SLIDE 1
The Aharonov-Bohm Effect and the Geometry of Connections Leo Tzou
Speaker is partially supported by the Academy of Finland Speaker is partially supported by NSF Grant DMS-386104 1
SLIDE 2 Double Slit Experiment
- ne shoots electrons through a double slit
2
SLIDE 3
Double Slit Experiment and obtains a symmetric distribution function
3
SLIDE 4
The Aharonov-Bohm Experiment One places a solenoid behind the slit
4
SLIDE 5
The Aharonov-Bohm Experiment without magnetic potential
5
SLIDE 6
The Aharonov-Bohm Experiment and scatter electrons
6
SLIDE 7
The Aharonov-Bohm Experiment and obtains again a symmetric distribution function
7
SLIDE 8 The Aharonov-Bohm Experiment
- We turn on the magnetic potential A in such a way that the magnetic
field E = ∇ × A is completely contained in the solenoid.
- Observe that electrons only pass through regions where E = 0
8
SLIDE 9 The Aharonov-Bohm Experiment
- Surprisingly, the distribution is no longer symmetrical.
- Remember that electrons only passed through regions of vanishing
magnetic field. So what happened?
- The region where E = ∇ × A vanishes is not simply connected.
- The potential A does not vanish up to gauge outside of the solenoid.
- What kind of A gives trivial interference patterns?
9
SLIDE 10 The Magnetic Schr¨
Quantum mechanical effects involving the magnetism is modeled by the magnetic Schr¨
LAu := (d + iA)∗(d + iA)
if A=0
u = 0
- A is real valued 1-form represents the magnetic potential.
- The curl dA is the magnetic field.
- How does A effect the boundary behaviour of the solution?
- Classically only dA matters and not A.
- But when there is topology the Aharonov-Bohm experiment suggests
- therwise.
10
SLIDE 11
This question was studied in the setting of Euclidean setting with cavities by Ballestero-Weder. In the geometric setting the time dependent hyperpolic and boundary spectral data case was done by Kurylev et al. Furthermore recovering the magnetic field from partial boundary measurement was done recently by Emanuvilov-Yamamoto-Uhlmann on planar domains. 11
SLIDE 12 Calder´
- n Problem for (d + iA)∗(d + iA)
- Let (M, g) be a Riemannian manifold with boundary and f ∈ C∞(∂M)
- Assume well-posedness, there exists a unique uf solving
(d + iA)∗(d + iA)uf = 0 uf = f ∂M
- Define the Dirichlet-Neumann map by
ΛA : f → iˆ
n(d + iA)uf
where ˆ n is the normal vector field along the boundary. Does ΛA uniquely determine A?
12
SLIDE 13 Calder´
- n Problem for (d + iA)∗(d + iA)
- Let (M, g) be a Riemannian manifold with boundary and f ∈ C∞(∂M)
- Assume well-posedness, there exists a unique uf solving
(d + iA)∗(d + iA)uf = 0 uf = f ∂M
- Define the Dirichlet-Neumann map by
ΛA : f → iˆ
n(d + iA)uf
where ˆ n is the normal vector field along the boundary. Does ΛA uniquely determine A? NO
13
SLIDE 14 Gauge Invariance
- Let φ ∈ C∞(M) be a real function with φ |∂M= 0.
- Consider the operator LA+dφ = (d + iA + idφ)∗(d + iA + idφ)
- If LAu = 0 then LA+dφe−iφu = 0.
- So ΛA = ΛA+dφ.
Natural Conjecture (false in general): If ΛA1 = ΛA2 then A1 − A2 is exact. This holds only on planar domains
14
SLIDE 15 Simply Connected Domains Lets suppose M =unit disk.
- Observe that LA = (d+iA)∗(d+iA) =(2nd order elliptic)
+idA+
|A|2.
