Bloch Electron in a Strong Magnetic Field
Adiabatic Derivation of the Harper ModelGiuseppe De Nittis
Mathematical Physics Sector
SISSA International School for Advanced Studies, Trieste
IIIrdMathematicalMethods inQuantumMechanics
Bressanone, February 16-21, 2009
supervisor:
prof.Gianfausto Dell’Antonio based on joint work with:
Gianluca Panati & Frédéric Faure
Outline
1 Introduction
An overview on our research project
2 The space adiabatic perturbation theory The philosophy
The ingredients
3 Adiabatic theory for the strong magnetic field regime
The physical model
Separation of the scales and adiabatic parameter
Formal expansion of the semiclassical symbol: orderδ4approximation
The statement of the main results
4 Is the periodic magnetic field important?
The research project consists in the solution of 3 different (at mathematical level) but related problems:
Adiabatic problem. Starting from the physical model (free electron in a 2D lattice with an orthogonal uniform magnetic field), and using the
space adiabatic perturbation theoryof Panati, Spohn and Teufel[PST1],
[PST2], one wants to deduce rigorously two effective models: theHarper
modelin the strong limit (B−1→0) and theHofstdater modelin the
weak limit (B→0).
Algebraic problem.Using theC∗-algebraic framework one wants to
show that the two models areisomorphicat algebraic level
(isospectrality) butnot unitarly equivalent(different spectral type).
Geometric duality.As a consequence of the non unitary equivalence and using the tools of the differential geometry one wants to show that the two models are characterized by a different topology. This amounts to
prove that the two models have different values of thefirst Chern class
related by theTKNN formulaMCHof+NCHar=1 when the adiabatic
parameter is rationalM/N. This explains the structure of the twocolor
coded quantum butterflies [Av].
The research project consists in the solution of 3 different (at mathematical level) but related problems:
Adiabatic problem. Starting from the physical model (free electron in a 2D lattice with an orthogonal uniform magnetic field), and using the
space adiabatic perturbation theoryof Panati, Spohn and Teufel[PST1],
[PST2], one wants to deduce rigorously two effective models: theHarper
modelin the strong limit (B−1→0) and theHofstdater modelin the
weak limit (B→0).
Algebraic problem.Using theC∗-algebraic framework one wants to
show that the two models areisomorphicat algebraic level
(isospectrality) butnot unitarly equivalent(different spectral type).
Geometric duality.As a consequence of the non unitary equivalence and using the tools of the differential geometry one wants to show that the two models are characterized by a different topology. This amounts to
prove that the two models have different values of thefirst Chern class
related by theTKNN formulaMCHof+NCHar=1 when the adiabatic
parameter is rationalM/N. This explains the structure of the twocolor
The research project consists in the solution of 3 different (at mathematical level) but related problems:
Adiabatic problem. Starting from the physical model (free electron in a 2D lattice with an orthogonal uniform magnetic field), and using the
space adiabatic perturbation theoryof Panati, Spohn and Teufel[PST1],
[PST2], one wants to deduce rigorously two effective models: theHarper
modelin the strong limit (B−1→0) and theHofstdater modelin the
weak limit (B→0).
Algebraic problem.Using theC∗-algebraic framework one wants to
show that the two models areisomorphicat algebraic level
(isospectrality) butnot unitarly equivalent(different spectral type).
Geometric duality.As a consequence of the non unitary equivalence and using the tools of the differential geometry one wants to show that the two models are characterized by a different topology. This amounts to
prove that the two models have different values of thefirst Chern class
related by theTKNN formulaMCHof+NCHar=1 when the adiabatic
parameter is rationalM/N. This explains the structure of the twocolor
coded quantum butterflies [Av].
Color-Coded Quantum Butterflies:
Outline
1 Introduction
An overview on our research project
2 The space adiabatic perturbation theory
The philosophy
The ingredients
3 Adiabatic theory for the strong magnetic field regime
The physical model
Separation of the scales and adiabatic parameter
Formal expansion of the semiclassical symbol: orderδ4approximation
The statement of the main results
4 Is the periodic magnetic field important?
Some comments The effective model
When the system shows aseparationintoslowandfast degrees of freedomcertain dynamical degrees of freedom lose their autonomous status. The fast modes quickly adapt to the slow modes which in turn are
governed by a suitableeffective Hamiltonian. This mechanism is called
adiabatic decoupling. The paradigm is the Born-Oppenheimer approximation for the motion of nuclei.
The slow degrees of freedom are “semiclassical” in the sense that the full
Hamiltonian can be seen as theWeyl quantizationof an operator-valued
symbol on the phase space of the classical degrees of freedom. The origin of decoupling can be traced to a spectral property of the symbol in
the sense that it can be decomposed in arelevant separated part. This can
be associated with analmost invariant subspace(invariant under the
evolution up to errors small to any order inε) of the Hilbert space of the
system.
The a. i. subspace depends onε and is not easily accessible. In order to
obtain a useful description of the effective intraband dynamics (effective
Hamiltonian) we need aunitarymap from the a. i. subspace into an
When the system shows aseparationintoslowandfast degrees of freedomcertain dynamical degrees of freedom lose their autonomous status. The fast modes quickly adapt to the slow modes which in turn are
governed by a suitableeffective Hamiltonian. This mechanism is called
adiabatic decoupling. The paradigm is the Born-Oppenheimer approximation for the motion of nuclei.
The slow degrees of freedom are “semiclassical” in the sense that the full
Hamiltonian can be seen as theWeyl quantizationof an operator-valued
symbol on the phase space of the classical degrees of freedom. The origin of decoupling can be traced to a spectral property of the symbol in
the sense that it can be decomposed in arelevant separated part. This can
be associated with analmost invariant subspace(invariant under the
evolution up to errors small to any order inε) of the Hilbert space of the
system.
