Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th...

102
Brownian couplings and applications Mihai N. Pascu Transilvania University of Bra¸ sov, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland, September 10 – 14, 2012 M. N. Pascu (Transilvania Univ) Couplings of RBM 1/31 14.09.2012 1 / 31

Transcript of Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th...

Page 1: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Brownian couplings and applications

Mihai N. Pascu

Transilvania University of Brasov, Romania

6th International Conference on Stochastic Analysis and Its Applications

Bedlewo, Poland, September 10 – 14, 2012

M. N. Pascu (Transilvania Univ) Couplings of RBM 1/31 14.09.2012 1 / 31

Page 2: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Plan of the talk

The method of coupling of reflecting Brownian motions (RBM) is a usefultechnique for proving results on various functionals associated to RBM.

In this talk, we will present two such couplings: the scaling coupling and themirror coupling.

As an application of the scaling coupling, we will prove a monotonicity of thelifetime of reflecting Brownian motion with killing, which implies the validityof the Hot Spots conjecture of J. Rauch for a certain class of domains.

As applications of the mirror coupling, we will present the proof of theLaugesen-Morpurgo conjecture, and a unifying proof of the results of I.Chavel and W. Kendall on Chavel’s conjecture.

Time-permitting, I will discuss some recent results on translation couplingand its applications.

M. N. Pascu (Transilvania Univ) Couplings of RBM 2/31 14.09.2012 2 / 31

Page 3: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Plan of the talk

The method of coupling of reflecting Brownian motions (RBM) is a usefultechnique for proving results on various functionals associated to RBM.

In this talk, we will present two such couplings: the scaling coupling and themirror coupling.

As an application of the scaling coupling, we will prove a monotonicity of thelifetime of reflecting Brownian motion with killing, which implies the validityof the Hot Spots conjecture of J. Rauch for a certain class of domains.

As applications of the mirror coupling, we will present the proof of theLaugesen-Morpurgo conjecture, and a unifying proof of the results of I.Chavel and W. Kendall on Chavel’s conjecture.

Time-permitting, I will discuss some recent results on translation couplingand its applications.

M. N. Pascu (Transilvania Univ) Couplings of RBM 2/31 14.09.2012 2 / 31

Page 4: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Plan of the talk

The method of coupling of reflecting Brownian motions (RBM) is a usefultechnique for proving results on various functionals associated to RBM.

In this talk, we will present two such couplings: the scaling coupling and themirror coupling.

As an application of the scaling coupling, we will prove a monotonicity of thelifetime of reflecting Brownian motion with killing, which implies the validityof the Hot Spots conjecture of J. Rauch for a certain class of domains.

As applications of the mirror coupling, we will present the proof of theLaugesen-Morpurgo conjecture, and a unifying proof of the results of I.Chavel and W. Kendall on Chavel’s conjecture.

Time-permitting, I will discuss some recent results on translation couplingand its applications.

M. N. Pascu (Transilvania Univ) Couplings of RBM 2/31 14.09.2012 2 / 31

Page 5: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Plan of the talk

The method of coupling of reflecting Brownian motions (RBM) is a usefultechnique for proving results on various functionals associated to RBM.

In this talk, we will present two such couplings: the scaling coupling and themirror coupling.

As an application of the scaling coupling, we will prove a monotonicity of thelifetime of reflecting Brownian motion with killing, which implies the validityof the Hot Spots conjecture of J. Rauch for a certain class of domains.

As applications of the mirror coupling, we will present the proof of theLaugesen-Morpurgo conjecture, and a unifying proof of the results of I.Chavel and W. Kendall on Chavel’s conjecture.

Time-permitting, I will discuss some recent results on translation couplingand its applications.

M. N. Pascu (Transilvania Univ) Couplings of RBM 2/31 14.09.2012 2 / 31

Page 6: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Plan of the talk

The method of coupling of reflecting Brownian motions (RBM) is a usefultechnique for proving results on various functionals associated to RBM.

In this talk, we will present two such couplings: the scaling coupling and themirror coupling.

As an application of the scaling coupling, we will prove a monotonicity of thelifetime of reflecting Brownian motion with killing, which implies the validityof the Hot Spots conjecture of J. Rauch for a certain class of domains.

As applications of the mirror coupling, we will present the proof of theLaugesen-Morpurgo conjecture, and a unifying proof of the results of I.Chavel and W. Kendall on Chavel’s conjecture.

Time-permitting, I will discuss some recent results on translation couplingand its applications.

M. N. Pascu (Transilvania Univ) Couplings of RBM 2/31 14.09.2012 2 / 31

Page 7: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

– Where should I put the bed, to keep warm in the long run?

M. N. Pascu (Transilvania Univ) Couplings of RBM 3/31 14.09.2012 3 / 31

Page 8: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Heuristics

Consider u (t, x) the solution of the Neumann heat equation in a smoothbounded domain D ⊂ Rd with generic initial condition u0.Let x+t be the hot spot at time t and x−t be the cold spot, i.e.

u(t, x+t

)= max

x∈Du (t, x) and u

(t, x−t

)= min

x∈Du (t, x)

If the second Neumann eigenvalue λ2 is simple, and ϕ2 is a correspondingsecond Neumann eigenfunction, for large t we have

u (t, x) =∫

Du0 + e−λ2tϕ2 (x)

∫D

u0ϕ2 + R2(t, x) ≈ c0 + c1e−λ2tϕ2 (x) ,

so x+t and x−t are close to the maximum/minimum points of ϕ2.Hot spots (x+t ) and cold spots (x−t ) repel each other, so the distance betweenthem tends to increase wrt t. In convex domains, the maximum distance isattained for points on the boundary.Together with the above, this suggests the following.

M. N. Pascu (Transilvania Univ) Couplings of RBM 4/31 14.09.2012 4 / 31

Page 9: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Heuristics

Consider u (t, x) the solution of the Neumann heat equation in a smoothbounded domain D ⊂ Rd with generic initial condition u0.Let x+t be the hot spot at time t and x−t be the cold spot, i.e.

u(t, x+t

)= max

x∈Du (t, x) and u

(t, x−t

)= min

x∈Du (t, x)

If the second Neumann eigenvalue λ2 is simple, and ϕ2 is a correspondingsecond Neumann eigenfunction, for large t we have

u (t, x) =∫

Du0 + e−λ2tϕ2 (x)

∫D

u0ϕ2 + R2(t, x) ≈ c0 + c1e−λ2tϕ2 (x) ,

so x+t and x−t are close to the maximum/minimum points of ϕ2.Hot spots (x+t ) and cold spots (x−t ) repel each other, so the distance betweenthem tends to increase wrt t. In convex domains, the maximum distance isattained for points on the boundary.Together with the above, this suggests the following.

M. N. Pascu (Transilvania Univ) Couplings of RBM 4/31 14.09.2012 4 / 31

Page 10: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Heuristics

Consider u (t, x) the solution of the Neumann heat equation in a smoothbounded domain D ⊂ Rd with generic initial condition u0.Let x+t be the hot spot at time t and x−t be the cold spot, i.e.

u(t, x+t

)= max

x∈Du (t, x) and u

(t, x−t

)= min

x∈Du (t, x)

If the second Neumann eigenvalue λ2 is simple, and ϕ2 is a correspondingsecond Neumann eigenfunction, for large t we have

u (t, x) =∫

Du0 + e−λ2tϕ2 (x)

∫D

u0ϕ2 + R2(t, x) ≈ c0 + c1e−λ2tϕ2 (x) ,

so x+t and x−t are close to the maximum/minimum points of ϕ2.Hot spots (x+t ) and cold spots (x−t ) repel each other, so the distance betweenthem tends to increase wrt t. In convex domains, the maximum distance isattained for points on the boundary.Together with the above, this suggests the following.

M. N. Pascu (Transilvania Univ) Couplings of RBM 4/31 14.09.2012 4 / 31

Page 11: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Heuristics

Consider u (t, x) the solution of the Neumann heat equation in a smoothbounded domain D ⊂ Rd with generic initial condition u0.Let x+t be the hot spot at time t and x−t be the cold spot, i.e.

u(t, x+t

)= max

x∈Du (t, x) and u

(t, x−t

)= min

x∈Du (t, x)

If the second Neumann eigenvalue λ2 is simple, and ϕ2 is a correspondingsecond Neumann eigenfunction, for large t we have

u (t, x) =∫

Du0 + e−λ2tϕ2 (x)

∫D

u0ϕ2 + R2(t, x) ≈ c0 + c1e−λ2tϕ2 (x) ,

so x+t and x−t are close to the maximum/minimum points of ϕ2.Hot spots (x+t ) and cold spots (x−t ) repel each other, so the distance betweenthem tends to increase wrt t. In convex domains, the maximum distance isattained for points on the boundary.Together with the above, this suggests the following.

M. N. Pascu (Transilvania Univ) Couplings of RBM 4/31 14.09.2012 4 / 31

Page 12: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Heuristics

Consider u (t, x) the solution of the Neumann heat equation in a smoothbounded domain D ⊂ Rd with generic initial condition u0.Let x+t be the hot spot at time t and x−t be the cold spot, i.e.

u(t, x+t

)= max

x∈Du (t, x) and u

(t, x−t

)= min

x∈Du (t, x)

If the second Neumann eigenvalue λ2 is simple, and ϕ2 is a correspondingsecond Neumann eigenfunction, for large t we have

u (t, x) =∫

Du0 + e−λ2tϕ2 (x)

∫D

u0ϕ2 + R2(t, x) ≈ c0 + c1e−λ2tϕ2 (x) ,

so x+t and x−t are close to the maximum/minimum points of ϕ2.Hot spots (x+t ) and cold spots (x−t ) repel each other, so the distance betweenthem tends to increase wrt t. In convex domains, the maximum distance isattained for points on the boundary.Together with the above, this suggests the following.