- So by analytic techniques we can show that
ΛA1 = ΛA2 ⇒ dA1 = dA2
- By the fact that M is simply connected,
φ(z) =
z
z0
A1 − A2 is path independent and well defined so A1 = A2 + dφ Thus ΛA1 = ΛA2 ⇔ A1 − A2 = dφ
15
SLIDE 16
This corresponds to the double slit experiment with no topology:
16
SLIDE 17 Topological Obstructions
- On a surface M with genus similar analytic techniques will obtain
ΛA1 = ΛA2 ⇒ d(A1 − A2) = 0
- However, this does not imply A1 − A2 is exact. So,
A1 − A2 is exact ⇒ ΛA1 = ΛA2 A1 − A2 is closed ⇐ ΛA1 = ΛA2
17
SLIDE 18 Topological Obstructions
- On a surface M with genus similar analytic techniques will obtain
ΛA1 = ΛA2 ⇒ d(A1 − A2) = 0
- However, this does not imply A1 − A2 is exact. So,
A1 − A2 is exact ⇒ ΛA1 = ΛA2
- A1 − A2 is closed ⇐ ΛA1 = ΛA2
18
SLIDE 19 Topological Obstructions
- On a surface M with genus similar analytic techniques will obtain
ΛA1 = ΛA2 ⇒ d(A1 − A2) = 0
- However, this does not imply A1 − A2 is exact. So,
A1 − A2 is exact ⇒ ΛA1 = ΛA2 (cohomology of M) A1 − A2 is closed ⇐ ΛA1 = ΛA2
19
SLIDE 20 Topological Obstructions
- On a surface M with genus similar analytic techniques will obtain
ΛA1 = ΛA2 ⇒ d(A1 − A2) = 0
- However, this does not imply A1 − A2 is exact. So,
A1 − A2 is exact ⇒ ΛA1 = ΛA2 ? ⇔ ? A1 − A2 is closed ⇐ ΛA1 = ΛA2
20
SLIDE 21 A Satisfactory Answer (Guillarmou - LT, GAFA 2011) ΛA1 = ΛA2 ⇔ (A1 − A2) ∈ H1(M, ∂M; N) This means that
- d(A1 − A2) = 0 (ie. (A1 − A2) ∈ H1(M, ∂M; R))
- γ(A1 − A2) ∈ 2πN for all closed loops γ.
Corollary ΛA = Λ0 IFF dA = 0 and
- γ A ∈ 2πN for all loops γ.
What motivated us to this condition? The answer is in the geometry of connection.
21
SLIDE 22 Point of View of Parallel Transport Let E = C × M be the trivial complex line bundle over M.
- ∇A := d + iA is a connection acting on this line bundle.
- Let γ be a closed loop and z0 ∈ γ
- Fix v ∈ Ez0
22
SLIDE 23 Point of View of Parallel Transport Let E = C × M be the trivial complex line bundle over M.
- ∇A := d + iA is a connection acting on this line bundle.
- Let γ be a closed loop and z0 ∈ γ
- Parallel transport v along γ by ∇A
23
SLIDE 24 Point of View of Parallel Transport Let E = C × M be the trivial complex line bundle over M.
- ∇A := d + iA is a connection acting on this line bundle.
- Let γ be a closed loop and z0 ∈ γ
- Parallel transport v along γ by ∇A
24
SLIDE 25 Point of View of Parallel Transport Let E = C × M be the trivial complex line bundle over M.
- ∇A := d + iA is a connection acting on this line bundle.
- Let γ be a closed loop and z0 ∈ γ
- Parallel transport v along γ by ∇A
25
SLIDE 26 Point of View of Parallel Transport Let E = C × M be the trivial complex line bundle over M.
- ∇A := d + iA is a connection acting on this line bundle.
- Let γ be a closed loop and z0 ∈ γ
- Parallel transport v along γ by ∇A
26
SLIDE 27 Point of View of Parallel Transport Let E = C × M be the trivial complex line bundle over M.
- ∇A := d + iA is a connection acting on this line bundle.
- Let γ be a closed loop and z0 ∈ γ
- Parallel transport v along γ by ∇A
27
SLIDE 28 Point of View of Parallel Transport Let E = C × M be the trivial complex line bundle over M.
- ∇A := d + iA is a connection acting on this line bundle.
- Let γ be a closed loop and z0 ∈ γ
- Parallel transport v along γ by ∇A
28
SLIDE 29 Point of View of Parallel Transport Let E = C × M be the trivial complex line bundle over M.
- ∇A := d + iA is a connection acting on this line bundle.
- Let γ be a closed loop and z0 ∈ γ
- Parallel transport v along γ by ∇A
29
SLIDE 30 Point of View of Parallel Transport Let E = C × M be the trivial complex line bundle over M.
- ∇A := d + iA is a connection acting on this line bundle.
- Let γ be a closed loop and z0 ∈ γ
- to obtain v′ ∈ Ez0
30
SLIDE 31 Point of View of Parallel Transport Let E = C × M be the trivial complex line bundle over M.
- ∇A := d + iA is a connection acting on this line bundle.
- Let γ be a closed loop and z0 ∈ γ
- to obtain v′ ∈ Ez0
- Solving the ODE for parallel transport yields v′ = (ei
γ A)v 31
SLIDE 32
- Therefore that holonomy of d + iA is equal to that of ∇0 = d iff
- γ A ∈ 2πN
for all closed loops γ
- From the geometric point of view, it is not the exactness of A but
rather the isomorphism of the connections ∇A = d+iA and ∇0 = d that matters.
32
SLIDE 33 Proof of Result ΛA = Λ0 IFF dA = 0 and
- γ A ∈ 2πN for all loops γ.
- Analytic methods show that ΛA = Λ0 ⇒ dA = 0 and A = 0∂M
33
SLIDE 34 Proof of Result ΛA = Λ0 IFF dA = 0 and
- γ A ∈ 2πN for all loops γ.