The a. i. subspace depends onε and is not easily accessible. In order to
obtain a useful description of the effective intraband dynamics (effective
Hamiltonian) we need aunitarymap from the a. i. subspace into an
easily accessible andε-independentreference subspace.
When the system shows aseparationintoslowandfast degrees of freedomcertain dynamical degrees of freedom lose their autonomous status. The fast modes quickly adapt to the slow modes which in turn are
governed by a suitableeffective Hamiltonian. This mechanism is called
adiabatic decoupling. The paradigm is the Born-Oppenheimer approximation for the motion of nuclei.
The slow degrees of freedom are “semiclassical” in the sense that the full
Hamiltonian can be seen as theWeyl quantizationof an operator-valued
symbol on the phase space of the classical degrees of freedom. The origin of decoupling can be traced to a spectral property of the symbol in
the sense that it can be decomposed in arelevant separated part. This can
be associated with analmost invariant subspace(invariant under the
evolution up to errors small to any order inε) of the Hilbert space of the
system.
The a. i. subspace depends onε and is not easily accessible. In order to
obtain a useful description of the effective intraband dynamics (effective
Hamiltonian) we need aunitarymap from the a. i. subspace into an
Outline
1 Introduction
An overview on our research project
2 The space adiabatic perturbation theory
The philosophy
The ingredients
3 Adiabatic theory for the strong magnetic field regime
The physical model
Separation of the scales and adiabatic parameter
Formal expansion of the semiclassical symbol: orderδ4approximation
The statement of the main results
4 Is the periodic magnetic field important?
Some comments The effective model
I) Thephysicalstate space of the system decomposes (up to a unitary
transformW), as
Hphy−→W L2(Rd)⊗Hf
whereL2(Rd) =:Hsis the state space of theslow degrees of freedom
andHfan (arbitrary separable) state space of thefast degrees of
freedom. Theclassical phase spaceof the slow degree of freedom is thus
T∗Rd'R2dwithz:= (p
s,xs)∈R2d.
II) Thephysical HamiltonianHb (up to a unitary transform), generating the
time-evolution, is given as the Weyl quantization of asemiclassical
symbolHε :T∗Rd→B(D,Hf)whereDis a suitablenormeddomain
on whichHε is bounded. “Semiclassical” means thatHε admits a formal
expansionHε(z)∑jεjHj(z)whoseprincipal symbolisH0(z). III) Constant Gap Condition (CGC):for anyz∈T∗Rdthe spectrum
σ(z)of
H0(z)contains arelevant subsetσ∗(z)which isuniformlyseparated from
the remainder by a gap, namely
inf
z∈T∗Rd dist(σ(z)
I) Thephysicalstate space of the system decomposes (up to a unitary
transformW), as
Hphy−→W L2(Rd)⊗Hf
whereL2(Rd) =:Hsis the state space of theslow degrees of freedom
andHfan (arbitrary separable) state space of thefast degrees of
freedom. Theclassical phase spaceof the slow degree of freedom is thus
T∗Rd'R2dwithz:= (p
s,xs)∈R2d.
II) Thephysical HamiltonianHb (up to a unitary transform), generating the
time-evolution, is given as the Weyl quantization of asemiclassical
symbolHε :T∗Rd→B(D,Hf)whereDis a suitablenormeddomain
on whichHε is bounded. “Semiclassical” means thatHε admits a formal
expansionHε(z)∑jεjHj(z)whoseprincipal symbolisH0(z).
III) Constant Gap Condition (CGC):for anyz∈T∗Rdthe spectrum
σ(z)of
H0(z)contains arelevant subsetσ∗(z)which isuniformlyseparated from
the remainder by a gap, namely
inf
z∈T∗Rd dist(σ(z)
\σ∗(z),σ∗(z)) =Cg>0.
I) Thephysicalstate space of the system decomposes (up to a unitary
transformW), as
Hphy−→W L2(Rd)⊗Hf
whereL2(Rd) =:Hsis the state space of theslow degrees of freedom
andHfan (arbitrary separable) state space of thefast degrees of
freedom. Theclassical phase spaceof the slow degree of freedom is thus
T∗Rd'R2dwithz:= (p
s,xs)∈R2d.
II) Thephysical HamiltonianHb (up to a unitary transform), generating the
time-evolution, is given as the Weyl quantization of asemiclassical
symbolHε :T∗Rd→B(D,Hf)whereDis a suitablenormeddomain
on whichHε is bounded. “Semiclassical” means thatHε admits a formal
expansionHε(z)∑jεjHj(z)whoseprincipal symbolisH0(z). III) Constant Gap Condition (CGC):for anyz∈T∗Rdthe spectrum
σ(z)of
H0(z)contains arelevant subsetσ∗(z)which isuniformlyseparated from
the remainder by a gap, namely
inf
Relevant separated part of the spectrum:
Outline
1 Introduction
An overview on our research project
2 The space adiabatic perturbation theory The philosophy
The ingredients
3 Adiabatic theory for the strong magnetic field regime
The physical model
Separation of the scales and adiabatic parameter
Formal expansion of the semiclassical symbol: orderδ4approximation
The statement of the main results
4 Is the periodic magnetic field important?
TheBloch-Landau Hamiltonian
HBL:=
1
2m
h
−i}~∇−q
c~AΓ(~x)− q c~A(~x)
i2
+VΓ(~x)
It acts on thephysicalHilbert spaceHphy:=L2(R2).Γ⊂R2is the lattice
spanned by{~a;~b}(non orthogonal in general) such that
ΩΓ:= (axby−bxay)>0 is the volume of thefundamental cellMΓof the
lattice.~AΓandVΓareΓ-periodicandsmooth.