M. N. Pascu (Transilvania Univ) Couplings of RBM 4/31 14.09.2012 4 / 31

Page 13: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Hot Spots conjecture (Jeffrey Rauch, 1974)

Conjecture 1 (Hot Spots conjecture)Maxima and minima of second Neumann eigenfunctions of convex boundeddomains are attained (only) on the boundary of the domain.

B. Kawohl: true for balls, annuli, parallelipipeds in Rd

K. Burdzy and W. Werner: counterexample (non-convex domain): minimum inside the domain,maximum on the boundary

R. Bass and K. Burdzy: stronger counterexample (non-convex domain): both minimum andmaximum inside the domain

D. Jerisson and N. Nadirashvili: true if the domain has two orthogonal axis of symmetry

R. Bañuelos and K. Burdzy: true if there are two orthogonal axes of symmetry, or just one axis ofsymmetry and ϕ2 symmetric wrt it

MNP: true if there are two orthogonal axes of symmetry, or just one axis of symmetry and ϕ2antisymmetric wrt it, or ... (some condition on the nodal set of ϕ2)

Other results: true for obtuse triangles, for some some doubly connected domains, for nearlycircular domains.

HS still open in its full generality! (e.g., proof for acute triangles?...)

M. N. Pascu (Transilvania Univ) Couplings of RBM 5/31 14.09.2012 5 / 31

Page 14: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Hot Spots conjecture (Jeffrey Rauch, 1974)

Conjecture 1 (Hot Spots conjecture)Maxima and minima of second Neumann eigenfunctions of convex boundeddomains are attained (only) on the boundary of the domain.

B. Kawohl: true for balls, annuli, parallelipipeds in Rd

K. Burdzy and W. Werner: counterexample (non-convex domain): minimum inside the domain,maximum on the boundary

R. Bass and K. Burdzy: stronger counterexample (non-convex domain): both minimum andmaximum inside the domain

D. Jerisson and N. Nadirashvili: true if the domain has two orthogonal axis of symmetry

R. Bañuelos and K. Burdzy: true if there are two orthogonal axes of symmetry, or just one axis ofsymmetry and ϕ2 symmetric wrt it

MNP: true if there are two orthogonal axes of symmetry, or just one axis of symmetry and ϕ2antisymmetric wrt it, or ... (some condition on the nodal set of ϕ2)

Other results: true for obtuse triangles, for some some doubly connected domains, for nearlycircular domains.

HS still open in its full generality! (e.g., proof for acute triangles?...)

M. N. Pascu (Transilvania Univ) Couplings of RBM 5/31 14.09.2012 5 / 31

Page 15: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Hot Spots conjecture (Jeffrey Rauch, 1974)

Conjecture 1 (Hot Spots conjecture)Maxima and minima of second Neumann eigenfunctions of convex boundeddomains are attained (only) on the boundary of the domain.

B. Kawohl: true for balls, annuli, parallelipipeds in Rd

K. Burdzy and W. Werner: counterexample (non-convex domain): minimum inside the domain,maximum on the boundary

R. Bass and K. Burdzy: stronger counterexample (non-convex domain): both minimum andmaximum inside the domain

D. Jerisson and N. Nadirashvili: true if the domain has two orthogonal axis of symmetry

R. Bañuelos and K. Burdzy: true if there are two orthogonal axes of symmetry, or just one axis ofsymmetry and ϕ2 symmetric wrt it

MNP: true if there are two orthogonal axes of symmetry, or just one axis of symmetry and ϕ2antisymmetric wrt it, or ... (some condition on the nodal set of ϕ2)

Other results: true for obtuse triangles, for some some doubly connected domains, for nearlycircular domains.

HS still open in its full generality! (e.g., proof for acute triangles?...)

M. N. Pascu (Transilvania Univ) Couplings of RBM 5/31 14.09.2012 5 / 31

Page 16: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Hot Spots conjecture (Jeffrey Rauch, 1974)

Conjecture 1 (Hot Spots conjecture)Maxima and minima of second Neumann eigenfunctions of convex boundeddomains are attained (only) on the boundary of the domain.

B. Kawohl: true for balls, annuli, parallelipipeds in Rd

K. Burdzy and W. Werner: counterexample (non-convex domain): minimum inside the domain,maximum on the boundary

R. Bass and K. Burdzy: stronger counterexample (non-convex domain): both minimum andmaximum inside the domain

D. Jerisson and N. Nadirashvili: true if the domain has two orthogonal axis of symmetry

R. Bañuelos and K. Burdzy: true if there are two orthogonal axes of symmetry, or just one axis ofsymmetry and ϕ2 symmetric wrt it

MNP: true if there are two orthogonal axes of symmetry, or just one axis of symmetry and ϕ2antisymmetric wrt it, or ... (some condition on the nodal set of ϕ2)

Other results: true for obtuse triangles, for some some doubly connected domains, for nearlycircular domains.

HS still open in its full generality! (e.g., proof for acute triangles?...)

M. N. Pascu (Transilvania Univ) Couplings of RBM 5/31 14.09.2012 5 / 31

Page 17: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Hot Spots conjecture (Jeffrey Rauch, 1974)

Conjecture 1 (Hot Spots conjecture)Maxima and minima of second Neumann eigenfunctions of convex boundeddomains are attained (only) on the boundary of the domain.

B. Kawohl: true for balls, annuli, parallelipipeds in Rd

K. Burdzy and W. Werner: counterexample (non-convex domain): minimum inside the domain,maximum on the boundary

R. Bass and K. Burdzy: stronger counterexample (non-convex domain): both minimum andmaximum inside the domain

D. Jerisson and N. Nadirashvili: true if the domain has two orthogonal axis of symmetry

R. Bañuelos and K. Burdzy: true if there are two orthogonal axes of symmetry, or just one axis ofsymmetry and ϕ2 symmetric wrt it

MNP: true if there are two orthogonal axes of symmetry, or just one axis of symmetry and ϕ2antisymmetric wrt it, or ... (some condition on the nodal set of ϕ2)

Other results: true for obtuse triangles, for some some doubly connected domains, for nearlycircular domains.

HS still open in its full generality! (e.g., proof for acute triangles?...)

M. N. Pascu (Transilvania Univ) Couplings of RBM 5/31 14.09.2012 5 / 31

Page 18: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Hot Spots conjecture (Jeffrey Rauch, 1974)

Conjecture 1 (Hot Spots conjecture)Maxima and minima of second Neumann eigenfunctions of convex boundeddomains are attained (only) on the boundary of the domain.

B. Kawohl: true for balls, annuli, parallelipipeds in Rd

K. Burdzy and W. Werner: counterexample (non-convex domain): minimum inside the domain,maximum on the boundary

R. Bass and K. Burdzy: stronger counterexample (non-convex domain): both minimum andmaximum inside the domain

D. Jerisson and N. Nadirashvili: true if the domain has two orthogonal axis of symmetry

R. Bañuelos and K. Burdzy: true if there are two orthogonal axes of symmetry, or just one axis ofsymmetry and ϕ2 symmetric wrt it

MNP: true if there are two orthogonal axes of symmetry, or just one axis of symmetry and ϕ2antisymmetric wrt it, or ... (some condition on the nodal set of ϕ2)

Other results: true for obtuse triangles, for some some doubly connected domains, for nearlycircular domains.

HS still open in its full generality! (e.g., proof for acute triangles?...)

M. N. Pascu (Transilvania Univ) Couplings of RBM 5/31 14.09.2012 5 / 31

Page 19: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Hot Spots conjecture (Jeffrey Rauch, 1974)

Conjecture 1 (Hot Spots conjecture)Maxima and minima of second Neumann eigenfunctions of convex boundeddomains are attained (only) on the boundary of the domain.

B. Kawohl: true for balls, annuli, parallelipipeds in Rd

K. Burdzy and W. Werner: counterexample (non-convex domain): minimum inside the domain,maximum on the boundary

R. Bass and K. Burdzy: stronger counterexample (non-convex domain): both minimum andmaximum inside the domain

D. Jerisson and N. Nadirashvili: true if the domain has two orthogonal axis of symmetry

R. Bañuelos and K. Burdzy: true if there are two orthogonal axes of symmetry, or just one axis ofsymmetry and ϕ2 symmetric wrt it

MNP: true if there are two orthogonal axes of symmetry, or just one axis of symmetry and ϕ2antisymmetric wrt it, or ... (some condition on the nodal set of ϕ2)

Other results: true for obtuse triangles, for some some doubly connected domains, for nearlycircular domains.

HS still open in its full generality! (e.g., proof for acute triangles?...)

M. N. Pascu (Transilvania Univ) Couplings of RBM 5/31 14.09.2012 5 / 31

Page 20: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Hot Spots conjecture (Jeffrey Rauch, 1974)

Conjecture 1 (Hot Spots conjecture)Maxima and minima of second Neumann eigenfunctions of convex boundeddomains are attained (only) on the boundary of the domain.

B. Kawohl: true for balls, annuli, parallelipipeds in Rd

K. Burdzy and W. Werner: counterexample (non-convex domain): minimum inside the domain,maximum on the boundary

R. Bass and K. Burdzy: stronger counterexample (non-convex domain): both minimum andmaximum inside the domain

D. Jerisson and N. Nadirashvili: true if the domain has two orthogonal axis of symmetry

R. Bañuelos and K. Burdzy: true if there are two orthogonal axes of symmetry, or just one axis ofsymmetry and ϕ2 symmetric wrt it

MNP: true if there are two orthogonal axes of symmetry, or just one axis of symmetry and ϕ2antisymmetric wrt it, or ... (some condition on the nodal set of ϕ2)

Other results: true for obtuse triangles, for some some doubly connected domains, for nearlycircular domains.

