The philosophical problems of implementing superselection

rules

Jorge Manero
arXiv:2310.03014v1 [physics.hist-ph] 4 Oct 2023

Abstract

Some physicists believe that superselection rules should be implemented to get rid of inconsistencies
when a theory is framed in terms of a new mathematical formulation, whilst others think that this new
formulation should be modified instead of implementing those rules, at the expense of introducing addi-
tional mathematical structure. The outcome, however, is that we are still uncertain whether these rules
should be implemented and how they should be interpreted and assessed from the philosophical point
of view. Based on a detailed examination of the group-theoretic reformulation of (relativistic and non-
relativistic) quantum mechanics that prompts physicists to impose superselection rules, I shall argue that
the implementation of these rules involves serious heuristic and epistemological concerns. Considering
this argument, I shall conclude that there are suitable philosophical reasons to claim that the imple-
mentation of superselection rules should be rejected and that there are certain circumstances when the
formulation of a theory should be modified.

1 Introduction
As demonstrated by many case studies in the history of physics, the mathematical reformulation of a theory
—initially formulated in terms of a different mathematical language—, may have important implications.
Among these, we are interested in two contrasted scenarios. According to the first scenario, the new for-
mulation may not be applicable to consistently describe all mathematical solutions (e.g., quantum wave
functions, classical state functions, etc.) allowed by the old formulation, so that some external readjust-
ments introduced by hand are required. The form of these readjustments may depend on the nature of the
solutions that cannot be described by the new formulation: if they are empirically meaningful, we may
think of incorporating auxiliary terms into this formulation to make the theory empirically adequate, and
if they are empirically meaningless, we may think of forbidding these solutions to get rid of possible in-
consistencies. According to the second scenario, however, the new formulation may introduce additional
mathematical structure to describe all mathematical solutions allowed by the old formulation, irrespective
of whether or not they are empirically meaningful. When physicists encounter a situation in which they
need to choose between both scenarios, some prefer the first one because they think that, for sake of sim-
plicity, additional structure should be avoided, whilst others prefer the second one because they think that,

1
for sake of naturaleness, the new formulation should not introduce external readjustments.
Interestingly, this situation is similar to the one encountered when physicists have to decide between
implementing (or not implementing) superselection rules at the moment they reformulate their theories in
terms of the language of Lie group theory. For example, in the framework of quantum mechanics (QM)
the first scenario occurs when this theory is reformulated in terms of a certain group-theoretic structure that
cannot consistently describe superpositions of wave functions with different parameter, such as different
mass, charge, and so on. Since, according to the creators of such a group-theoretic reformulation, these
wave functions are empirically meaningless, superselection rules are implemented to forbid them.1 On the
other hand, the second scenario occurs when QM is reformulated in terms of a group-theoretic structure that
can consistently describe superpositions of wave functions with different parameter. Since, according to the
creators of this alternative group-theoretic reformulation, these wave functions are also empirically mean-
ingless, they are described by surplus mathematical structure without implementing superspection rules.
Thus, we have to decide between the task of implementing superselection rules and forbid superpositions
of wave functions with different parameter or that of introducing surplus mathematical structure without
forbiding these superpositions and without implementing superselection rules.
As regards this two-fold dilemma, Steven Weinberg writes:

[...] it may or it may not be possible to prepare physical systems in arbitrary superpositions
of states [wave functions], but one cannot settle the question by reference to symmetry prin-
ciples, because whatever one thinks the symmetry group of nature may be, there is always
another group whose consequences are identical except for the absence of superselection rules
(Weinberg, 1995, 90-91).

According to Weinberg, there seems to be no physical way to resolve the above two-fold dilemma as regards
the implementation of superselection rules. However, the fact that there are no physical reasons to resolve
this dispute does not mean that there could not be philosophical reasons to do that.
In this spirit, I shall support the claim that the implementation of these rules involves serious concerns
of heuristic and epistemological kind. Firstly, I shall argue that they involve a constraint on theory de-
velopment by virtue of the fact that superselection rules remove surplus structure that might be empirically
fruitful for further developments; and secondly, I shall support the claim that they involve a stronger form of
underdetermination, as there is less underdetermination of the formulation by the theory when superselec-
tion rules are excluded and the formulation is modified. A brief explanation and examples will be provided
for both arguments. As a consequence of this analysis, I shall conclude that there are suitable philosophical
reasons to claim that the implementation of superselection rules should be rejected and that there are certain
circumstances when the formulation of a theory should be modified.
My methodological plan shall be as follows: in Section 2, I shall give an overview of Lie group theory;
in Section 3, I shall state the dilemma arising from the two scenarios described above; in Section 4, I shall
1 As
emphasised by (Jammer, 1974), the presence of superselection rules in QM undermines the one-to-one correspondence that
Von Neummann assumes between quantum states and wave functions in his original formulation of this theory.

2
address the heuristic concerns, whilst in Section 5 the epistemological concerns. Finally, I shall present
some concluding remarks in Section 6.

2 Theoretical background

2.1 The effectiveness of Lie group theory in physics

As argued by (French, 2014, ch.5), Lie group theory has been effective in elaborating formulations or
reformulations of many physical theories capable of generating new empirical predictions and of improving
our understanding about these theories. Considering the work of (Weyl, 1931; Wigner, 1939), I shall start
by introducing the basic technicalities of Lie group theory, followed by a brief explanation of the way this
mathematical language has been applied to the physical sciences and of how it underpins the fundamental
structure of some physical theories.
Simply put, a group G consists in a set of (finite or infinite) elements {x1 , x2 , ...} with an operation * de-
fined between two of these elements to form a third element within the group by means of a ‘multiplication
rule’ xi ∗ x j = xk , where i, j, k ∈ {1, 2, ...}. The set and the operation must satisfy the associative property
and must have both an identity element and an inverse element. If the group is a differentiable manifold
and the operation * is smooth (i.e., it defines a continuous mapping with derivatives of all orders) then it is
called a Lie group. As one can easily note, this abstract definition does not entail a straightforward physical
interpretation. However, this interpretation can be elaborated by means of the following two observations:
Firstly, there are geometrical objects upon which a Lie group acts. These can be figures composed of
spatial points, such as triangles or squares, but they can also be vectors, differential forms, tensors, etc.
The action of a Lie group on a set of these objects (e.g., a vector space, an affine space, etc.) is, in formal
terms, a Lie representation. This is a ‘copy’ of the Lie group defined in that set, in the same way a self-
portrait is a copy of the artist represented in a canvas. In the particular case of a Lie group G acting on a
vector space V (a case which will be our main focus in this paper), the Lie representation of G in V is a
function that goes from G to the group of automorphisms of V. Those automorphisms are realised as a set
of linear transformations or mappings between initial and final points in V. For example, the set of rotations
in three-dimensional space are linear transformations induced by the Lie representation in the set of 3 × 3
orthogonal matrices with determinant equal to one, called the special orthogonal group SO(3), on R3 .
Thus, in general we have two conceptually different mathematical objects that are equivalent (up to a
Lie representation): the abstract Lie groups, on the one hand, and the induced transformations in the math-
ematical space in which these Lie groups are Lie-represented, on the other. Considering this equivalence
relation, (Weyl, 1931) discovered that Lie groups are abstract mathematical objects that have no physical
interpretation unless they act upon the state space V of a physical theory T (hereinafter, vector space V of
T). As demonstrated by him, the way this interpretation is revealed is through the concept of invariance: the
Lie representation of a Lie group G in V takes the form of linear transformations that leave certain sorts of
objects invariant, which are interpreted as physical symmetries. In particular, the induced transformations

