Mathematical Physics

The University of Chicago Press, Chicago 60637

The University of Chicago Press, Ltd., London

© 1985 by The University of Chicago

All rights reserved. Published 1985

Printed in the United States of America

03 02 01 00 99 98      6 7 8 9

ISBN 978-0-226-22306-3 (e-book)

Library of Congress Cataloging in Publication Data

Geroch, Robert.

Mathematical physics.

(Chicago lectures in physics)

1. Mathematical physics.   I. Title.   II. Series.

QC20.G47      1985         530.1'5         85-17764

ISBN 0-226-28862-5 (pbk.)

Mathematical Physics

Robert Geroch

The University of Chicago Press

Chicago and London

Chicago Lectures in Physics

Robert M. Wald, Editor

Hellmut Fritzsche

Riccardo Levi-Setti

Roland Winston

Contents

1. Introduction

2. Categories

3. The Category of Groups

4. Subgroups

5. Normal Subgroups

6. Homomorphisms

7. Direct Products and Sums of Groups

8. Relations

9. The Category of Vector Spaces

10. Subspaces

11. Linear Mappings; Direct Products and Sums

12. From Real to Complex Vector Spaces and Back

13. Duals

14. Multilinear Mappings; Tensor Products

15. Example: Minkowski Vector Space

16. Example: The Lorentz Group

17. Functors

18. The Category of Associative Algebras

19. The Category of Lie Algebras

20. Example: The Algebra of Observables

21. Example: Fock Vector Space

22. Representations: General Theory

23. Representations on Vector Spaces

24. The Algebraic Categories: Summary

25. Subsets and Mappings

26. Topological Spaces

27. Continuous Mappings

28. The Category of Topological Spaces

29. Nets

30. Compactness

31. The Compact-Open Topology

32. Connectedness

33. Example: Dynamical Systems

34. Homotopy

35. Homology

36. Homology: Relation to Homotopy

37. The Homology Functors

38. Uniform Spaces

39. The Completion of a Uniform Space

40. Topological Groups

41. Topological Vector Spaces

42. Categories: Summary

43. Measure Spaces

44. Constructing Measure Spaces

45. Measurable Functions

46. Integrals

47. Distributions

48. Hilbert Spaces

49. Bounded Operators

50. The Spectrum of a Bounded Operator

51. The Spectral Theorem: Finite-dimensional Case

52. Continuous Functions of a Hermitian Operator

53. Other Functions of a Hermitian Operator

54. The Spectral Theorem

55. Operators (Not Necessarily Bounded)

56. Self-Adjoint Operators

Notes

Index of Defined Terms

1

Introduction

One sometimes hears expressed the view that some sort of uncertainty principle operates in the interaction between mathematics and physics: the greater the mathematical care used to formulate a concept, the less the physical insight to be gained from that formulation. It is not difficult to imagine how such a viewpoint could come to be popular. It is often the case that the essential physical ideas of a discussion are smothered by mathematics through excessive definitions, concern over irrelevant generality, etc. Nonetheless, one can make a case that mathematics as mathematics, if used thoughtfully, is almost always useful—and occasionally essential—to progress in theoretical physics.

What one often tries to do in mathematics is to isolate some given structure for concentrated, individual study: what constructions, what results, what definitions, what relationships are available in the presence of a certain mathematical structure—and only that structure? But this is exactly the sort of thing that can be useful in physics, for, in a given physical application, some particular mathematical structure becomes available naturally, namely, that which arises from the physics of the problem. Thus mathematics can serve to provide a framework within which one deals only with quantities of physical significance, ignoring other, irrelevant things. One becomes able to focus on the physics. The idea is to isolate mathematical structures, one at a time, to learn what they are and what they can do. Such a body of knowledge, once established, can then be called upon whenever it makes contact with the physics.

An everyday example of this point is the idea of a derivative. One could imagine physicists who do not understand, as mathematics, the notion of a derivative and the properties of derivatives. Such physicists could still formulate physical laws, for example, by speaking of the rate of change of . . . with . . . They could use their physical intuition to obtain, as needed in various applications, particular properties of these rates of change. It would be more convenient, however, to isolate the notion derivative once and for all, without direct reference to later physical applications of this concept. One learns what a derivative is and what its properties are: the geometrical significance of a derivative, the rule for taking the derivative of a product, etc. This established body of knowledge then comes into play automatically when the physics requires the use of derivatives. Having mastered the abstract concept rate of change all by itself, the mind is freed for the important, that is, the physical, issues.