HS still open in its full generality! (e.g., proof for acute triangles?...)

M. N. Pascu (Transilvania Univ) Couplings of RBM 5/31 14.09.2012 5 / 31

Page 21: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Hot Spots conjecture (Jeffrey Rauch, 1974)

Conjecture 1 (Hot Spots conjecture)Maxima and minima of second Neumann eigenfunctions of convex boundeddomains are attained (only) on the boundary of the domain.

B. Kawohl: true for balls, annuli, parallelipipeds in Rd

K. Burdzy and W. Werner: counterexample (non-convex domain): minimum inside the domain,maximum on the boundary

R. Bass and K. Burdzy: stronger counterexample (non-convex domain): both minimum andmaximum inside the domain

D. Jerisson and N. Nadirashvili: true if the domain has two orthogonal axis of symmetry

R. Bañuelos and K. Burdzy: true if there are two orthogonal axes of symmetry, or just one axis ofsymmetry and ϕ2 symmetric wrt it

MNP: true if there are two orthogonal axes of symmetry, or just one axis of symmetry and ϕ2antisymmetric wrt it, or ... (some condition on the nodal set of ϕ2)

Other results: true for obtuse triangles, for some some doubly connected domains, for nearlycircular domains.

HS still open in its full generality! (e.g., proof for acute triangles?...)

M. N. Pascu (Transilvania Univ) Couplings of RBM 5/31 14.09.2012 5 / 31

Page 22: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Hot Spots conjecture (Jeffrey Rauch, 1974)

Conjecture 1 (Hot Spots conjecture)Maxima and minima of second Neumann eigenfunctions of convex boundeddomains are attained (only) on the boundary of the domain.

B. Kawohl: true for balls, annuli, parallelipipeds in Rd

K. Burdzy and W. Werner: counterexample (non-convex domain): minimum inside the domain,maximum on the boundary

R. Bass and K. Burdzy: stronger counterexample (non-convex domain): both minimum andmaximum inside the domain

D. Jerisson and N. Nadirashvili: true if the domain has two orthogonal axis of symmetry

R. Bañuelos and K. Burdzy: true if there are two orthogonal axes of symmetry, or just one axis ofsymmetry and ϕ2 symmetric wrt it

MNP: true if there are two orthogonal axes of symmetry, or just one axis of symmetry and ϕ2antisymmetric wrt it, or ... (some condition on the nodal set of ϕ2)

Other results: true for obtuse triangles, for some some doubly connected domains, for nearlycircular domains.

HS still open in its full generality! (e.g., proof for acute triangles?...)

M. N. Pascu (Transilvania Univ) Couplings of RBM 5/31 14.09.2012 5 / 31

Page 23: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Definition 2 (Reflecting Brownian motion)

Reflecting Brownian motion in D ⊂ Rd starting at x0 ∈ D: a solution to

Xt = x0 + Bt +

∫ t

0νD (Xs) dLX

s , t ≥ 0, (1)

where Bt is a d-dimensional Brownian motion starting at origin, νD is theinward unit vector field on ∂D, LX

t is the local time of X on ∂D.

Bt

XtνD(Xs)

Bs

Xs

X0 = B0

D

M. N. Pascu (Transilvania Univ) Couplings of RBM 6/31 14.09.2012 6 / 31

Page 24: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Definition 3 (Reflecting Brownian motion with killing)

Reflecting Brownian motion in D killed on hitting S ⊂ ∂D, starting at x0 ∈ D:

Yt =

{Xt, t < τ†, t ≥ τ , (2)

where Xt is RBM in D starting at x0, τ = τ S = inf{t > 0 : Xt ∈ S} is thekilling time, and † /∈ D is the cemetery state.

Ys Xs

Bs

X0 = Y0

D

Yt= † , t ≥ τ

XτS⊂ ∂D

M. N. Pascu (Transilvania Univ) Couplings of RBM 7/31 14.09.2012 7 / 31

Page 25: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Couplings of RBM

BM is invariant under translation, rotation/symmetry and scaling (almost).

This gives rise to:

Synchronous coupling: (Bt,Bt + v)

Mirror coupling: (Bt,RBt)

Scaling coupling: (Bt, cBt/c2)

M. N. Pascu (Transilvania Univ) Couplings of RBM 8/31 14.09.2012 8 / 31

Page 26: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Couplings of RBM

BM is invariant under translation, rotation/symmetry and scaling (almost).

This gives rise to:

Synchronous coupling: (Bt,Bt + v)

Mirror coupling: (Bt,RBt)

Scaling coupling: (Bt, cBt/c2)

M. N. Pascu (Transilvania Univ) Couplings of RBM 8/31 14.09.2012 8 / 31

Page 27: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Couplings of RBM

BM is invariant under translation, rotation/symmetry and scaling (almost).

This gives rise to:

Synchronous coupling: (Bt,Bt + v)

Mirror coupling: (Bt,RBt)

Scaling coupling: (Bt, cBt/c2)

M. N. Pascu (Transilvania Univ) Couplings of RBM 8/31 14.09.2012 8 / 31

Page 28: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Couplings of RBM

BM is invariant under translation, rotation/symmetry and scaling (almost).

This gives rise to:

Synchronous coupling: (Bt,Bt + v)

Mirror coupling: (Bt,RBt)

Scaling coupling: (Bt, cBt/c2)

M. N. Pascu (Transilvania Univ) Couplings of RBM 8/31 14.09.2012 8 / 31

Page 29: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Couplings of RBM

BM is invariant under translation, rotation/symmetry and scaling (almost).

This gives rise to:

Synchronous coupling: (Bt,Bt + v)

Mirror coupling: (Bt,RBt)

Scaling coupling: (Bt, cBt/c2)

The above can be extended to the case of reflecting Brownian motion.

M. N. Pascu (Transilvania Univ) Couplings of RBM 8/31 14.09.2012 8 / 31

Page 30: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Couplings of RBM

BM is invariant under translation, rotation/symmetry and scaling (almost).

Couplings of RBM:

Synchronous coupling: (R. Atar, K. Burdzy, R. Bañuelos, Z. Q. Chen, M.Cranston)

M. N. Pascu (Transilvania Univ) Couplings of RBM 8/31 14.09.2012 8 / 31

Page 31: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Couplings of RBM

BM is invariant under translation, rotation/symmetry and scaling (almost).

Couplings of RBM:

Synchronous coupling: (R. Atar, K. Burdzy, R. Bañuelos, Z. Q. Chen, M.Cranston)

Mirror coupling : (W. S. Kendal, M. Cranston, R. Atar, K. Burdzy, R.Bañuelos, MNP)

M. N. Pascu (Transilvania Univ) Couplings of RBM 8/31 14.09.2012 8 / 31

Page 32: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Couplings of RBM

BM is invariant under translation, rotation/symmetry and scaling (almost).

Couplings of RBM:

Synchronous coupling: (R. Atar, K. Burdzy, R. Bañuelos, Z. Q. Chen, M.Cranston)

Mirror coupling : (W. S. Kendal, M. Cranston, R. Atar, K. Burdzy, R.Bañuelos, MNP)

Scaling coupling : (MNP)

M. N. Pascu (Transilvania Univ) Couplings of RBM 8/31 14.09.2012 8 / 31

Page 33: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Lemma 4 (“Multiplicative Skorokhod lemma” in the unit disk, MNP)

If Bt is a 2-dimensional BM, Mt = 1 ∨ sups≤t|Bs| and α−1

t = At =

∫ t

0

1M2

sds,

Xt =1

Mαt

Bαt , t ≥ 0

is a RBM in U = {z ∈ R2 : |z| < 1}.

U

b(t)

x(t)

x(0) = b(0)

Proof: Itô formula with f (x, y) = xy , Bt and Mt (and a time change).

M. N. Pascu (Transilvania Univ) Couplings of RBM 9/31 14.09.2012 9 / 31

Page 34: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Lemma 4 (“Multiplicative Skorokhod lemma” in the unit disk, MNP)

If Bt is a 2-dimensional BM, Mt = 1 ∨ sups≤t|Bs| and α−1

t = At =

∫ t

0

1M2

sds,

Xt =1

Mαt

Bαt , t ≥ 0

is a RBM in U = {z ∈ R2 : |z| < 1}.U

b(t)

x(t)

x(0) = b(0)

Proof: Itô formula with f (x, y) = xy , Bt and Mt (and a time change).

M. N. Pascu (Transilvania Univ) Couplings of RBM 9/31 14.09.2012 9 / 31

Page 35: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Lemma 4 (“Multiplicative Skorokhod lemma” in the unit disk, MNP)

If Bt is a 2-dimensional BM, Mt = 1 ∨ sups≤t|Bs| and α−1

t = At =

∫ t

0

1M2

sds,

Xt =1

Mαt

Bαt , t ≥ 0

is a RBM in U = {z ∈ R2 : |z| < 1}.U

b(t)

x(t)

x(0) = b(0)

Proof: Itô formula with f (x, y) = xy , Bt and Mt (and a time change).

M. N. Pascu (Transilvania Univ) Couplings of RBM 9/31 14.09.2012 9 / 31

Page 36: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Scaling coupling of RBM in U starting at (xeiθ, yeiθ) (0 < x ≤ y ≤ 1):

a pair (Xt,Yt), where Xt RBM in U starting at xeiθ, Yt =1

Mαt

Xαt ,

Mt =xy ∨ sup

s≤t|Xs|, and α−1

t = At =

∫ t

0

1M2

sds.

