction of mild solution via ZDuhamel formula:
. ∂t u = ∆u
⇒ u(t, x) = G (t, x − y )u0 (y ) dy =: (e∆t u0 )(x)
where G (t, x): heat kernel (compatible with bc)
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
4/18
2. Mild solutions of SPDE
∂t u = ∆u + F (u) + ξ
u(0, x) = u0 (x)
Construction of mild solution via ZDuhamel formula:
. ∂t u = ∆u
⇒ u(t, x) = G (t, x − y )u0 (y ) dy =: (e∆t u0 )(x)
.
where G (t, x): heat kernel (compatible with bc)
Z t
∆t
∂t u = ∆u + f
⇒ u(t, x) = (e u0 )(x) +
e∆(t−s) f (s, ·)(x) ds
0
Notation: u = Gu0 + G ∗ f
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
4/18
2. Mild solutions of SPDE
∂t u = ∆u + F (u) + ξ
u(0, x) = u0 (x)
Construction of mild solution via ZDuhamel formula:
. ∂t u = ∆u
⇒ u(t, x) = G (t, x − y )u0 (y ) dy =: (e∆t u0 )(x)
.
where G (t, x): heat kernel (compatible with bc)
Z t
∆t
∂t u = ∆u + f
⇒ u(t, x) = (e u0 )(x) +
e∆(t−s) f (s, ·)(x) ds
0
Notation: u = Gu0 + G ∗ f
.
∂t u = ∆u + ξ
⇒
u = Gu0 + G ∗ ξ
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
(stochastic convolution)
11 April 2016
4/18
2. Mild solutions of SPDE
∂t u = ∆u + F (u) + ξ
u(0, x) = u0 (x)
Construction of mild solution via ZDuhamel formula:
. ∂t u = ∆u
⇒ u(t, x) = G (t, x − y )u0 (y ) dy =: (e∆t u0 )(x)
.
where G (t, x): heat kernel (compatible with bc)
Z t
∆t
∂t u = ∆u + f
⇒ u(t, x) = (e u0 )(x) +
e∆(t−s) f (s, ·)(x) ds
0
Notation: u = Gu0 + G ∗ f
⇒
.
∂t u = ∆u + ξ
.
∂t u = ∆u + ξ + F (u)
u = Gu0 + G ∗ ξ
⇒
(stochastic convolution)
u = Gu0 + G ∗ [ξ + F (u)]
Aim: use Banach’s fixed-point theorem — but which function space?
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
4/18
Hölder spaces
Definition of C α for f : I → R, with I ⊂ R a compact interval:
.
0 < α < 1: |f (x) − f (y )| 6 C |x − y |α ∀x 6= y
.
α > 1: f ∈ C bαc and f 0 ∈ C α−1
.
α < 0: f distribution, |hf , ηxδ i| 6 C δ α
−bαc
where ηxδ (y ) = 1δ η( x−y
δ ) for all test functions η ∈ C
Property: f ∈ C α , 0 < α < 1
⇒
f 0 ∈ C α−1 where hf 0 , ηi = −hf , η 0 i
Remark: f ∈ C 1+α 6⇒ |f (x) − f (y )| 6 C |x − y |1+α . See e.g f (x) = x + |x|3/2
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
5/18
Hölder spaces
Definition of C α for f : I → R, with I ⊂ R a compact interval:
.
0 < α < 1: |f (x) − f (y )| 6 C |x − y |α ∀x 6= y
.
α > 1: f ∈ C bαc and f 0 ∈ C α−1
.
