Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Eur. Phys. J.

Plus (2011) 126: 4


DOI 10.1140/epjp/i2011-11004-2
THE EUROPEAN
PHYSICAL JOURNAL PLUS
Regular Article

On the Weyl algebra for a particle on a sphere

L. Braccia and L.E. Picassob


Dipartimento di Fisica, Università di Pisa and INFN, Sezione di Pisa, Largo B. Pontecorvo 3, 56127 Pisa, Italy

Received: 12 July 2010


Published online: 18 January 2011 – 
c Società Italiana di Fisica / Springer-Verlag 2011

Abstract. We study the Weyl algebra A pertaining to a particle constrained on a sphere, which is generated
by the coordinates n and by the angular momentum J. A is the algebra E3 of the Euclidean group in
space. We find its irreducible representations by a novel approach, by showing that they are the irreducible
representations (l0 , 0) of so(3, 1), with l0 or −l0 being equal to the Casimir operator J · n. Any integer or
half-integer l0 is allowed. The Hilbert space of a particle of spin S hosts 2S + 1 such representations. J
can be analyzed into the sum L + S, i.e. pure spin states can be identified, provided 2S + 1 irreducible
representations of A are glued together. These results apply to any surface which is diffeomorphic to S2 .

1 Introduction
Quantum mechanics has at its very foundation the representation of the canonical variables of the system which is
investigated, that is a set of operators which are associated to the variables which describe the configuration space and
to the generators of the movements allowed to the system. For example, for a system of n particles in R3 we choose
(j) (j)
3n coordinate operators qi , 1 ≤ j ≤ n, 1 ≤ i ≤ 3, and the associated momenta pi , which obey the commutation
relations
r , qs ] = −iδjk δrs .
[p(j) (k)
(1)
Being among the observables of the theory, coordinates and momenta are required to be self-adjoint operators. By
exponentiation we get unitary operators obeying the Weyl form of commutation relations. In this case it is well known
(von Neumann’s theorem [1,2]) that all the irreducible representations of these relations are equivalent.
In this note we address the same problem for a particle constrained to a sphere. The position of the particle can
be described by the vector r issuing from the centre of the sphere, in terms of which the movements, generated by the
angular momentum J which obeys
[Ji , Jj ] = ieijk Jk , (2)
are described by the commutation rules
[Ji , rj ] = ieijk rk . (3)
If we require that the components of r commute,

[ri , rj ] = 0, (4)

the algebra is closed. Noting that r2 commutes with all the generators, and that αr obeys eqs. (3) and (4) if r does,
we can choose r such that r2 = 1. In the following we denote such r by n and denote by A the algebra defined by
eqs. (2), (3) and (4) with n in place of r. The operator Σ ≡ J · n is a Casimir operator for A.
Our goal is the classification of the irreducible representations of n and J as self-adjoint operators, subject to the
commutation relations (2), (3) and (4). Our results will hold true for any surface S which is diffeomorphic to the
sphere S2 , S = h(S2 ), since the vectors Ji ≡ h∗ Ji generate the movements over S and obey eq. (2), while if we choose
in S the coordinates xi = xi ◦ h−1 , Ji and xj obey the commutation relations (3) and (4).
Equations (2), (3) and (4) define the Lie algebra E3 of the Euclidean group in space. Its irreducible representations
are known, and can be found either by purely algebraic methods [3] or by Frobenius’ method or by the method of
a
e-mail: bracci@df.unipi.it
b
e-mail: picasso@df.unipi.it
Page 2 of 6 The European Physical Journal Plus