Xt

Yt

Y0

X0

U

M. N. Pascu (Transilvania Univ) Couplings of RBM 10/31 14.09.2012 10 / 31

Page 37: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Lifetime of RBM in the unit disk, killed on a diameter

Xt

Yt

Y0 = iy

X0 = ix

U

S

Mt =xy∨sups≤t|Xs| ≤ 1 =⇒ At =

t∫0

1M2

sds ≥ t =⇒ αt = A−1

t ≤ t =⇒ τX = ατY≤ τY

(τX, τY denote the lifetime of Xt, Yt killed on the diameter S).

M. N. Pascu (Transilvania Univ) Couplings of RBM 11/31 14.09.2012 11 / 31

Page 38: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Lifetime of RBM in the unit disk, killed on a diameter

Xt

Yt

Y0 = iy

X0 = ix

U

S

Mt =xy∨sups≤t|Xs| ≤ 1 =⇒ At =

t∫0

1M2

sds ≥ t =⇒ αt = A−1

t ≤ t =⇒ τX = ατY≤ τY

(τX, τY denote the lifetime of Xt, Yt killed on the diameter S).

M. N. Pascu (Transilvania Univ) Couplings of RBM 11/31 14.09.2012 11 / 31

Page 39: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Lifetime of RBM in the unit disk, killed on a diameter

Xt

Yt

Y0 = iy

X0 = ix

U

S

Mt =xy∨sups≤t|Xs| ≤ 1 =⇒ At =

t∫0

1M2

sds ≥ t =⇒ αt = A−1

t ≤ t =⇒ τX = ατY≤ τY

(τX, τY denote the lifetime of Xt, Yt killed on the diameter S).

M. N. Pascu (Transilvania Univ) Couplings of RBM 11/31 14.09.2012 11 / 31

Page 40: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Lifetime of RBM in the unit disk, killed on a diameter

Xt

Yt

Y0 = iy

X0 = ix

U

S

Mt =xy∨sups≤t|Xs| ≤ 1 =⇒ At =

t∫0

1M2

sds ≥ t =⇒ αt = A−1

t ≤ t =⇒ τX = ατY≤ τY

(τX, τY denote the lifetime of Xt, Yt killed on the diameter S).

M. N. Pascu (Transilvania Univ) Couplings of RBM 11/31 14.09.2012 11 / 31

Page 41: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Lifetime of RBM in the unit disk, killed on a diameter

Xt

Yt

Y0 = iy

X0 = ix

U

S

Mt =xy∨sups≤t|Xs| ≤ 1 =⇒ At =

t∫0

1M2

sds ≥ t =⇒ αt = A−1

t ≤ t =⇒ τX = ατY≤ τY

(τX, τY denote the lifetime of Xt, Yt killed on the diameter S).

M. N. Pascu (Transilvania Univ) Couplings of RBM 11/31 14.09.2012 11 / 31

Page 42: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Lifetime of RBM in the unit disk, killed on a diameter

Xt

Yt

Y0 = iy

X0 = ix

U

S

Mt =xy∨sups≤t|Xs| ≤ 1 =⇒ At =

t∫0

1M2

sds ≥ t =⇒ αt = A−1

t ≤ t =⇒ τX = ατY≤ τY

(τX, τY denote the lifetime of Xt, Yt killed on the diameter S).

M. N. Pascu (Transilvania Univ) Couplings of RBM 11/31 14.09.2012 11 / 31

Page 43: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Scaling coupling and applications

Corollary 5 (Monotonicity of lifetime in the disk)

For any t > 0, P(τ x > t) is a radially increasing function in U(τ x is the lifetime of RBM in U starting at x, killed on a diameter).

Remark: for t large, P(τ x > t) ≈ ce−µ1tψ1(x) = ce−λ2tϕ2(x).

Theorem 6 (Monotonicity of antisymmetric second Neumann eigenfunctions)If ϕ is a second Neumann eigenfunction of the Laplacian on U, antisymmetricwith respect to a diameter, then ϕ is a radially monotone function.

Remark: any second Neumann eigenfunction is antisymmetric in the disk!

Corollary 7 (Hot Spots for the unit disk)The Hot Spots conjecture holds for the unit disk U, that is for any secondNeumann eigenfunction ϕ of the laplacian on U we have

min∂U

ϕ = minUϕ < max

Uϕ = max

∂Uϕ,

M. N. Pascu (Transilvania Univ) Couplings of RBM 12/31 14.09.2012 12 / 31

Page 44: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Scaling coupling and applications

Corollary 5 (Monotonicity of lifetime in the disk)

For any t > 0, P(τ x > t) is a radially increasing function in U(τ x is the lifetime of RBM in U starting at x, killed on a diameter).

Remark: for t large, P(τ x > t) ≈ ce−µ1tψ1(x) = ce−λ2tϕ2(x).

Theorem 6 (Monotonicity of antisymmetric second Neumann eigenfunctions)If ϕ is a second Neumann eigenfunction of the Laplacian on U, antisymmetricwith respect to a diameter, then ϕ is a radially monotone function.

Remark: any second Neumann eigenfunction is antisymmetric in the disk!

Corollary 7 (Hot Spots for the unit disk)The Hot Spots conjecture holds for the unit disk U, that is for any secondNeumann eigenfunction ϕ of the laplacian on U we have

min∂U

ϕ = minUϕ < max

Uϕ = max

∂Uϕ,

M. N. Pascu (Transilvania Univ) Couplings of RBM 12/31 14.09.2012 12 / 31

Page 45: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Scaling coupling and applications

Corollary 5 (Monotonicity of lifetime in the disk)

For any t > 0, P(τ x > t) is a radially increasing function in U(τ x is the lifetime of RBM in U starting at x, killed on a diameter).

Remark: for t large, P(τ x > t) ≈ ce−µ1tψ1(x) = ce−λ2tϕ2(x).

Theorem 6 (Monotonicity of antisymmetric second Neumann eigenfunctions)If ϕ is a second Neumann eigenfunction of the Laplacian on U, antisymmetricwith respect to a diameter, then ϕ is a radially monotone function.

Remark: any second Neumann eigenfunction is antisymmetric in the disk!

Corollary 7 (Hot Spots for the unit disk)The Hot Spots conjecture holds for the unit disk U, that is for any secondNeumann eigenfunction ϕ of the laplacian on U we have

min∂U

ϕ = minUϕ < max

Uϕ = max

∂Uϕ,

M. N. Pascu (Transilvania Univ) Couplings of RBM 12/31 14.09.2012 12 / 31

Page 46: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Scaling coupling and applications

Corollary 5 (Monotonicity of lifetime in the disk)

For any t > 0, P(τ x > t) is a radially increasing function in U(τ x is the lifetime of RBM in U starting at x, killed on a diameter).

Remark: for t large, P(τ x > t) ≈ ce−µ1tψ1(x) = ce−λ2tϕ2(x).

Theorem 6 (Monotonicity of antisymmetric second Neumann eigenfunctions)If ϕ is a second Neumann eigenfunction of the Laplacian on U, antisymmetricwith respect to a diameter, then ϕ is a radially monotone function.

Remark: any second Neumann eigenfunction is antisymmetric in the disk!

Corollary 7 (Hot Spots for the unit disk)The Hot Spots conjecture holds for the unit disk U, that is for any secondNeumann eigenfunction ϕ of the laplacian on U we have

min∂U

ϕ = minUϕ < max

Uϕ = max

∂Uϕ,

M. N. Pascu (Transilvania Univ) Couplings of RBM 12/31 14.09.2012 12 / 31

Page 47: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Scaling coupling and applications

Corollary 5 (Monotonicity of lifetime in the disk)

For any t > 0, P(τ x > t) is a radially increasing function in U(τ x is the lifetime of RBM in U starting at x, killed on a diameter).

Remark: for t large, P(τ x > t) ≈ ce−µ1tψ1(x) = ce−λ2tϕ2(x).

Theorem 6 (Monotonicity of antisymmetric second Neumann eigenfunctions)If ϕ is a second Neumann eigenfunction of the Laplacian on U, antisymmetricwith respect to a diameter, then ϕ is a radially monotone function.

Remark: any second Neumann eigenfunction is antisymmetric in the disk!

Corollary 7 (Hot Spots for the unit disk)The Hot Spots conjecture holds for the unit disk U, that is for any secondNeumann eigenfunction ϕ of the laplacian on U we have

min∂U

ϕ = minUϕ < max

Uϕ = max

∂Uϕ,

M. N. Pascu (Transilvania Univ) Couplings of RBM 12/31 14.09.2012 12 / 31

Page 48: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

More applications of scaling coupling...

The previous result is known (B. Kawohl, [6])... _Conformal invariance of RBM + geometric characterization of a convex maps⇒ the same is true for any smooth bounded convex domain D ⊂ R2! ¨

Theorem 8 (MNP)

If D ⊂ R2 is a convex C1,α domain (0 < α < 1), and at least one of thefollowing hypothesis hold,

i) D is symmetric with respect to both coordinate axes;

ii) D is symmetric with respect to the horizontal axis and the diameter towidth ratio dD/lD is larger than 4j0

π ≈ 3.06;

then Hot Spots conjecture holds for the domain D.

M. N. Pascu (Transilvania Univ) Couplings of RBM 13/31 14.09.2012 13 / 31

Page 49: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

More applications of scaling coupling...

The previous result is known (B. Kawohl, [6])... _Conformal invariance of RBM + geometric characterization of a convex maps⇒ the same is true for any smooth bounded convex domain D ⊂ R2! ¨

Theorem 8 (MNP)

If D ⊂ R2 is a convex C1,α domain (0 < α < 1), and at least one of thefollowing hypothesis hold,

i) D is symmetric with respect to both coordinate axes;

ii) D is symmetric with respect to the horizontal axis and the diameter towidth ratio dD/lD is larger than 4j0

π ≈ 3.06;

then Hot Spots conjecture holds for the domain D.

