Math

Prove non-necessity without logical impasse

Step-by-step mathematics solution: Prove non-necessity without logical impasse

As an Amazon Associate, I earn from qualifying purchases. For more practice problems like this, see Schaum’s Outline of Calculus, 7th Edition.

1. What the problem is asking (in plain language)

In a category a morphism (f\colon A\to B) can have two special properties

  • Monic (or monomorphism): (f) can be cancelled on the left; i.e. if (g_1,g_2\colon X\to A) satisfy (f\circ g_1 = f\circ g_2) then (g_1=g_2).
  • Epic (or epimorphism): (f) can be cancelled on the right; i.e. if (h_1,h_2\colon B\to Y) satisfy (h_1\circ f = h_2\circ f) then (h_1=h_2).

In many familiar categories (sets, groups, vector spaces, …) a morphism that is both monic and epic is automatically an isomorphism (it has a two‑sided inverse).

The exercise asks us to show that this is not true in general: give a concrete category together with a morphism that is both monic and epic but fails to be an isomorphism.

The second part of the question is more philosophical: instead of proving “(P) does not imply (Q)” by a reductio ad absurdum (assume (P\Rightarrow Q) and derive a contradiction), we should give a direct counterexample. The solution will illustrate how a single example settles the “non‑necessity’’ claim.

2. Step‑by‑step construction of a counterexample

Step 1 – Choose a simple category where intuition fails

The easiest place to find a morphism that is monic and epic but not invertible is the category (\mathbf{Set}_{!*}) of pointed sets and base‑point‑preserving functions.

  • An object is a pair ((X,x_0)) where (X) is a set and (x_0\in X) is a distinguished element (the base point).
  • A morphism (\;f\colon (X,x_0)\to (Y,y_0)) is a function (f\colon X\to Y) such that (f(x_0)=y_0).

Why this category?

  • It contains many morphisms that are not bijections, yet the usual “cancellation” properties can still hold.
  • It is small enough to write down an explicit map.

Step 2 – Define the objects and the morphism

Take the following two pointed sets:

[ A=({0,1},0),\qquad B=({0},0). ]

Thus

  • (A) has two elements, with (0) the base point.
  • (B) has a single element, which is necessarily the base point.

Define a morphism

[ f\colon A\longrightarrow B,\qquad f(0)=0,\;f(1)=0 . ]

Because both source and target base points are sent to the target base point, (f) is indeed a morphism in (\mathbf{Set}_{!*}).

Step 3 – Show that (f) is monic

Take any two morphisms (g_1,g_2\colon (X,x_0)\to A) such that

[ f\circ g_1 = f\circ g_2 . ]

Both composites are maps (X\to B) that must send every element of (X) to the unique element (0) of (B); thus they are identical functions.

But the only way two base‑point‑preserving maps into (A) can give the same composite is that they agree on every element of (X). Indeed, for any (x\in X),

[ f\bigl(g_1(x)\bigr)=0 = f\bigl(g_2(x)\bigr), ]

and the only pre‑image of (0) under (f) is the whole set ({0,1}). However, the base‑point condition forces (g_i(x_0)=0).
If there existed an (x) with (g_1(x)=1) while (g_2(x)=0), the composites would still be equal (both give (0)). Hence we need a more careful argument:

Because the target (B) has only one element, the equality (f\circ g_1 = f\circ g_2) tells us nothing about the values of (g_1) and (g_2).
In fact, any two morphisms into (A) satisfy the equality of composites, because the composite lands in the one‑point set.

Consequently, the cancellation condition for monomorphisms is vacuously true: there is no pair (g_1\neq g_2) that could violate it, because the premise “(f\circ g_1 = f\circ g_2)” is always satisfied, but the conclusion “(g_1=g_2)” is not required in the definition of a monomorphism.

Wait—this is a subtle point. In the definition of monomorphism we require:

For all objects (X) and all morphisms (g_1,g_2\colon X\to A),
(f\circ g_1 = f\circ g_2 \implies g_1 = g_2).

Because the implication’s antecedent is always true, the only way the implication could fail is if we could find (g_1\neq g_2) with the same composite. But in (\mathbf{Set}_{!}) we *can find such a pair: take (X=A) and let (g_1=\operatorname{id}_A), (g_2) be the constant map sending both elements to the base point (0). Both are base‑point‑preserving, and

[ f\circ g_1 = f = f\circ g_2 . ]

Thus the implication fails; consequently (f) is not monic.

Our earlier attempt was wrong! We need a different category where the monic condition really holds.

Step 4 – Switch to a better category: (\mathbf{Ring}) (commutative rings with unity)

In the category (\mathbf{Ring}) a morphism is a unital ring homomorphism.

Consider the inclusion

[ i\colon \mathbb{Z} \longrightarrow \mathbb{Q}, ]

the canonical embedding of the integers into the rational numbers.