~A(~x):=B
2~ez∧~x=
B
2(−y,x) (symmetric gauge)
where~ezis the normalized vector orthogonal toΓ.
TheBloch-Landau Hamiltonian
HBL:=
1
2m
h
−i}~∇−q
c~AΓ(~x)− q c~A(~x)
i2
+VΓ(~x)
It acts on thephysicalHilbert spaceHphy:=L2(R2).Γ⊂R2is the lattice
spanned by{~a;~b}(non orthogonal in general) such that
ΩΓ:= (axby−bxay)>0 is the volume of thefundamental cellMΓof the
lattice.~AΓandVΓareΓ-periodicandsmooth.
~A(~x):=B
2~ez∧~x=
B
2(−y,x) (symmetric gauge)
Experimental setting
Outline
1 Introduction
An overview on our research project
2 The space adiabatic perturbation theory The philosophy
The ingredients
3 Adiabatic theory for the strong magnetic field regime
The physical model
Separation of the scales and adiabatic parameter
Formal expansion of the semiclassical symbol: orderδ4approximation
The statement of the main results
4 Is the periodic magnetic field important?
Let (for the moment)~AΓ=0and denotePx:=−i}∂xandQx:=multiplication
byx(and similarly fory):
HBL:=
1
2m
"
Px+
qB 2cQy
2
+
Py−
qB 2cQx
2#
+VΓ(Qx,Qy)
Define the new variables~X(fast):= (K1,K2)(magnetic momenta),
~X(slow):= (G1,G2)(centre of the cyclotron orbit)
K1:=−
~b∗·~Q
2√ε −σq
√ ε
} ~a·
~P
K2:= ~a∗·~Q
2√ε −σq`
√ ε
}
~b·~P
G1:=
~b∗·~Q
2 −σq
ε
}~a·
~P
G2:=
~a∗·~Q
2 +σq
ε
}
~b·~P
σqis thesign of the chargeq,ε:=|q|Ωc} ΓB =
1
Z
Φ0
ΦB is thesemiclassical
parameter, withZ:=|qe|,Φ0thequantum of fluxandΦBtheflux ofBper unit
cell.~a∗and~b∗satisfies~a·~a∗=1=~b·~b∗and~b·~a∗=0=~a·~b∗(dual vectors).
Let (for the moment)~AΓ=0and denotePx:=−i}∂xandQx:=multiplication
byx(and similarly fory):
HBL:=
1
2m
"
Px+
qB 2cQy
2
+
Py−
qB 2cQx
2#
+VΓ(Qx,Qy)
Define the new variables~X(fast):= (K1,K2)(magnetic momenta),
~X(slow):= (G1,G2)(centre of the cyclotron orbit)
K1:=−
~b∗·~Q
2√ε −σq
√ ε
} ~
a·~P
K2:= ~a∗·~Q
2√ε −σq`
√ ε
}
~b·~P
G1:=
~b∗·~Q
2 −σq
ε
}~a·
~P
G2:=
~a∗·~Q
2 +σq
ε
}
~b·~P
σqis thesign of the chargeq,ε:=|q|Ωc} ΓB =
1
Z
Φ0
ΦB is thesemiclassical
parameter, withZ:=|qe|,Φ0thequantum of fluxandΦBtheflux ofBper unit
With respect the new variables we have:
HBL0 := 1
E0
HBL=
1
εΞ(K1,K2) +V G2+ √
ε K2,G1−
√ ε K1
.
The new variables make evident theadiabatic separationbetween slow
and fast degrees of freedom, indeed
[K1;K2] =iσq1, [G1;G2] =iσqε1, [Kj;Gk] =0, j,k=1,2
They are dimensionless, then we can factorize all the physical constants
withE0:= }
2
mΩΓ which fixes anatural unit of energy. Fixed the energy
scale, the dynamics is governed by thedimensionlessHamiltonianH0BL.
The functionV is related with theΓ-periodic potentialVΓby the relation
V(~a∗·~x,~b∗·~x):=E1
0VΓ(~x); it is dimensionless andbi-periodicwith
periods 1, i.e.V(x+1,y) =V(x,y+1) =V(x,y).
Ξ:= 1
2ΩΓ
|~a|2K
22+|~b|2K12−~a·~b{K1;K2}
with discrete spectrum
{λn:= (n+12) : n∈N}(Landau levels).
With respect the new variables we have:
HBL0 := 1
E0
HBL=
1
εΞ(K1,K2) +V G2+ √
ε K2,G1−
√ ε K1
.
The new variables make evident theadiabatic separationbetween slow
and fast degrees of freedom, indeed
[K1;K2] =iσq1, [G1;G2] =iσqε1, [Kj;Gk] =0, j,k=1,2
They are dimensionless, then we can factorize all the physical constants
withE0:= }
2
mΩΓ which fixes anatural unit of energy. Fixed the energy
scale, the dynamics is governed by thedimensionlessHamiltonianH0BL.
The functionV is related with theΓ-periodic potentialVΓby the relation
V(~a∗·~x,~b∗·~x):=E1
0VΓ(~x); it is dimensionless andbi-periodicwith
periods 1, i.e.V(x+1,y) =V(x,y+1) =V(x,y).
Ξ:= 1
2ΩΓ
|~a|2K
22+|~b|2K12−~a·~b{K1;K2}
with discrete spectrum
With respect the new variables we have:
HBL0 := 1
E0
HBL=
1
εΞ(K1,K2) +V G2+ √
ε K2,G1−
√ ε K1
.
The new variables make evident theadiabatic separationbetween slow
and fast degrees of freedom, indeed
[K1;K2] =iσq1, [G1;G2] =iσqε1, [Kj;Gk] =0, j,k=1,2
They are dimensionless, then we can factorize all the physical constants
withE0:= }
2
mΩΓ which fixes anatural unit of energy. Fixed the energy
scale, the dynamics is governed by thedimensionlessHamiltonianHBL0 .
