Categorical Morphisms

Patrick Reany

21 September 2025

Note: This chat I had with Copilot is likely edited for relevance and brevity. I usually omit Copilot's emphasis (italics or bolding) to save a lot of time on my part, post chat. You'll forgive me if I, on rare occasion, highlight a word or phrase that Copilot wrote, for didactic emphasis. (Naturally, I get to correct my spelling, grammar, and other minor errors for free.)

Patrick:
Just a wading into category theory. I'm presently watching a YouTube video in which the instructor is teaching beginning category theory. He gives the example of morphisms in the category of groups. He claims it's homomorphisms. Okay, then he moves on. What?! We've talked about similar things before, but I want the reason WHY it's homomorphisms. Is it because

1. Group homomorphisms are a simple instantiation of morphism,
2. Group homorphisms give mathematicians what they want in practical terms, or
3. Only group homomorphisms will work as a morphism?

[It's as if the usual way that category theory is taught is gaslighhting the new student to the subject. On the one hand we are told that morphisms downplay the internals of the objects within a category, but on the other hand, we will find that the crucial exception is that the morphism must retain knowlegde of what makes the objects in a particularly category distinguishable from the objects of every other category. I know that what I just said is not particularly clear, but I think that the tension between them will be made clear by the end of the chat.]

Copilot:

You're asking exactly the kind of question that separates passive watching from active understanding---bravo. Let's tackle your three possibilities head-on, because each touches on a different layer of why group homomorphisms are the morphisms in the category Grp.

1. Are group homomorphisms just a simple instantiation of morphism?

Yes---but not just that. Group homomorphisms are a natural instantiation of morphisms in the context of groups.

In category theory, morphisms are meant to preserve the structure that defines the objects.

This makes homomorphisms the canonical choice. It's not arbitrary---it's baked into what it means to be a group.

2. Do group homomorphisms give mathematicians what they want in practical terms?

Absolutely. Homomorphisms allow mathematicians to:

• Compare groups in meaningful ways
• Transfer structure from one group to another
• Study kernels and images, which lead to concepts like normal subgroups and quotient groups

In short, homomorphisms are the workhorses of group theory. They preserve enough structure to be useful, but are flexible enough to allow rich behavior. So yes, they're chosen because they're useful.

3. Are group homomorphisms the only thing that will work as morphisms?

Yes---if you're talking about the category Grp. That category is defined so that:
• Objects are groups
• Morphisms are group homomorphisms

If you tried to use arbitrary functions between groups, they wouldn't preserve the group structure, and you'd break the rules of the category. So in Grp, only homomorphisms qualify.

But here's the twist: you could define a different category with groups as objects and, say, injective homomorphisms or even group isomorphisms as morphisms. That would be a different category---like Grp_inj or Grp_iso. Category theory lets you play with these definitions, but each category has to be internally consistent.

So Why Homomorphisms?

Because:

• They preserve the structure that defines groups (essential for morphisms)
• They're practically useful for studying and comparing groups
• They're the only functions that satisfy the axioms required for the category Grp

You nailed it by asking all three questions. Want to explore how this idea generalizes to other categories---like rings, topological spaces, or vector spaces?

Patrick:
Okay, so what is the proof that the group homomorphism is the most general mapping between groups in the category of Group that will act as the morphism (or arrow) between groups? (I suppose the proof would have to show that any other choice would have to satisfy all the properties of homomorphism. And this must follow because categorical morphisms must preserve all the algebraic structure inherent in groups. But this claim suddenly feels like circular reasoning, because the "structure that is inherent in groups" is whatever is defined by group homomorphims, if you see my point.)

Copilot:
The non-circular reason morphisms in Grp are homomorphisms

What you're sensing is real: "morphisms preserve structure" can feel circular unless we say what the structure is, independently of morphisms. Here's the clean way to break the circle.

What "structure" a group has, independently of morphisms

A group is a model of an equational theory with:

• Operations: a binary operation \(\cdot\), a unary inverse \((\cdot)^{-1}\), and a nullary constant \(e\).
• Equations: associativity, identities, and inverses.

This is specified before we talk about maps between groups. In universal algebra, a morphism of models must be a function that preserves the given operations and hence all the equations built from them.

• Definition (homomorphism): A function \(f:G\to H\) is a group homomorphism iff \[ f(x\cdot y)=f(x)\cdot f(y),\quad f(e_G)=e_H,\quad f(x^{-1})=f(x)^{-1}. \] Any one of these beyond the first is redundant: from \(f(xy)=f(x)f(y)\) you get preservation of \(e\) and inverses automatically.

This is not defined by category theory; it's defined by the algebraic signature and axioms of groups. Category theory then forms the category whose objects are models and whose arrows are exactly these structure-preserving functions.

"Most general" in three precise senses

There are three rigorous ways to understand "most general," and each singles out homomorphisms.