M. N. Pascu (Transilvania Univ) Couplings of RBM 13/31 14.09.2012 13 / 31

Page 50: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

More applications of scaling coupling...

The previous result is known (B. Kawohl, [6])... _Conformal invariance of RBM + geometric characterization of a convex maps⇒ the same is true for any smooth bounded convex domain D ⊂ R2! ¨

Theorem 8 (MNP)

If D ⊂ R2 is a convex C1,α domain (0 < α < 1), and at least one of thefollowing hypothesis hold,

i) D is symmetric with respect to both coordinate axes;

ii) D is symmetric with respect to the horizontal axis and the diameter towidth ratio dD/lD is larger than 4j0

π ≈ 3.06;

then Hot Spots conjecture holds for the domain D.

M. N. Pascu (Transilvania Univ) Couplings of RBM 13/31 14.09.2012 13 / 31

Page 51: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

– Where should I put the radiator, to feel warmest at all times?

M. N. Pascu (Transilvania Univ) Couplings of RBM 14/31 14.09.2012 14 / 31

Page 52: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Laugesen-Morpurgo conjecture

Conjecture 9 (R. Laugesen, C. Morpurgo, 1998)

For any t > 0, pU (t, x, x) is a strictly increasing radial function in the unitball U, that is

pU(t, x, x) < pU(t, y, y), (3)

for all x, y ∈ U with ‖x‖ < ‖y‖.

Remark: Laugesen-Morpugo conjecture⇒ Hot spots conjecture for the disk.

M. N. Pascu (Transilvania Univ) Couplings of RBM 15/31 14.09.2012 15 / 31

Page 53: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Laugesen-Morpurgo conjecture

Conjecture 9 (R. Laugesen, C. Morpurgo, 1998)

For any t > 0, pU (t, x, x) is a strictly increasing radial function in the unitball U, that is

pU(t, x, x) < pU(t, y, y), (3)

for all x, y ∈ U with ‖x‖ < ‖y‖.

Remark: Laugesen-Morpugo conjecture⇒ Hot spots conjecture for the disk.

M. N. Pascu (Transilvania Univ) Couplings of RBM 15/31 14.09.2012 15 / 31

Page 54: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Mirror coupling of BM/RBM

Was introduced by Kendall ([7]) and Cranston ([Cr]) (BM on manifolds), anddeveloped by Burdzy et al. ([1], [2], [3], ...) (RBM in a smooth domains).

Loosely speaking, in both cases the idea is that the increments of oneBM/RBM is the mirror image of the other.

A bit more precise: let Xt,Yt be RBM is a smooth domain D ⊂ Rd, withdriving BM Bt,Zt, and consider the SDE:

Zt = Bt − 2∫ t

0

Xs − Ys

‖Xs − Ys‖2 (Xs − Ys) · dBs. (4)

Burdzy et al. proved the existence of a strong solution and pathwiseuniqueness of the above SDE for t < τ = inf {s > 0 : Xs = Ys}.We let Xt = Yt for t ≥ τ , and refer to (Xt,Yt) as a mirror coupling in Dstarting at x, y ∈ D.

M. N. Pascu (Transilvania Univ) Couplings of RBM 16/31 14.09.2012 16 / 31

Page 55: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Mirror coupling of BM/RBM

Was introduced by Kendall ([7]) and Cranston ([Cr]) (BM on manifolds), anddeveloped by Burdzy et al. ([1], [2], [3], ...) (RBM in a smooth domains).

Loosely speaking, in both cases the idea is that the increments of oneBM/RBM is the mirror image of the other.

A bit more precise: let Xt,Yt be RBM is a smooth domain D ⊂ Rd, withdriving BM Bt,Zt, and consider the SDE:

Zt = Bt − 2∫ t

0

Xs − Ys

‖Xs − Ys‖2 (Xs − Ys) · dBs. (4)

Burdzy et al. proved the existence of a strong solution and pathwiseuniqueness of the above SDE for t < τ = inf {s > 0 : Xs = Ys}.We let Xt = Yt for t ≥ τ , and refer to (Xt,Yt) as a mirror coupling in Dstarting at x, y ∈ D.

M. N. Pascu (Transilvania Univ) Couplings of RBM 16/31 14.09.2012 16 / 31

Page 56: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Mirror coupling of BM/RBM

Was introduced by Kendall ([7]) and Cranston ([Cr]) (BM on manifolds), anddeveloped by Burdzy et al. ([1], [2], [3], ...) (RBM in a smooth domains).

Loosely speaking, in both cases the idea is that the increments of oneBM/RBM is the mirror image of the other.

A bit more precise: let Xt,Yt be RBM is a smooth domain D ⊂ Rd, withdriving BM Bt,Zt, and consider the SDE:

Zt = Bt − 2∫ t

0

Xs − Ys

‖Xs − Ys‖2 (Xs − Ys) · dBs. (4)

Burdzy et al. proved the existence of a strong solution and pathwiseuniqueness of the above SDE for t < τ = inf {s > 0 : Xs = Ys}.We let Xt = Yt for t ≥ τ , and refer to (Xt,Yt) as a mirror coupling in Dstarting at x, y ∈ D.

M. N. Pascu (Transilvania Univ) Couplings of RBM 16/31 14.09.2012 16 / 31

Page 57: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Mirror coupling of BM/RBM

Was introduced by Kendall ([7]) and Cranston ([Cr]) (BM on manifolds), anddeveloped by Burdzy et al. ([1], [2], [3], ...) (RBM in a smooth domains).

Loosely speaking, in both cases the idea is that the increments of oneBM/RBM is the mirror image of the other.

A bit more precise: let Xt,Yt be RBM is a smooth domain D ⊂ Rd, withdriving BM Bt,Zt, and consider the SDE:

Zt = Bt − 2∫ t

0

Xs − Ys

‖Xs − Ys‖2 (Xs − Ys) · dBs. (4)

Burdzy et al. proved the existence of a strong solution and pathwiseuniqueness of the above SDE for t < τ = inf {s > 0 : Xs = Ys}.We let Xt = Yt for t ≥ τ , and refer to (Xt,Yt) as a mirror coupling in Dstarting at x, y ∈ D.

M. N. Pascu (Transilvania Univ) Couplings of RBM 16/31 14.09.2012 16 / 31

Page 58: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Mirror coupling of BM/RBM

Was introduced by Kendall ([7]) and Cranston ([Cr]) (BM on manifolds), anddeveloped by Burdzy et al. ([1], [2], [3], ...) (RBM in a smooth domains).

Loosely speaking, in both cases the idea is that the increments of oneBM/RBM is the mirror image of the other.

A bit more precise: let Xt,Yt be RBM is a smooth domain D ⊂ Rd, withdriving BM Bt,Zt, and consider the SDE:

Zt = Bt − 2∫ t

0

Xs − Ys

‖Xs − Ys‖2 (Xs − Ys) · dBs. (4)

Burdzy et al. proved the existence of a strong solution and pathwiseuniqueness of the above SDE for t < τ = inf {s > 0 : Xs = Ys}.We let Xt = Yt for t ≥ τ , and refer to (Xt,Yt) as a mirror coupling in Dstarting at x, y ∈ D.

M. N. Pascu (Transilvania Univ) Couplings of RBM 16/31 14.09.2012 16 / 31

Page 59: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

What does this mean?For a unitary vector m, let H(m) = I − 2mm′ (reflection in the hyperplanethrough the origin and perpendicular to m), and define G(x) = H

(x‖x‖

)if

x 6= 0 and G(0) = I.

(4)⇐⇒ dZt = G(

Xt − Yt

||Xt − Yt||

)dWt,

so, (4) says that the increments dZt and dBt are mirror images wrt hyperplaneof symmetryMt between Xt and Yt).

Yt

Xt

H

0

Xξ = Yξ

Xt = Yt

m

Figure: Mirror coupling of Brownian motions (no reflection).M. N. Pascu (Transilvania Univ) Couplings of RBM 17/31 14.09.2012 17 / 31

Page 60: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

What does this mean?For a unitary vector m, let H(m) = I − 2mm′ (reflection in the hyperplanethrough the origin and perpendicular to m), and define G(x) = H

(x‖x‖

)if

x 6= 0 and G(0) = I.

(4)⇐⇒ dZt = G(

Xt − Yt

||Xt − Yt||

)dWt,

so, (4) says that the increments dZt and dBt are mirror images wrt hyperplaneof symmetryMt between Xt and Yt).

Yt

Xt

H

0

Xξ = Yξ

Xt = Yt

m

Figure: Mirror coupling of Brownian motions (no reflection).M. N. Pascu (Transilvania Univ) Couplings of RBM 17/31 14.09.2012 17 / 31

Page 61: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

What does this mean?For a unitary vector m, let H(m) = I − 2mm′ (reflection in the hyperplanethrough the origin and perpendicular to m), and define G(x) = H

(x‖x‖

)if

x 6= 0 and G(0) = I.

(4)⇐⇒ dZt = G(

Xt − Yt

||Xt − Yt||

)dWt,

so, (4) says that the increments dZt and dBt are mirror images wrt hyperplaneof symmetryMt between Xt and Yt).

Yt

Xt

H

0

Xξ = Yξ

Xt = Yt

m

Figure: Mirror coupling of Brownian motions (no reflection).M. N. Pascu (Transilvania Univ) Couplings of RBM 17/31 14.09.2012 17 / 31

Page 62: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

What does this mean?For a unitary vector m, let H(m) = I − 2mm′ (reflection in the hyperplanethrough the origin and perpendicular to m), and define G(x) = H

(x‖x‖

)if

x 6= 0 and G(0) = I.

