The Topological Orthogonal Group and the Spin Group

Introduction

The isometries of a quadratic form form a group, the orthogonal group, and over the Clifford algebra they are covered two-to-one by the Pin and Spin groups of the units generated by vectors. Both groups are studied algebraically in Part I: the isometry group is generated by reflections, the Clifford group is the set of invertibles that conjugate the space of vectors to itself, the twisted conjugation action of the Clifford group on the vectors is the isometry, and the kernel on the invertibles of norm one is $\{\pm1\}$, which is the double cover. This article reads those groups with a topology, and the content is entirely about the topology: the orthogonal group is closed in the group of invertible operators and is a topological group; it is compact for a definite form over the real numbers and only locally compact for an indefinite one; the determinant separates two components and the indefinite forms have two more; the covering by Pin and Spin is a covering map of topological groups, not only an algebraic covering; the spinor norm is a continuous homomorphism into a finite group when the field is the real numbers; and the low-dimensional isomorphisms are homeomorphisms of compact groups.

The two facts that carry the article are the closedness of the isometry group and the covering property of the Pin–Spin map. The first is a one-line argument — an isometry is an invertible operator satisfying continuous equations — but it has two consequences that matter later: the group is a topological group for the induced operations, and it is locally compact exactly when the ambient algebra is, which over a complete valued field and in finite dimension happens always and gives the open-mapping theorem for its continuous homomorphisms. The second fact is what makes the words double cover a topological statement: the Clifford group is a group of invertibles of a finite-dimensional algebra, hence is locally compact; the action map on the vectors is a continuous surjective homomorphism onto the orthogonal group, hence is open by the open-mapping theorem for locally compact groups; its kernel is the group of scalars, and on the elements of norm one the kernel is the two-element group $\{\pm1\}$, which is discrete. A continuous open surjective homomorphism with discrete kernel is a covering map, and the Pin group is therefore a covering space of the orthogonal group, with the Spin group covering the special orthogonal group.

The article treats the isometry group and its topology, its compactness in the definite case, the determinant and the components, the generation by reflections and the topology of the set of reflections, the covering by Pin and Spin, the spinor norm, and the low-dimensional isomorphisms. The symplectic group of a continuous alternating form is the counterpart of The Continuous Symplectic Form and the Poisson Bracket, and the unitary group of a Hermitian form is the counterpart of Topology on Sesqualgebras with a degree-2 form; both are named where the arguments are identical. The smooth structure, the Lie algebra, the exponential map and the homogeneous space structure are Geometry; the fundamental group and the theory of covering spaces are Algebraic Topology; the algebraic generation of the orthogonal group by reflections and the algebraic definition of the Clifford group are the Part I entries The Orthogonal Lie Algebra, The Cartan Involution and the Cartan Decomposition and The Two-Sided Operators and the Spin Group. Throughout, $F$ is a topological field of characteristic not $2$ or $0$ as the antisymmetrisation requires, $V$ is a finite-dimensional space over $F$ of dimension $n$ with a non-degenerate quadratic form $\mathfrak{Q}$, and the Clifford algebra, the Clifford group and the norm are those of Topological Clifford Algebras and of the Part I entries.

The Isometry Group as a Closed Subgroup

Definition. The orthogonal group of the quadratic space is

$$ \mathrm{O}(V,\mathfrak{Q}) = \{T\in GL(V) : \mathfrak{Q}(Tv)=\mathfrak{Q}(v)\ \text{for all}\ v\in V\} , $$

and the special orthogonal group is $\mathrm{SO}(V,\mathfrak{Q})=\{T\in\mathrm{O}(V,\mathfrak{Q}) : \det T=1\}$, as in The Orthogonal Lie Algebra.

Theorem (the isometry group is closed). The group $\mathrm{O}(V,\mathfrak{Q})$ is a closed subgroup of $GL(V)$ with the topology induced from $\operatorname{End}(V)$, and it is a topological group for the induced operations. The group $\mathrm{SO}(V,\mathfrak{Q})$ is closed in $\mathrm{O}(V,\mathfrak{Q})$ and open in it.