3
of G can be associated with the dynamical evolution of a physical system described by the laws of T. Based
on this association, we say that the abstract Lie group G is a symmetry group of a theory T describing some
physical system if the fundamental laws of T are invariant under the induced continuous transformations
of G in the vector space V of T (i.e., the Lie representations of G in V). Since the laws of a theory T can
be reformulated in this way —as invariants of a certain symmetry group G—, the identification of G (to-
gether with the vector space V of T in which G is Lie-represented) is sufficient to determine its dynamical
structure, and furthermore, the complete laws of T can be obtained by Lie-representing G in V.
Secondly, there is a very important concept associated with the empirical consequences of Lie group
theory. Let us define the Lie algebra of a Lie group G as the tangent space of G at the identity (i.e., a
group-theoretic structure which allows one to define the local group structure of G in a vector space). More
specifically, the associated real Lie algebra G of a Lie group G is a real vector space (of the same dimension
to that of G) with a bilinear, alternating, antisymmetric product map [·, ·] : G × G → G , called Lie bracket,
that satisfies some commutation relations: the Jacobi identity.2 As demonstrated by (Wigner, 1939), the Lie
algebra G associated with the symmetry group G of a theory T is relevant to the predictive component of T
by virtue of the fact that the Lie representation of G in the vector space V of T is associated with infinitesimal
transformations that represent the physical properties relevant to T. For example, in classical mechanics and
QM, the infinitesimal transformations induced by the generators of the associated Lie algebras (i.e., the
basis operators TA of G , where A = {1, 2, ...}) are spelled out in terms of physical properties of the system
(e.g., momentum, energy, etc.).
Note that both observations imply that the formulation or reformulation of a theory in terms of Lie
representations of abstract symmetry groups and their associated Lie algebras permits not only to reveal the
fundamental mathematical structure of that theory, but also to formally represent its complex connection
with the empirical domain. It is the illuminating bridge between abstract mathematics, dynamics, and
empirical predictions in physics (realised by these two observations) that makes Lie group theory both
predictably and theoretically effective. Once we have introduced the central technical elements of Lie
group theory, let us proceed to define two different kinds of Lie representations.

2.2 Two Lie group representations

One should note that not all Lie representations define a Lie group homomorphism, i.e., a function µ : G → H
that, given two elements g, g′ ∈ G and another two elements µ (g), µ (g′ ) ∈ H, preserves its group operations:

µ (g)µ (g′ ) = µ (gg′ ). (1)

Lie group homomorphisms µ that send a Lie group G to the group of automorphisms H = Aut(V ) of a
vector space V are called True Lie representations of G in V . Conversely, homomorphisms ω that send a
2
The Jacobi identity is [X, [Y,Z]] + [Y, [Z,X]] + [Z, [X,Y ]] = 0; for all X,Y,Z ∈ G .

4
Lie group G to the group of automorphisms H = Aut(V ) of a vector space V that satisfy:

ω (g)ω (g′ ) = ψ (g, g′ )ω (gg′ ) (2)

are called Projective Lie representations of G in V (or Lie representations ‘up to a phase’), where ψ (g, g′ )
is a scalar transformation of the elements of the Lie group that satisfies:

ψ (g, g′ )ψ (g′′ , gg′ ) = ψ (g′′ , g)ψ (g′′ g, g′ ), (3)

with g, g′ , g′′ ∈ G. Among the entire zoo of Lie representation categories, these two deserve our attention.
The important aspect to be remarked here is that, unlike true Lie representations, projective Lie represen-
tations are homomorphisms that involve structural losses in the form of a scalar transformation ψ (g, g′ )
meaning that the induced transformations in the Lie representation space do not preserve all mathemati-
cal features that are constitutive of the associated abstract Lie group. As we shall see in a moment, these
structural losses arising from projective Lie representations are directly associated with the notion of a
superselection rule.
An illustrative example of this distinction is the action of SO(3) on two vector spaces. The action of
SO(3) on R3 is a true Lie representation, whilst its action on the Hilbert space H is a projective one, the
difference being that vectors that are equivalent up to 2π rotations in the former space are not equivalent
in the latter space (they differ up to a minus sign); vectors are, in fact, equivalent in H up to 4π rotations.
In physical terms, this means that for every two successive 2π rotations induced by SO(3) along each
coordinate axis of R3 , there are exactly two invariant vectors in H (differing up to a minus sign) associated
with two discrete 1/2-spin values. Let us proceed to define the concept of a superselection rule.

2.3 Superselection rules

With one significant exception, one may demonstrate that for any element of the Lie group g ∈ G, the phase
ψ (g, g′ ) arising from projective Lie representations of the form ω (g) is independent of any element of the
vector space v ∈ V to which the transformation ω (g) is applied. To prove this, let us follow (Weinberg, 1995,
52-53) and consider two linearly independent vectors vA ,vB ∈ V , and assume that there exists a superposed
vector vAB ∈ V , such that vAB = vA + vB . Applying ω (gg′ )ψ (g, g′ ) to vAB , we obtain:

ω (gg′ )ψ (g, g′ )AB (vA + vB ) = ω (g)ω (g′ )(vA + vB)

= ω (g)ω (g′ )vA + ω (g)ω (g′ )vB

= ω (gg′ )ψ (g, g′ )A vA + ω (gg′)ψ (g, g′ )B vB .

We then obtain:
ψ (g, g′ )AB (vA + vB) = ψ (g, g′ )A vA + ψ (g, g′ )B vB .

5
Since vA ,vB are linearly independent, the last expression means that:

ψ (g, g′ )AB = ψ (g, g′ )A = ψ (g, g′ )B .

Thus, we have proved that the phase is independent of any element of V, a result which is compatible
with our definition of projective Lie representation in (2), whose vector elements to which it applies are
not included. Note, however, that this demonstration fails for only one exceptional case: when it is not
possible to find a projective Lie representation of a wave function superposition vAB represented by vA + vB .
In such a case, for any pair of phases ψ (g, g′ )A and ψ (g, g′ )B , there is no way to obtain a different global
phase ψ (g, g′ )AB capable of satisfying the form (2) of a projective Lie representation. The only way to
obtain in general a projective Lie representation seems to forbid superpositions of the form vAB = vA + vB .
Considering this exceptional case, we are ready to introduce the definition of a superselection rule:

We say that there is a superselection rule between different vectors vA , vB the operators ω (g)ω (g′ )
and ω (gg′ ) act upon, when there exists no vector vAB , such that vAB = vA + vB , and therefore,
the phases ψ (g, g′ )A and ψ (g, g′ )B depend on these classes of vectors.