group contraction [4]. In this note we classify the irreducible representations by means of a so far unnoticed relation
between E3 and so(3, 1). The operators D ≡ 2i [J2 , n] and J generate a so(3, 1) algebra for which the Casimir operator
D · J vanishes, while the other Casimir operator J2 − D2 is connected to the Casimir operator J · n of E3 by the
equation (J · n)2 = J2 − D2 + I. By exploiting this connection we prove that an E3 -irreducible space is irreducible with
respect to J and D as well. Conversely, given a so(3,√1)-irreducible space for which J · D = 0, with D and J we can
construct two realizations of E3 , such that J · n = ± J2 − D2 + 1, respectively, the space being irreducible for each
of them. We summarize this situation by saying, somewhat improperly, that an irreducible representation of E3 is an
irreducible representation of so(3, 1) with J · D = 0 and conversely. By this connection, we recover the well-known
result that the irreducible representations of E3 are labelled by the eigenvalue σ of the central operator Σ, 2σ ∈ Z,
and that each representation hosts all the angular momenta j = |σ| + n, n ∈ N (once for each representation).
It may be a surprise that for a simply connected configuration space as the sphere there are non-equivalent
representations [5]. The explanation is that the algebra A is a subalgebra of the algebra AS generated by n, L and S,
with S the spin angular momentum and J = L + S. Therefore, it may be expected that an irreducible representation
of AS contains several non-equivalent irreducible representations of A. Actually, for a spinless particle we must have
J · n = 0, and the only allowed representation is the one with σ = 0. Instead, for S = 12 we have σ = ± 12 , and
correspondingly there are two non-equivalent representations of eqs. (2), (3) and (4) and H ≡ H 12 ⊕ H− 12 is irreducible
with respect to AS . For a generic spin S, an irreducible representation of AS contains 2S +1 irreducible representations
of A, with values of σ ranging from S to −S.
In sect. 2 we determine all the irreducible representations of the algebra generated by n, J, subject to eqs. (2), (3)
and (4). Section 3 discusses the case when the algebra pertains to a particle of spin S, and sect. 4 contains our
conclusions.

2 The irreducible representations


We introduce the vector operator D,
i 2 J∧n−n∧J
D≡ [J , n] = , (5)
2 2
such that
[Ji , Dj ] = ieijk Dk . (6)
D is self-adjoint since n and J are.
Since D · J = 0, for the commutator [Di , Dj ] ≡ eijk Ak we find Ai Jr = Ji Ar , whence Ai ∝ Ji and the J, D algebra
is closed. From the explicit calculation we find

[Di , Dj ] = −ieijk Jk . (7)

Hence, J and D generate the Lie algebra so(3, 1). For later use, we also note that the following relation can be verified:

D ∧ J = n J2 − J(J · n). (8)

Since the vectors J and D generate the Lie algebra so(3, 1) of the proper Lorentz group SO(3, 1) (see [6], where J
and D are denoted as H and F, respectively), an irreducible representation of A can be decomposed into irreducible
representations of so(3, 1).
In order to understand which representations of so(3, 1) can appear in a representations of A, we consider the
Casimir operators of so(3, 1), C1 ≡ J · D and C2 ≡ J2 − D2 . As noted above, C1 vanishes,

J · D = 0, (9)

while for C2 we have


J2 − D2 = −I + (J · n)2 = −I + Σ 2 . (10)
Hence, within an irreducible representation of A we must have Σ = σI and J − D = (−1 + σ )I.
2 2 2

In order to exploit eqs. (9) and (10), we recall the relevant properties of the irreducible representations of so(3, 1) [6].
An irreducible representation of so(3, 1) in a Hilbert space V is labelled by a pair of numbers l0 and l1 . l0 can be
assumed to be nonnegative. If V is reduced according to the irreducible representations of the algebra su(2) generated
by J, l0 denotes the minimum value of j appearing in this reduction, and the allowed j’s increase by unit steps up to
a maximum value or to infinity, depending on l1 . If l1 ∈ R and |l1 | − l0 ∈ N+ , the maximum value of j appearing in H
is |l1 | − 1, otherwise all the values j ≥ l0 are present. No value of j appears more than once in V. The representation
of SO(3, 1) is unitary if l1 is an imaginary number, or l0 = 0, l1 ∈ R, |l1 | ≤ 1.
L. Bracci and L.E. Picasso: On the Weyl algebra for a particle on a sphere Page 3 of 6