(4)⇐⇒ dZt = G(

Xt − Yt

||Xt − Yt||

)dWt,

so, (4) says that the increments dZt and dBt are mirror images wrt hyperplaneof symmetryMt between Xt and Yt).

Yt

Xt

H

0

Xξ = Yξ

Xt = Yt

m

Figure: Mirror coupling of Brownian motions (no reflection).M. N. Pascu (Transilvania Univ) Couplings of RBM 17/31 14.09.2012 17 / 31

Page 63: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

What does this mean?For a unitary vector m, let H(m) = I − 2mm′ (reflection in the hyperplanethrough the origin and perpendicular to m), and define G(x) = H

(x‖x‖

)if

x 6= 0 and G(0) = I.

(4)⇐⇒ dZt = G(

Xt − Yt

||Xt − Yt||

)dWt,

so, (4) says that the increments dZt and dBt are mirror images wrt hyperplaneof symmetryMt between Xt and Yt).

Yt

Xt

H

0

Xξ = Yξ

Xt = Yt

m

Figure: Mirror coupling of Brownian motions (no reflection).M. N. Pascu (Transilvania Univ) Couplings of RBM 17/31 14.09.2012 17 / 31

Page 64: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Lemma 10 (“MirrorMt moves towards origin”, MNP)

Let Xt, Yt be a mirror coupling of RBM in U starting at x, y ∈ U, and letτ = inf{t > 0 : Xt = Yt} and τ 1 = inf{t > 0 : 0 ∈Mt}.

For all times t < τ ∧ τ 1, the mirrorMt moves towards the origin, in such a way that ifa point P ∈ U and the origin are separated byMt1 for t1 ∈ [0, τ ∧ τ 1), then the point

P and the origin are separated byMt2 for all t2 ∈ [t1, τ ∧ τ 1).

0

Yt

Xt

xy

Mt M0

At

Bt

P

M. N. Pascu (Transilvania Univ) Couplings of RBM 18/31 14.09.2012 18 / 31

Page 65: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Inequalities for the Neumann heat kernel pU(t, x, y) of the unit ball U

Theorem 11 (MNP)

For any points x, y, z ∈ U such that ‖y‖ ≤ ‖x‖ and ‖x− z‖ ≤ ‖y− z‖, andany t > 0 we have:

pU (t, y, z) ≤ pU (t, x, z) . (5)

Corollary 12

For any x ∈ U− {0}, r ∈ (0,min {‖x‖, 1− ‖x‖}) and t > 0 we have:∫∂U

pU (t, x + ru, x) dσ(u) ≤ pU(t, x+r x‖x‖ , x) ≤ pU(t, x+r x

‖x‖ , x+r x‖x‖), (6)

where σ is the normalized surface measure on ∂U.

M. N. Pascu (Transilvania Univ) Couplings of RBM 19/31 14.09.2012 19 / 31

Page 66: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Inequalities for the Neumann heat kernel pU(t, x, y) of the unit ball U

Theorem 11 (MNP)

For any points x, y, z ∈ U such that ‖y‖ ≤ ‖x‖ and ‖x− z‖ ≤ ‖y− z‖, andany t > 0 we have:

pU (t, y, z) ≤ pU (t, x, z) . (5)

Corollary 12

For any x ∈ U− {0}, r ∈ (0,min {‖x‖, 1− ‖x‖}) and t > 0 we have:∫∂U

pU (t, x + ru, x) dσ(u) ≤ pU(t, x+r x‖x‖ , x) ≤ pU(t, x+r x

‖x‖ , x+r x‖x‖), (6)

where σ is the normalized surface measure on ∂U.

M. N. Pascu (Transilvania Univ) Couplings of RBM 19/31 14.09.2012 19 / 31

Page 67: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Resolution of the Laugesen-Morpurgo conjecture

Theorem 13 (MNP)

For any t > 0, pU (t, x, x) is a strictly increasing radial function in U, that is

pU(t, x, x) < pU(t, y, y), (7)

for all x, y ∈ U with ‖x‖ < ‖y‖.

Proof.d

d‖x‖pU (t, x, x) = limr↘0

pU(t, x + r x‖x‖ , x + r x

‖x‖)− pU (t, x, x)

r

≥ limr↘0

∫∂U pU (t, x + ru, x) dσ(u)− pU (t, x, x)

r

=

∫∂U

limr↘0

pU (t, x + ru, x)− pU (t, x, x)r

dσ(u)

=

∫∂U∇pU (t, x, x) · u dσ(u)

= 0.

M. N. Pascu (Transilvania Univ) Couplings of RBM 20/31 14.09.2012 20 / 31

Page 68: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Resolution of the Laugesen-Morpurgo conjecture

Theorem 13 (MNP)

For any t > 0, pU (t, x, x) is a strictly increasing radial function in U, that is

pU(t, x, x) < pU(t, y, y), (7)

for all x, y ∈ U with ‖x‖ < ‖y‖.

Proof.d

d‖x‖pU (t, x, x) = limr↘0

pU(t, x + r x‖x‖ , x + r x

‖x‖)− pU (t, x, x)

r

≥ limr↘0

∫∂U pU (t, x + ru, x) dσ(u)− pU (t, x, x)

r

=

∫∂U

limr↘0

pU (t, x + ru, x)− pU (t, x, x)r

dσ(u)

=

∫∂U∇pU (t, x, x) · u dσ(u)

= 0.

M. N. Pascu (Transilvania Univ) Couplings of RBM 20/31 14.09.2012 20 / 31

Page 69: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Resolution of the Laugesen-Morpurgo conjecture

Theorem 13 (MNP)

For any t > 0, pU (t, x, x) is a strictly increasing radial function in U, that is

pU(t, x, x) < pU(t, y, y), (7)

for all x, y ∈ U with ‖x‖ < ‖y‖.

Proof.d

d‖x‖pU (t, x, x) = limr↘0

pU(t, x + r x‖x‖ , x + r x

‖x‖)− pU (t, x, x)

r

≥ limr↘0

∫∂U pU (t, x + ru, x) dσ(u)− pU (t, x, x)

r

=

∫∂U

limr↘0

pU (t, x + ru, x)− pU (t, x, x)r

dσ(u)

=

∫∂U∇pU (t, x, x) · u dσ(u)

= 0.

M. N. Pascu (Transilvania Univ) Couplings of RBM 20/31 14.09.2012 20 / 31

Page 70: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Resolution of the Laugesen-Morpurgo conjecture

Theorem 13 (MNP)

For any t > 0, pU (t, x, x) is a strictly increasing radial function in U, that is

pU(t, x, x) < pU(t, y, y), (7)

for all x, y ∈ U with ‖x‖ < ‖y‖.

Proof.d

d‖x‖pU (t, x, x) = limr↘0

pU(t, x + r x‖x‖ , x + r x

‖x‖)− pU (t, x, x)

r

≥ limr↘0

∫∂U pU (t, x + ru, x) dσ(u)− pU (t, x, x)

r

=

∫∂U

limr↘0

pU (t, x + ru, x)− pU (t, x, x)r

dσ(u)

=

∫∂U∇pU (t, x, x) · u dσ(u)

= 0.

M. N. Pascu (Transilvania Univ) Couplings of RBM 20/31 14.09.2012 20 / 31

Page 71: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Resolution of the Laugesen-Morpurgo conjecture

Theorem 13 (MNP)

For any t > 0, pU (t, x, x) is a strictly increasing radial function in U, that is

pU(t, x, x) < pU(t, y, y), (7)

for all x, y ∈ U with ‖x‖ < ‖y‖.

Proof.d

d‖x‖pU (t, x, x) = limr↘0

pU(t, x + r x‖x‖ , x + r x

‖x‖)− pU (t, x, x)

r

≥ limr↘0

∫∂U pU (t, x + ru, x) dσ(u)− pU (t, x, x)

r

=

∫∂U

limr↘0

pU (t, x + ru, x)− pU (t, x, x)r

dσ(u)

=

∫∂U∇pU (t, x, x) · u dσ(u)

= 0.M. N. Pascu (Transilvania Univ) Couplings of RBM 20/31 14.09.2012 20 / 31

Page 72: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

– Are we going to be warmer or colder in a bigger apartment??

M. N. Pascu (Transilvania Univ) Couplings of RBM 21/31 14.09.2012 21 / 31

Page 73: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Chavel’s conjecture on domain monotonicity of Neumann heat kernel

Conjecture 14 (I. Chavel, 1986)If D1 ⊂ D2 are convex domains then for all t > 0 and x, y ∈ D1 we have

pD1 (t, x, y) ≥ pD2 (t, x, y) .

I. Chavel: TRUE, if D2 is a ball centered at x (or y) and D1 is convex(integration by parts).W. S. Kendall: TRUE, if D1 is a ball centered at x (or y) and D2 is convex(coupling arguments).

Theorem 15 (Chavel + Kendall)If D1 ⊂ D2 are convex domains then for all t > 0 and x, y ∈ D1 we have

pD1 (t, x, y) ≥ pD2 (t, x, y) ,

whenever there exists a ball B centered at either x or y such thatD1 ⊂ B ⊂ D2.

M. N. Pascu (Transilvania Univ) Couplings of RBM 22/31 14.09.2012 22 / 31

Page 74: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Chavel’s conjecture on domain monotonicity of Neumann heat kernel

Conjecture 14 (I. Chavel, 1986)If D1 ⊂ D2 are convex domains then for all t > 0 and x, y ∈ D1 we have

pD1 (t, x, y) ≥ pD2 (t, x, y) .

I. Chavel: TRUE, if D2 is a ball centered at x (or y) and D1 is convex(integration by parts).W. S. Kendall: TRUE, if D1 is a ball centered at x (or y) and D2 is convex(coupling arguments).