Proof. For each $v$ the map $T\mapsto \mathfrak{Q}(Tv)-\mathfrak{Q}(v)$ is continuous, being the composite of the evaluation, the action of $T$ and the continuous form; the set where all these vanish is an intersection of closed sets, hence closed, and the condition that a spanning set of vectors be preserved suffices. The determinant $\det : \operatorname{End}(V)\to F$ is polynomial, hence continuous, so $\mathrm{SO}$ is the closed preimage of $\{1\}$; it is also the preimage of the complement of $\{1\}$ under the determinant together with $\mathrm{O}\setminus\mathrm{SO}$, hence open in $\mathrm{O}$, because the determinant of an isometry has square one and so takes at most two values different from $0$, which makes the two sets $\{1\}$ and its complement clopen in the image. For the topological group statement, the multiplication and the inversion of $GL(V)$ are continuous for the topology induced from $\operatorname{End}(V)$, as in Topological Clifford Algebras, §The Continuity of the Structure, and the closed subgroup inherits them.

Corollary (local compactness and its consequences). Let $F$ be a complete valued field and $V$ finite-dimensional. Then $\mathrm{O}(V,\mathfrak{Q})$ is a locally compact Hausdorff topological group, and every continuous surjective homomorphism from it onto a locally compact group is open.

Proof. A closed subset of the finite-dimensional space $\operatorname{End}(V)$ is locally compact, and a closed subgroup of a locally compact group is locally compact; the open-mapping theorem for continuous surjective homomorphisms of locally compact groups is classical.

Remark (the infinite-dimensional case). For an infinite-dimensional normed space the same proof gives the closedness of $\mathrm{O}(V,\mathfrak{Q})$ in $GL(V)$ and produces a topological group, but the group is not locally compact and the open-mapping theorem is not available. The consequence is that the covering property of the next sections is a finite-dimensional statement, and the article says so; the infinite-dimensional analogues use the completion and the CAR algebra of The Completion of a Clifford Algebra and Infinite-Dimensional Clifford Algebras and CAR with Inner Conjugation.

Compactness and the Components

Theorem (the definite case is compact). Let $F=\mathbb{R}$ and let $q$ be positive definite. Then $\mathrm{O}(V,\mathfrak{Q})$ is compact.

Proof. The group is closed and it is bounded: an isometry preserves the norm $\lVert v\rVert=\mathfrak{Q}(v)^{1/2}$ and hence its operator norm is one, since $\lVert T\rVert=\sup\{\mathfrak{Q}(Tv)^{1/2} : \mathfrak{Q}(v)=1\}=1$. A closed bounded subset of a finite-dimensional real space is compact.

Theorem (the determinant and the components). The determinant is a continuous homomorphism $\det : \mathrm{O}(V,\mathfrak{Q})\to\{\pm1\}$ onto a discrete group, so $\mathrm{SO}(V,\mathfrak{Q})$ is a clopen normal subgroup of index at most two. If $F=\mathbb{R}$ and $q$ is definite then $\mathrm{O}(V,\mathfrak{Q})$ has exactly two connected components, which are $\mathrm{SO}$ and the coset of a reflection, and $\mathrm{SO}$ is connected. If $q$ has signature $(n,m)$ with $p\ge1$ and $q\ge1$, then $\mathrm{O}(V,\mathfrak{Q})$ has four connected components, indexed by the determinant and by the sign of the determinant of the restriction to a maximal positive-definite subspace, and $\mathrm{SO}$ has two, the identity component being the component of the identity of $\mathrm{O}$.

Proof. The determinant of an isometry satisfies $\det(T)^{2}=1$, since $T$ preserves the form up to the discriminant, so its image lies in $\{\pm1\}$, a discrete group in which every point is clopen; the preimage of a clopen set is clopen, and the kernel of a homomorphism onto $\{\pm1\}$ has index at most two. For the definite case the connectedness of $\mathrm{SO}$ is proved by the polar decomposition, whose home is the Hermitian layer, and the count of components there follows; see the remark below. For the indefinite case the extra invariant is the sign of the determinant of the restriction to a maximal positive subspace, which is locally constant and separates the two components of $\mathrm{SO}$; the details of the count are those of the classical theory of the orthogonal groups.

Remark (where the proof of connectedness is). The connectedness of $\mathrm{SO}(V,\mathfrak{Q})$ in the definite case uses the polar decomposition and the exponential, which are proved in the Hermitian articles Unitary Equivalence and Congruence of Operators of the Form and The Polar Decomposition of an Operator of the Form; the article therefore states the result and defers its proof, rather than duplicating it. What is proved here is the closedness, the compactness, the clopenness of $\mathrm{SO}$ and the local constancy of the extra invariant, which is what the covering theory below needs.