Within an irreducible representation (l0 , l1 ) the Casimir operators C1 and C2 are expressed in terms of l0 and l1
as follows:

D · J = 2il0 l1 I, (11)
J − D2 = (−1 + l02 + l12 )I.
2
(12)

Equations (9) and (11) imply l0 l1 = 0. Equations (10) and (12) imply l1 = 0. In fact, if l1 = 0, then necessarily
l0 = 0. In this case the space V which hosts an irreducible representation of so(3, 1) contains the vector ξ00 such
that Ji ξ00 = 0, hence in V J · n = 0. Then, by comparing eqs. (10) and (12), we see that l1 vanishes as well. In
conclusion, given an irreducible representation of A in a space H, if the representation is decomposed into irreducible
representations of so(3, 1), only representations of type (l0 , 0) can be present.
It is still to be understood how many irreducible representations of so(3, 1) can appear in an irreducible represen-
tation of A. The answer is simply that only one representation is present, as proved in the following theorem.

Theorem 1. Any irreducible representation of A is an irreducible representation of so(3, 1) of type (l0 , 0). The operator
Σ is either Σ = l0 I or Σ = −l0 I. Conversely, given an irreducible representation (l0 , 0) of so(3, 1), with J and D we
can construct an irreducible representation of A with σ = l0 or σ = −l0 .

Proof. It suffices to prove that a subspace V irreducible with respect to the generators D, J of a (l0 , 0) representation
of so(3, 1) is invariant with respect to n as well, hence it is irreducible with respect to A.
If ξjm is a basis vector for V, from eq. (8) we find

D ∧ Jξjm = j(j + 1)nξjm − σJξjm . (13)

Since V is invariant with respect to J, D, if j = 0 we see that nξjm ∈ V. On the other hand, j = 0 can appear only if
l0 = 0. In this case, by applying D to ξ00 and observing that nξ00 is an eigenvector of J2 with j = 1 and using eq. (5),
we find
−2nξ00 = 2iDξ00 , (14)
hence nξ00 ∈ V. Finally, any (l0 , 0) representation of so(3, 1) can be read as a representation of A, where the action of
n on the vectors of the basis is defined according to eqs. (13) and (14):
1
nξjm = [D ∧ J + σJ]ξjm , j = 0, (15)
j(j + 1)
nξ00 = −iDξ00 , (16)

where σ = ±l0 . As shown in the Appendix, eqs. (15) and (16) yield two vector operators n± which satisfy eqs. (13)
and (14) as well as the commutation relations (2) and (3) and the condition n2± = 1. The construction of D± by
eq. (5) yields D. Thus, a representation of so(3, 1) of type (l0 , 0) is a representation of the algebra A with n = n+ or
n = n− , depending on the choice of σ = l0 or σ = −l0 in eq. (15). 

In conclusion, the irreducible representations of A can be labelled by a signed integer or half-integer σ. |σ| = l0
denotes the irreducible representation (l0 , 0) of so(3, 1), σ specifies the value of Σ = J · n = ±l0 I. Since A = E3 , the
result yields the classification of the unitary irreducible representations of E3 . Since J2 , Σ and Jz commute, the basis
vectors of the irreducible space Hσ can be labelled as ξjm σ
, with j ≥ |σ| and m the eigenvalue of Jz .