Although we shall see some examples illustrating this definition later, it is important to remark that supers-
election rules represent external impositions put by hand to prevent the existence of superpositions of wave
functions with different parameter. In some cases, as we shall see, superselection rules are equivalent to the
external imposition that vector spaces vA , vB should be linearly dependent, meaning that there cannot be a
vector element of the form vAB , and that the parameters involved in the neglected superposition should be
globally defined (i.e., equal for any subspace of V ).
There is, however, one way to avoid the implementation of superselection rules and to allow the pos-
sibility of defining in a suitable way Lie representations of superpositions of the form vAB = vA + vB. The
suggestion is to change the algebraic and topological structure of the Lie group in question to yield true
Lie representations instead of projective ones. This is motivated by the fact that true Lie representations
eliminate the phase associated with projective Lie representations, and without the presence of this phase,
the above problem does not arise, which means that there is always the possibility of defining Lie represen-
tations of superpositions of wave functions with different parameter. Let us briefly introduce this suggestion
by describing the way true Lie representations can be constructed.

2.4 Lifting theorem

It is a remarkable mathematical result proved by (Weinberg, 1995, 83-84) that, thanks to certain algebraic
and topological features of Lie groups, the scalar transformation associated with a projective Lie represen-
tation is nil ψ (g, g′ ) = 1 iff the following conditions are met:

(i) The generators of the Lie group in this Lie representation can be redefined so as to eliminate all
central charges from the associated Lie algebra.3
3 Central charges are operators that commute with all the other operators of a theory. By eliminating these operators, one is simply

6
 (ii) The Lie group is simply connected (the first fundamental group equals the identity element).4

One way of formulating this theorem is that any Lie group G that only admits true Lie representations (i.e.,
ψ (g, g′ ) = 1) is simply connected and its associated Lie algebra does not have central charges. Conversely,
this theorem also shows that there are two (non-exclusive) ways in which projective Lie representations
arise —either algebraically, by virtue of the presence of central charges associated with the corresponding
Lie algebras, or topologically, by virtue of the fact that this Lie group is not simply connected. Going back
to the previous example, let us note that SO(3) is not simply connected because the set of rotations around
any fixed direction by angles ranging from −π to π forms a closed loop in the manifold S3 /{−I, I} (i.e.,
the manifold ball S3 with antipodal points identified) that cannot be shrunk continuously to a point. This
means, according to this theorem, that SO(3) accepts projective Lie representations, as corroborated by the
fact that it is Lie-represented projectively in H .
As regards the problem of the phase in superpositions of wave functions with different parameter, one
central result of the previous theorem is that if we have a Lie group G that admits projective Lie representa-
tions and, therefore, cannot account for the existence of these superpositions, the only way of obtaining true
Lie representations that may solve the alleged problem is via two separated procedures, the first algebraic,
and the second topological. Let us analyse these procedures in detail.
Given a projective Lie representation ω of G, there is no way of ‘lifting’ ω to a true Lie representation
of the same Lie group G: this is possible only if ω ‘lifts’ to the true Lie representation of a different Lie
group, namely, the maximal central extension of G. This means that to define a true Lie representation for
G one must not only appeal to the algebraic central extension of G but also to its universal cover (called the
topological central extension). While the algebraic central extension allows us to redefine the generators of
G in this Lie representation and eliminate the central charges from the associated Lie algebra, the universal
cover of G permits the definition of a unique simply connected Lie group that is equivalent to G up to
a group homomorphism. In this way, when both algebraic and topological extensions are applied to the
original Lie group G, we can prescind from having projective Lie representations and we can solely obtain
true ones. Consider again the example of the special orthogonal group SO(3). One can demonstrate that
the double cover of SO(3) is the special unitary group SU(2), which is diffeomorphic to the ball manifold
S3 . This topological correspondence is illustrated by the observation that there are two classes of points
in S3 for every rotation (because every rotation is a point in S3 /{−I, I}), so that SU(2) has twice as many
elements as SO(3) itself). Since any loop in the ball manifold S3 can be contracted continuously to a point,
SU(2) is simply connected and thus defines the universal cover of SO(3). As such, SU(2) only accepts true
Lie representations, as corroborated by the fact that its action on H is a true Lie representation.
Thus, opposed to the implementation of superselection rules, we can solve the problem of the phase
in superpositions of wave functions with different parameter by constructing a true Lie Representation
through these procedures and having in turn the possibility of defining suitable Lie representations of these
superpositions.
redefining all the generators of the Lie algebra in such a way that they do not appear on the commutation relations.
4 More simply put, this means that any loop that starts and ends at the same group element may be shrunk continuously to a point.

7
3 The two-fold dilemma
Since there are physical theories that allow the possibility of describing superpositions of wave functions
with different parameter (either real or nonexistent), one expects that the formulation or reformulation of
these theories in group-theoretic terms should also account for these superpositions. Unfortunately, there
are symmetry groups compatible with certain theories that only accept projective Lie representations in the
corresponding vector space. As demonstrated above, the problem associated with these Lie representations
is that they are not capable of describing superpositions of wave functions with different parameter in a
consistent way. Therefore, we end up with a problem of adequacy as there are elements (either real or
nonexistent) described by certain theories, such as superpositions of wave functions with different parame-
ter, that have no counterpart in the group-theoretic structure of these theories. In response to this problem
of adequacy, however, we have two different options on the table:

Option 1: One might either implement superselection rules to forbid the existence of superpositions
of wave functions with different parameter without changing the underlying projective Lie represen-
tations; or

Option 2: One might modify the projective Lie representations to yield true Lie representations with-
out forbidding the existence of superpositions of wave functions with different parameter and without
implementing superselection rules.

Considering these options, the following dilemma emerges (hereinafter the two-fold dilemma): Which is
the most reasonable option?
In the following two sections, we shall see that, although there seems to be no physical reason to prefer
one option instead of the other, there are at least two philosophical reasons to prefer the second option
rather than the first one. These reasons correspond to some heuristic and epistemological concerns that only
arise when superselection rules are implemented in the framework of projective Lie representations. Let us
analyse each of these concerns in the corresponding order.

4 The heuristic concern

Let us note that the effectiveness of Lie group theory in physics does not consist in its mere notational
capability or its possibility of rewriting a set of physical statements in group-theoretic terms; it enables us
to have a set of heuristic strategies for theoretical development at our disposal that contributes to generate
new predictions. In this spirit, let us present a heuristic strategy of that sort, which (i) is framed in the
context of any physical theory capable of being formulated in group-theoretic terms; and (ii) only works
when the symmetry groups of that theory admit true (as opposed to projective) Lie representations. As we
shall see, the presentation of this strategy shall suggest us that the implementation of superselection rules
in the framework of projective Lie representations is not only unnecessary but also is an impediment that
prevents us from unveiling all the heuristic virtues associated with true Lie representations.