The only problem is that it takes a certain amount of effort to learn mathematics. Fortunately, two circumstances here intervene. First, the mathematics one needs for theoretical physics can often be mastered simply by making a sufficient effort. This activity is quite different from, and far more straightforward than, the originality and creativity needed in physics itself. Second, it seems to be the case in practice that the mathematics one needs in physics is not of a highly sophisticated sort. One hardly ever uses elaborate theorems or long strings of definitions. Rather, what one almost always uses, in various areas of mathematics, is the five or six basic definitions, some examples to give the definitions life, a few lemmas to relate various definitions to each other, and a couple of constructions. In short, what one needs from mathematics is a general idea of what areas of mathematics are available and, in each area, enough of the flavor of what is going on to feel comfortable. This broad and largely shallow coverage should in my view be the stuff of mathematical physics.

There is, of course, a second, more familiar role of mathematics in physics: that of solving specific physical problems which have already been formulated mathematically. This role encompasses such topics as special functions and solutions of differential equations. This second role has come to dominate the first in the traditional undergraduate and graduate curricula. My purpose, in part, is to argue for redressing the balance.

We shall here take a brief walking tour through various areas of mathematics, providing, where appropriate and available, examples in which this mathematics provides a framework for the formulation of physical ideas.

By way of general organization, chapters 2–24 deal with things algebraic and chapters 25–42 with things topological. In chapters 43–50 we discuss some special topics: structures which combine algebra and topology, Lebesque integrals, Hilbert spaces. Lest the impression be left that no difficult mathematics can ever be useful in physics, we provide, in chapters 51–56, a counterexample: the spectral theorem. Strictly speaking, the only prerequisites are a little elementary set theory, algebra, and, in a few places, some elementary calculus. Yet some informal contact with such objects as groups, vector spaces, and topological spaces would be most helpful.

The following texts are recommended for additional reading: A. H. Wallace, Algebraic Topology (Elmsford, NY: Pergamon, 1963), and C. Goffman and G. Pedrick, First Course in Functional Analysis (Englewood Cliffs, NJ: Prentice-Hall, 1965). Two examples of more advanced texts, to which the present text might be regarded as an introduction, are: M. Reed and B. Simon, Methods of Modern Mathematical Physics (New York: Academic, 1972), and Y. Choquet-Bruhat, C. DeWitt-Morette, and M. Dillard-Bleick, Analysis, Manifolds and Physics (Amsterdam: North-Holland, 1982).

2

Categories

In each area of mathematics (e.g., groups, topological spaces) there are available many definitions and constructions. It turns out, however, that there are a number of notions (e.g., that of a product) that occur naturally in various areas of mathematics, with only slight changes from one area to another. It is convenient to take advantage of this observation. Category theory can be described as that branch of mathematics in which one studies certain definitions in a broader context—without reference to the particular area to which the definition might be applied. It is the mathematics of mathematics. Although this subject takes a little getting used to, it is, in my opinion, worth the effort. It provides a systematic framework that can help one to remember definitions in various areas of mathematics, to understand what many constructions mean and how they can be used, and even to invent useful definitions when needed. We here summarize a few facts from category theory.

A category consists of three things—i) a class O (whose elements will be called objects), ii) a set Mor(A, B) (whose elements will be called morphisms from A to B), where A and B are any two¹ objects, and iii) a rule which assigns, given any objects A, B, and C and any morphism φ from A to B and morphism ψ from B to C, a morphism, written ψ ∘ φ, from A to C (this ψ ∘ φ will be called the composition of φ with ψ)—subject to the following two conditions:

1. Composition is associative. If A, B, C, and D are any four objects, and φ, ψ, and λ are morphisms from A to B, from B to C, and from C to D, respectively, then

(Note that each side of this equation is a morphism from A to D.)

2. Identities exist. For each object A, there is a morphism iA from A to A (called the identity morphism on A) with the following property: if φ is any morphism from A to B, then

if μ is any morphism from C to A, then

That is the definition of a category. It all seems rather abstract. In order to see what is really going on with this definition—why it is what it is—one has to look at a few examples. We shall have abundant opportunity to do this: almost every mathematical structure we look at will turn out to be an example of a category. In order to fix ideas for the present, we consider just one example (the simplest, and probably the best).