The functionV is related with theΓ-periodic potentialVΓby the relation
V(~a∗·~x,~b∗·~x):=E1
0VΓ(~x); it is dimensionless andbi-periodicwith
periods 1, i.e.V(x+1,y) =V(x,y+1) =V(x,y).
Ξ:= 1
2ΩΓ
|~a|2K
22+|~b|2K12−~a·~b{K1;K2}
with discrete spectrum
{λn:= (n+12) : n∈N}(Landau levels).
With respect the new variables we have:
HBL0 := 1
E0
HBL=
1
εΞ(K1,K2) +V G2+ √
ε K2,G1−
√ ε K1
.
The new variables make evident theadiabatic separationbetween slow
and fast degrees of freedom, indeed
[K1;K2] =iσq1, [G1;G2] =iσqε1, [Kj;Gk] =0, j,k=1,2
They are dimensionless, then we can factorize all the physical constants
withE0:= }
2
mΩΓ which fixes anatural unit of energy. Fixed the energy
scale, the dynamics is governed by thedimensionlessHamiltonianHBL0 .
The functionV is related with theΓ-periodic potentialVΓby the relation
V(~a∗·~x,~b∗·~x):=E1
0VΓ(~x); it is dimensionless andbi-periodicwith
periods 1, i.e.V(x+1,y) =V(x,y+1) =V(x,y).
Ξ:= 1
2ΩΓ
|~a|2K
22+|~b|2K12−~a·~b{K1;K2}
with discrete spectrum
With respect the new variables we have:
HBL0 := 1
E0
HBL=
1
εΞ(K1,K2) +V G2+ √
ε K2,G1−
√ ε K1
.
The new variables make evident theadiabatic separationbetween slow
and fast degrees of freedom, indeed
[K1;K2] =iσq1, [G1;G2] =iσqε1, [Kj;Gk] =0, j,k=1,2
They are dimensionless, then we can factorize all the physical constants
withE0:= }
2
mΩΓ which fixes anatural unit of energy. Fixed the energy
scale, the dynamics is governed by thedimensionlessHamiltonianHBL0 .
The functionV is related with theΓ-periodic potentialVΓby the relation
V(~a∗·~x,~b∗·~x):=E1
0VΓ(~x); it is dimensionless andbi-periodicwith
periods 1, i.e.V(x+1,y) =V(x,y+1) =V(x,y).
Ξ:= 1
2ΩΓ
|~a|2K
22+|~b|2K12−~a·~b{K1;K2}
with discrete spectrum
{λn:= (n+12) : n∈N}(Landau levels).
Lettbe the (slow)microscopic time-scale. The physic of the problem has a
natural (fast)ultramicroscopic time-scalefixed by thecyclotron frequency
ωc:=|mcq|B by the relationτ :=ωct. With respect this new scale the
time-dependent (dimensionless) Schrödinger equation reads
HBL0 ψ=i}
E0
∂
∂tψ=i
}ωc
E0
∂
∂ τψ=i
1
ε ∂
∂ τψ
then therelevant(dimensionless)Hamiltonianfrom the physical viewpoint in
the limit of a strong magnetic field is
εHBL0 =Ξ(K1,K2) +ε V G2+
√
εK2,G1−
√ εK1
.
TheStone-von Neumann uniqueness Theoremassures that there exists a unitary mapW:Hphy−→L2(R)s⊗L2(R)f:=Hs⊗Hfsuch that W:(G1,G2)→(Qs:=σqxs,Ps:−iε ∂s)onHsand
W:(K1,K2)→(Qf:=σqxf,Pf:−i∂f)onHf, moreover the relevant
Hamiltonian reads
b
H:=WεHBL0 W−1=1s⊗Ξ(Qf,Pf) +ε V Ps+
√
ε Pf,Qs−
√ εQf
Lettbe the (slow)microscopic time-scale. The physic of the problem has a
natural (fast)ultramicroscopic time-scalefixed by thecyclotron frequency
ωc:=|mcq|B by the relationτ :=ωct. With respect this new scale the
time-dependent (dimensionless) Schrödinger equation reads
HBL0 ψ=i}
E0
∂
∂tψ=i
}ωc
E0
∂
∂ τψ=i
1
ε ∂
∂ τψ
then therelevant(dimensionless)Hamiltonianfrom the physical viewpoint in
the limit of a strong magnetic field is
εHBL0 =Ξ(K1,K2) +ε V G2+
√
εK2,G1−
√ εK1
.
TheStone-von Neumann uniqueness Theoremassures that there exists a unitary mapW:Hphy−→L2(R)s⊗L2(R)f:=Hs⊗Hfsuch that W:(G1,G2)→(Qs:=σqxs,Ps:−iε ∂s)onHsand
W:(K1,K2)→(Qf:=σqxf,Pf:−i∂f)onHf, moreover the relevant
Hamiltonian reads
b
H:=WεHBL0 W−1=1s⊗Ξ(Qf,Pf) +ε V Ps+
√
ε Pf,Qs−
√ εQf
Outline
1 Introduction
An overview on our research project
2 The space adiabatic perturbation theory The philosophy
The ingredients
3 Adiabatic theory for the strong magnetic field regime
The physical model
Separation of the scales and adiabatic parameter
Formal expansion of the semiclassical symbol: orderδ4approximation
The statement of the main results
4 Is the periodic magnetic field important?
Summarizing:
the von NeumannWprovides a decomposition of the physical Hilbert
spaceHphy→Hs⊗Hfrelated to the existence of a separation of scales;
the physically relevant HamiltonianHb, acting onHs⊗Hf, can be seen as
theWeyl-quantizationof the symbol
Hδ(ps,xs):=Ξ(Qf,Pf) +δ2V(ps+δ Pf,qs−δ Qf)
whereδ :=
√
ε and the quantization rule isps→Ps,xs→Qs;
the symbolHδ is a function onT∗Rwith values in theunbounded
operators onHfbut can be see as a symbol with values inB(D,Hf)
which is the Banach space ofbounded operatorsfrom the Hilbert space
D(the domain ofΞwith thegraph-norm) into the Hilbert spaceHf;
the principal symbol isH0=Ξwhich is “morally” an harmonic
oscillator and its spectrum satisfies theCGC.