Theorem 15 (Chavel + Kendall)If D1 ⊂ D2 are convex domains then for all t > 0 and x, y ∈ D1 we have

pD1 (t, x, y) ≥ pD2 (t, x, y) ,

whenever there exists a ball B centered at either x or y such thatD1 ⊂ B ⊂ D2.

M. N. Pascu (Transilvania Univ) Couplings of RBM 22/31 14.09.2012 22 / 31

Page 75: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Chavel’s conjecture on domain monotonicity of Neumann heat kernel

Conjecture 14 (I. Chavel, 1986)If D1 ⊂ D2 are convex domains then for all t > 0 and x, y ∈ D1 we have

pD1 (t, x, y) ≥ pD2 (t, x, y) .

I. Chavel: TRUE, if D2 is a ball centered at x (or y) and D1 is convex(integration by parts).W. S. Kendall: TRUE, if D1 is a ball centered at x (or y) and D2 is convex(coupling arguments).

Theorem 15 (Chavel + Kendall)If D1 ⊂ D2 are convex domains then for all t > 0 and x, y ∈ D1 we have

pD1 (t, x, y) ≥ pD2 (t, x, y) ,

whenever there exists a ball B centered at either x or y such thatD1 ⊂ B ⊂ D2.

M. N. Pascu (Transilvania Univ) Couplings of RBM 22/31 14.09.2012 22 / 31

Page 76: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Chavel’s conjecture on domain monotonicity of Neumann heat kernel

Conjecture 14 (I. Chavel, 1986)If D1 ⊂ D2 are convex domains then for all t > 0 and x, y ∈ D1 we have

pD1 (t, x, y) ≥ pD2 (t, x, y) .

I. Chavel: TRUE, if D2 is a ball centered at x (or y) and D1 is convex(integration by parts).W. S. Kendall: TRUE, if D1 is a ball centered at x (or y) and D2 is convex(coupling arguments).

Theorem 15 (Chavel + Kendall)If D1 ⊂ D2 are convex domains then for all t > 0 and x, y ∈ D1 we have

pD1 (t, x, y) ≥ pD2 (t, x, y) ,

whenever there exists a ball B centered at either x or y such thatD1 ⊂ B ⊂ D2.

M. N. Pascu (Transilvania Univ) Couplings of RBM 22/31 14.09.2012 22 / 31

Page 77: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Extension of the mirror coupling

Question: can one define a mirror coupling of Brownian motions living indifferent domains D1 and D2?Answer: YES, for example if D1 = R and D2 = (0,∞):Construction (Tanaka formula): Xt BM in R, Yt = |Xt| RBM on (0,∞) and

dYt = sgn(Xt)dXt + dLXt , t ≥ 0. (8)

YES, if:

D1 = R2 and D2 half plane

D1 = R2 and D2 convex polygonal domain

D1 = R2 and D2 convex domain

D2 convex and D2 ⊂⊂ D1

M. N. Pascu (Transilvania Univ) Couplings of RBM 23/31 14.09.2012 23 / 31

Page 78: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Extension of the mirror coupling

Question: can one define a mirror coupling of Brownian motions living indifferent domains D1 and D2?Answer: YES, for example if D1 = R and D2 = (0,∞):Construction (Tanaka formula): Xt BM in R, Yt = |Xt| RBM on (0,∞) and

dYt = sgn(Xt)dXt + dLXt , t ≥ 0. (8)

YES, if:

D1 = R2 and D2 half plane

D1 = R2 and D2 convex polygonal domain

D1 = R2 and D2 convex domain

D2 convex and D2 ⊂⊂ D1

M. N. Pascu (Transilvania Univ) Couplings of RBM 23/31 14.09.2012 23 / 31

Page 79: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Extension of the mirror coupling

Question: can one define a mirror coupling of Brownian motions living indifferent domains D1 and D2?Answer: YES, for example if D1 = R and D2 = (0,∞):Construction (Tanaka formula): Xt BM in R, Yt = |Xt| RBM on (0,∞) and

dYt = sgn(Xt)dXt + dLXt , t ≥ 0. (8)

YES, if:

D1 = R2 and D2 half plane

D1 = R2 and D2 convex polygonal domain

D1 = R2 and D2 convex domain

D2 convex and D2 ⊂⊂ D1

M. N. Pascu (Transilvania Univ) Couplings of RBM 23/31 14.09.2012 23 / 31

Page 80: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Extension of the mirror coupling

Question: can one define a mirror coupling of Brownian motions living indifferent domains D1 and D2?Answer: YES, for example if D1 = R and D2 = (0,∞):Construction (Tanaka formula): Xt BM in R, Yt = |Xt| RBM on (0,∞) and

dYt = sgn(Xt)dXt + dLXt , t ≥ 0. (8)

YES, if:

D1 = R2 and D2 half plane

D1 = R2 and D2 convex polygonal domain

D1 = R2 and D2 convex domain

D2 convex and D2 ⊂⊂ D1

M. N. Pascu (Transilvania Univ) Couplings of RBM 23/31 14.09.2012 23 / 31

Page 81: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Extension of the mirror coupling

Question: can one define a mirror coupling of Brownian motions living indifferent domains D1 and D2?Answer: YES, for example if D1 = R and D2 = (0,∞):Construction (Tanaka formula): Xt BM in R, Yt = |Xt| RBM on (0,∞) and

dYt = sgn(Xt)dXt + dLXt , t ≥ 0. (8)

YES, if:

D1 = R2 and D2 half plane

D1 = R2 and D2 convex polygonal domain

D1 = R2 and D2 convex domain

D2 convex and D2 ⊂⊂ D1

M. N. Pascu (Transilvania Univ) Couplings of RBM 23/31 14.09.2012 23 / 31

Page 82: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Extension of the mirror coupling

Question: can one define a mirror coupling of Brownian motions living indifferent domains D1 and D2?Answer: YES, for example if D1 = R and D2 = (0,∞):Construction (Tanaka formula): Xt BM in R, Yt = |Xt| RBM on (0,∞) and

dYt = sgn(Xt)dXt + dLXt , t ≥ 0. (8)

YES, if:

D1 = R2 and D2 half plane

D1 = R2 and D2 convex polygonal domain

D1 = R2 and D2 convex domain

D2 convex and D2 ⊂⊂ D1

M. N. Pascu (Transilvania Univ) Couplings of RBM 23/31 14.09.2012 23 / 31

Page 83: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Extension of the mirror coupling

Question: can one define a mirror coupling of Brownian motions living indifferent domains D1 and D2?Answer: YES, for example if D1 = R and D2 = (0,∞):Construction (Tanaka formula): Xt BM in R, Yt = |Xt| RBM on (0,∞) and

dYt = sgn(Xt)dXt + dLXt , t ≥ 0. (8)

YES, if:

D1 = R2 and D2 half plane

D1 = R2 and D2 convex polygonal domain

D1 = R2 and D2 convex domain

D2 convex and D2 ⊂⊂ D1

M. N. Pascu (Transilvania Univ) Couplings of RBM 23/31 14.09.2012 23 / 31

Page 84: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Extension of the mirror coupling

Question: can one define a mirror coupling of Brownian motions living indifferent domains D1 and D2?Answer: YES, if D1,2 ⊂ Rd smooth, with non-tangential boundaries, andD1 ∩ D2 convex

D1

D2

Figure: Typical domains for the extended mirror coupling.

M. N. Pascu (Transilvania Univ) Couplings of RBM 23/31 14.09.2012 23 / 31

Page 85: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

An application of mirror coupling:a unifying proof of Chavel’s conjecture

Theorem 16

If D1 ⊂ D2 ⊂ Rd are smooth and D1 is a convex domain, then for all t > 0and x, y ∈ D1 we have

pD1 (t, x, y) ≥ pD2 (t, x, y) ,

whenever there exists a ball B centered at either x or y such thatD1 ⊂ B ⊂ D2.

M. N. Pascu (Transilvania Univ) Couplings of RBM 24/31 14.09.2012 24 / 31

Page 86: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Sketch of the proof

Consider a mirror coupling of RBM Xt,Yt in D2,D1, starting at X0 = Y0 = x.For all t > 0, the mirrorMt of the coupling cannot separate Yt and y.

y

D1 D2

B(y, r)

Mt

Xt

Yt

‖Yt− y‖ ≤ ‖Xt− y‖, t > 0 =⇒ pD1 (t, x, y) > pD2 (t, x, y) .

M. N. Pascu (Transilvania Univ) Couplings of RBM 25/31 14.09.2012 25 / 31

Page 87: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Sketch of the proof

Consider a mirror coupling of RBM Xt,Yt in D2,D1, starting at X0 = Y0 = x.For all t > 0, the mirrorMt of the coupling cannot separate Yt and y.

y

D1 D2

B(y, r)

Mt

Xt

Yt

‖Yt− y‖ ≤ ‖Xt− y‖, t > 0 =⇒ pD1 (t, x, y) > pD2 (t, x, y) .

M. N. Pascu (Transilvania Univ) Couplings of RBM 25/31 14.09.2012 25 / 31

Page 88: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Sketch of the proof

Consider a mirror coupling of RBM Xt,Yt in D2,D1, starting at X0 = Y0 = x.For all t > 0, the mirrorMt of the coupling cannot separate Yt and y.

y

D1 D2

B(y, r)

Mt

Xt

Yt

‖Yt− y‖ ≤ ‖Xt− y‖, t > 0 =⇒ pD1 (t, x, y) > pD2 (t, x, y) .