Reflections and the Topology of their Set

Definition. For $a\in V$ with $\mathfrak{Q}(a)\neq0$ the reflection in the hyperplane orthogonal to $a$ is the isometry $s_{a}(v)=v-2B(v,a)\mathfrak{Q}(a)^{-1}a$, as in The Two-Sided Operators and the Spin Group.

Proposition. The map $a\mapsto s_{a}$ is continuous from the open set $\{a : \mathfrak{Q}(a)\neq0\}$ to $\mathrm{O}(V,\mathfrak{Q})$, its image is the set of reflections, and in the definite real case the set of reflections is compact, being the continuous image of the sphere $\{N(a)=1\}=\{\mathfrak{Q}(a)=-1\}$.

Proof. The formula for $s_{a}$ is a rational expression in $a$ and $\mathfrak{Q}(a)$, and $\mathfrak{Q}(a)\neq0$ on the domain, so the map is continuous; the image is the set of reflections by definition of the latter. In the definite case the map $a\mapsto s_{a}$ is invariant under $a\mapsto\lambda a$ for $\lambda\neq0$, so it factors through the level set $\{N(a)=1\}=\{\mathfrak{Q}(a)=-1\}$, a sphere, hence compact, and the image of a compact set is compact.

Theorem (generation in the topological form). Every element of $\mathrm{O}(V,\mathfrak{Q})$ is a product of at most $n$ reflections, and in the definite real case the map $(s_{1},\dots,s_{n})\mapsto s_{1}\cdots s_{n}$ from the $n$-fold product of the set of reflections onto $\mathrm{O}(V,\mathfrak{Q})$ is continuous and surjective; hence $\mathrm{O}(V,\mathfrak{Q})$ is a continuous image of a compact set, which reproves its compactness.

Proof. The algebraic statement is the theorem of Cartan–Dieudonné as in The Two-Sided Operators and the Spin Group; the continuity is the continuity of the product in the topological group, and the surjectivity is the algebraic theorem. The image of a compact set under a continuous map is compact, and the image is the whole group.

Remark (the parity of the length). The parity of the number of reflections in a decomposition is well defined, since the determinant of a reflection is $-1$; the map $\mathrm{O}(V,\mathfrak{Q})\to\{\pm1\}$ it defines is the determinant, and the article does not use it beyond the clopenness above. The Dickson invariant of The Two-Sided Operators and the Spin Group refines the determinant when the field is not the real numbers, and it is a homomorphism into a discrete group, hence also locally constant.

The Two-Fold Covering by Pin and Spin

Definition. The Clifford group is the group of versors $\Gamma(V,\mathfrak{Q})=\{x\in \mathrm{Cl}(V,\mathfrak{Q})^{\times} : \chi_{x}(V)\subseteq V\}$ with $\chi_{x}(y)=x\,y\,\alpha(x)^{-1}$, where $\alpha$ is the parity involution; the Clifford norm is $N(x)=x\,x^{\natural}$ for the Clifford conjugation $x^{\natural}=\alpha(x^{r})$, so that $N(u)=-\mathfrak{Q}(u)$ on a vector; the Pin group is $\mathrm{Pin}(V,\mathfrak{Q})=\{x\in\Gamma(V,\mathfrak{Q}) : N(x)=1\}$ and the Spin group is its even part $\mathrm{Spin}(V,\mathfrak{Q})=\mathrm{Pin}(V,\mathfrak{Q})\cap\mathrm{Cl}^{0}(V,\mathfrak{Q})$, as in The Two-Sided Operators and the Spin Group and The Clifford, Pin and Spin Groups with Signed Inner Conjugation. The action of the Clifford group on the vectors is

$$ \boxed{\;\rho(x) : v\mapsto \chi_{x}(v)=x\,v\,\alpha(x)^{-1}\;} . $$

Lemma (the groups of the cover are locally compact closed subsets). Let $F$ be a complete valued field and $V$ finite-dimensional. Then the Clifford group is locally closed in the finite-dimensional algebra and the Pin and Spin groups are closed in the group of invertibles; each is a locally compact topological group.