The last step is to show in what senseHδ is “semiclassical”.
Summarizing:
the von NeumannWprovides a decomposition of the physical Hilbert
spaceHphy→Hs⊗Hfrelated to the existence of a separation of scales;
the physically relevant HamiltonianHb, acting onHs⊗Hf, can be seen as
theWeyl-quantizationof the symbol
Hδ(ps,xs):=Ξ(Qf,Pf) +δ2V(ps+δ Pf,qs−δ Qf)
whereδ :=
√
ε and the quantization rule isps→Ps,xs→Qs;
the symbolHδ is a function onT∗Rwith values in theunbounded
operators onHfbut can be see as a symbol with values inB(D,Hf)
which is the Banach space ofbounded operatorsfrom the Hilbert space
D(the domain ofΞwith thegraph-norm) into the Hilbert spaceHf;
the principal symbol isH0=Ξwhich is “morally” an harmonic
oscillator and its spectrum satisfies theCGC.
Summarizing:
the von NeumannWprovides a decomposition of the physical Hilbert
spaceHphy→Hs⊗Hfrelated to the existence of a separation of scales;
the physically relevant HamiltonianHb, acting onHs⊗Hf, can be seen as
theWeyl-quantizationof the symbol
Hδ(ps,xs):=Ξ(Qf,Pf) +δ2V(ps+δ Pf,qs−δ Qf)
whereδ :=
√
ε and the quantization rule isps→Ps,xs→Qs;
the symbolHδ is a function onT∗Rwith values in theunbounded
operators onHfbut can be see as a symbol with values inB(D,Hf)
which is the Banach space ofbounded operatorsfrom the Hilbert space
D(the domain ofΞwith thegraph-norm) into the Hilbert spaceHf;
the principal symbol isH0=Ξwhich is “morally” an harmonic
oscillator and its spectrum satisfies theCGC.
The last step is to show in what senseHδ is “semiclassical”.
Summarizing:
the von NeumannWprovides a decomposition of the physical Hilbert
spaceHphy→Hs⊗Hfrelated to the existence of a separation of scales;
the physically relevant HamiltonianHb, acting onHs⊗Hf, can be seen as
theWeyl-quantizationof the symbol
Hδ(ps,xs):=Ξ(Qf,Pf) +δ2V(ps+δ Pf,qs−δ Qf)
whereδ :=
√
ε and the quantization rule isps→Ps,xs→Qs;
the symbolHδ is a function onT∗Rwith values in theunbounded
operators onHfbut can be see as a symbol with values inB(D,Hf)
which is the Banach space ofbounded operatorsfrom the Hilbert space
D(the domain ofΞwith thegraph-norm) into the Hilbert spaceHf;
the principal symbol isH0=Ξwhich is “morally” an harmonic
oscillator and its spectrum satisfies theCGC.
Summarizing:
the von NeumannWprovides a decomposition of the physical Hilbert
spaceHphy→Hs⊗Hfrelated to the existence of a separation of scales;
the physically relevant HamiltonianHb, acting onHs⊗Hf, can be seen as
theWeyl-quantizationof the symbol
Hδ(ps,xs):=Ξ(Qf,Pf) +δ2V(ps+δ Pf,qs−δ Qf)
whereδ :=
√
ε and the quantization rule isps→Ps,xs→Qs;
the symbolHδ is a function onT∗Rwith values in theunbounded
operators onHfbut can be see as a symbol with values inB(D,Hf)
which is the Banach space ofbounded operatorsfrom the Hilbert space
D(the domain ofΞwith thegraph-norm) into the Hilbert spaceHf;
the principal symbol isH0=Ξwhich is “morally” an harmonic
oscillator and its spectrum satisfies theCGC.
The last step is to show in what senseHδ is “semiclassical”.
“Formally” one has the following expansion:
δ2V(ps+δ Pf,qs−δ Qf) =
+∞
∑
n,m=−∞
δ2vn,mei2π(nps+mxs)ei2π δ(nPf−mQf)
= +∞
∑
n,m=−∞
δ2vn,mei2π(nps+mxs)
+∞
∑
j=0
(i2π δ)j
j! (nPf−mQf)
j ! = +∞
∑
j=0δj+2 (i2π)
j
j!
+∞
∑
n,m=−∞vn,mei2π(nps+mxs)(nPf−mQf)j
!
= +∞
∑
j=0
δj+2Hj+2(ps,xs)
the derivation is rigorous since:(i)Vis bi-periodic and smooth (Fourier
expansion),(ii)there exists a dense setF⊂Hfofanalytic vectors
(exponential expansion),(iii)the double sum converges in norm on the dense
setF.Danger!!whenjincreasesHj+2becomes “more unbounded” and the
domains of definition “shrink”. The domain ofHk withk>4 is too small
compared to the domainDofΞand so we have problems with the definition
of a common domain of selfadjointness.
“Formally” one has the following expansion:
δ2V(ps+δ Pf,qs−δ Qf) =
+∞
∑
n,m=−∞
δ2vn,mei2π(nps+mxs)ei2π δ(nPf−mQf)
= +∞
∑
n,m=−∞
δ2vn,mei2π(nps+mxs)
+∞
∑
j=0
(i2π δ)j
j! (nPf−mQf)
j ! = +∞
∑
j=0δj+2 (i2π)
j
j!