α < 0: f distribution, |hf , ηxδ i| 6 C δ α
−bαc
where ηxδ (y ) = 1δ η( x−y
δ ) for all test functions η ∈ C
Property: f ∈ C α , 0 < α < 1
⇒
f 0 ∈ C α−1 where hf 0 , ηi = −hf , η 0 i
Remark: f ∈ C 1+α 6⇒ |f (x) − f (y )| 6 C |x − y |1+α . See e.g f (x) = x + |x|3/2
Case of the heat kernel: (∂t − ∆)u = f
u =G ∗f
P
Parabolic scaling Csα : |x − y | −→ |t − s|1/2 + di=1 |xi − yi |
Parabolic scaling Csα :
x−y
1
δ η( δ )
−→
⇒
1
η( t−s
, x−y
δ )
δ2
δ d+2
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
5/18
Schauder estimates and fixed-point equation
Schauder estimate
α∈
/ Z, f ∈ Csα
⇒
G ∗ f ∈ Csα+2
Fact: in dimension d, space-time white noise ξ ∈ Csα a.s. ∀α < − d+2
2
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
6/18
Schauder estimates and fixed-point equation
Schauder estimate
α∈
/ Z, f ∈ Csα
⇒
G ∗ f ∈ Csα+2
Fact: in dimension d, space-time white noise ξ ∈ Csα a.s. ∀α < − d+2
2
Fixed-point equation: u = Gu0 + G ∗ [ξ + F (u)]
−3/2−
1/2−
⇒ G ∗ ξ ∈ Cs
⇒ F (u) defined
.
d = 1: ξ ∈ Cs
.
d = 3: ξ ∈ Cs
.
d = 2: ξ ∈ Cs−2 ⇒ G ∗ ξ ∈ Cs0 ⇒ F (u) not defined
−5/2−
−
−1/2−
⇒ G ∗ ξ ∈ Cs
⇒ F (u) not defined
−
Boundary case, can be treated with Besov spaces
[Da Prato & Debussche 2003]
Why not use mollified noise? Limit ε → 0 does not exist
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
6/18
Regularity structures
Basic idea of Martin Hairer [Inventiones Math. 198, 269–504, 2014]:
Lift mollified fixed-point equation
u = Gu0 + G ∗ [ξ ε + F (u)]
to a larger space called a Regularity structure
M(u0 , Z ε )
SM
RM
MΨ
(u0 , ξ ε )
UM
S̄M
uε
.
u ε = S̄(u0 , ξ ε ): classical solution of mollified equation
.
U = S(u0 , Z ε ): solution map in regularity structure
.
S and R are continuous (in suitable topology)
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
7/18
Regularity structures
Basic idea of Martin Hairer [Inventiones Math. 198, 269–504, 2014]:
Lift mollified fixed-point equation
u = Gu0 + G ∗ [ξ ε + F (u)]
to a larger space called a Regularity structure
(u0 , MZ ε )
SM
RM
MΨ
(u0 , ξ ε )
UM
S̄M
û ε
.
u ε = S̄(u0 , ξ ε ): classical solution of mollified equation
.
U = S(u0 , Z ε ): solution map in regularity structure
.
S and R are continuous (in suitable topology)
.
Renormalisation: modification of the lift Ψ
Aternative approaches for d = 3: [Catellier & Chouk ’13], [Kupiainen ’15]
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
7/18
Regularity structures
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
7/18
Basic idea: Generalised Taylor series
f : I → R, 0 < α < 1
f ∈ C 2+α
⇔
f ∈ C 2 and f 00 ∈ C α
Associate with f the triple (f , f 0 , f 00 )
When does a triple (f0 , f1 , f2 ) represent a function f ∈ C 2+α ?
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
8/18
Basic idea: Generalised Taylor series
f : I → R, 0 < α < 1
f ∈ C 2+α
⇔
f ∈ C 2 and f 00 ∈ C α
Associate with f the triple (f , f 0 , f 00 )
When does a triple (f0 , f1 , f2 ) represent a function f ∈ C 2+α ?
When there is a constant C such that for all x, y ∈ I
|f0 (y ) − f0 (x) − (y − x)f1 (x) − 12 (y − x)2 f2 (x)| 6 C |x − y |2+α
|f1 (y ) − f1 (x) − (y − x)f2 (x)| 6 C |x − y |1+α
|f2 (y ) − f2 (x))| 6 C |x − y |α
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
8/18
Basic idea: Generalised Taylor series
f : I → R, 0 < α < 1
f ∈ C 2+α
⇔
f ∈ C 2 and f 00 ∈ C α
Associate with f the triple (f , f 0 , f 00 )
When does a triple (f0 , f1 , f2 ) represent a function f ∈ C 2+α ?
