This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.Lecture 3 Representations of the Heisenberg Group
The Heisenberg group, as many other things, is worth to be seen in
action. We are going to describe various form of its actions on linear
spaces, that is, representations of ℍn.
3.1 Left Regular Representations and Its Subrepresentations
As usual, we can extend geometrical action of ℍ1 on
itself by left
shift to a linear representation
Λ(g): f(g′) ↦ f(g−1g′), g,g′∈ℍ1
(1) |
on a certain linear space of functions.
Exercise 1
Check that the Lebesgue measure dg=
ds dx dy on
ℍ
1∼ℝ
3 is invariant under the left
shift (1).
Thus the (left invariant) Haar measure on ℍ1
coincides with the Lebesgue measure. The same measure is also
invariant under the right shifts, thus ℍ1 is
unimodular. Consequently, the
action (1) on
L2(ℍ,dg) is unitary, it is called the
left regular representation. Moreover, the
action (1) on the Banach space
Lp(ℍ,dg) for any 1≤ p < ∞ is also
an isometry.
The left regular representation (1) is
reducible on L2(ℍ,dg). To find its
(possibly irreducible) subrepresentations,
we can construct linear representations of ℍ1 by
induction from a character χ
of the centre Z, see Section 1.4 and [159]*§ 13. There are several equivalent forms for the
construction, here we prefer the following one,
cf. § 3.2
and [159]*§ 13 [321]*Ch. 5.
Let χ be a character of Z, that is, a group homomorphism of
Z to the group of unimodular complex numbers. Let
F2χ(ℍ1) be the space of functions on
ℍ1 having the properties:
f(gh)=χ(h)f(g), for all g∈ ℍn,
h∈ Z
(2) |
and
| ∫ | | | ⎪
⎪ | f(0,x,y) | ⎪
⎪ | 2dx dy<∞.
(3) |
Then F2χ(ℍ1) is invariant under the left
shifts (1) and those shifts restricted to
F2χ(ℍ1) make a representation
χ of ℍ1 induced by χ.
Exercise 2
Check that the induced representation is unitary if
F2χ(ℍ
1)
is considered as a Hilbert
space with the norm defined by the integral in (3).
However, the representation χ is not
necessarily irreducible. Indeed, left shifts are commuting with the
right action of the group. Thus, any subspace of null-solutions of a
linear combination aS+∑j=1n (bjXj+cjYj) of
left-invariant vector fields is left-invariant and we can restrict
χ to this subspace. The left-invariant differential
operators define analytic condition for functions, cf. Cor. 6.
Example 3
The function f0(
s,
x,
y)=
ei h s −h(x2 +y2)/4, where
h=2πℏ
is a real number, belongs to
F2χ(ℍ
n)
for the character
χ(
s)=
ei h s. It is also a null solution for all the
operators Xj−
i Yj. The closed linear span of functions
fg=Λ(
g)
f0 is invariant under left shifts and provide a
model for the representation Heisenberg
group, cf. below (7).
3.2 Induced Representations on Homogeneous Spaces
We can also build representations on the homogeneous spaces
using the formula (5). We briefly remind this
alternative construction of induced representations
here [159]*§ 13.2. Consider a subgroup H of a
group G. Let a smooth section s:G/H→ G be a
right inverse of the natural projection p:G→
G/H. Thus any element g∈ G can be uniquely decomposed as
g=s(p(g))*r(g) where the map
r:G→ H is defined by the previous identity. For a
character χ of H we can define a
lifting
Lχ: L2(G/H) →
L2χ(G) as follows:
[Lχ f](g)=χ(r(g))f(p(g))
where f(x)∈ L2(G/H).
(4) |
The image space of the lifting Lχ satisfies
to (2) and is invariant under
left shifts. We also define the pulling
P:L2χ(G) → L2(G/H),
which is a left inverse of the lifting and explicitly can be given,
for example, by [PF](x)=F(s(x)). Then the induced
representation on L2(G/H) is generated by the formula
It implies the formula (5) for induced representations
on the homogeneous spaces:
[χ(g) f](x)= χ(r(g−1 * s(x))) f(g−1· x).