8
 Following (Redhead, 1975; Da Costa & French, 2003; Bueno & French, 2018), one suggestive heuristic
strategy is to focus not only on mathematical structures that are empirically relevant by virtue of having
counterparts at the empirical level, but also on those structures that are surplus, in the sense that they are
elements of the mathematical formalism that do not entail some sort of phenomena in the observable world
directly relevant to the theory. As corroborated by two particular case studies, I shall support the second
option of the two-fold dilemma (i.e., to have true Lie representations without implementing superselection
rules) based on the heuristic fact that there are theories that can be formulated or reformulated in terms of
symmetry groups that initially contain surplus structure in the form of true (as opposed to projective) Lie
representations that, nevertheless, turn out to be empirically fruitful at a later point in their development.
More specifically, considering a more robust notion of surplus structure in accordance with actual practice,
I will seek to demonstrate that true Lie representations are significant mathematical sources of potential
developments by virtue of the fact that their presence as an initially surplus structure in the mathematical
domain suggests a way of searching for counterpart relations at the empirical domain, some of which hold
between phenomena that are completely undetectable or unknown for physicists of the time. While the
first case study is associated with superpositions of wave functions with different mass and how they are
interpreted in the transition from standard QM to the low-velocity approximation of the relativistic domain
of QM, the second one is about superpositions of wave functions with different spin and how they are
interpreted in the transition from standard QM to related domains of quantum field theory (i.e., nuclear
physics and supersymmetry).
First and foremost, let us start with establishing, from the perspective of the partial structures frame-
work, an appropriate definition of surplus structure in accordance with actual practice, and then proceed to
describe the corresponding case studies.

4.1 Surplus structure and actual practice

We cannot define the notion of surplus mathematical structure in connection with Lie group theory if we
do not consider that a physical theory T can be reformulated in terms of a given mathematical structure,
and for the present purposes, in terms of a group-theoretic structure M’. Since physical theories are already
mathematised when they come from the hands of their creators, with such mathematisation not necessarily
being group-theoretic, we have to consider the case in which T, initially framed by an arbitrary mathematical
structure M, is reformulated in terms of a new group-theoretic structure, M’.
As mentioned by (Da Costa & French, 2003; Bueno & French, 2018), I think that the most appropriate
way to do so —when actual practices are considered— is in terms of the partial structures representational
framework, and particularly in terms of the notions of partial isomorphism and partial homomorphism
deployed according to this framework. The basic idea behind these notions is that they can represent in
formal terms the complex structural relationships that take place on the frontier between the mathematical
and the physical domains, in turn supplying the formal means of articulating the openness and partially
informative status of our theories. In particular, they can capture the partial way in which some structures

9
are transferred all the way down from the mathematical domain to the physical domain, and then from this
level to the empirical domain of our accessible experience (Bueno et al., 2002). Considering these notions,
we can suppose that a theory T, initially formulated in terms of a mathematical structure M that stands at
the physical level, can certainly be reformulated in terms of the mathematical structure M’ that stands at the
abstract mathematical level.5 This can be done by bringing to the physical domain those relations that hold
at M’ that have empirical counterparts relevant to T.6 Thus, we can define surplus mathematical structure as
the remaining structures pertaining to M’ that are not brought from the mathematical to the physical domain
because they have no counterparts at the empirical level relevant to T.
However, even if the aforementioned standard definition is appropriate for specific contexts, it ignores
certain aspects that are relevant when actual scientific practices are investigated in their full complexity.
Considering that physical theories are not only sets of formal linguistic statements in direct correspondence
with the data, but also incorporate other relevant elements, including hypotheses that are indirectly con-
firmed by phenomena investigated by other theories, we can identify at least one more general notion of
surplus structure that will be relevant in revealing the heuristic role of Lie group theory in one of the two
case studies to be investigated below. Thus, we can define a relativised notion of surplus structure with
respect to a given theory or set of hypotheses, for example, T’. This means that we can interpret surplus
structure as the part of the structure M’ that does not have a counterpart at the empirical domain of T’ (as is
defined according to the standard notion) but which it may have at the empirical domain of another theory
or set of hypotheses that are not part of T’. This possible case occurs when T’ defines relations that are
brought from M’ to some set of physical hypotheses of T’ that are confirmed by phenomena that pertain
to the empirical domain of another theory and do not enter into the empirical scope of T’, known in the
literature as non-entailed evidence.7 Any element of mathematics is, according to this definition, surplus
with respect to a given theory or set of hypotheses T’, but at the same time it may be fruitful relative to other
theories or sets of hypotheses that are also entailed by M’. Defining surplus structure in this more general
way not only reveals the wide scope of its fruitful application but also presupposes the central heuristic role
played by non-entailed empirical evidence in developing and confirming new hypotheses in the context of
actual practice.
Having appropriately defined surplus structure, l shall proceed to present our case studies. While the first
case study will only appeal to the notion of surplus structure associated with entailed evidence, the second
will consider, in addition to the foregoing, the notion of surplus structure associated with non-entailed
evidence.
5 Of course, in the case in which the theory is originally formulated group-theoretically, then M=M’.
6
A set of relations is said to have empirical counterparts if an isomorphic mapping (framed in terms of the partial structures mode
of representation) holds between these mathematical relations and the relations that hold among representations of the phenomena
(usually known as data models).
7 Non-entailed empirical evidence is defined by (Laudan & Leplin, 1991) in terms of the following principle for confirmation:

suppose that a mathematical theory M’ implies two logically independent physical hypotheses, H1 and H2 , and in turn H1 entails the
observational consequence e. Then, if e is true it counts as empirical evidence for H1 , for M’, and also for H2 , even though e is not a
logical consequence of H2 .

10
4.2 Case study 1: The spacetime symmetry group of quantum mechanics

It is well-known by physicists that the Schrödinger equation is invariant under unitary transformations of
the form
ψ ′ (r′ ,t ′ ) = exp[i∆m (r,t)/h̄]ψ (r,t), (4)

where r′ and r are three-dimensional vectors and ∆m (r,t) is a linear transformation in the fourth-dimensional
Galilean spacetime (r,t) (e.g., ∆m = m(v2t/2 − v · r) for boosts in (r,t)). Since these transformations are
induced by projective Lie Representations of the standard Galilean group G in Hilbert space H , then G can
be associated with the spacetime symmetry of QM.8 However, after Bargmann’s contribution (Bargmann,
1954), extensive literature has acknowledged that not all wave function superpositions that are solutions
to the Schrödinger equation can be consistently described in terms of projective Lie representations. One
example is the set of solutions resulting from superposing wave functions of particles with different mass.
In this particular situation, one can show that there are transformations induced by projective Lie represen-
tations, called Bargmann transformations, which produce a non-trivial interference term when acting on
superpositions of wave functions with different mass, while they act as the identity on Galilean spacetime
coordinates. To make this example explicit, let us suppose that ψ (r,t) = ψ1 (m1 , r,t) + ψ2 (m2 , r,t) is a
wave function superposition with two masses in an arbitrary inertial reference frame. If one performs the
transformation g−1 −1
2 g1 g2 g1 = e (i.e., the identity by the action on Galilean spacetime coordinates) where

g1 is a translation by r and g2 is a pure boost by v ∈ R3 , then the action of this transformation on H is:

ψg−1 g−1 g2 g1 (r,t) = exp[im1 (v · r)/h̄]ψ1 (r,t) + exp[im2 (v · r)/h̄]ψ2 (r,t). (5)
2 1