To give an example of a category, one must say what the objects are, what the morphisms are, what composition of morphisms is—and one must verify that conditions 1 and 2 above are indeed satisfied. Let the objects be ordinary sets. For two objects (now, sets) A and B, let Mor(A, B) be the set of all mappings from the set A to the set B. (Recall that a mapping from set A to set B is a rule that assigns, to each element of A, some element of B.) Finally, let composition of morphisms, in this example, be ordinary composition of mappings. (That is, if φ is a mapping from set A to set B and ψ is a mapping from set B to set C, then ψ ∘ φ is the mapping from set A to set C which sends the element a of A to the element ψ(φ(a)) of C.) We now have the objects, the morphisms, and the composition law. We must check that conditions 1 and 2 are satisfied. Condition 1 is indeed satisfied in this case: it is precisely the statement that composition of mappings on sets is associative. Condition 2 is also satisfied: for any set A, let iA be the identity mapping (i.e., for each element a of A, iA(a) = a) from A to A. Thus we have here an example of a category. It is called the category of sets.

This example is in some sense typical. It is helpful to think of the objects as being really sets (perhaps, as in later examples, with additional structure) and of the morphisms as really mappings (which, in these later examples, will be structure preserving). With this mental picture, it is easy to remember the definition of a category—and to follow the constructions we shall shortly introduce on categories.

This example suggests the introduction of the following notation for categories. We shall write A B to mean "A and B are objects, and φ is a morphism from A to B."

We now wish to give a few examples of how one carries over notions from categories in general to specific categories.

Let φ be a morphism from A to B. This φ is said to be a monomorphism if the following property is satisfied: given any object X and any two morphisms, α and α′, from X to A such that φ ∘ α = φ ∘ α′, it follows that α = α′ (figure 1). This φ is said to be an epimorphism if the following property is satisfied: given any object X and any two morphisms, β and β′, from B to X such that β ∘ φ = β′ ∘ φ, it follows that β = β′ (figure 2). (That is, monomorphisms are the things that can be canceled out of morphism equations on the left; epimorphisms can be canceled out of morphism equations on the right.)

Figure 1

Figure 2

As usual, one makes sense out of these definitions by appealing to our example, the category of sets.

THEOREM 1. In the category of sets, a morphism is a monomorphism if and only if it is one-to-one.

(Recall that a mapping from set A to set B is said to be one-to-one if no two distinct elements of A are mapped to the same element of B.)

Proof. Let φ be a mapping from set A to set B, which is one-to-one. We show that this φ is a monomorphism. Let X be any set, and let α and α′ be mappings from X to A such that φ ∘ α = φ ∘ α′. We must show that α = α′. If α and α′ were different, they would differ on some element of X; that is, there would be an x in X such that α(x) would be different from a′(x). Then, since φ is one-to-one, we would have φ(α(x)) different from φ(a′(x)). But this contradicts φ ∘ α = φ ∘ α′. Hence φ is a monomorphism.

Let φ be a mapping from set A to set B which is a monomorphism. We show that this φ is one-to-one. Let a and a′ be elements of A such that φ(a) = φ(a′). We must show that a = a′. Let X be the set having only one element, x. Let α be the mapping from X to A with α(x) = a, and let α′ be the mapping from X to A with α′(x) = a′. Then, since φ(a) = φ(a′), φ ∘ α(x) = φ ∘ α′(x). That is, φ ∘ α = φ ∘ α′. But φ is supposed to be a monomorphism; hence α = α′. In particular, we must have α(x) = α′(x); that is, we must have a = a′. Hence, φ is one-to-one. ◻

THEOREM 2. In the category of sets, a morphism is an epimorphism if and only if it is onto.

(Recall that a mapping from set A to set B is said to be onto if every element of B is the image, under the mapping, of some element of A.)

Proof. Let φ be a mapping from set A to set B, which is onto. We show that this φ is an epimorphism. Let X be any set, and let β and β′ be mappings from B to X such that β ∘ φ = β′ ∘ φ. We must show that β = β′. If β and β′ were different, they would differ on some element of B; that is, there would be a b in B such that β(b) would be different from β′(b). But, since φ is onto, there is an a in A such that φ = b. Hence β ∘ φ(a) would be different from β′ ∘ φ(a). This contradicts β ∘ φ = β′ ∘ φ. Hence φ is an epimorphism.