(6) |
The corresponding forms of the induced representations are:
- For H=Z the map r: ℍ1→ Z is
r(s,x,y)=(s,0,0). For the character
χℏ(s,0,0)=e2π i ℏ s, the representation
of ℍ1 on L2(ℝ2) is,
cf. (21):
[ℏ(s,x,y)f] (x′,y′) = e | | f(x′−x,y′−y).
(7) |
The Fourier transform maps this representation to the
Fock–Segal–Bargmann (FSB)
representation (13), see Exercise 5.
- For H=Hx the map r(s,x,y)=(0,x,0). For the
character χ(0,x,0)=e2π i ℏ x, the
representation ℍ1 on L2(ℝ2)
is, cf. (22):
[ℏ(s,x,y)f] (s′,y′) = e−2π i ℏ
x f(s′−s−xy′+ | | xy,y′−y).
(8) |
- For H=H′x={(s,0,y)∈ℍ1} the map
r: ℍ1→ H′x is
r(s,x,y)=(s−1/2xy,0,y). For the character
χℏ(s,0,y)=e2π i (ℏ s+qy), the
representation of ℍ1 on
L2(ℝ1) is, cf. (24):
[ℏ(s,x,y)f](x′)= exp(2π i (ℏ
(−s+yx′− | | xy)−qy)) f(x′−x).
(9) |
For q=0, this a key to the Schrödinger representation of
the Heisenberg group, which is obtaned by the Fourier transform.
We will see below that all these (and many others) representations
with the same value ℏ are unitary equivalent.
3.3 Co-adjoint Representation and Method of Orbits
Let h1* be the dual space to h1, that
is the space of all linear functional on
h1∼ℝ3. For physical reasons, we use letter
(h ,q,p) to denote bi-orthonormal coordinates to the
exponential ones (s,x,y) on hn.
The inner automorphism from the Exercise 1 map the
unit (0,0,0) of ℍ1 to itself. Thus inner automorphisms generate a
transformation of the tangent space at (0,0,0) to itself, which is
a linear map given by the same formula (16).
Exercise 4
Check that the adjoint of the linear
transformation (16) is:
ad*(s,x,y): (h ,q,p) ↦ (h , q+h y,
p−h x),
(10) |
where (
s,
x,
y)∈ ℍ
n.
The above map is called the co-adjoint
representation [159, § 15.1] Ad*:
hn* → hn* of ℍn.
There are two types of orbits in (10) for
Ad*, i.e. Euclidean spaces ℝ2n and single
points:
|
Oh | = | {(h, q,p): for a fixed
h≠ 0, (q,p) ∈ ℝ2n}, | (11) |
O(q,p) | = | {(0,q,p): for a fixed
(q,p)∈ ℝ2n}.
| (12) |
|
All complex representations are induced [159, § 13] by a
character χh(s,0,0)=ei h s of the
centre of ℍn generated by
(h,0,0)∈hn* and
shifts (10) from the left on
orbits (11). The explicit
formula respecting physical units [181] is:
χ(s,x,y): fh(q,p) ↦
e −i( h s+qx+py)
fh(q− | | y, p+ | | x).
(13) |
This is the Fock–Segal–Bargmann (FSB) type representation
Exercise 5
Check that the Fourier transform
f(q,p)= | ∫ | | f(x′,y′) e−ℏ(x′p+y′q) d
x′ d y′
|
intertwines representations (7)
and (13).
The Stone–von Neumann Theorem 6 describes
all unitary irreducible representations of ℍn
parametrised up to equivalence by two classes of
orbits (11) and (12):
- The infinite dimensional representations by transformation
χ (13) for h ≠ 0 in
Fock [104, 139] space
F2(Oh)⊂L2(Oh)
of null solutions of Cauchy–Riemann type operators [181].
- The one-dimensional representations as multiplication by
a constant on ℂ=L2(O(q,p))
which drops out from (13) for h =0:
(q,p)(s,x,y): c ↦ e−i(qx+py)c.
(14) |
3.4 Stone–von Neumann Theorem
The following result [159, § 18.4], [104, Chap. 1,
§ 5] reduces
Theorem 6 (The Stone–von Neumann)
Let be a unitary representation of ℍ
n on
a Hilbert space H, such that (
s,0,0)=
e2i h
sI for a non-zero real h. Then H=⊕
Hα
where the Hα’s are mutually orthogonal subspaces of H,
each invariant under , such that that the restriction
|
Hα is unitary equivalent to
h for each α
. In particular, if
is irreducible then is equivalent to h.