  • It is monic: ring homomorphisms in (\mathbf{Ring}) are always monomorphisms iff they are injective as functions, and (i) is injective.
  • It is epic: surprisingly, (i) is an epimorphism in (\mathbf{Ring}) even though it is not surjective.
    • Proof of epicness: let (\varphi,\psi\colon \mathbb{Q}\to R) be ring homomorphisms (with (R) any ring) such that (\varphi\circ i = \psi\circ i).
      This means (\varphi(n)=\psi(n)) for every integer (n).
      Any rational number can be written as (n/m) with (n,m\in\mathbb{Z}), (m\neq 0). Because homomorphisms preserve addition and multiplication and send (1) to (1), the value of (\varphi) on (n/m) is forced to be (\varphi(n)\,\varphi(m)^{-1}) (the inverse of (\varphi(m)) exists because (\varphi(m)) must be a unit in any ring that receives a homomorphism from a field). The same holds for (\psi). Since (\varphi) and (\psi) agree on all integers, they agree on all fractions, hence (\varphi=\psi).
      Therefore (i) is epic.
  • Yet (i) is not an isomorphism: it is not surjective, so no inverse homomorphism (\mathbb{Q}\to\mathbb{Z}) exists.

Thus we have a morphism that is both monic and epic but not an isomorphism.

Step 5 – Verify each property formally

Property Verification
Monic In (\mathbf{Ring}), a morphism (f) is monic ⇔ the underlying function of sets is injective. The map (\mathbb{Z}\to\mathbb{Q}) sends distinct integers to distinct rationals, so it is injective ⇒ monic.
Epic Let (\alpha,\beta:\mathbb{Q}\to R) be ring homomorphisms with (\alpha\circ i = \beta\circ i). For any (q\in\mathbb{Q}) write (q=n/m). Then (\alpha(q)=\alpha(n)\,\alpha(m)^{-1} = \beta(n)\,\beta(m)^{-1} = \beta(q)). Hence (\alpha=\beta). Thus (i) is epic.
Not an isomorphism An isomorphism in (\mathbf{Ring}) must be a bijective ring homomorphism. The inclusion (\mathbb{Z}\hookrightarrow\mathbb{Q}) is not surjective, therefore it cannot have a two‑sided inverse.

Step 6 – Concluding the logical point

Because we have explicitly produced an object (the morphism (i:\mathbb{Z}\to\mathbb{Q})) that satisfies the antecedent “monic and epic’’ while failing the consequent “isomorphism”, we have directly demonstrated that the implication

[ \text{(monic (\wedge) epic)} \;\Longrightarrow\; \text{isomorphism} ]

is not universally valid in the category (\mathbf{Ring}). No proof by contradiction is needed; a single counterexample suffices.

3. Final answer

A morphism can be both monic and epic without being an isomorphism.

Counterexample: In the category (\mathbf{Ring}) of (commutative) rings with unity, the canonical inclusion

[ i\colon \mathbb{Z}\hookrightarrow \mathbb{Q} ]

is

  • monic (injective as a function),
  • epic (any two ring homomorphisms out of (\mathbb{Q}) that agree on the integers must agree everywhere),

but it is not an isomorphism (it is not surjective, so no inverse homomorphism exists).

Therefore the statement “every morphism that is both monic and epic is an isomorphism’’ is false in general. The existence of this single example settles the “non‑necessity’’ claim without any need for a proof‑by‑contradiction.

4. Common Mistakes

Mistake Why it is wrong How to avoid it
Using a set‑theoretic intuition (“monic = injective, epic = surjective”) and concluding that monic + epic ⇒ bijective. In many concrete categories (Set, Group, Vect) this equivalence holds, but it fails in categories where epimorphisms are not surjective (e.g., (\mathbf{Ring}), (\mathbf{Top})). Remember that “epic’’ is a categorical cancellation property, not necessarily surjectivity.
Choosing a category where the morphism is not monic (as happened with the pointed‑set example). If the morphism fails one of the two required properties, the example does not address the claim. Verify both monic and epic conditions before declaring a counterexample.
Thinking that a counterexample must be “exotic’’ (very large or abstract). Simple, well‑known algebraic categories often contain the needed example. Look first at standard algebraic categories (rings, modules, topological spaces) before resorting to more sophisticated constructions.
Relying on a proof by contradiction to show “not every monic‑epic is iso”. While valid, it does not provide the concrete information the problem asks for. Provide an explicit morphism that satisfies the hypothesis but not the conclusion; that directly exhibits the failure of the implication.
Confusing “epic in a subcategory’’ with “epic in the ambient category”. A morphism may be epic in a subcategory (e.g., monoids) but not in the larger one. State clearly which category you are working in and verify the epic property in that category.

By checking each property carefully and presenting the explicit inclusion (\mathbb{Z}\to\mathbb{Q}) as above, we avoid these pitfalls and give a clean, rigorous answer.

Original question: Prove non-necessity without logical impasse on Mathematics Stack Exchange, licensed CC BY-SA.

More Math Problems