1) Preservation of all terms

• Claim: The weakest reasonable requirement is that for every group term \(t(x_1,\dots,x_n)\) built from \(\cdot,(\cdot)^{-1},e\), a morphism must satisfy \[ f\big(t_G(a_1,\dots,a_n)\big)=t_H\big(f(a_1),\dots,f(a_n)\big). \]
• Consequence: Requiring this for just the primitive operations already forces the above equation for all terms. That is exactly the definition of a homomorphism. Any weaker requirement fails to preserve some equation of the theory; any stronger requirement (e.g., injectivity) is extra structure, giving a proper subcategory.

2) Closure plus free-forgetful adjunction

• Setup: Consider the forgetful functor \(U:\mathbf{Grp}\to\mathbf{Set}\).
• Fact: There is a left adjoint \(F:\mathbf{Set}\to\mathbf{Grp}\) (the free group functor) with a natural bijection \[ \mathrm{Hom}_{\mathbf{Grp}}(F(X),G)\cong \mathrm{Hom}_{\mathbf{Set}}(X,UG). \]
• Force: For this adjunction to exist, morphisms in \(\mathbf{Grp}\) must be exactly the maps determined by where generators go, extended multiplicatively---that is, homomorphisms. If you enlarge the class of arrows to include any non-homomorphism, the universal property of the free group breaks; if you shrink it (e.g., only isomorphisms), the left adjoint ceases to be left adjoint to the same forgetful functor.

This characterizes homomorphisms as the largest class of functions between groups that:

• Contain identities

• Are closed under composition

• Make \(U\) admit the canonical left adjoint \(F\)

3) Morphisms of the equational theory

• Viewpoint: Groups are algebras for a finitary signature with equations. The general notion of "morphism of algebras" in this setting is: preserve the operations, hence all equations. That produces homomorphisms---no more, no less. This is the standard, theory-independent recipe for forming "the category of models."

Why alternatives fail or change the category

• Arbitrary functions: Don't preserve \(\cdot\); kernels, images, quotients, and the isomorphism theorems collapse. Composition closure is fine, but you're no longer relating group structure.
• Weaker preservation (e.g., only \(e\) or only inverses): Not closed under composition in a way that interacts properly with group laws, and you can't even state core results (kernel/image factorization) coherently.
• Stronger arrows (injective/surjective/isomorphisms): Perfectly valid, but they define different categories (wide subcategories of \(\mathbf{Grp}\)) that are too restrictive to support the usual constructions (e.g., cokernels with surjective-only arrows).

Not circular: structure first, then morphisms

• Step 1: Fix the signature \((\cdot,(\cdot)^{-1},e)\) and equations; that defines what a group is.
• Step 2: By the general principle for algebraic theories, morphisms are the functions that preserve those operations (equivalently, all terms). That yields homomorphisms.
• Step 3: Form the category with those objects and arrows. Its good behavior (free groups, products, quotients, isomorphism theorems) validates the choice.

Two crisp propositions to keep in mind

• Proposition A: A function \(f:G\to H\) is a homomorphism iff for every word \(w(x_1,\dots,x_n)\) in the free group, \[ f\big(w_G(a_1,\dots,a_n)\big)=w_H\big(f(a_1),\dots,f(a_n)\big). \] Preserving the primitive operations is equivalent to preserving all words.
• Proposition B: The class of arrows in a category of groups that makes the forgetful functor admit the free left adjoint and contains all identities and is closed under composition is exactly the class of group homomorphisms.

If you want, we can unpack either proposition in detail with proofs and examples.

Patrick:
So, there does appear to be some hidden desiderata in the "definition of a categorical morphism," which is that it should always allow for the maximal amount of algebraic, topological, metrical, etc, information allowable in an arbitrary object of the category to be mirrored in the image object --- a sort of technical framing constraint.

Copilot:
Yes---that's a very sharp observation. What you're noticing is that the choice of morphisms in a category isn't arbitrary; it's guided by a meta-principle that mathematicians rarely state outright but always rely on. Let's make that principle explicit.

The Hidden Desideratum of Morphisms

When we define a category, we want its morphisms to:

1. Preserve the defining structure of the objects.

• In Grp, that means the group law.
• In Top, that means the topology.
• In Vect, that means linear structure.

2. Be as general as possible while still preserving that structure.

• If you demand too much (e.g., only isomorphisms), you lose richness: the category becomes too rigid.
• If you demand too little (e.g., arbitrary functions between groups), you lose the ability to talk about structure at all.

3. Support the universal constructions mathematicians care about.

• Free objects, products, quotients, limits, colimits, adjunctions---all of these only work if morphisms are chosen to preserve exactly the right amount of structure.

This is what you're calling a "technical framing constraint": morphisms must preserve all and only the structure that defines the objects, no more and no less.

Why "maximal amount of information" is the right intuition

Think of it this way:

• A morphism should mirror everything that is invariant under the theory.
• For groups, the invariants are exactly the group equations.
• For topological spaces, the invariants are open-set relationships.
• For metric spaces, you could choose either distance-preserving maps (isometries), or the weaker class of non-expansive maps, or even continuous maps if you only care about topology. Each choice gives a different category, but in each case the morphisms are the maximal class of maps that still respect the chosen structure.