+∞
∑
n,m=−∞vn,mei2π(nps+mxs)(nPf−mQf)j
!
= +∞
∑
j=0
δj+2Hj+2(ps,xs)
the derivation is rigorous since:(i)V is bi-periodic and smooth (Fourier
expansion),(ii)there exists a dense setF⊂Hfofanalytic vectors
(exponential expansion),(iii)the double sum converges in norm on the dense
setF.Danger!!whenjincreasesHj+2becomes “more unbounded” and the
domains of definition “shrink”. The domain ofHk withk>4 is too small
compared to the domainDofΞand so we have problems with the definition
To solve this problem we define theorderδ4approximate symbol
e
Hδ(ps,xs):=Ξ+ 4
∑
j=2
δjHj(ps,xs)
which is selfadjoint on the domainDofΞand we define theremainderas
Rδ(ps,xs):=Hδ(ps,xs)−Heδ(ps,xs).
Proposition(key argument)
The remainderRδ is aB(D,Hf)-valued symbol of orderO(δ4), namely
kRδ(ps,xs)kB(D,Hf)6δ4Cfor all(ps,xs)∈R2and for a suitable constant
C>0. Moreover ifπr:=∑mi=1|ψkiihψki|is the projection on the subspace spanned by the finite family{ψki}m
i=1of eigenfunctions ofΞ, then one has
kRδ(ps,xs)πrkB(Hf)=kπrRδ(ps,xs)kB(Hf)6δ 5C
k,for all(ps,xs)∈R2and
for a suitable constantCk>0, namelyRδπr,πrRδ and[Rδ;πr]are a
B(Hf)-valued symbols of orderO(δ5).
To solve this problem we define theorderδ4approximate symbol
e
Hδ(ps,xs):=Ξ+ 4
∑
j=2
δjHj(ps,xs)
which is selfadjoint on the domainDofΞand we define theremainderas
Rδ(ps,xs):=Hδ(ps,xs)−Heδ(ps,xs).
Proposition(key argument)
The remainderRδ is aB(D,Hf)-valued symbol of orderO(δ4), namely
kRδ(ps,xs)kB(D,Hf)6δ4Cfor all(ps,xs)∈R2and for a suitable constant
C>0. Moreover ifπr:=∑mi=1|ψkiihψki|is the projection on the subspace spanned by the finite family{ψki}m
i=1of eigenfunctions ofΞ, then one has
kRδ(ps,xs)πrkB(Hf)=kπrRδ(ps,xs)kB(Hf)6δ 5C
k,for all(ps,xs)∈R2and
for a suitable constantCk>0, namelyRδπr,πrRδ and[Rδ;πr]are a
Outline
1 Introduction
An overview on our research project
2 The space adiabatic perturbation theory The philosophy
The ingredients
3 Adiabatic theory for the strong magnetic field regime
The physical model
Separation of the scales and adiabatic parameter
Formal expansion of the semiclassical symbol: orderδ4approximation The statement of the main results
4 Is the periodic magnetic field important?
Some comments The effective model
Now we can give the statement of the space adiabatic theorem for the strong field limit:
I) Existence of the almost invariant subspace
There exists an orthogonal projectionΠε ∈B(Hs⊗Hf)such that
[Hb
∼
;Πδ] =O0(δ∞), [Hb;Πδ] =O0(δ5) for δ→0.
MoreoverΠδ =πb+O0(δ
∞), where
b
πis the Weyl quantization of a
semiclassical symbolπ(ps,xs)∑jδj πj(ps,xs). The principal partπ0of the
symbolπis the spectral projection ofΞcorresponding to the given isolated
family of Landau bands{λki}m
Now we can give the statement of the space adiabatic theorem for the strong field limit:
I) Existence of the almost invariant subspace
There exists an orthogonal projectionΠε ∈B(Hs⊗Hf)such that
[Hb
∼
;Πδ] =O0(δ∞), [Hb;Πδ] =O0(δ5) for δ →0.
MoreoverΠδ =πb+O0(δ
∞), where
b
πis the Weyl quantization of a
semiclassical symbolπ(ps,xs)∑jδjπj(ps,xs). The principal partπ0of the
symbolπis the spectral projection ofΞcorresponding to the given isolated
family of Landau bands{λki}m
i=1, namelyπ0:=∑mi=1|ψkiihψki|whereψki is thekitheigenfunction ofΞ.
II) Reference subspace and intertwining unitary operator
Letπr:=π0(ps,xs)for all(ps,xs)∈R2, letΠr:=1Hs⊗πr∈B(Hs⊗Hf)be
its Weyl quantization andK:=RanΠr. EvidentlyK'L2(R)s⊗CmsinceΠr
is am-dimensional projection. There exists a unitary operator
Uδ ∈B(Hs⊗Hf)such that
Πr:=Uδ Πδ Uδ −1
andUδ =
b
u+O0(δ∞),
whereu∑jδjujhas principal symbolu0=1Hf.
For the last part of the theorem we need to define the following second order differential operator
D2
(~a,~b):=
|~a|2 ∂
2
∂x2s −2~a·
~b ∂2
∂xs∂ps
+|~b|2 ∂
2
∂p2s
,
when~a·~b=0 and|~a|=|~b|=1 (square lattice) thenD2
II) Reference subspace and intertwining unitary operator
Letπr:=π0(ps,xs)for all(ps,xs)∈R2, letΠr:=1Hs⊗πr∈B(Hs⊗Hf)be
its Weyl quantization andK:=RanΠr. EvidentlyK'L2(R)s⊗CmsinceΠr
is am-dimensional projection. There exists a unitary operator
Uδ ∈B(Hs⊗Hf)such that
Πr:=Uδ Πδ Uδ −1
andUδ =
b
u+O0(δ∞),
whereu∑jδjujhas principal symbolu0=1Hf.