Proof. The group of invertibles of a finite-dimensional algebra is open, as in The Continuous Volume Element, Duality and the Hodge Star, §The Volume Element and its Continuity; the condition $\alpha(x)vx^{-1}\in V$ for a spanning set of vectors is a closed condition in the group of invertibles, since $x\mapsto x^{-1}$ is continuous there, so the Clifford group is closed in the invertibles and hence locally closed in the algebra; the norm condition is closed, being the preimage of $\{1\}$; and the intersection with the even part is closed. Closed subgroups of locally compact groups are locally compact, and a locally closed subset of a finite-dimensional space is locally compact.

Theorem (the covering). The action $\rho$ is a continuous surjective homomorphism of $\Gamma$ onto $\mathrm{O}(V,\mathfrak{Q})$ with kernel the scalars of the algebra, and on $\mathrm{Pin}(V,\mathfrak{Q})$ it has kernel $\{\pm1\}$. Consequently $\rho : \mathrm{Pin}(V,\mathfrak{Q})\to\mathrm{O}(V,\mathfrak{Q})$ is a two-fold covering map of topological groups, $\rho^{-1}(\mathrm{SO}(V,\mathfrak{Q}))=\mathrm{Spin}(V,\mathfrak{Q})$, and $\rho : \mathrm{Spin}(V,\mathfrak{Q})\to\mathrm{SO}(V,\mathfrak{Q})$ is a two-fold covering map onto the special orthogonal group.

Proof. The algebraic statements, that $\rho$ is multiplicative, that $\rho(x)$ is an isometry and that the map onto $\mathrm{O}(V,\mathfrak{Q})$ is surjective by the generation of the orthogonal group by reflections, are the Part I theorems; the kernel on the Clifford group is the set of central scalars, and on the elements of norm one the condition $N(x)=1$ leaves only $\{\pm1\}$, since a central scalar $\lambda$ has $N(\lambda)=\lambda^{2}$ and the two roots of $\lambda^{2}=1$ are $\pm1$ in the characteristic not two of the layer. The map is continuous, and it is open by the open-mapping theorem for continuous surjective homomorphisms of locally compact groups, applied to the locally compact groups of the lemma; a continuous open surjective homomorphism whose kernel is discrete is a covering map, and $\{\pm1\}$ is discrete. The identification of the preimage of $\mathrm{SO}$ with the even Pin group is the algebraic statement of the parity of the length of a reflection decomposition.

Corollary (the covering has two sheets and is nontrivial). The fibre $\rho^{-1}(T)$ has two elements for every $T$. The covering is nontrivial, that is, it admits no continuous section into the Clifford group mapping the identity to the identity, whenever $-1$ is not a square in the algebra's centre and the even Clifford group is connected; in particular $\mathrm{Spin}(V,\mathfrak{Q})\to\mathrm{SO}(V,\mathfrak{Q})$ does not split over the real numbers for a definite form of dimension at least two.

Proof. A covering with a global continuous section is a product cover; a section of a group covering with $\sigma(1)=1$ is an algebraic splitting, and the existence of a splitting is an algebraic statement about the group, which fails when $-1$ is not a square and the group is connected. The classical instance is the nontriviality of the double cover of $\mathrm{SO}(n)$; the article records the criterion and leaves the case-by-case verification to the low-dimensional lists.

Remark (the universal cover and where it is proved). For a definite real form of dimension at least three the covering $\mathrm{Spin}(V,\mathfrak{Q})\to\mathrm{SO}(V,\mathfrak{Q})$ is the universal covering, a topological statement whose proof belongs to Algebraic Topology: it requires the connectedness and the vanishing of the fundamental group of the Spin group beyond its two-element centre, the computation $\pi_{1}(\mathrm{SO}(n))=\mathbb{Z}/2$ of The Fundamental Group and Covering Spaces and the theory of covering spaces of the same part. The present article stops at the covering property; the identification with the universal cover is stated in The Low-Dimensional Spin Groups and the Exceptional Isomorphisms with Inner Conjugation and proved there by the topological means of that layer.

The Spinor Norm

Definition. The spinor norm of an isometry $\sigma=\rho(x)$ expressed by a versor $x$ of length $k$ is

$$ \theta : \mathrm{O}(V,\mathfrak{Q})\to F^{\times}/(F^{\times})^{2}, \qquad \theta(\sigma) = (-1)^{k}N(x)\ (F^{\times})^{2}, $$

as in The Spinor Norm and the Structure of the Orthogonal Group with Inner Conjugation; it is well defined because a second versor differs from the first by a scalar, whose norm is a square, and because the parity of the length is the determinant. On a reflection it returns $\theta(s_{a})=\mathfrak{Q}(a)$ modulo squares.