Let φ be a mapping from set A to set B, which is an epimorphism. We show that this φ is onto. Suppose, on the contrary, that there were some element b of B which was not the image, under φ, of any element of A. Let X be the set having just two elements, x and y. Let β be the mapping from B to X which sends b to x and the rest of B to y. Let β′ be the mapping from B to X which sends all of B to y. Then, since φ sends no element of A to b, we have β ∘ φ = β′ ∘ φ (for both of these mappings from A to X send all of A to y). Since φ was assumed an epimorphism, we must have β = β′. But, by construction, β does not equal β′. We have a contradiction. Hence φ is onto. ◻

It should be clear that there is no real content to these proofs: all one has to do to obtain a proof is keep from getting confused. One should think of a monomorphism as a fancy way of saying one-to-one and of an epimorphism as a fancy way of saying onto. Why does one bother to invent fancy words and fancy ways of saying these simple things? The point is that the definition of, for example, monomorphism is different in an important way from the definition of one-to-one. The latter refers directly to the sets themselves (i.e., to the elements of the sets and what happens to those elements under mappings). The former, however, refers only to objects, morphisms, and composition of morphisms. That is, monomorphism is a statement one can make about a category, while one-to-one is a statement one can only make about sets. As we shall see, there are lots of categories other than the category of sets: in every such category, the notion of a monomorphism will be available. Theorems 1 and 2 are good examples of the sort of activity that takes place in category theory. One takes a notion (e.g., one-to-one or onto) that refers directly to the detailed nature of the objects to which it is applied and finds a categorical version of that notion, a version that refers only to the things that go into making a category (objects, morphisms, composition). One thus acquires the ability to carry over this same notion to many different areas of mathematics.

Recall that any composition of one-to-one mappings on sets is itself one-to-one, and similarly for onto. This observation about sets and mappings suggests a theorem in category theory.

THEOREM 3. Let A B C. Then ψ ∘ φ is a monomorphism if both φ and ψ are monomorphisms; ψ ∘ φ is an epimorphism if both φ and ψ are epimorphisms.

Proof. Let φ and ψ be monomorphisms. We show that ψ ∘ φ is a monomorphism. Let X be any object, and α and a′ morphisms from X to A such that (ψ ∘ φ) ∘ α = (ψ ∘ φ) ∘ α′ (figure 3). We must show that α = α′. By condition 1 for a category, ψ ∘ (φ ∘ α) = ψ ∘ (φ ∘ α′). Since ψ is a monomorphism, we have φ ∘ α = φ ∘ α′. But now, since φ is a monomorphism, we have α = α′. Hence ψ ∘ φ is a monomorphism. Similarly for epimorphism. ◻

Figure 3

Note that theorem 3 is actually easier for categories in general than it is for the special case of sets. This phenomenon is by no means rare.

Everyone knows what a subset is. We formulate this notion categorically. A subobject of an object A is an object A′ along with a monomorphism A′ → A. Since, in the category of sets, monomorphisms are just one-to-one mappings, it is clear that, in the category of sets, subobjects are just subsets.

There is, in set theory, the notion of a one-to-one correspondence between sets. We formulate categorically. A morphism φ from A to B is said to be an isomorphism if there is a morphism φ′ from B to A such that φ′ ∘ φ = iA and φ ∘ φ′ = iB (figure 4). (In words, a morphism is an isomorphism if it has an inverse, which works on both the left and the right.) It is clear that, in the category of sets, an isomorphism is just a one-to-one, onto mapping. In fact, the statement of what an isomorphism is, when applied to the case of sets, is just what one really means by a correspondence between sets.

Figure 4

Example. A set is said to be countable if there exists an isomorphism between that set and the set of positive integers. Thus the set of real numbers is not countable.

We emphasize that these notions—monomorphism, epimorphism, isomorphism, subobject—while of some importance in themselves, are important primarily as examples of the point of view of category theory. We shall shortly give two additional, somewhat richer, examples. We first need a little terminology. By a diagram we mean any collection of objects along with a collection of morphisms between various of those objects. (E.g., a diagram is what is pictured.) A diagram is said to commute if it has the following property: given any two objects in the diagram, and any two morphisms between those objects, obtained by composition of the morphisms in the diagram, those two morphisms are equal. Thus the statement that the diagram of figure 5 commutes is the statement that α = α′, γ ∘ β = δ ∘ α (both morphisms from A to D), and ν ∘ δ = μ (both morphisms from B to E). It turns out that many statements in category theory can be formulated as the statement that an appropriate diagram commutes. For example, one could (with no gain in clarity) define a monomorphism as follows: A B is a monomorphism if, whenever the first diagram of figure 6 commutes, so does the second.