Proof.
3.5 Shale–Weil Representation
The Shale–Weil theorem [104]*§ 4.2
[138]*p. 830 states that any representation
ℏ of the Heisenberg groups generates a unitary
oscillator (or metaplectic) representation
SWℏ of the Sp′(2),
the two-fold cover of the symplectic group [104]*Thm. 4.58.
The Shale–Weil theorem allows us also to expand any representation
ℏ of the Heisenberg group to the representation
2ℏ=ℏ⊕SWℏ of the
group Schrödinger group G.
Of course, there is the derived form of the Shale–Weil representation
for g. It can often be explicitly written in contrast
to the Shale–Weil representation.
Example 7
Let ℏ
be the Schrödinger
representation [104]*§ 1.3 of ℍ
1 in
L2(ℝ)
, that is [199]*(3.5):
[χ(s,x,y) f ](q)=e2πiℏ (s−xy/2)
+2πi x q f(q−ℏ y).
|
Thus the action of the derived representation on the Lie algebra
h1 is:
ℏ(X)=2πi q, ℏ(Y)=−ℏ | | ,
ℏ(S)=2πiℏ I.
(15) |
Then the associated Shale–Weil representation of [2]
in
L2(ℝ)
has the
derived action, cf. [322]*(2.2) [104]*§ 4.3:
SWℏ(A) =− | | | − | | ,
SWℏ(B)=− | | | − | | ,
SWℏ(Z)= | | | − | | .
(16) |
We can verify commutators (12) and
(13), (19) for
operators (15–16).
It is also obvious that in this representation the following
algebraic relations hold:
|
SWℏ(A) | = | | (17) |
| = | | |
SWℏ(B) | = | | (18) |
SWℏ(Z) | = | | (19) |
|
Thus it is common in quantum optics to name g as a Lie
algebra with quadratic generators, see [109]*§ 2.2.4.
Note that SWℏ(Z) is the Hamiltonian of the
harmonic oscillator (up to a factor). Then we can consider
SWℏ(B) as the Hamiltonian of a repulsive
(hyperbolic) oscillator. The operator
SWℏ(B−Z/2)=ℏi/4πd2/dq2
is the parabolic analog. A graphical representation of all three
transformations defined by those Hamiltonian is given in
Fig. 11.1 and a further discussion of these
Hamiltonians can be found in [338]*§ 3.8.
An important observation, which is often missed, is that the
three linear symplectic transformations are unitary rotations in the
corresponding hypercomplex algebra, cf. [194]*§ 3. This
means, that the symplectomorphisms generated by operators Z,
B−Z/2, B within time t coincide with the
multiplication of hypercomplex number q+ι p by eι
t, see Section 3.1 and
Fig. 11.1, which is just another illustration of the
Similarity and Correspondence Principle 5.
Example 8
There are many advantages of considering representations of the
Heisenberg group on the phase
space [139]*§ 1.7
[104]*§ 1.6 [74]. A convenient
expression for Fock–Segal–Bargmann
(FSB) representation on the phase space is,
cf. § 4.2.1 and [181]*(2.9)
[74]*(1):
[F(s,x,y) f] (q,p)=
e−2πi(ℏ s+qx+py)
f | ⎛
⎜
⎜
⎝ | q− | | y, p+ | | x | ⎞
⎟
⎟
⎠ | .
(20) |
Then the derived representation of h1 is:
F(X)=−2πi q+ | | ∂p,
F(Y)=−2πi p− | | ∂q,
F(S)=−2πiℏ I.
(21) |
This produces the derived form of the Shale–Weil representation:
SWF(A) = | | ⎛
⎝ | q∂q−p∂p | ⎞
⎠ | ,
SWF(B)=− | | ⎛
⎝ | p∂q+q∂p | ⎞
⎠ | ,
SWF(Z)=p∂q−q∂p.
(22) |
Note that this representation does not contain the parameter
ℏ
unlike the equivalent
representation (16). Thus the FSB model explicitly shows the equivalence
of SWℏ
1 and SWℏ
2
if ℏ
1 ℏ
2>0
[104]*Thm. 4.57.As we will also see below the FSB-type representations in
hypercomplex numbers produce almost the same Shale–Weil
representations.
Last modified: October 28, 2024.