3 The case of spin S


For a particle of spin S constrained to the unit sphere the Hilbert space H is the tensor product HL ⊗ HS , with HL
spanned by the eigenvectors of L2 and Lz and HS spanned by the eigenvectors of Sz . In this case J = L + S, Σ = S · n.
In this section first we discuss how H decomposes into subspaces Hσ irreducible with respect to A. Next, we show how
the pure spin states ζSm , i.e. the vectors of HL ⊗ HS whose orbital part has L = 0, can be constructed in terms of
σ
the vectors ξjm transforming according to the irreducible representations of A. This will enable us to define the action
of a spin operator S, which is straightforward when H is viewed as HL ⊗ HS , also on the vectors ξjm σ
. The outcome
of this investigation will be that a spin operator can be defined by gluing together 2S + 1 irreducible representations
of A.
In order to tell the representations of A which are present in H, we note that, since [Si , nj ] = 0, Σ has 2S + 1
eigenvalues σ ranging from −S to S. Thus, we have 2S+1 irreducible representations of A. Alternatively, the irreducible
representations can be counted by observing that the states with angular momentum j ≤ S can be constructed by
Page 4 of 6 The European Physical Journal Plus

composing spin and orbital momentum in 2j +1 different ways, while the states with j ≥ S can be constructed in 2S +1
different ways. Altogether, we find again 2S + 1 irreducible representations. More precisely, if S is integer, we have the
representation σ = 0 and two additional representations σ = ±l0 for any l0 such that 1 ≤ l0 ≤ S; if S is half-integer
we have two representations σ = ±l0 for any l0 such that 12 ≤ l0 ≤ S. As a consequence, the Hilbert space H can be
decomposed into the direct sum of spaces Hσ which host irreducible representations of A: H = ⊕l0 ≤S ⊕σ=±l0 Hσ .
If in HL ⊗ HS we choose a basis consisting of vectors ηjm
L
which are eigenvectors of L2 , J2 and Jz , the pure spin
0 σ
states ζSm are the vectors ηSm . The search for ζSm in terms of ξjm is best illustrated in the simple cases S = 12 and
S = 1.
σ L
The identification of the vectors ξSm in terms of the vectors ηjm amounts to a simple eigenvalue problem for Σ,
1
and for S = 2 we easily find

−1 1   1 1  
ξ 1 m2 = √ η 01 m + η 11 m , ξ 12 m = √ η 01 m − η 11 m , (17)
2 2 2 2 2 2 2 2

while, for S = 1 we have

−1 1 √ 0 √ 1 
ξ1m = √ 2η1m + 3η1m 2
+ η1m , (18)
6
1  0 √ 2 
0
ξ1m = √ η1m − 2η1m , (19)
3

1 √ 0 √ 1 
1
ξ1m = √ 2η1m − 3η1m 2
+ η1m . (20)
6
1 −1 1
For S = from eq. (17) there follows ζ 12 m = √1 [ξ 1 2 + ξ 12 m ], while for S = 1, from eqs. (18), (19) and (20), there
2 2 2m 2
−1
follows ζ1m = √13 [ξ1m 0
+ ξ1m 1
+ ξ1m ].
σ
We state that, by appropriately choosing the phases of the states ξSm , we have
1 
ζSm = √ σ
ξSm . (21)
2S + 1 σ

In order to prove eq. (21), we note that the operator

A = J · (n ∧ S) + i(J · S) − i(J · n)(n · S) (22)

obeys the following commutation relations:

[J · n, A] = A, (23)

[A, A ] = 2{(J · n)[J + S ] − 2(J · n) }.
2 2 3
(24)

The term (J · n)(n · S) = (J · n)2 in eq. (22), which commutes with J · n, makes the commutator [A, A† ] diagonal in
σ
the ξSm basis.
By eq. (23), A† and A act as lowering and raising operators for σ, A† ξSm
σ+1
∝ ξSm
σ
, while eq. (24) implies

A† ξSm
σ+1
A† ξSm
σ+1
σ
ξSm =  ≡ , (25)
f (σ + 1) + . . . + f (S) D(σ)
f (σ) ≡ 4{σS(S + 1) − σ 3 }. (26)