Figure 5

Figure 6

We now proceed to the final two examples of definitions in category theory. Let A and B be objects. A product of A and B is an object C, together with morphisms from C to A and from C to B, such that the following property holds: if C′ is any object, and α′ and β′ any morphisms from C′ to A and from C′ to B, respectively, there is a unique morphism γ from C′ to C such that the diagram of figure 7 commutes. A direct sum of A and B is an object C, together with morphisms from A to C and from B to C, such that the following property holds: if C′ is any object, and α′ and β′ any morphisms from A to C′ and from B to C′, respectively, there is a unique morphism γ from C to C′ such that the diagram of figure 8 commutes.

Figure 7

Figure 8

Note that a direct product consists not only of an object but also of a certain pair of morphisms, and similarly for a direct sum. We emphasize that there is no guarantee that, given two objects, either a direct product or a direct sum will exist: in fact, there are examples in which one or both do not exist. However, most of the categories one commonly deals with have the property that any two objects in the category do have both a direct product and a direct sum. Finally, we remark that, although we have here defined the product and sum of only two objects, the definition has an obvious extension to an arbitrary collection of objects (including infinite collections). The corresponding diagrams for the direct product and the direct sum are shown in figure 9. Nothing of consequence happens in the passage from the case of two objects to the case of an arbitrary collection of objects: proofs, for example, can be repeated almost word for word. We treat the case of only two objects because it makes arguments less confusing and because it is this case which is normally needed in practice.

Figure 9

We now find out what these general definitions mean in the specific context of the category of sets. Recall a construction from set theory. The Cartesian product, A × B, of sets A and B is the set of all pairs (a, b) with a an element of A and b an element of B. There is a natural mapping, α, from the Cartesian product A × B to A, given by α(a, b) = a. (That is, α is the mapping which ignores the second entry of an element, (a, b), of the Cartesian product.) Similarly, β(a, b) = b is a mapping from A × B to B.

THEOREM 4. In the category of sets, (A × B,α, β) is a direct product of sets A and B, where A × B is the Cartesian product, and α and β are the mappings above.

Proof. Let C′ be any set, and α′ and β′ any mappings from C′ to A and from C′ to B, respectively. We define a mapping γ from C′ to A × B as follows: for c′ in C′, set γ(c′) = (α′(c), β′(c)). (Note that the right side is indeed an element of A × B.) We must show that this γ makes the diagram of figure 10 commute (i.e., that α ∘ γ = α′ and β ∘ γ = β′), and that this γ is the only mapping from C′ to A × B which makes it commute. To see that α ∘ γ = α′, apply the mapping on the left to a typical element, c′, of C′: α(γ(c′)) = α(α′(c′), β′(c′)) = α′(c′). Hence α ∘ γ = α′ and, similarly, β ∘ γ = β′. Thus the diagram of figure 10 indeed commutes. All that remains is to show that this γ be another mapping from C′ to A × B which gives a commuting diagram. Fix an element c′ of C(c′) of A × B. Since α = α′, we must have α(c′)) = α′(c′); since β = β′, we must have β(c′)) = β′(c′). But the only element x of A × B such that α(x) = α′(c′) and β(x) = β′(c′) is the element (α′(c′), β(c(c′) = (α′(c′), β′(c= γ, establishing uniqueness of the mapping. ◻

To paraphrase this proof, "γ decides where (in A × B) it will send an element c′ of C′ as follows: it sees where (in A) α′ sends c′ and where (in B) β′ sends c′ and puts these two into a pair to get an element of A × B."