When there is a constant C such that for all x, y ∈ I
|f0 (y ) − f0 (x) − (y − x)f1 (x) − 12 (y − x)2 f2 (x)| 6 C |x − y |2+α
|f1 (y ) − f1 (x) − (y − x)f2 (x)| 6 C |x − y |1+α
|f2 (y ) − f2 (x))| 6 C |x − y |α
Notation: f = f0 1 + f1 X + f2 X 2
Regularity structure: Generalised Taylor basis whose basis elements can
also be singular distributions
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
8/18
Definition of a regularity structure
Definition [M. Hairer, Inventiones Math 2014]
A Regularity structure is a triple (A, T , G) where
1. Index set: A ⊂ R, bdd below, locally finite, 0 ∈ A
M
2. Model space: T =
Tα , each Tα Banach space, T0 = span(1) ' R
α∈A
3. Structure group: G group of linear maps Γ : T → T such that
M
Γτ − τ ∈
Tβ
∀τ ∈ Tα
and Γ1 = 1 ∀Γ ∈ G.
β<α
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
9/18
Definition of a regularity structure
Definition [M. Hairer, Inventiones Math 2014]
A Regularity structure is a triple (A, T , G) where
1. Index set: A ⊂ R, bdd below, locally finite, 0 ∈ A
M
2. Model space: T =
Tα , each Tα Banach space, T0 = span(1) ' R
α∈A
3. Structure group: G group of linear maps Γ : T → T such that
M
Γτ − τ ∈
Tβ
∀τ ∈ Tα
and Γ1 = 1 ∀Γ ∈ G.
β<α
Polynomial regularity structure on R:
.
A = N0
.
Tk ' R, Tk = span(X k )
.
Γh (X k ) = (X − h)k ∀h ∈ R
Polynomial reg. structure on Rd : X k = X1k1 . . . Xdkd ∈ T|k| , |k| =
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
Pd
i=1 ki
9/18
Regularity structure for ∂t u = ∆u − u 3 + ξ
New symbols: Ξ, representing ξ, Hölder exponent |Ξ|s = α0 = − d+2
2 −κ
New symbols: I(τ ), representing G ∗ f , Hölder exponent |I(τ )|s = |τ |s + 2
New symbols: τ σ, Hölder exponent |τ σ|s = |τ |s + |σ|s
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
10/18
Regularity structure for ∂t u = ∆u − u 3 + ξ
New symbols: Ξ, representing ξ, Hölder exponent |Ξ|s = α0 = − d+2
2 −κ
New symbols: I(τ ), representing G ∗ f , Hölder exponent |I(τ )|s = |τ |s + 2
New symbols: τ σ, Hölder exponent |τ σ|s = |τ |s + |σ|s
|τ |s
α0
3α0 + 6
2α0 + 4
d =3
− 52 − κ
− 32 − 3κ
−1 − 2κ
d =2
−2 − κ
0 − 3κ
0 − 2κ
I(I(Ξ)3 )I(Ξ)2
I(Ξ)
5α0 + 12
α0 + 2
− 12 − 5κ
− 12 − κ
2 − 5κ
0−κ
I(I(Ξ)3 )I(Ξ)
I(I(Ξ)2 )I(Ξ)2
I(Ξ)2 Xi
1
4α0 + 10
4α0 + 10
2α0 + 5
0
0 − 4κ
0 − 4κ
0 − 2κ
0
2 − 4κ
2 − 4κ
1 − 2κ
0
3α0 + 8
...
1
2
2 − 3κ
...
τ
Ξ
I(Ξ)3
I(Ξ)2
I(I(Ξ)3 )
...
Symbol
Ξ
Xi
1
...
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
− 3κ
...
11 April 2016
10/18
Fixed-point equation for ∂t u = ∆u − u 3 + ξ
u = G ∗ [ξ ε − u 3 ] + Gu0
U = I(Ξ − U 3 ) + ϕ1 + . . .
⇒
U0 = 0
U1 =
+ ϕ1
U2 =
+ ϕ1 −
− 3ϕ
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
+ ...
11 April 2016
11/18
Fixed-point equation for ∂t u = ∆u − u 3 + ξ
u = G ∗ [ξ ε − u 3 ] + Gu0
U = I(Ξ − U 3 ) + ϕ1 + . . .
⇒
U0 = 0
U1 =
+ ϕ1
U2 =
+ ϕ1 −
− 3ϕ
+ ...
To prove convergence, we need
.
A model (Π, Γ): ∀z ∈ Rd+1 , Πz τ is distribution describing τ near z
Γzz̄ ∈ G describes translations: Πz̄ = Πz Γzz̄
.