M. N. Pascu (Transilvania Univ) Couplings of RBM 25/31 14.09.2012 25 / 31

Page 89: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Sketch of the proof

Consider a mirror coupling of RBM Xt,Yt in D2,D1, starting at X0 = Y0 = x.For all t > 0, the mirrorMt of the coupling cannot separate Yt and y.

y

D1 D2

B(y, r)

Mt

Xt

Yt

‖Yt− y‖ ≤ ‖Xt− y‖, t > 0 =⇒ pD1 (t, x, y) > pD2 (t, x, y) .

M. N. Pascu (Transilvania Univ) Couplings of RBM 25/31 14.09.2012 25 / 31

Page 90: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Sketch of the proof

Consider a mirror coupling of RBM Xt,Yt in D2,D1, starting at X0 = Y0 = x.For all t > 0, the mirrorMt of the coupling cannot separate Yt and y.

y

D1 D2

B(y, r)

Mt

Xt

Yt

‖Yt− y‖ ≤ ‖Xt− y‖, t > 0 =⇒ pD1 (t, x, y) > pD2 (t, x, y) .

M. N. Pascu (Transilvania Univ) Couplings of RBM 25/31 14.09.2012 25 / 31

Page 91: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Lion and Man problem (R. Rado, 1953)

 

 

 

tX tY

D

Given x, y ∈ D, does there exist X,Y with X0 = x, Y0 = y, |X′t | = |Y ′t | = 1 and

dist(Xt,Yt) > c, for all t > 0?

(Littlewood, Besicovitch, Croft, Bollobas et. al., Nahin, ...)M. N. Pascu (Transilvania Univ) Couplings of RBM 26/31 14.09.2012 26 / 31

Page 92: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Shy couplings (Benjamini, Burdzy and Chen – PTRF, 2007)

ε-shy coupling: P(dist(Xt,Yt) > ε for all t ≥ 0) > 0.

RBM case: no co-adapted shy coupling in

bounded convex planar domains with C2 boundary, without linesegments (BBC, 2007).

bounded convex domains in Rd, without line segments in the boundary ifd ≥ 3 (Kendall, 2009).

bounded CAT(0) spaces with boundary satisfying uniform exteriorsphere and interior cone conditions, e.g. simply connected boundedplanar domains with C2 boundary (Bramson, Burdzy & Kendall,preprint).

M. N. Pascu (Transilvania Univ) Couplings of RBM 27/31 14.09.2012 27 / 31

Page 93: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Shy couplings (Benjamini, Burdzy and Chen – PTRF, 2007)

ε-shy coupling: P(dist(Xt,Yt) > ε for all t ≥ 0) > 0.

RBM case: no co-adapted shy coupling in

bounded convex planar domains with C2 boundary, without linesegments (BBC, 2007).

bounded convex domains in Rd, without line segments in the boundary ifd ≥ 3 (Kendall, 2009).

bounded CAT(0) spaces with boundary satisfying uniform exteriorsphere and interior cone conditions, e.g. simply connected boundedplanar domains with C2 boundary (Bramson, Burdzy & Kendall,preprint).

M. N. Pascu (Transilvania Univ) Couplings of RBM 27/31 14.09.2012 27 / 31

Page 94: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Shy couplings (Benjamini, Burdzy and Chen – PTRF, 2007)

ε-shy coupling: P(dist(Xt,Yt) > ε for all t ≥ 0) > 0.

RBM case: no co-adapted shy coupling in

bounded convex planar domains with C2 boundary, without linesegments (BBC, 2007).

bounded convex domains in Rd, without line segments in the boundary ifd ≥ 3 (Kendall, 2009).

bounded CAT(0) spaces with boundary satisfying uniform exteriorsphere and interior cone conditions, e.g. simply connected boundedplanar domains with C2 boundary (Bramson, Burdzy & Kendall,preprint).

M. N. Pascu (Transilvania Univ) Couplings of RBM 27/31 14.09.2012 27 / 31

Page 95: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Shy couplings (Benjamini, Burdzy and Chen – PTRF, 2007)

ε-shy coupling: P(dist(Xt,Yt) > ε for all t ≥ 0) > 0.

RBM case: no co-adapted shy coupling in

bounded convex planar domains with C2 boundary, without linesegments (BBC, 2007).

bounded convex domains in Rd, without line segments in the boundary ifd ≥ 3 (Kendall, 2009).

bounded CAT(0) spaces with boundary satisfying uniform exteriorsphere and interior cone conditions, e.g. simply connected boundedplanar domains with C2 boundary (Bramson, Burdzy & Kendall,preprint).

M. N. Pascu (Transilvania Univ) Couplings of RBM 27/31 14.09.2012 27 / 31

Page 96: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

BMan and BLion problem on S2 (joint with I. Popescu) 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

tX tY

2S

Theorem 17

For any X0, Y0 ∈ S2, there exist couplings of BM on S2 s.t. for all t ≥ 0 we have

a) |Xt − Yt| =√

4− |y + x|2 e−t ↗ 2 (distance-increasing coupling)

b) |Xt − Yt| = |y− x| e−t/2 ↘ 0 (distance-decreasing coupling)

c) |Xt − Yt| = |y− x| = const (fixed-distance coupling= “translation coupling”).

M. N. Pascu (Transilvania Univ) Couplings of RBM 28/31 14.09.2012 28 / 31

Page 97: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

BMan and BLion problem on S2 (joint with I. Popescu) 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

tX tY

2S

Theorem 17

For any X0, Y0 ∈ S2, there exist couplings of BM on S2 s.t. for all t ≥ 0 we have

a) |Xt − Yt| =√

4− |y + x|2 e−t ↗ 2 (distance-increasing coupling)

b) |Xt − Yt| = |y− x| e−t/2 ↘ 0 (distance-decreasing coupling)

c) |Xt − Yt| = |y− x| = const (fixed-distance coupling= “translation coupling”).

M. N. Pascu (Transilvania Univ) Couplings of RBM 28/31 14.09.2012 28 / 31

Page 98: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

BMan and BLion problem on S2 (joint with I. Popescu) 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

tX tY

2S

Theorem 17

For any X0, Y0 ∈ S2, there exist couplings of BM on S2 s.t. for all t ≥ 0 we have

a) |Xt − Yt| =√

4− |y + x|2 e−t ↗ 2 (distance-increasing coupling)

b) |Xt − Yt| = |y− x| e−t/2 ↘ 0 (distance-decreasing coupling)

c) |Xt − Yt| = |y− x| = const (fixed-distance coupling= “translation coupling”).

M. N. Pascu (Transilvania Univ) Couplings of RBM 28/31 14.09.2012 28 / 31

Page 99: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

BMan and BLion problem on S2 (joint with I. Popescu) 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

tX tY

2S

Theorem 17

For any X0, Y0 ∈ S2, there exist couplings of BM on S2 s.t. for all t ≥ 0 we have

a) |Xt − Yt| =√

4− |y + x|2 e−t ↗ 2 (distance-increasing coupling)

b) |Xt − Yt| = |y− x| e−t/2 ↘ 0 (distance-decreasing coupling)

c) |Xt − Yt| = |y− x| = const (fixed-distance coupling= “translation coupling”).

M. N. Pascu (Transilvania Univ) Couplings of RBM 28/31 14.09.2012 28 / 31

Page 100: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

Thank you!

M. N. Pascu (Transilvania Univ) Couplings of RBM 29/31 14.09.2012 29 / 31

Page 101: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

References

R. Atar, K. Burdzy, On Neumann eigenfunctions in lip domains, J. Amer. Math. Soc. 17 (2004) , pp. 243 – 265.

R. Atar, K. Burdzy, Mirror couplings and Neumann eigenfunctions, Indiana Univ. Math. J. 57 (2008), No. 3, pp. 1317 – 1351.

K. Burdzy, W. S. Kendall, Efficient Markovian couplings: Examples and counterexamples, Ann. Appl. Probab. 10 (2000), No. 2, pp.

362 – 409.

I. Chavel, Heat diffusion in insulated convex domains, J. London Math. Soc. (2) 34 (1986), No. 3, pp. 473 – 478.

M. Cranston, Gradient estimates on manifolds using coupling, J. Funct. Anal. 99 (1991), No.1, 110 – 124.

W. Doeblin, Expose de la Theorie des Chaines simples constantes de Markoff aun nombre fini d’Etats, Rev. Math. de l’Union

lnterbalkanique 2 (1938), pp. 77 – 105.

B. Kawohl, Rearrangements and convexity of level sets in PDE, Lecture Notes in Mathematics, Vol. 1150, Springer-Verlag, 1985.

W. S. Kendall, Nonnegative Ricci curvature and the Brownian coupling property, Stochastics 19 (1986), No. 1-2, pp. 111 – 129.

W. S. Kendall, Coupled Brownian motions and partial domain monotonicity for the Neumann heat kernel, J. Funct. Anal. 86 (1989),

No. 2, pp. 226 – 236.

M. N. Pascu (Transilvania Univ) Couplings of RBM 30/31 14.09.2012 30 / 31

Page 102: Mihai N. Pascu - impan.pl · Mihai N. Pascu Transilvania University of Braso¸ v, Romania 6th International Conference on Stochastic Analysis and Its Applications Bedlewo, Poland,

References

M. N. Pascu, Scaling coupling of reflecting Brownian motions and the hot spots problem, Trans. Amer. Math. Soc. 354 (2002), No. 11,

pp. 4681 – 4702.

M. N. Pascu, Monotonicity properties of Neumann heat kernel in the ball, J. Funct. Anal. 260 (2011), No. 2, pp. 490 – 500.

M. N. Pascu, Mirror coupling of reflecting Brownian motion and an application to Chavel’s conjecture, Electron. J. Probab. 16 (2011),

No. 18, pp. 504 – 530.

M. N. Pascu (Transilvania Univ) Couplings of RBM 31/31 14.09.2012 31 / 31