Figure 10

In order to obtain the analogous result for the direct sum, we must again recall a construction from set theory. The disjoint union of two sets, A1 and A2, is the set, written A1 ∪d A2, consisting of all pairs (x, n), where n is either the number 1 or the number 2, and where x is an element of An. (It is convenient here to call the sets A1 and A2 rather than A and B.) This definition requires a little explanation. The elements of A1 ∪d A2 of the form (x,1) form a "copy of A1," for the x in (x,1) is allowed to range over A1. Similarly, the elements of the form (x,2) form a copy of A2. But the elements of the form (x,1) together with the elements of the form (x,2) exhaust the elements of A1 ∪d A2. Thus A1 ∪d A2 is "the union of a copy of A1 with a copy of A2." Why this use of copies? Why not just take the union of A1 and A2? This would be fine if A1 and A2 were disjoint. If, however, A1 and A2 have some elements in common, then these elements would be included only once in A1 ∪ A2, whereas they are included twice (once as an element of A1 and once as an element of A2) in A1 ∪d A2. Denote by α1 the mapping from A1 to A1 ∪d A2 which sends a typical element, a1, of A1 to the element (a1,1) of A1 ∪d A2; similarly for α2.

THEOREM 5. In the category of sets, (A1 ∪d A2, α1, α2) is a direct sum of the sets A1 and A2, where A1 ∪d A2 is the disjoint union, and α1 and α2 are the mappings above.

Proof. Let C′ be any set, and α1′ and α2′ any mappings from A1 to C′ and from A2 to C′, respectively. We define a mapping γ from A1 ∪d A2 to C′ as follows: γ(x,1) = α1′(x), γ(x,2) = α2′(x). (Note that, in the first equation, x must be an element of A1 while, in the second, it must be an element of A2.) We must show that this γ is the only mapping which makes the diagram of figure 11 commute (i.e., that this γ, and only this one, satisfies γ ∘ α1 = α1′ and γ ∘ α2 = α2′). To see that γ ∘ α1 = α1′, apply the mapping on the left to a typical element, x, of A1: γ(α1(x)) = γ(x,1) = α1′(x). Hence γ ∘ α1 = α1′, and, similarly, γ ∘ α2 = α2′. Thus the diagram of figure 11 indeed commutes. To show that this γ ∘ α1 = α1′, we have, applying this to an arbitrary element x of A(α1(x)) = α1′(x). But α1(x) = (x(x,1) = α1′(x(x,2) = α2′(x= γ. ◻

Figure 11

Theorems 4 and 5 show, in particular, that direct products and direct sums always exist in the category of sets. (It is interesting to note that the definition of a direct sum, in particular, is perhaps more straightforward than that of the disjoint union.) One expects from these theorems that, even in other categories, direct products will be product-like, and direct sums union-like. This will turn out in some sense to be the case. However, for objects in more complicated categories, one cannot just go taking naive products and unions and expect to get, as a result, things having the structure appropriate for objects in that category. These categorical definitions will force us within each category to do the right thing and will ensure, in particular, that we always end up with an object in that category. Whereas there is nothing very subtle in monomorphisms and epimorphisms (they are always the obvious things), direct products and direct sums can, as we shall see, be very clever in combining two objects, and making all the structure work out right, to get new objects.

One of the things that makes direct products and direct sums interesting is that they are unique in a certain sense. This is not completely obvious even for sets. One could, offhand, imagine working very hard to find, for two sets A and B, a set C and mappings from C to A and from C to B such that the definition of a direct product is satisfied but such that this C is essentially different from A × B. In fact, direct products and direct sums are, in an appropriate sense, unique, and, in fact, this is true in any category. The appropriate sense is the following:

THEOREM 6. Let A and B be objects, and let (C,α,β) and (C′,α′,β′) be two direct products of these objects. Then there is one and only one isomorphism from C′ to C for which the diagram of figure 12 commutes. Similarly for the direct sum.

(That is, not only are C and C′ isomorphic as objects, but there is a unique isomorphism between them which preserves their "relationship with A and B.")

Figure 12

Proof. Since C is a product, there is a (unique) morphism γ such that the diagram of figure 12 commutes. Since C′ is a product, there is a (unique) morphism γ′ such that the diagram of figure 13 commutes. The first sentence implies α ∘ γ = α′; the second implies α′ ∘ γ′ = α. Hence α ∘ (γ ∘ γ′) = (α ∘ γ) ∘ γ′ = α′ ∘ γ′ = α. Similarly, β ∘ (γ ∘ γ′) = β. Now consider the diagram of figure 14. (Note the two (C’s.) The result just derived is precisely the statement that this diagram commutes if μ is replaced by γ ∘ γ′. Clearly, this diagram commutes if μ is replaced by iC. But C is a direct product, so there can be only one μ which makes this diagram commute. Hence γ ∘ γ′ = iC. Similarly, γ′ ∘ γ = iC′. Thus γ is an isomorphism. Similarly for the direct sum. ◻

Figure 13

Figure 14

This completes our brief survey of category theory. We emphasize again that theorems 3 and 6 are about category theory while theorems 1, 2, 4, and 5 are about the category of sets. There are, of course, many more definitions and theorems in category theory. We shall introduce a few of these, when needed, later.