So the "maximality" isn't about preserving all possible information---it's about preserving exactly the information encoded in the definition of the object. That's why it doesn't collapse into circularity: the structure is defined first (operations, axioms, topology, metric), and then morphisms are chosen to be the most general maps that preserve that structure.

A Guiding Principle

A nice way to phrase the hidden desideratum is:

> Morphisms in a category are the most general maps that preserve the structure singled out in the definition of the objects, so that the category supports the natural constructions and theorems of that branch of mathematics.

Would you like me to illustrate this with a side-by-side comparison of how this principle plays out in different categories (Sets, Groups, Rings, Topological spaces, Metric spaces)? That way you can see the "framing constraint" in action across contexts.

Patrick:
Okay, so, given the previous discussion and your Guiding Principle, it's no wonder that the notion of a categorical morphism is all of: 1) what mathematicians want (desiderata), 2) the natural/canonical choice , and 3) the only possible choice (at least in the category of group, the homomorphism). So, we're back at the beginning, except that now I think I have an explanation of it. Bravo!

Copilot:
Exactly! You've closed the loop beautifully. What felt at first like a suspicious circle---"morphisms preserve structure, and structure is what morphisms preserve"---turns out to be a well‑founded triangle:

• Desiderata (what mathematicians want): morphisms should let us compare objects in a way that respects the theory we're studying.
• Natural/Canonical choice: given the operations and axioms that define the objects, there's a unique "most general" way to preserve them.
• Only possible choice (within that category): once the structure is fixed, the morphisms are forced---any weaker choice breaks the theory, any stronger choice defines a different category.

That's why in Grp, homomorphisms aren't just convenient---they're inevitable. And the same pattern repeats across mathematics: continuous maps in Top, linear maps in Vect, ring homomorphisms in Ring, etc. Each time, the morphisms are exactly those arrows that preserve the defining structure, no more and no less.

So yes---you've now got the explanation that ties it all together. That "aha" moment is the real beginning of seeing why category theory is so powerful: it doesn't just unify mathematics, it explains why the choices mathematicians have been making all along were the only ones that could have worked.

Would you like me to sketch how this principle extends to non‑algebraic categories (like Topology or Metric Spaces), where the structure isn't just equations but something subtler? That's often where the idea of "morphisms as the most general structure‑preserving maps" really shines.

Patrick:
Yes, provide by illustration this side-by-side comparison of how this principle plays out in different categories (Sets, Groups, Rings, Topological spaces, Metric spaces). That way we can see the "framing constraints" in action across contexts.

Copilot:
Perfect---let's make the "framing constraint" visible by laying out several familiar categories side by side. You'll see how the objects are defined by some structure, and the morphisms are then the most general maps that preserve exactly that structure.

Key Takeaways

• Set is the baseline: morphisms are just functions.
• Grp, Ring, Vect, etc.: morphisms are the most general maps that preserve the algebraic operations.
• Top: morphisms are the most general maps that preserve the open-set structure.
• Met: here you see flexibility---different communities choose different morphisms depending on what "structure" they care about (exact distances vs. inequalities).

This illustrates your point: morphisms are simultaneously

1. What mathematicians want (they preserve the intended structure),
2. The natural/canonical choice (they follow from the definition of the structure), and
3. The only possible choice (if you want a full, rich category that supports the constructions of that theory).

Would you like me to extend this table to vector spaces, posets, and manifolds as well? Those examples show even more clearly how the "framing constraint" guides the choice of morphisms.

Patrick:
Yes, provide similar examples for vector spaces, posets, and manifolds, and can we include the Hilbert space category and the symmetric moniodal category as well?

Copilot:
Excellent---let's expand the table to include vector spaces, posets, manifolds, Hilbert spaces, and symmetric monoidal categories. This will show how the "framing constraint" principle applies across algebra, order theory, geometry, analysis, and higher‑level categorical structures.

Observations

• Vect and Hilb: morphisms are linear maps, but Hilbert spaces add the analytic constraint of boundedness.
• Posets: the category structure is literally the order relation---morphisms are determined by the objects.
• Manifolds: smooth maps are the "just right" choice---continuous maps are too weak, diffeomorphisms too strong.
• Symmetric monoidal categories: here the morphisms are not just functors, but functors that preserve the extra tensorial structure. This shows how category theory scales up: once you enrich the structure, morphisms must preserve it too.

The Meta‑Principle in Action

Across all these examples, the framing constraint is visible:

• Morphisms are chosen to preserve exactly the structure that defines the objects.
• They are the most general maps that do so.
• Choosing weaker or stronger morphisms gives you a different category, often too trivial or too rigid to be useful.

Would you like me to push this one step further and show how this principle works in functor categories (where morphisms are natural transformations) and higher categories (where morphisms themselves have morphisms)? That's where the idea of "structure‑preserving maps" really blossoms into the higher‑dimensional world.

[I'll leave this step to the interested reader.]

Afterword:

Why does this chat feel like an episode of Untold Stories of the ER --- where the hidden realities finally get revealed?