Spacesnof modelled distributions
o
M
Dγ = f : Rd+1 →
Tβ : kf (z) − Γzz̄ f (z̄)kβ . kz − z̄kγ−β
s
β<γ
equipped with a seminorm
.
The Reconstruction theorem: provides a unique map R : Dγ → Csα∗
δ i| . δ γ
(α∗ = inf A) s.t. |hRf − Πz f (z), ηs,z
(constructed using wavelets)
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
11/18
Canonical model Z ε = (Πε , Γε )
Defined inductively by
(Πεz Ξ)(z̄) = ξ ε (z̄)
(Πεz X k )(z̄) = (z̄ − z)k
(Πεz τ σ)(z̄) = (Πεz τ )(z̄)(Πεz σ)(z̄)
Z
(Πεz I(τ ))(z̄) = G (z̄ − z 0 )(Πεz τ )(z 0 ) dz 0 − polynomial term
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
12/18
Canonical model Z ε = (Πε , Γε )
Defined inductively by
(Πεz Ξ)(z̄) = ξ ε (z̄)
(Πεz X k )(z̄) = (z̄ − z)k
(Πεz τ σ)(z̄) = (Πεz τ )(z̄)(Πεz σ)(z̄)
Z
(Πεz I(τ ))(z̄) = G (z̄ − z 0 )(Πεz τ )(z 0 ) dz 0 − polynomial term
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
12/18
Canonical model Z ε = (Πε , Γε )
Defined inductively by
(Πεz Ξ)(z̄) = ξ ε (z̄)
(Πεz X k )(z̄) = (z̄ − z)k
(Πεz τ σ)(z̄) = (Πεz τ )(z̄)(Πεz σ)(z̄)
Z
(Πεz I(τ ))(z̄) = G (z̄ − z 0 )(Πεz τ )(z 0 ) dz 0 − polynomial term
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
12/18
Canonical model Z ε = (Πε , Γε )
Defined inductively by
(Πεz Ξ)(z̄) = ξ ε (z̄)
(Πεz X k )(z̄) = (z̄ − z)k
(Πεz τ σ)(z̄) = (Πεz τ )(z̄)(Πεz σ)(z̄)
Z
(Πεz I(τ ))(z̄) = G (z̄ − z 0 )(Πεz τ )(z 0 ) dz 0 − polynomial term
Then ∃K s.t. RKf = G ∗ Rf and the following diagrams commute:
Dγ
R
Csα∗
K
Dγ+2
R
Csα∗ +2
(u0 , Z ε )
S
R
U
R
(u0 , ξ ε )
uε
G∗
S̄
where α∗ = inf A and Kf = If + polynomial term + nonlocal term
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
12/18
Why do we need to renormalise?
Let Gε = G ∗ %ε where %ε is the mollifier
(Πεz̄
ε
)(z) = (G ∗ ξ )(z) = (Gε ∗ ξ)(z) =
Z
Gε (z − z1 )ξ(z1 ) dz1
belongs to first Wiener chaos, limit ε → 0 well-defined
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
13/18
Why do we need to renormalise?
Let Gε = G ∗ %ε where %ε is the mollifier
(Πεz̄
ε
)(z) = (G ∗ ξ )(z) = (Gε ∗ ξ)(z) =
Z
Gε (z − z1 )ξ(z1 ) dz1
belongs to first Wiener chaos, limit ε → 0 well-defined
ZZ
ε
2
ε
(Πz̄ )(z) = (G ∗ ξ )(z) =
Gε (z − z1 )Gε (z − z2 )ξ(z1 )ξ(z2 ) dz1 dz2
diverges as ε → 0
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
13/18
Why do we need to renormalise?