Exercise 1. Prove that every isomorphism is both a monomorphism and an epimorphism. (The converse is false, for example, in the category of Hausdorff topological spaces.)

Exercise 2. Let A B C. Prove that, if ψ ∘ φ is a monomorphism, so is φ; that, if ψ ∘ φ is an epimorphism, so is ψ. Find examples, in the category of sets, to show that the converses are false.

Exercise 3. Let the objects consist of pairs, (A, A′), where A is a set and A′ is a subset of A. Let a morphism from (A, A′) to (B, B′) consist of a mapping φ from set A to set B such that, whenever a′ is in A′, φ(a′) is in B′. Let composition of morphisms be composition of mappings. Prove that this is a category. Discuss monomorphisms, epimorphisms, direct products, and direct sums.

Exercise 4. Let the objects be sets with exactly 17 elements, the morphisms mappings of such sets, and composition composition. Verify that this is a category. Prove that in this category no two objects have either a direct product or a direct sum.

Exercise 5. Prove that the φ′ in the definition of an isomorphism is unique.

Exercise 6. Fix a set A. For A′ and A″ subsets of A, write A′ ≤ A″ if A′ is a subset of A″. It is immediately evident that i) A′ ≤ A′, ii) A′ ≤ A″ ≤ A′′′ implies A′ ≤ A′′′, and iii) A′ ≤ A″ and A″ ≤ A′ imply A′ = A″. Define a similar ≤ for subobjects of a fixed object in an arbitrary category and prove that these three properties again hold.

Exercise 7. In the category of sets, the two morphisms (α and β) in a direct product are monomorphisms and the two morphisms in a direct sum are epimorphisms. Is this true in every category?

Exercise 8. In the category of sets, there is a natural isomorphism, given three sets A, B, and C, from the set (A × B) × C to the set A × (B × C), where × is a Cartesian product. This observation suggests a theorem in category theory. State it and prove it. Similarly for disjoint union.

Exercise 9. For purposes of this exercise, call an object A in a category atomic if it has the following property: given any object B, there is one and only one morphism, φB, from B to A. Prove that any two atomic objects are isomorphic. Prove that any morphism from an atomic object to another object is a monomorphism. Prove that, given any objects A and B, with A atomic, (B, φB, iB) is a direct product of A and B. What are the atomic objects in the category of sets? Do the arrows reversed version of all this.

Exercise 10. In the category of sets, A × (B ∪d C) is naturally isomorphic with (A × B) d (A × C). Is this a special case of some theorem in category theory?

Exercise 11. Fix two categories. Introduce a new category that can be thought of as the product of these. (Hint: choose, for the objects in this category, pairs consisting of one object from each of the given categories.)

3

The Category of Groups

A group consists of two things—i) a set G, and ii) a rule which assigns, given two elements (in a specific order) g and g′ of G, an element (normally written gg′ and called the product of g with g′) of G—subject to the following three conditions:

1. The product is associative. For any three elements, g, g′, and g″, of G,

2. An identity exists. There is an element of G (called the identity, and normally written e) with the following property: for any element g of G,

3. Inverses exist. Given any element g of G, there is an element of G (normally written g–1 and called the inverse of g) such that

It is immediate that the identity is unique [proof: if e′ were another, then ee′ = e′, since e is an identity, while ee′ = e, since e′ is an identity, whence e = e′] and that inverses are unique [proof: if h and h′ are both inverses of g, then h′ = h′ e = h′ (gh) = (h′ g)h = eh = h] and that the inverse of e is e [proof: ee = ee = e].

To give an example of a group, one must, of course, say what the set is and what the product rule is, and then one must verify that conditions 1, 2, and 3 above are satisfied.

Example. Denote by Z the collection of all integers (positive or negative, including zero). The rule associate with any two integers their sum assigns to any two integers another. This is a group—condition 1 is associativity of addition of integers; the integer zero will serve as an identity for condition 2; the integer –n will serve as an inverse for the integer n—called the additive group of integers.

Example. The collection of all real numbers, R, with the rule, again, being addition is a group, called the additive group of reals.