0
Since ηSm σ
=  σ cσ ξSm , if |cσ | = |(ξSm
σ 0
, ηSm σ
)| is independent of σ and the phases of ξSm are so chosen that
cσ = k > 0, then σ ξSmσ
∝ ηSm
0
. Since
0
D(σ)(ηSm σ
, ξSm 0
) = (AηSm σ+1
, ξSm 0
) = (ηSm σ+1
, ξSm )[−iS(S + 1) − i(σ + 1) + i(σ + 1)2 ],

the condition |(ηSm


0 σ
, ξSm )| = |(ηSm
0 σ+1
, ξSm )| requires

D(σ)2 = f (σ + 1) + . . . + f (S) = [S(S + 1) − σ(σ + 1)]2 , (27)

with f (σ) defined in eq. (26). Condition (27) holds for σ = S, and it is easily seen that if it holds for σ + 1 it holds for
σ as well. In conclusion, eq. (21) holds true.
L. Bracci and L.E. Picasso: On the Weyl algebra for a particle on a sphere Page 5 of 6

It is possible to introduce the spin operator S starting from 2S + 1 representations σ = ±l0 , with l0 ≤ S, of A. If
H ≡ ⊕l0 ≤S ⊕σ=±l0 Hσ and ζSm are defined as in eq. (21), the vectors

L
ηjm = L
Pm−p (ni )ζSp (L, m − p; S, p|jm), (28)
p

L
where (L, a; S, b|jm) are Clebsch-Gordan coefficients and Pm−p (ni ) are polynomials in the components of n which
have with J the commutation rules of a tensor operator with j = L, Jz = m − p, the set {ηjm L
} is a basis for H, and
L 2
the vectors ηjm play the role of the eigenstates of J and Jz constructed by composing eigenstates of spin and of the
orbital angular momentum. If S is defined on this basis as an operator commuting with n, such that Si acts onto ζSm
as the i-th component of the spin operator acts onto a state with spin S and Sz = m, clearly S behaves as a vector
and satisfies the commutation relations of an angular momentum. Thus, an analysis of the angular momentum J into
a spin part S and an orbital part L = J − S is possible, provided the appropriate irreducible representations of A are
glued together.

4 Conclusions
We have studied the algebra A of the observables pertaining to a particle constrained to a sphere. The algebra is
generated by the coordinate operators n and by the generators of the movements in the configuration space, that is
the angular momentum J. A coincides with the Lie algebra E3 of the Euclidean group in space. This algebra possesses
the Casimir operator Σ = J · n.
We have classified its irreducible representations by showing that they can be reduced into irreducible representa-
tions of so(3, 1) and that, actually, only one representation, of a definite type, can be present. Indeed, the irreducible
representations of A are irreducible representations (l0 , 0) of the Lie algebra so(3, 1) such that Σ = ±l0 I. Hence, the
irreducible representations of A can be labelled by σ = ±l0 . Any non-negative integer or half-integer l0 is allowed.
Conversely, in any (l0 , 0) irreducible representation of so(3, 1) one can construct an operator n± such that J, D and
n± generate an algebra A for which J · n± = ±l0 I. These conclusions apply to any surface S which is diffeomorphic
to the sphere as well.
The representations described above occur in the Hilbert space H of a particle of spin S, which contains 2S + 1
irreducible representations σ = ±l0 , with l0 ranging from 0 or 12 , depending on S, to S. If the total angular momentum
J is analyzed into J = L + S, the algebra A is a subalgebra of the algebra AS generated by n, L and S. The space
H is irreducible with respect to AS , but with respect to A it is reduced into 2S + 1 irreducible subspaces Hσ where
the representations σ = ±l0 live. Rather than the direct sum of these spaces, H can be also described as the tensor
product H = HL ⊗ HS . The passage from a basis of H consisting of a basis for each Hσ to a basis consisting of vectors
with definite orbital momentum L requires the identification of the pure spin states ζSm , which are found to be the
σ
sum of ξSm .