Proposition (continuity of the spinor norm). Let $F=\mathbb{R}$. Then the spinor norm takes values in the two-element group $\mathbb{R}^{\times}/(\mathbb{R}^{\times})^{2}=\{\pm1\}$ and is a continuous homomorphism $\mathrm{O}(V,\mathfrak{Q})\to\{\pm1\}$; consequently it is locally constant and its kernel is a clopen subgroup of index at most two, containing the identity component of $\mathrm{O}(V,\mathfrak{Q})$.

Proof. Over the real numbers every nonzero scalar is a square or the negative of a square according as it is positive or negative, so the target is the two-element group $\{\pm1\}$; the map is a homomorphism by the algebra of the spinor norm. A homomorphism into a discrete group is continuous if and only if its kernel is open; the kernel is the preimage of the identity, and it is open because the map is locally constant, which in turn follows because the reflection formula depends continuously on $a$ and the sign of $\mathfrak{Q}(a)$ is locally constant on $\{a : \mathfrak{Q}(a)\neq0\}$; and the image of a connected set is connected, so the identity component lies in the kernel.

Remark (the relation to the covering). The spinor norm measures the obstruction to lifting an isometry to the Clifford group: an isometry $T\in\mathrm{O}(V,\mathfrak{Q})$ lifts to $\mathrm{Pin}(V,\mathfrak{Q})$ if and only if $\mathrm{N}(T)=1$, and then there are exactly two lifts. The statement is algebraic in form and topological in content, since it says that the covering of the theorem is the one classified by a continuous homomorphism into a two-element group; the classification of the two-fold coverings of a connected group by its continuous homomorphisms into $\mathbb{Z}/2$ is the theory of The Fundamental Group and Covering Spaces.

The Low-Dimensional Cases

Example ($\mathrm{Spin}(3)\cong SU(2)$ and the three-sphere). For the definite real form of dimension three the Spin group is isomorphic to the group $SU(2)$ of unitary quaternions, which is homeomorphic to the three-dimensional sphere; the covering $\mathrm{Spin}(3)\to\mathrm{SO}(3)$ is the standard two-fold covering of the rotation group by the sphere, with fibre $\{\pm1\}$, and $\mathrm{SO}(3)$ is homeomorphic to the real projective space of dimension three, which is the quotient of the sphere by the antipodal map. Every statement here is a homeomorphism of compact groups and is developed in The Low-Dimensional Spin Groups and the Exceptional Isomorphisms with Inner Conjugation; the article records the topological shape because it is the model of the covering.

Example ($\mathrm{Spin}(4)\cong SU(2)\times SU(2)$). For the definite real form of dimension four the Spin group is the product of the two copies of the unit quaternions, one acting on the self-dual and the other on the anti-self-dual bivectors of The Continuous Volume Element, Duality and the Hodge Star, §The Two Cases as Topological Decompositions; the covering is the two-fold covering of the compact group $\mathrm{SO}(4)$, and the splitting of the covering over the real numbers for even dimension four is the algebraic reflection of the fact that the two factors act independently. The example is the one where the covering is closest to a product, and it shows that the nontriviality criterion of the corollary above is a criterion and not a theorem about all dimensions.

Example (the indefinite case). For a real form of signature $(1,3)$ the identity component of the orthogonal group is not all of $\mathrm{SO}$ and is not compact; the Spin group covering it is the universal cover of the identity component under the same topological hypotheses as in the definite case, and the covering of $\mathrm{SO}$ has four components above the two components of $\mathrm{SO}$. The example is the Lorentzian one of The Low-Dimensional Spin Groups and the Exceptional Isomorphisms with Inner Conjugation, and it shows that the count of components of the cover is the count of the base, doubled.

Summary

The orthogonal group $\mathrm{O}(V,\mathfrak{Q})$ is a closed subgroup of $GL(V)$, hence a topological group for the induced operations, and in finite dimension over a complete valued field it is locally compact; it is compact for a definite real form, being closed and bounded. The determinant is a continuous homomorphism onto $\{\pm1\}$, so $\mathrm{SO}(V,\mathfrak{Q})$ is a clopen subgroup of index at most two; the definite real case has exactly two components and the indefinite case four, the extra invariant being the locally constant sign of the determinant on a maximal positive subspace. The map $a\mapsto s_{a}$ is continuous from $\{q\neq0\}$ onto the set of reflections, and in the definite real case that set is compact, being the image of the unit sphere; the generation of the group by at most $n$ reflections makes $\mathrm{O}(V,\mathfrak{Q})$ a continuous image of a compact set.