Example. Let S be any set. Denote by G the collection of all one-to-one, onto mappings from S to S. Associate with any two elements, μ and ν, of G their composition, ν ∘ μ, another element of G. This G, together with this product rule, is a group: condition 1 follows from associativity of composition of mappings; a suitable identity for condition 2 is the identity mapping from S to S; condition 3 follows from the fact that one-to-one, onto mappings have inverses. This group is called the permutation group on the set S.

Example. Let A be any object in any category, and denote by G the collection of all isomorphisms from A to A. Choose, for the product rule, composition of isomorphisms (noting that the composition of two isomorphisms is an isomorphism). The result is a group: conditions 1 and 2 for a group are immediate from conditions 1 and 2, respectively, for a category; condition 3 for a group is immediate from the definition of an isomorphism (in a category).

Most of the groups which occur naturally in physics seem to arise by an application of this last example in some special case. To describe a physical situation, one introduces some space, with an appropriate structure. One makes these spaces with this structure into a category, so the particular physical situation is described by some object A in that category. Then the corresponding group, as in the last example, becomes the group of "structure-preserving transformations on A, that is, the symmetry group." The last example above is, if you like, a reason why groups are important in physics.

A group G is said to be abelian if, for any two elements g and g′ of G, we have

(Thus the additive group of integers and the additive group of reals are abelian. The permutation group on a set S is abelian when and only when S has no more than two elements.) One almost always uses the following special notation for abelian groups: instead of gg′, one writes g + g′ (and calls it the sum); instead of e, one writes 0; instead of g–1, one writes –g; instead of gh–1, one writes g – h.

Let G and H be groups. A homomorphism from G to H is a mapping φ from the set G to the set H such that, for any two elements, g and g′, of G,

(That is, "given two elements of G, one gets the same element of H whether one i) first takes the product of these elements in G and sends the result, by φ to H or ii) first sends these elements to H, by φ, and there takes the product.") It is immediate that, for φ a homomorphism, φ(e) = e (the e on the left is in G; the e on the right in H) [proof: φ(e) = φ(e)φ(e)[φ(e)]–1 = φ(ee)[φ(e)]–1 = φ(e)[φ(e)]–1 = e] and φ(g–1) = [φ(g)]–1 for any g in G [proof: φ(g–1) = φ(g–1)φ(g)[φ(g)]–1 = φ(g–1g)[φ(g)]–1 = φ(e)[φ(g)]–1 = e[φ(g)]–1 = [φ(g)]–1]. (If either of these properties did not follow from the definition of a homomorphism, one would change the definition so that they would follow.) Thus a homomorphism from one group to another is a mapping which preserves, in the strongest sense, all the structure available. Note also that the composition of two homomorphisms is another: if G H and H K are homomorphisms of groups, then G K is also a homomorphism [proof: for g and g′ elements of G, ψ ∘ φ(gg′) ψ[φ(gg′)] = ψ[φ(g)φ(g′)] = ψ[φ(g)]ψ[φ(g′)] = ψ ∘ φ(g)ψ ∘ φ(g′)].

Let the objects be groups, the morphisms homomorphisms from groups to groups, and the composition composition of homomorphisms. We thus obtain a category—composition of morphisms (here, homomorphisms) is associative because composition of mappings is associative; the identity morphism from group G to itself is the identity mapping (obviously a homomorphism) from G to G. This is called the category of groups. Replacing group everywhere above by abelian group, we obtain the category of abelian groups. Note the way that additional structure (in this case a product structure) is incorporated in the passage from sets to groups, in the passage from mappings to homomorphisms, and in the passage from the category of sets to the category of groups.

Categorical definitions now become applicable, in particular, to groups. For monomorphisms, the explicit meaning is the same for groups as for sets.

THEOREM 7. In the category of groups, monomorphisms are one-to-one homomorphisms.

Proof. Let G H be a one-to-one homomorphism, and consider K G H with φ ∘ α = φ ∘ α′. If, for some k in K, we had α(k) ≠ α′(k), then, since φ is one-to-one, we would have φ ∘ α(k) ≠ φ ∘ α′(k), contradicting φ ∘ α = φ ∘ α′. Hence φ is a monomorphism. Suppose, conversely, that φ is a monomorphism. Let g and g′ be elements of G with φ(g) = φ(g′). Then φ(x) = e, where x = g–1g′. Let α be the homomorphism from Z, the additive group