This is, in fact, a projective Lie representation of the identity transformation iff m1 = m2 , in which case
it has the form of (2). However, in general it is not a Lie representation at all. This implies that if the
masses of the particles are not equal, Bargmann transformations cannot be described by projective Lie
representations, meaning that there is no group-theoretic counterpart of a particular solution of the theory
(i.e., its group-theoretic reformulation is not adequate with respect to the observed phenomena). There was
a time when Bargmann’s followers did not care about this issue because they thought that superpositions
of wave functions with different mass were surplus elements of the formalism. The way they could make
the theory consistent within the Galilean group-theoretic formulation was to neglect the possibility of those
superpositions by imposing a superselection rule by hand that makes the mass equal for those cases.
Looking at this situation from the perspective of the partial structures approach, we first have to define
the partial structure SG corresponding to QM, as standardly formulated. This structure can be (extensionally,
crudely) expressed as:
SG = DG , R(G) i , f(G) j , a(G) k i∈I, j∈J,k∈K
, (6)
8 The Galilean group G is a Lie group defined by G ≃ R4 ⊗ (R3 ⊗ SO(3)). The group elements are labelled by g = (b,r,v,R),
S S
where b ∈ R;r,v ∈ R3 , and R ∈ SO(3). These elements are naturally Lie-represented in Galilean spacetime as time translations, space
translations, velocity boosts, and rotations, respectively.

11
where DG is a non-empty set that stands for the domain of SG representing equivalent classes of eigenvectors
(i.e., quantum properties, such as momentum, energy, total angular momentum, etc.), and the family of
unitary operators that leave them invariant (e.g., momentum operator, Hamiltonian operator, total angular
momentum operator, etc.); RG is a non-empty set of n-place partial relations9 holding among the elements
of the domain DG , which encode the group multiplication law satisfied by the corresponding projective Lie
representation of G in Hilbert space G(H ); the elements fG are full and partial homomorphisms,10 some
of which hold between SG and the partial structure of G(H ) (which we denote by SG(H ) ). This means that
SG is a substructure of SG(H ) in the sense that only a subfamily of all the known relations holding for SG(H )
turns out to be the set of those known relations holding for SG . Those non-shared known relations that hold
for SG(H ) but do not hold for SG are given by the multiplicative law of transformations between vectors that
are not invariant with respect to unitary operators in H (i.e., they are not eigenvectors and have no physical
interpretation); finally, aG is the set of n-place partial relations associated with surplus structure with respect
to SG , namely, the relations that are not transferred from the mathematical domain, given by SG(H ) , to the
physical domain, given by SG .11 However, with the exception of non-invariant vectors, this set is empty
for the only way of having projective Lie representations in H induced by G is through the imposition
of superselection rules that forbid surplus structure in the form of superpositions of wave functions with
different mass.
One can expect to have true Lie Representations induced by an analogue of G that consistently describe
superpositions of wave functions with different mass without the need to change the empirical implications
of QM. As discussed in Section 2, we can just lift G to the extended Galilean group G’ by means of
its algebraic and topological central extension to obtain a true Lie representation in H , namely, G’(H ).
Since, according to Bargmann’s followers, the choice between G and G’ does not amount to any empirical
difference, the additional structure acquired by extending the standard Galilean group is, therefore, a set
of surplus relations within the partial structure SG’(H ) of the form aG’ (i.e., surplus structure with respect
to SG ) that is transferred by means of a partial homomorphism to SG as unknown relations holding (or
not holding) for the latter partial structure. Specifically, these relations correspond to new multiplicative
laws induced by the Lie representation of an additional group element that is incorporated into the group
extension G’. Physically speaking, these multiplicative laws involve mass transformations that allow us to
describe superpositions of wave functions with different mass that, according to Bargmann’s followers, are
empirically meaningless.
However, various contributions, such as (Greenberger, 2001; Hernandez, 2012), have demonstrated that,
9 To give a more general definition, an n-place partial relation R over D is a triple (R ,R ,R ), where R , R , and R are mutually
1 2 2 1 2 3
disjointed sets, such that: R1 is the set of n-tuples that (we know) belong to R, R2 is the set of n-tuples that (we know) do not
belong to R, and R3 is the set of n-tuples for which it is not known whether or not they belong to R. With respect to the notation,
I is an index set that labels each of the n-place partial relations; J is an index set corresponding to the labels of ‘vertical’ full and
partial homomorphisms; and K is another index set that labels each partial relation associated with surplus structure. ‘Vertical’
homomorphisms express intra-theoretical relations holding between models of the same theory.
10 We say that a partial function f : D → D′ is a partial homomorphism from S to S′ if for every x and every y in D, R xy → R′ f (x) f (y)
1 1
and R2 xy → R′2 f (x) f (y).
11 Strictly speaking, a does not belong to S , which represents a physical structure, but to S
G G G(H ) , which represents a mathematical
structure. However, SG needs to contain these surplus relations in order to represent the situation in which they might become part of
the known relations R1 holding for SG .

12
from the perspective of the low-velocity approximation of the relativistic domain of QM, superpositions of
wave functions with different mass do have empirical significance. Specifically, they have argued that the in-
terference term giving rise to Bargmann’s rule exhibits physical effects, and that these effects occur because
there is a relativistic time difference between the corresponding inertial observers at the low velocity limit.
This fact strongly suggests that, contrary to G, G’ is the fundamental symmetry group of non-relativistic
QM. Again, all this can be captured in terms of the partial structures approach. So, assuming that we are
relativistic physicists willing to describe quantum systems at the low-velocity limit, then we can express the
partial structure corresponding to the low-velocity relativistic domain of QM as:

SG’ = DG’ , R(G’) i , f(G’) j , a(G’) k i∈I, j∈J,k∈K

, (7)

where DG’ are equivalent classes of eigenvectors representing quantum properties and the family of uni-
tary operators that leave them invariant, and RG’ encodes the group multiplication law satisfied by G’(H )
(including mass transformations). The rest of the elements in the structure ( fG’ and aG’ ) have the same
meaning as in SG . In this case, surplus partial relations aG’ are transferred from SG’(H ) into known relations
holding for the partial structure SG’ through a full homomorphism. This means that what initially was a
partial homomorphism that transferred known relations holding for SG’(H ) to unknown relations holding
(or not holding) for SG becomes a full homomorphism that transfers all known relations holding for SG’(H )
to known relations holding for SG’ .
With this formal discussion we can conclude that the surplus mathematical structure associated with
the extended Galilean group G’ can be seen as the source of potential developments in the realm of QM.
In this case, the mere fact of reformulating QM in terms of G’ that admits true Lie representations in H
encourages us to look at the theory from the perspective of a low-velocity relativistic setting, in which case
the phase-terms included in the group extension provide suitable grounds for describing empirically mean-
ingful superpositions of wave functions with different mass. Looking at this case study from an ahistorical
standpoint, if we were situated before the moment relativistic physics was discovered, the possibility of
framing the theory in terms of true Lie representations would prompt us to look for novel phenomena at
the empirical level (i.e., superpositions of wave functions with different mass), given their group-theoretic
counterparts.

4.3 Case study 2: The spin symmetry group

It is a historical fact that (Weyl, 1931) developed a comprehensive analysis of the notion of spin by identi-
fying its deep group-theoretic foundations. Specifically, he demonstrated that this notion can be associated
with an intrinsic form of angular momentum that comes into play when the special orthogonal group SO(3)
is Lie-represented in H . As described in Section 2.4, SO(3) is not a simply connected Lie group (i.e., there
exists at least a loop that cannot be shrunk to a point), so its Lie representation in H is projective. This
fact, as we are about to discover, has important implications.