For the last part of the theorem we need to define the following second order differential operator
D2
(~a,~b):=
|~a|2 ∂
2
∂x2s −2~a·
~b ∂2
∂xs∂ps
+|~b|2 ∂
2
∂p2s
,
when~a·~b=0 and|~a|=|~b|=1 (square lattice) thenD2
(~a,~b)=4.
III) The effective Hamiltonian in a single band reference subspace
Consider the case in which the isolated family of Landau bands reduces to a
single Landau bandλn, namelyπr=|ψnihψn|. Lethandehbe the
resummations of the formal symbolsu]π]Hδ]π]u
−1andu]
π]Heδ]π]u−1
respectively and denote bybh=:Hbeff
δ (effective Hamiltonian) andbh
∼their
quantization. One has thatHbδeffandbh∼are inB(Hs⊗Hf)and
b
h∼−Hbeff
δ =O0(δ
5). Moreover[
b
h∼;Πr] = [Hbeff
δ ;Πr] =0 hence they can be
seen as elements ofB(K), namely as a bounded operators onL2(R)s. Then
B(Hs⊗Hf)3ΠδHb∼δΠδ Uδ
−→ bh∼+O0(δ∞) =Hbeff
δ +O0(δ
5)∈B(L2(R)s).
Up to the orderδ4on has that
b
Hδeff=λn1Hs+ε Vb(Ps,Qs) +ε2
1 ΩΓ λn 2 \ D2
(~a,~b)(V)(Ps,Qs) +O0
ε
5 2
whereδ2=ε.VbandD\2
(~a,~b)(V)are the Weyl quantization of the bi-periodic
symbolsV(ps,xs)andD2
Outline
1 Introduction
An overview on our research project
2 The space adiabatic perturbation theory The philosophy
The ingredients
3 Adiabatic theory for the strong magnetic field regime
The physical model
Separation of the scales and adiabatic parameter
Formal expansion of the semiclassical symbol: orderδ4approximation
The statement of the main results
4 Is the periodic magnetic field important?
Some comments
The effective model
This result extends (non-orthogonal lattice,periodic magnetic field) and reinforce (local asymptotic equivalence vs. local isospectrality) a
previous result of Helffer and Sj˝ostrand[HS].
The first non-trivial order in the expansion ofHbeff
δ isδ 2=
ε.
The term of orderδ vanishes sinceH1(ps,xs) =0identically and the
term of orderδ3vanishes for technical reasons which can be
summarized in the fact thatH3(ps,xs)is anodd polynomialinPfandQf.
If the periodic vector potential is switched; i.e.~AΓ6=0, then
H1(ps,xs)6=0but since it is odd inPfandQfit gives no contribution to
b
Hδeffin the theory for asingle decoupled Landau band.
If we consider apair of consecutive decoupled Landau bandthen also the
odd terms give contribution toHbeff
δ , hence the leading order in the
perturbation isδ =: √
ε if we consider also the effects due to the
This result extends (non-orthogonal lattice,periodic magnetic field) and reinforce (local asymptotic equivalence vs. local isospectrality) a
previous result of Helffer and Sj˝ostrand[HS].
The first non-trivial order in the expansion ofHbeff
δ isδ 2=
ε.
The term of orderδ vanishes sinceH1(ps,xs) =0identically and the
term of orderδ3vanishes for technical reasons which can be
summarized in the fact thatH3(ps,xs)is anodd polynomialinPfandQf.
If the periodic vector potential is switched; i.e.~AΓ6=0, then
H1(ps,xs)6=0but since it is odd inPfandQfit gives no contribution to
b
Hδeffin the theory for asingle decoupled Landau band.
If we consider apair of consecutive decoupled Landau bandthen also the
odd terms give contribution toHbeff
δ , hence the leading order in the
perturbation isδ =: √
ε if we consider also the effects due to the
microscopical magnetic field caused by the the crystal nuclei.
This result extends (non-orthogonal lattice,periodic magnetic field) and reinforce (local asymptotic equivalence vs. local isospectrality) a
previous result of Helffer and Sj˝ostrand[HS].
The first non-trivial order in the expansion ofHbeff
δ isδ 2=
ε.
The term of orderδ vanishes sinceH1(ps,xs) =0identically and the
term of orderδ3vanishes for technical reasons which can be
summarized in the fact thatH3(ps,xs)is anodd polynomialinPfandQf.
If the periodic vector potential is switched; i.e.~AΓ6=0, then
H1(ps,xs)6=0but since it is odd inPfandQfit gives no contribution to
b
Hδeffin the theory for asingle decoupled Landau band.
If we consider apair of consecutive decoupled Landau bandthen also the
odd terms give contribution toHbeff
δ , hence the leading order in the
perturbation isδ =: √
ε if we consider also the effects due to the
This result extends (non-orthogonal lattice,periodic magnetic field) and reinforce (local asymptotic equivalence vs. local isospectrality) a
previous result of Helffer and Sj˝ostrand[HS].
The first non-trivial order in the expansion ofHbeff
δ isδ 2=
ε.
The term of orderδ vanishes sinceH1(ps,xs) =0identically and the
term of orderδ3vanishes for technical reasons which can be
summarized in the fact thatH3(ps,xs)is anodd polynomialinPfandQf.
If the periodic vector potential is switched; i.e.~AΓ6=0, then
H1(ps,xs)6=0but since it is odd inPfandQfit gives no contribution to
b
Heff
δ in the theory for asingle decoupled Landau band.