- Analytic methods show that ΛA = Λ0 ⇒ dA = 0 and A = 0∂M
Consider a closed loop γ on M:
34
SLIDE 35 Proof of Result ΛA = Λ0 IFF dA = 0 and
- γ A ∈ 2πN for all loops γ.
- Analytic methods show that ΛA = Λ0 ⇒ dA = 0 and A = 0∂M
Consider a closed loop γ on M:
Since dA = 0 we can choose any representative of the homology class. 35
SLIDE 36 Proof of Result ΛA = Λ0 IFF dA = 0 and
- γ A ∈ 2πN for all loops γ.
- Analytic methods show that ΛA = Λ0 ⇒ dA = 0 and A = 0∂M
Consider a closed loop γ on M:
so we deform the curve as such so that γ = Γ1 + Γ2 with Γ2 ⊂ ∂M 36
SLIDE 37 Proof of Result ΛA = Λ0 IFF dA = 0 and
- γ A ∈ 2πN for all loops γ.
- Analytic methods show that ΛA = Λ0 ⇒ dA = 0 and A = 0∂M
Consider a closed loop γ on M: We need that
A
=0
∈ 2πN
37
SLIDE 38 Proof of Result ΛA = Λ0 IFF dA = 0 and
- γ A ∈ 2πN for all loops γ.
- Analytic methods show that ΛA = Λ0 ⇒ dA = 0 and A = 0∂M
Consider z0, z1 ∈ ∂M ∩ Γ1 : We need that
A
=0
∈ 2πN
38
SLIDE 39 Proof of Result ΛA = Λ0 IFF dA = 0 and
- γ A ∈ 2πN for all loops γ.
- Analytic methods show that ΛA = Λ0 ⇒ dA = 0 and A = 0∂M
Consider z0, z1 ∈ ∂M ∩ Γ1 Let f ∈ C∞(∂M) such that f(z1) = f(z0) = 1 : We need that
A
=0
∈ 2πN
39
SLIDE 40 Proof of Result
- Let L0w = 0 and LAv = 0 such that
v = w = f ∂M
∂νw = ∂νv
40
SLIDE 41 Proof of Result
- Let ∆w = 0 and LAv = 0 such that
v = w = f ∂M
∂νw = ∂νv Consider an open neighbourhood U of Γ1 We will show that in U one can transform v to w via parallel transport.
41
SLIDE 42 Proof of Result
- Let ∆w = 0 and LAv = 0 such that
v = w = f ∂M
∂νw = ∂νv Consider an open neighbourhood U of Γ1 We will show that in U one can transform v to w via parallel transport.
42
SLIDE 43 L0w = LAv = 0, w = v = f on ∂M, ∂νw = ∂νv
φ(z) =
z
z0
A
43
SLIDE 44 L0w = LAv = 0, w = v = f on ∂M, ∂νw = ∂νv
- Since U is simply connected and dA = 0, we can extend φ to U
with dφ = A
44
SLIDE 45 L0w = LAv = 0, w = v = f on ∂M, ∂νw = ∂νv
- Since U is simply connected and dA = 0, we can extend φ to U
with dφ = A Therefore, if L0w = 0 then e−iφw solves LA(e−iφw) = 0 in U.
45
SLIDE 46 L0w = LAv = 0, w = v = f on ∂M, ∂νw = ∂νv
- Since U is simply connected and dA = 0, we can extend φ to U
with dφ = A Therefore, if L0w = 0 then e−iφw solves LA(e−iφw) = 0 in U.
46
SLIDE 47
47
SLIDE 48
48
SLIDE 49
49
SLIDE 50
50
SLIDE 51
I claim that e−iφw = v in U
51
SLIDE 52
Since A = 0 on ∂M
52
SLIDE 53
Since A = 0 on ∂M
53
SLIDE 54
Since A = 0 on ∂M
54
SLIDE 55
Since A = 0 on ∂M
55
SLIDE 56
Both v and e−iφw solve LA(v) = LA(e−iφw) = 0 in U.
56
SLIDE 57
Unique continuation gives e−iφw = v in U.
57
SLIDE 58
In particular, e−iφ(z1)w(z1) = v(z1).
58
SLIDE 59
In particular, e−iφ(z1)f(z1) = f(z1).
59
SLIDE 60
In particular, e−iφ(z1)1 = 1.
60
SLIDE 61
Therefore, φ(z1) =
z1
z0 A ∈ 2πN
61
SLIDE 62
Generalization to Higher Rank Bundles Let E be a complex bundle of rank n over a Riemann surface M with boundary. Let ∇ be a connection acting on sections of E and V an endomorphism of E. Define the elliptic operator L := ∇∗∇ + V and consider the boundary value problem Luf = 0 M, u = f ∂M Let Λ∇,V : f → ∇νuf |∂M be the Dirichlet-Neumann operator. Thm (Albin-Guillarmou-LT) If Λ∇1,V1 = Λ∇2,V2 then there exists a unitary endomorphism F : E → E such that F −1∇1F = ∇2, F −1V1F = V2
62