13
 Similar to the Galilean case (something which one would expect since SO(3) is the non-normal subgroup
of the standard Galilean group), this projective Lie representation cannot consistently describe superposi-
tions of wave functions with integer and half-integer spin. As proven in (Hegerfeld et al., 1968), the only
consistent way of framing the theory in terms of this rotational group-theoretic framework is by imposing
a superselection rule by hand (widely known as the univalence superspection rule) that neglects the possi-
bility of these superpositions. According to many textbooks, superpositions of wave functions with integer
and half-integer spin have never been observed. This implies that there exist phenomenological grounds for
imposing superselection rules, which have to be applied by hand to make the group-theoretic reformulation
of the theory adequate with respect to the observed phenomena. From this perspective, if we were to extend
the symmetry group to admit true Lie representations, it would not amount to any empirical difference. True
Lie representations would not be empirically meaningful and their introduction would generate a consider-
able amount of surplus structure at the level of the abstract Lie group formalism that would simply make
the theory unnatural.
However, after the publication of (Mirman, 1970), there have been various doubts regarding the validity
of the univalence superselection rule. This sceptical position is supported by experimental results that
have presumably demonstrated that superpositions of wave functions with integer and half-integer spin
are manifested through certain observable effects. (Rauch et al., 1975) is a relevant example in this regard.
Specifically, it is known that the superposition of a half-integer spin wave function, represented by the vector
ψ , with an integer spin wave function, represented by the vector φ , is not invariant under complete rotations
2π of the coordinate axes, because ψ changes its sign, while φ is not affected. It follows from this simple
fact that even if the expectation values of quantum properties are quadratic in the wave function (which
means that changes of sign cannot be detected by ordinary experiments), physical effects produced by
these rotational transformations can be observed through sophisticated experiments performed with crystal
neutron-interferometers. However, it is fair to acknowledge that the way these experiments are designed,
together with the physical implications drawn by (Rauch et al., 1975), have not been free from criticism and
should be taken with strong reservations.12
Under these circumstances, we can provide a better example that contributes to revealing, on more solid
ground, the heuristic fruitfulness associated with true Lie representations, namely, the theory of supersym-
metry (Wess & Zumino, 1971). This theory is an extension to the standard model of particle physics that
seeks to solve various problems and plug theoretical gaps where this model is still incomplete. Although it
is currently supported by some prominent physicists, the legitimacy of the theory of supersymmetry is not
based on the confirmation of new predictions, but on the evidence already predicted and explained by the
standard model (whose scope is confined to well-established limits), and on the fact that this model is known
to be incomplete. The essential aspect of this theory is that it postulates a large number of partner-particles
(called superparticles), in addition to the standard ones, that differ by a half-integer spin. In particular,
12 According to (Giulini, et al., 1996, 287-288), the experiments conducted by (Rauch et al., 1975) did not measure the relative
phase between bosonic and fermionic degrees of freedom, and the possibility of measuring this phase can lead to serious problems
with causality within the framework of local quantum field theory.

14
the postulation of these particles opens up the possibility of describing superpositions of wave functions
with integer and half-integer spin as physical (although not observable) processes. The way this theory
successfully describes those superpositions is by extending the symmetry group SO(3) to its topological
and algebraic extension. As is demonstrated by the lifting theorem in Section 2.4, given that the Lie algebra
associated with this Lie group does not have central charges, only a topological extension of it is required
by calculating its double cover (in this case, its universal cover). As mentioned earlier, this extension is the
special unitary group SU(2). One of the special features associated with this group extension is, in essence,
the fact that it only admits true Lie Representations in H , meaning that superpositions of wave functions
with integer and half-integer spin can be consistently described without the need to impose a superselection
rule. So, from an ahistorical point of view, if we were to travel backwards in time before supersymmetry the-
ory was conceived and if we were to use SU(2) instead of SO(3), then, we would be encouraged to look for
elements at the physical level, given their mathematical counterparts in the form of surplus group-theoretic
structure. However, in contrast with the first example of crystal neutron-interferometer experiments, the
postulation of superparticles is a theoretical hypothesis that has not been directly observed but which is
indirectly confirmed by other empirical implications associated with the theory of supersymmetry (i.e., the
phenomena explained by the standard model). It is this contrast that makes this theory an example sufficient
to associate the notion of surplus structure with non-entailed evidence (already defined in the first part of
the present section). Let us inquire into some details of this association. In this and the ensuing analysis, the
form of the corresponding partial structures is analogous to (6) and (7), so I shall not repeat the description.
The set of true Lie representations of SU(2) forms part of a wider mathematical theory M’ (used to
formulate the theory of supersymmetry) that entails a set of physical hypotheses H1 associated with the
presence of superparticles. However, M’ also entails many well-confirmed phenomena e through other
hypotheses H2 logically independent of H1 . Since M’ is an extension to the standard model with no further
predictions (i.e., M’ includes it at the mathematical level), the phenomena denoted by e range over some of
the successful predictions covered beneath the umbrella of particle physics. Thus, thanks to the development
of supersymmetry, true Lie representations of SU(2) can now be said to be theoretically fruitful because if
e is true, it counts as empirical evidence for H2 and also for H1 , even though e is not a logical consequence
of H1 . Going back to the previously stated definition of surplus structure, even if true Lie representations of
SU(2) do not have direct empirical counterparts and they are surplus with respect to the set of hypotheses
H1 (i.e., hypotheses associated with undetectable superparticles), they are fruitful with respect to the wider
theory of supersymmetry, where the phenomena within the scope of the standard model of particle physics
e counts as non-entailed empirical evidence of H1 . This may also be represented in terms of the partial
structures framework. Before supersymmetry theory was developed, the set of relations associated with
true Lie representations of SU(2) was part of surplus structure with respect to the standard model of particle
physics, contained within the partial structure associated with M’. The point is that even if we did not know
whether these relations did or did not hold at the physical level, they were eventually transferred to this
level as known relations holding within the partial structure associated with superparticles.

15
 I will proceed now to discussing the second part of the paper, the one associated with the epistemological
concern of implementing superselection rules.