If we consider apair of consecutive decoupled Landau bandthen also the
odd terms give contribution toHbeff
δ , hence the leading order in the
perturbation isδ =: √
ε if we consider also the effects due to the
microscopical magnetic field caused by the the crystal nuclei.
This result extends (non-orthogonal lattice,periodic magnetic field) and reinforce (local asymptotic equivalence vs. local isospectrality) a
previous result of Helffer and Sj˝ostrand[HS].
The first non-trivial order in the expansion ofHbeff
δ isδ 2=
ε.
The term of orderδ vanishes sinceH1(ps,xs) =0identically and the
term of orderδ3vanishes for technical reasons which can be
summarized in the fact thatH3(ps,xs)is anodd polynomialinPfandQf.
If the periodic vector potential is switched; i.e.~AΓ6=0, then
H1(ps,xs)6=0but since it is odd inPfandQfit gives no contribution to
b
Heff
δ in the theory for asingle decoupled Landau band.
If we consider apair of consecutive decoupled Landau bandthen also the
odd terms give contribution toHbeff
δ , hence the leading order in the
perturbation isδ =: √
ε if we consider also the effects due to the
Outline
1 Introduction
An overview on our research project
2 The space adiabatic perturbation theory The philosophy
The ingredients
3 Adiabatic theory for the strong magnetic field regime
The physical model
Separation of the scales and adiabatic parameter
Formal expansion of the semiclassical symbol: orderδ4approximation
The statement of the main results
4 Is the periodic magnetic field important?
Some comments
The effective model
Now suppose that~AΓ6=0 and assume (to simplify the notation) thatΓ=Z2
(square lattice). LetgAbe thecoupling constantof the periodic magnetic field
andgV thecoupling constantof the periodic electric field.
In this case the semiclassical symbol is given by
Hδ(ps,xs) =Ξ+δgA f(ps,xs)ba+f(ps,xs)ba
†
+δ2gVV(ps,xs) +O(δ2gA)
wheref =Ax−iAyandba,ba
†are the usualannihilationandcreation operators
of the harmonic oscillatorΞ.
Letπr:=|ψ0ihψ0|+|ψ1ihψ1|,Πrits quantization andK'L2(R)s⊗C2the
reference subspace. The effective dynamics is given (up to a constant term) by
b
Heffδ =
1
21 δgAbf(Ps,Qs) †
δgAbf(Ps,Qs) −121
+O0 δ2gV
.
In generalgAgV, then the approximation is meaningful ifδ ggA
V. The
HamiltonianHbeff
δ is aDirac-likeoperator and
σ(Hbδeff) =± r
1
4+σ(bfbf
†)∪ ±
r
1 4+σ(bf
†
bf)
Now suppose that~AΓ6=0 and assume (to simplify the notation) thatΓ=Z2
(square lattice). LetgAbe thecoupling constantof the periodic magnetic field
andgV thecoupling constantof the periodic electric field.In this case the semiclassical symbol is given by
Hδ(ps,xs) =Ξ+δgA f(ps,xs)ba+f(ps,xs)ba
†
+δ2gVV(ps,xs) +O(δ2gA)
wheref =Ax−iAyandba,ba
†are the usualannihilationandcreation operators
of the harmonic oscillatorΞ.
Letπr:=|ψ0ihψ0|+|ψ1ihψ1|,Πrits quantization andK'L2(R)s⊗C2the
reference subspace. The effective dynamics is given (up to a constant term) by
b
Heffδ =
1
21 δgAbf(Ps,Qs) †
δgAbf(Ps,Qs) −121
+O0 δ2gV
.
In generalgAgV, then the approximation is meaningful ifδ ggA
V. The
HamiltonianHbeff
δ is aDirac-likeoperator and
σ(Hbδeff) =± r
1
4+σ(bfbf
†)∪ ±
r
1 4+σ(bf
†
bf)
.
Now suppose that~AΓ6=0 and assume (to simplify the notation) thatΓ=Z2
(square lattice). LetgAbe thecoupling constantof the periodic magnetic field
andgV thecoupling constantof the periodic electric field.In this case the semiclassical symbol is given by
Hδ(ps,xs) =Ξ+δgA f(ps,xs)ba+f(ps,xs)ba
†
+δ2gVV(ps,xs) +O(δ2gA)
wheref =Ax−iAyandba,ba
†are the usualannihilationandcreation operators
of the harmonic oscillatorΞ.
Letπr:=|ψ0ihψ0|+|ψ1ihψ1|,Πrits quantization andK'L2(R)s⊗C2the
reference subspace. The effective dynamics is given (up to a constant term) by
b
Heffδ =
1
21 δgAbf(Ps,Qs)†
δgAbf(Ps,Qs) −121
+O0 δ2gV
.
In generalgAgV, then the approximation is meaningful ifδ ggA
V. The
HamiltonianHbeff
δ is aDirac-likeoperator and
σ(Hbδeff) =± r
1
4+σ(bfbf
†)∪ ±
r
1 4+σ(bf
†bf)
References
[Av] J. Avron.Colored Hofstadter butterflies.
http://arxiv.org/abs/math-ph/0308030.
[HS] H. Helffer and J. Sj˝ostrand.Équation de Schrödinger avec
champ magnétique et équation de HarperinLecture Notes in
Physics. Schrödinger operators (Sønderborg, 1988),345,
118–198 (1989). Springer.
[PST1] G. Panati, H. Spohn and S. Teufel.Effective dynamics for Bloch
electrons: Peierls substitution and beyond. Comm. Math. Phys.
242, 547-578 (2003).
[PST2] G. Panati, H. Spohn and S. Teufel.Space-Adiabatic
Perturbation Theory. Adv. Theor. Math. Phys. 7, 145-204
(2003).