Appendix A. Construction of n
Given a representation of so(3, 1) of type (l0 , 0) in a space H with basis {ξjm }, the operators n± defined in eq. (15)
(with σ = ±l0 for n± ) for j = 0
1
n± ξjm = [D ∧ J ± l0 J]ξjm (A.1)
j(j + 1)
and in eq. (16) for j = 0, are such that J, D and n+ (or n− ) generate an algebra A.
The operators n± are vector operators. In addition they are self-adjoint, as can be verified calculating (ξjm , n± ξj  m )
and (n± ξjm , ξj  m ). From eq. (A.1), if j = 0 it follows that (n± · J)ξjm = ±l0 ξjm . If ξ00 is present, then l0 = 0,
(n± · J)ξ00 = 0. Hence
n± · J = ±l0 I (A.2)
and n± obeys eq. (8).
In order to prove that [n±i , n±j ] = 0, note that the commutation rule of a vector V with J2 , [J2 , V] = 2iV∧J+2V,
for the vector n± ∧n± yields [J2 , n± ∧n± ] = 0, since (n± ∧n± )∧J = in± ∧n± , because of eq. (A.2). As a consequence,
from eq. (8) we get (J 4 ≡ (J2 )2 )
n± ∧ n± J 4 = iJ2 n± ∧ D ± l0 J2 D − J2 (n± · D)J ± l0 J2 n± ∧ J. (A.3)
In eq. (A.3) J2 n± can be expressed in terms of J and D by taking the adjoint of eq. (8), which yields
J2 n± = −J ∧ D ± l0 J. (A.4)
Page 6 of 6 The European Physical Journal Plus

The outcome is n± ∧ n± J 4 = 0, hence n± ∧ n± ξjm = 0 if j = 0. As for the case j = 0, which is present only if l0 = 0,
we have
i 1
n±i n±j ξ00 = − (D ∧ J)i Dj ξ00 = [δij D2 − Dj Di ]ξ00 . (A.5)
2 2
Recalling eq. (7), we see that [n±i , n±j ]ξ00 = 0. Thus, the components of n+ (n− ) commute with each other.
The equation n2± = I is proved by scalar left multiplication of eq. (8) with J2 n± and use of eq. (A.4), which yields

n2± J 4 = −(J ∧ D) · (D ∧ J) + l02 J2 = −J2 (1 − D2 ) + J2 l02 = J 4 .

Thus, if j = 0 n2± ξjm = ξjm . As for n2± ξ00 , recalling that, if l0 = 0 then D2 = J2 + I, we have

i 1
n2± ξ00 = − (D ∧ J) · Dξ00 = (J2 + 2D2 )ξ00 = ξ00 .
2 2
Finally, 2D± ≡ J ∧ n± − n± ∧ J = 2D. Indeed, by eq. (8) one finds D2± J2 = D2 J2 , hence (D2± − D2 )ξjm = 0 if
j = 0. As for ξ00 we have

2D± ξ00 = J ∧ n± ξ00 = −iJ ∧ Dξ00 = −i[J ∧ D + D ∧ J]ξ00 = 2Dξ00 .

In conclusion, D± = D.

References
1. J. von Neumann, Math. Ann. 104, 570 (1931).
2. W. Thirring, Quantum Mathematical Physics (Springer-Verlag, Berlin) 2002, pp. 85-86.
3. W. Miller, Commun. Pure Appl. Math. 17, 527 (1964).
4. J.D. Talman, Special Functions (Benjamin, New York) 1968, pp. 215-225.
5. G. Morchio, F. Strocchi, Lett. Math. Phys. 82, 219 (2007).
6. I.M. Gelfand, R.A. Minlos, Z.Ya. Shapiro, Representations of the Rotation and Lorentz Groups and Their Applications
(Pergamon Press, Oxford) 1963, pp. 186-201.

You might also like