5 The epistemological concern

Another way to recognise the effectiveness of Lie group theory in physics without having to implement
superselection rules, is to present one epistemological virtue offered by this language, which (i) is again
framed in the context of any physical theory capable of being formulated in group-theoretic terms; and (ii)
only works when the symmetry groups of that theory admit true (as opposed to projective) Lie represen-
tations. In a similar way to the last section, we shall see that the implementation of superselection rules
in the framework of projective Lie Representations is not only unnecessary but also is an impediment that
prevents us from unveiling all the epistemological virtues associated with true Lie representations.
Let us first note that there are certainly well-documented formulations compatible with a single theory
all of which are mathematically inequivalent but empirically equivalent —as illustrated by the conflict
between the Hamiltonian and the Lagrangian frameworks in classical mechanics (Pooley, 2006). This
familiar situation enables us to claim that there is in fact an underdetermination of the formulation by the
theory, where no theoretical and empirical ground is available to choose one formulation instead of another.
As we know, this particular case of underdetermination at the level of the mathematical structure may cast
doubts on the truth of scientific theories, and hence may be interpreted as a problem of epistemological kind.
In response to this problem, some philosophers demand that we should engage in a broader inter-theoretical
analysis capable of identifying some commonalities between the possible, underdetermined formulations
of the theory, and of framing these commonalities in terms of a single, unifying formulation that entails all
the empirical consequences of that theory. In other words, they intend to demonstrate that there exists a
single and unifying mathematical framework that approximately embraces all the available formulations of
a theory and is capable of accounting for the empirical data obtained by any of these formulations (Pooley,
2006; French, 2014, 43-47) —as the identification of a symplectic structure unifying the Hamiltonian and
Lagrangian approaches (Belot, 2006).
Following this analysis, one suggestive strategy concerned with the epistemological virtues of Lie group
theory is to focus on formulations or reformulations based on unifying group-theoretic structures that are
less underdetermined by the same theory and by the corresponding evidence. The resulting problem is
to find a way to identify those structures. As corroborated by one case study, I will support the second
option of the two-fold dilemma (i.e., to have true Lie representations without implementing superselection
rules) based on the mathematical fact that there are theories that can be formulated or reformulated in
terms of a single symmetry group that admits true Lie representations that, nevertheless, turn out to be less
underdetermined compared to other symmetry groups that admit projective ones. This means that the way
to identify unifying group-theoretic structures that are less underdetermined is via symmetry groups that
accept true (as opposed to projective) Lie representations. In this context, our case study shall be associated
with the symmetry group of electroweak phenomena.

16
5.1 Underdetermination of the formulation

Let us start with establishing, from the perspective of the partial structures framework, an appropriate defi-
nition of structural equivalence that shall help us to formally frame and overcome the underdetermination
of the formulation by the theory, and then proceed to describe the corresponding case study.
Suppose that S and S′ are different set-theoretic models corresponding to two different mathematical
structures compatible with the same theory and the same body of empirical evidence. We say that two math-
ematical structures are structurally equivalent iff there is a full isomorphism (full bijective homomorphism)
between S and S′ . This means that there is an equivalence class of mathematical structures differing up to
a bijective, structure-preserving mapping regardless of the diversity of models that lie at the interpretative
(ontological) level. Considering this definition, we can define the underdetermination of two formulations
compatible with a theory T (and a single body of evidence), represented by S and S′ , as the situation in
which S and S′ are not fully isomorphic, provided their corresponding mathematical structures entail the
same empirical consequences of T.
In the context of this contribution, however, the last definition should be framed in terms of any given
theory T that is formulated or can be reformulated in group-theoretic terms. To do that, we must charac-
terise, firstly, the way a group-theoretic structure G is compatible with a single theory T; and secondly, the
way G entails a body of empirical evidence. The first question can be answered by means of identifying
the symmetry group G of T. It has been mentioned that the identification of G (together with the vector
space of T in which G is Lie-represented) is sufficient to determine the dynamical structure of T (since
the laws of T can be interpreted as invariants of G). As regards the second question, it is through the Lie
algebra associated with the symmetry group G of T that a body of empirical evidence can be entailed by T.
Let us remind us that the group-theoretic element responsible for the predictive and empirical component
of a physical theory T corresponds to the Lie algebra associated with the symmetry group G of T. This is
because the Lie representation of the Lie algebra in the vector space of T is associated with infinitesimal
transformations, some of which represent the physical properties relevant to T.
As already mentioned, the resulting underdetermination at the level of the mathematical structure can
be overcome by identifying a single, unifying formulation that approximately embraces all the empirically-
equivalent formulations of the theory in question. This can formally achieved by means of establishing a
set of conditions for structural equivalence in the group-theoretic context, and then show that the alleged
formulations satisfy those conditions. More specifically, if we consider that the structural equivalence
holding between different group-theoretic structures defines a Lie group homeomorphism between them
(hereinafter isomorphism, as interpreted under the set-theoretic representation), we must show that for any
possible symmetry group G of T whose associated Lie algebra is isomorphic to GU , there exists a symmetry
group GU of T with the same associated Lie algebra GU , such that G is isomorphic to GU . As we shall see
now, an approximation to this group-theoretic structure can be identified by means of the following claim:

For any possible symmetry group G of T whose associated Lie algebra is isomorphic to GU ,
there exists a symmetry group GU of T with associated Lie algebra GU , such that G is isomor-

17
 phic to GU , if G and GU admit true (as opposed to projective) Lie representations.

Since there could be certain formulation S of T that is not framed in group-theoretic terms and is not
structurally equivalent to other formulations S′ , S′′ , · · · of T, this claim only solves the underdetermination
problem approximately, but it suffices to demonstrate that there are formulations or reformulations based
on unifying group-theoretic structures that are less underdetermined by the same theory and by the corre-
sponding evidence.
In order to justify the above claim, it would be wise to ask the following question initially proposed by
John M. Lee: To what extent is the correspondence between Lie groups and Lie algebras (or at least between
their isomorphism classes) one-to-one? As demonstrated by (Lee, 2012, 195), the mapping that sends a Lie
group to its Lie algebra is a functor that takes Lie group isomorphisms to Lie algebra isomorphisms. It
follows from the existence of this functor that isomorphic Lie groups have isomorphic Lie algebras (Lee,
2012, 196). The converse, however, is not true: one counterexample is the real group and the torus group
which are non-isomorphic as Lie groups but their Lie algebras are isomorphic. However, by restricting
our attention to the simply connected Lie groups, the Lie group-Lie algebra correspondence turns out to be
one-to-one. Considering a theorem proved by (Lee, 2012, 530), one may demonstrate that if G and H are
simply connected Lie groups with isomorphic Lie algebras, then G and H are isomorphic. Therefore, it can
be shown that “there is a one-to-one correspondence between isomorphism classes of finite-dimensional
Lie algebras and isomorphism classes of simply connected Lie groups, given by associating each simply
connected Lie group with its Lie algebra” (Lee, 2012, 531).
In addition to this result, we can come back to the lifting theorem stated in Section 2.4 and corroborate
that Lie groups that admit true (as opposed to projective) Lie representations are simply connected as a
necessary condition. It follows from both theorems that Lie groups that admit true Lie representations have
a one-to-one correspondence with their associated Lie algebras. Thus, the above claim is finally justified.
The important outcome that follows from justifying this claim is that Lie groups accepting true Lie rep-
resentations are the kind of group theoretic-structures that can offer an approximate solution to problems of
theoretical underdetermination. In summary, we can always find a symmetry group GU of a theory T that
accepts true Lie representations, and therefore, possesses a one-to-one correspondence with its associated
Lie algebra. This implies that there exists a single symmetry group GU that accepts true Lie representations
for every Lie algebra, whilst there are many symmetry groups that only accept projective Lie representations
for every Lie algebra. Since Lie algebras are the group-theoretic elements responsible for the empirical con-
sequence of the theory in question, we conclude that the Lie group GU that accepts true Lie representations
is less underdetermined by the same Lie algebra and the same body of empirical evidence, compared to any
other symmetry group that only accepts projective Lie representations.