Let Gε = G ∗ %ε where %ε is the mollifier
(Πεz̄
Z
ε
)(z) = (G ∗ ξ )(z) = (Gε ∗ ξ)(z) =
Gε (z − z1 )ξ(z1 ) dz1
belongs to first Wiener chaos, limit ε → 0 well-defined
ZZ
ε
2
ε
(Πz̄ )(z) = (G ∗ ξ )(z) =
Gε (z − z1 )Gε (z − z2 )ξ(z1 )ξ(z2 ) dz1 dz2
diverges as ε → 0
Wick product: ξ(z1 ) ξ(z2 ) = ξ(z1 )ξ(z2 ) − δ(z1 − z2 )
(Πεz̄
ZZ
)(z) =
Z
Gε (z − z1 )Gε (z − z2 )ξ(z1 ) ξ(z2 ) dz1 dz2 +
´¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¸ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¶
in 2nd Wiener chaos, bdd
bε
Renormalised model: (Π
z̄
)(z) = (Πεz̄
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
Gε (z − z1 )2 dz1
´¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹¸ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¶
C1 (ε)→∞
)(z) − C1 (ε)
11 April 2016
13/18
3. The case of the FitzHugh–Nagumo equations
Fixed-point equation
u(t, x) = G ∗ [ξ ε + u − u 3 + v ](t, x) + Gu0 (t, x)
Z t
v (t, x) =
u(s, x) e(t−s)a2 a1 ds + eta2 v0
0
Lifted version
U = I[Ξ + U − U 3 + V ] + Gu0
V = EU + Qv0
where E is an integration map which is not regularising in space
New symbols E(I(Ξ)) = , etc. . .
We expect U, and thus also V to be α-Hölder for α < − 12
Thus I(U − U 3 + V ) should be well-defined
The standard theory has to be extended, because E does not correspond
to a smooth kernel
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
14/18
Details on implementing E
Problems:
.
.
Fixed-point equation requires diagonal identity (Πt,x τ )(t, x) = 0
Usual definition of K would contain Taylor series
X Xk Z
D k G (z − z̄)(Πz τ )(dz̄)
J (z)τ =
k!
|k|s <α
X Xk Z
N f (z) =
D k G (z − z̄)(Rf − Πz f (z))(dz̄)
k!
|k|s <γ
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
15/18
Details on implementing E
Problems:
.
.
Fixed-point equation requires diagonal identity (Πt,x τ )(t, x) = 0
Usual definition of K would contain Taylor series
X Xk Z
D k G (z − z̄)(Πz τ )(dz̄)
J (z)τ =
k!
|k|s <α
X Xk Z
N f (z) =
D k G (z − z̄)(Rf − Πz f (z))(dz̄)
k!
|k|s <γ
Solution:
.
.
.
Define ΠEτ only if −2 < |τ |s < 0 (otherwise Eτ = 0) ⇒ J (z)τ = 0
P
P
Define K only for f = |τ |s <0 cτ τ + |τ |s >0 cτ (t, x)τ =: f− + f+
⇒ can take Rf = Πt,x f (t, x) and thus N f = 0 for these f
Time-convolution with Q lifted to
(KQ f )(t, x) =
X
|τ |s <0
cτ Eτ +
X Z
Q(t−s)cτ (s, x) ds τ =: (Ef− +Qf+ )(t, x)
|τ |s >0
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
Conclusion
11 April 2016
15/18
Fixed-point equation
Consider ∂t u = ∆u + F (u, v ) + ξ with F a polynomial of degree 3
If (U, V ) satisfies fixed-point equation
U = I[Ξ + F (U, V )] + Gu0 + polynomial term
V = EU− + QU+ + Qv0
then (RU, RV ) is solution, provided RF (U, V ) = F (RU, RV )
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
16/18
Fixed-point equation
Consider ∂t u = ∆u + F (u, v ) + ξ with F a polynomial of degree 3
If (U, V ) satisfies fixed-point equation
U = I[Ξ + F (U, V )] + Gu0 + polynomial term
V = EU− + QU+ + Qv0
then (RU, RV ) is solution, provided RF (U, V ) = F (RU, RV )
Fixed point is of the form
U=
V =
+ ϕ1 + a1
+ ψ1 + â1
+ a2
+ â2
+ a3
+ â3
+ a4
+ â4
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
+ b1
+ b̂1
+ b2
+ b̂2
11 April 2016
+ b3
+ ...
+ b̂3
+ ...
16/18
Fixed-point equation
Consider ∂t u = ∆u + F (u, v ) + ξ with F a polynomial of degree 3
If (U, V ) satisfies fixed-point equation
U = I[Ξ + F (U, V )] + Gu0 + polynomial term
V = EU− + QU+ + Qv0
then (RU, RV ) is solution, provided RF (U, V ) = F (RU, RV )
Fixed point is of the form
U=
V =
+ ϕ1 + a1
+ ψ1 + â1
+ a2
+ â2
+ a3
+ â3
+ a4
+ â4
+ b1
+ b̂1
+ b2
+ b̂2
+ b3
+ ...