The Clifford group is locally closed in the finite-dimensional algebra and the Pin and Spin groups are closed in it, hence locally compact; the action $\rho(x)v=\alpha(x)vx^{-1}$ is a continuous surjective homomorphism onto $\mathrm{O}(V,\mathfrak{Q})$ with kernel the scalars, and with kernel exactly $\{\pm1\}$ on the Pin group. By the open-mapping theorem for locally compact groups a continuous open surjective homomorphism with discrete kernel is a covering map, so Pin is a two-fold covering of $\mathrm{O}$ and Spin a two-fold covering of $\mathrm{SO}$, with two elements in every fibre; the covering is nontrivial when $-1$ is not a square and the even group is connected, and it is the universal covering in the definite case of dimension at least three — a statement owned by Algebraic Topology.

The spinor norm $\theta$ is a homomorphism $\mathrm{O}(V,\mathfrak{Q})\to F^{\times}/(F^{\times})^{2}$ sending a reflection to $\mathfrak{Q}(a)$ modulo squares; over $\mathbb{R}$ it takes values in the two-element group, it is continuous, hence locally constant, with kernel a clopen subgroup of index at most two containing the identity component, and an isometry lifts to the Pin group exactly when its spinor norm is one. The low-dimensional cases exhibit the covering concretely: $\mathrm{Spin}(3)\cong SU(2)$ covers $\mathrm{SO}(3)$ two-to-one with $\mathrm{SO}(3)$ the real projective space, $\mathrm{Spin}(4)$ is a product of two copies of $SU(2)$ and the covering splits, and in signature $(1,3)$ the cover has twice the components of the base. What is not here: the Lie group structure, the exponential and the homogeneous spaces are Geometry; the fundamental group and the classification of coverings are Algebraic Topology; and in infinite dimension the groups are not locally compact, so the covering argument awaits the completions of The Completion of a Clifford Algebra.

Summary of Notation

symbol meaning
$\mathrm{O}(V,\mathfrak{Q})$, $\mathrm{SO}(V,\mathfrak{Q})$ the isometry group and its determinant-one subgroup, clopen
$\Gamma$ the Clifford group $\{x : \alpha(x)vx^{-1}\in V\}$
$\mathrm{Pin}(V,\mathfrak{Q})$, $\mathrm{Spin}(V,\mathfrak{Q})$ the versors of norm $1$ and their even part
$\rho(x)v=\alpha(x)vx^{-1}$ the action onto $\mathrm{O}(V,\mathfrak{Q})$, a two-fold covering on Pin
$s_{a}$ the reflection, the map $a\mapsto s_{a}$ continuous on $\{q\neq0\}$
$\theta$ the spinor norm, continuous into $\{\pm1\}$ over $\mathbb{R}$
$\{\pm1\}$ the discrete kernel, the two sheets of the covering
$\mathrm{Spin}(3)\cong SU(2)\cong S^{3}$ the model of the covering

Further Reading

  • Claude Chevalley, The Algebraic Theory of Spinors and Clifford Algebras (Springer, 1997), for the Clifford group, the norm and the action on the vectors.
  • Pertti Lounesto, Clifford Algebras and Spinors, 2nd edition (Cambridge University Press, 2001), for the Pin and Spin groups and the low-dimensional isomorphisms.
  • Ian R. Porteous, Clifford Algebras and the Classical Groups (Cambridge University Press, 1995), for the spinor norm and the generation by reflections.
  • Nicolas Bourbaki, Topological Groups (Springer, 1998), for the open-mapping theorem for locally compact groups and the covering properties of continuous homomorphisms with discrete kernel.
  • John Stillwell, Naive Lie Theory (Springer, 2008), for the concrete covering $\mathrm{Spin}(3)\to\mathrm{SO}(3)$ and the shape of $\mathrm{SO}(3)$.
  • Allen Hatcher, Algebraic Topology (Cambridge University Press, 2002), for the fundamental group of the orthogonal and spin groups and the classification of covering spaces, named here and deferred.