5.2 Case study

There is a general recipe to obtain some examples of symmetry groups of a theory T (sharing the same
Lie algebra) that only admit projective Lie representations, and therefore, are underdetermined by T and

18
its empirical consequences. This recipe is cast in the form of the following theorem demonstrated by (Lee,
2012, 557):

Let G be a finite-dimensional Lie algebra. The connected Lie groups whose Lie algebras are
isomorphic to G are (up to isomorphism) precisely those of the form G/H, where G is the
simply connected Lie group with Lie algebra G , and H is a discrete normal subgroup of G.

This means that if G is the symmetry group of a theory T with associated Lie algebra G , there could be
many non-isomorphic, non simply-connected symmetry groups of T with the same Lie algebra G , each one
corresponding to the quotient Lie group G/H for any discrete normal subgroup H of G. This can be rephrase
as stating that non simply-connected Lie groups differing up to a covering mapping form a general case
where the same Lie algebra does not necessarily mean the same Lie group.
Coming back to the problem of underdetermination, we may conclude that any quotient group of the
form G/H that is not simply-connected (and therefore, does not admit true Lie representations) is under-
determined by T and by its empirical consequences (given by their shared Lie algebras). Examples of
Lie groups differing up to a covering mapping are many, but we can provide one that is relevant to many
contemporary physicists.
As is well known by physicists, a major success for gauge invariant theories was the discovery that
the Lie group SU(2)×U(1), resulting from juxtaposing the circle group U(1) and the special unitary group
SU(2), corresponds to the symmetry group of the electroweak sector of the standard model of particle
physics. It is from this unified theory of electromagnetism and the weak nuclear force that photons and W,
Z bosons arise as elementary particles.
Considering this important observation, it can be demonstrated that the unitary group U(2) is covered by
the compact group SU(2)×U(1), where the corresponding covering homomorphism is given by p(z, A) =
zA, where z ∈ {−I, I}. This means that the Lie group U(2) is a Z2 quotient of SU(2)×U(1), where Z2 is
generated by minus the identity element of SU(2)×U(1). In other words,

U(2) = (SU(2) × U(1))/Z2.

Since both Lie groups differ by a covering mapping, they share the same underlying Lie algebra U , and
therefore, seem to have the same empirical consequences.
However, there is an important mathematical result that is associated with this fundamental symmetry
group. It can also be demonstrated that the universal cover of U(2) is not SU(2)×U(1), but the Lie group
(SU(2)×R), which also shares the same Lie algebra U . This means that the Lie algebra U uniquely
corresponds to the symmetry group SU(2)×R. As a result, we conclude that if our focus is on electroweak
phenomena, we should appeal to the symmetry group SU(2)×R that accepts true Lie representations in
order to have a less underdetermined group-theoretic structure, as opposed to U(2) or SU(2)×U(1) which
only accept projective Lie representations and, therefore, are underdetermined by the same theory and the
same body of evidence.

19
6 Conclusion
As mentioned in the introduction, there seems to be no physical way to resolve the two-fold dilemma as
regards the implementation of superselection rules. Indeed, one might either implement superselection
rules to forbid the existence of arbitrary superpositions of wave functions with different parameter without
changing the underlying projective Lie representations or one might modify the projective Lie representa-
tion to yield a true Lie representation without forbidding the existence of these superpositions and without
implementing superselection rules.
However, the fact that there are no physical reasons to resolve this dispute does not mean that there
could not be philosophical reasons to do that. In this spirit, the conclusion of this contribution is that it is
reasonable to think that the second option is better by virtue of the fact that the elimination of superselection
rules leads to heuristic and epistemological virtues. As regards the heuristic virtues, we have demonstrated
that, although true Lie representations incorporate surplus mathematical structure as a result of forbidding
the implementation of superselection rules, they are significant mathematical sources of potential develop-
ments. As regards the epistemological virtues, we have demonstrated that there are theories that can be
formulated or reformulated in terms of a single symmetry group that admits true Lie representations that,
nevertheless, turn out to be less underdetermined compared to other symmetry groups that admit projective
ones, and lead to the implementation of superselection rules.

References
Bargmann,V. (1954). On unitary ray representations of continuous groups. Annals of Mathematics, 59(1),
1-46.

Belot, G. (2006). The representation of time and change in mechanics. In J. Butterfield, & J. Earman (Eds.),
Philosophy of Physics (pp. 133-228). Amsterdam: North-Holland.

Bueno, O., French, S., & Ladyman, J. (2002). On representing the relationship between the mathematical
and the empirical. Philosophy of Science, 69, 452-473.

Bueno, O., & French, S. (2018). Applying mathematics. Oxford: Oxford University Press.

da Costa, N., & French, S. (2003). Science and partial truth: A unitary approach to models and scientific
reasoning. Oxford: Oxford University Press.

French, S. (2014). The structure of the World: Metaphysics and representation. Oxford: Clarendon Press.

Giulini, D., Joos, E., Kiefer, C., Kupsch, J., Stamatescu, I., & Zeh, D. (1996). Decoherence and the appear-
ance of a classical world in quantum theory. Heidelberg: Springer-Verlag.

Greenberger, D. (2001). Inadequacy of the usual Galilean transformation in quantum mechanics. Physical
Review Letters, 87(10), 1004051-1004054.

20
Hegerfeld, G., Kraus, K., & Wigner, E. (1968). Proof of the fermion superselection rule without the as-
sumption of time-reversal invariance. Journal of Mathematical Physics, 9, 2029-2031.

Hernandez, H. (2012). From Bargmann’s superselection rule to quantum Newtonian spacetime. Foundations
of Physics, 42(10), 1350-1364.

Jammer, M. (1974). Philosophy of quantum mechanics. The interpretations of quantum mechanics in his-
torical perspective. New York: John Wiley and Sons.

Laudan, L., & Leplin, J. (1991). Empirical equivalence and underdetermination. The Journal of Philosophy,
88, 449-472.

Lee, J. (2012). Introduction to smooth manifolds. Graduate Texts in Mathematics. Vol. 218 (Second ed.).
New York London: Springer-Verlag.

Mirman, R. (1970). Analysis of the experimental meaning of coherent superposition and the nonexistence
of superselection rules. Physical Review D, 1, 3349-3363.

Pooley, O. (2006). Points, particles and structural realism. In D. Rickles, S. French, & J. Saatsi (Eds.),
Structural foundations of quantum gravity (pp. 83-120). Oxford: Oxford University Press.

Rauch, H., Zeilinger, A., Badurek, G., & Wilfing, A. (1975). Verification of coherent spinor rotation of
fermions. Physical Letters, 54A(6), 425-427.

Redhead, M. (1975). Symmetry in intertheory relations. Synthese, 32(1), 77-112.

Weinberg, S. (1995). The quantum theory of fields I. Cambridge: Cambridge University Press.

Wess, J., & Zumino, B. (1971). Consequences of anomalous Ward identities. Physical Letters B, 37, 95-97.

Weyl, H. (1931). The theory of groups and quantum mechanics. New York: Dover.

Wigner, E. (1939). On unitary representations of the inhomogeneous Lorentz group. Annals of Mathematics,
40, 149-204.

21