+ b̂3
+ ...
.
Prove existence of fixed point in (modification of) Dγ with γ = 1 + κ̄
.
Extend from small interval [0, T ] up to first exit from large ball
.
Deal with renormalisation procedure
Conclusion
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
16/18
Renormalisation
.
Renormalisation group: group of linear maps M : T → T
M
Associated model: ΠM
z s.t. Π τ = ΠMτ where Πz = ΠΓfz
Allen–Cahn eq.: M = e−C1 L1 −C2 L2 with L1 :
→ 1, L2 :
FHN eq.: the same group suffices because Q is smoothing
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
→1
17/18
Renormalisation
.
Renormalisation group: group of linear maps M : T → T
M
Associated model: ΠM
z s.t. Π τ = ΠMτ where Πz = ΠΓfz
Allen–Cahn eq.: M = e−C1 L1 −C2 L2 with L1 :
→ 1, L2 :
FHN eq.: the same group suffices because Q is smoothing
.
→1
(ε) Mε
b z τ s.t. if Π
b (ε)
Look for r.v. Π
then ∃κ, θ > 0 s.t.
z = (Πz )
λ 2
2θ 2|τ |s +κ
b z τ, η λ i2 . λ2|τ |s +κ
bz τ − Π
b (ε)
EhΠ
EhΠ
z τ, ηz i . ε λ
z
p
b (ε)
b(ε)
Then (Π
z , Γz ) converges to limiting model, with explicit L bounds
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
17/18
Renormalisation
.
Renormalisation group: group of linear maps M : T → T
M
Associated model: ΠM
z s.t. Π τ = ΠMτ where Πz = ΠΓfz
Allen–Cahn eq.: M = e−C1 L1 −C2 L2 with L1 :
→ 1, L2 :
FHN eq.: the same group suffices because Q is smoothing
.
→1
(ε) Mε
b z τ s.t. if Π
b (ε)
Look for r.v. Π
then ∃κ, θ > 0 s.t.
z = (Πz )
λ 2
2θ 2|τ |s +κ
b z τ, η λ i2 . λ2|τ |s +κ
bz τ − Π
b (ε)
EhΠ
EhΠ
z τ, ηz i . ε λ
z
p
b (ε)
b(ε)
Then (Π
z , Γz ) converges to limiting model, with explicit L bounds
.
Renormalised equations have nonlinearity Fb s.t.
Fb (MU, MV ) = MF (U, V ) + terms of Hölder exponent > 0
FHN eq. with cubic nonlinearity
F = α1 u + α2 v + β1 u 2 + β2 uv + β3 v 2 + γ1 u 3 + γ2 u 2 v + γ3 uv 2 + γ4 v 3
Fb (u, v ) = F (u, v ) − c0 (ε) − c1 (ε)u − c2 (ε)v
with the ci (ε) depending on C1 , C2 , provided either d = 2 or γ2 = 0
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
17/18
Concluding remarks
.
Models with ∂t u of order u 4 + v 4 and ∂t v of order u 2 + v should be
renormalisable
Current approach does not work when singular part (t, x)-dependent
.
Global existence: recent progress by J.-C. Mourrat and H. Weber on
2D and 3D Allen–Cahn
.
More quantitative results?
References
.
Martin Hairer, A theory of regularity structures, Invent. Math. 198 (2), pp
269–504 (2014)
.
Martin Hairer, Introduction to Regularity Structures, lecture notes (2013)
.
Ajay Chandra, Hendrik Weber, Stochastic PDEs, regularity structures, and
interacting particle systems, preprint arXiv/1508.03616
.
N. B., Christian Kuehn, Regularity structures and renormalisation of
FitzHugh-Nagumo SPDEs in three space dimensions, Elec J Probab 21
(18):1-48 (2016)
.
N. B., Mayonnaise et élections américaines, Dossier Pour la Science (2016)
Regularity structures and renormalisation of FitzHugh-Nagumo SPDEs
11 April 2016
18/18