The Gold Standard of Universal Algebra | Choosing Quotient Representatives That Preserve Operations

Let q:A→B be a surjective homomorphism. We can think of B as the quotient determined by q's kernel congruence, or as an isomorphic target representing those classes. A section is a map s:B→A satisfying q(s(b))=b for every b in B.

This equation says s selects an element in the fibre over each b. It does not say s(q(a))=a for every a in A. When several elements share a fibre, s can select only one of them. Returning the correct quotient value and recovering the original element are different requirements.

For a simple set example, let q identify red and orange as warm, and blue and green as cool. A section may choose red for warm and blue for cool. Starting at orange, applying q and then s returns red, not orange. Starting at warm, applying s and then q returns warm as required.

The two round trips have different domains. The equation q∘s=id_B is the section condition. The other composite s∘q is a map A→A that replaces each element by its selected representative. It is generally not the identity on A.

Now restore algebraic operations. To call s a homomorphic section, we need both the section equation and operation preservation. Merely belonging to the correct fibre proves the first condition only. Our central question is whether representatives can be chosen so that all their operation equations hold simultaneously.

Contents · Next chapter

For a binary operation m, the section must satisfy s(m_B(b,c))=m_A(s(b),s(c)). The operation on the left is first performed in the quotient; the one on the right is performed on the selected representatives. Both routes must return the same actual element of A, not just equivalent elements.

The quotient construction already guarantees that these two results have the same q-value whenever q∘s=id. Applying q to the right-hand side gives m_B(b,c), as does applying q to the left. But equal quotient values are weaker than equal representatives upstairs.

This is the source of many plausible mistakes. Someone checks that representative sums land in the correct classes and concludes that the representative map is a homomorphism. That check confirms compatibility of q with the operations; it does not prove exact equality needed for s.

Constants also matter. A homomorphic section must send each designated constant of B to the corresponding designated constant of A. Choosing a different element in the constant's fibre may satisfy q∘s=id while failing the nullary-operation requirement immediately.

For additive groups, list the obligations as preservation of zero, addition and negation. Some follow from others once the group structure is used, but stating the full signature prevents accidental switches to a setting with fewer requirements. Later, the same selected subset will behave differently if multiplication and one are added.

Contents · Previous chapter · Next chapter

Let q:C12→C4 send a residue modulo 12 to its residue modulo 4. This is well-defined because 4 divides 12, and it preserves the additive-group operations. Its fibres are {0,4,8}, {1,5,9}, {2,6,10} and {3,7,11}.

The familiar representative choice sends 0,1,2,3 in C4 to the identically named residues in C12. It satisfies q∘s=id_C4 because every chosen value lies in the required fibre. It also preserves zero. At first glance the small labels make the choice look especially natural.

But in C4, 1+3=0. Applying the proposed section gives s(0)=0. Adding the selected representatives in C12 gives 1+3=4. These are different elements of C12, although q sends both to zero. Therefore this representative choice is not a homomorphism.

The failure concerns exact operation preservation upstairs. It does not mean q was an invalid quotient or that representative arithmetic gives the wrong quotient class. Reducing the sum again does recover the correct class. The stronger claim that the representatives themselves form a faithful algebraic copy has failed.

To repair the situation, we must either change the representative choice or prove that no choice can work. Trying more examples with the same failed choice will not settle the second question. The next chapter uses the cyclic structure to search systematically rather than guessing another four-element list.

Contents · Previous chapter · Next chapter

A homomorphism from the additive group C4 is determined by the image k of its generator 1. The image of residue b must be b·k modulo 12. Because four copies of 1 equal zero in C4, we need 4k=0 modulo 12. This is the condition that makes the formula well-defined.

The section condition also requires q(k)=1, meaning k≡1 modulo 4. The candidates in C12 are 1,5,9. Multiplying them by 4 modulo 12 gives 4,8,0 respectively. Only k=9 satisfies the required order condition.

Thus define s(b)=9b modulo 12. Its values on 0,1,2,3 are 0,9,6,3. These belong to the four correct fibres because 9b≡b modulo 4. The formula preserves addition, and the relation 4·9=36≡0 modulo 12 ensures that changing a representative of b by 4 leaves the result unchanged.

Check the earlier failure again. In C4, 1+3=0. The new representatives add to 9+3=12≡0 in C12, exactly s(0). Another check is 1+1=2: the representatives give 9+9=18≡6, which is s(2).

The successful list looks less familiar than {0,1,2,3}, but its operation behaviour is correct. Algebraic usefulness comes from the preservation equations, not from choosing the smallest numerical labels. The calculation also proves uniqueness of the homomorphic section for this particular quotient, because all possible generator images were checked.

Contents · Previous chapter · Next chapter

The section's image is S={0,3,6,9} inside C12. It is closed under addition and negation and contains zero. The restriction q|S sends these four elements bijectively onto C4, with inverse s. Thus the quotient has an actual subgroup copy inside the original group.

The order of labels in that copy matters for the map. The generator 1 of C4 corresponds to 9, not to 3 in this chosen section, because q(9)=1 while q(3)=3. A subset being abstractly isomorphic to C4 is not by itself enough; the isomorphism must fit the specified q.

Each fibre contains exactly one element of S. Existence follows from the section condition. For uniqueness, suppose s(b) and s(c) lie in the same fibre. Applying q gives b=c, so the selected elements are equal. This is an algebraic transversal: one representative per class forming a subalgebra.

Conversely, suppose a subalgebra S of A meets each q-fibre exactly once. The restriction q|S is a bijective homomorphism onto B. Its inverse is a homomorphism in the ordinary total-operation setting, because preservation can be checked using the unique preimages of the operation's inputs.

Therefore finding a homomorphic section and finding such a subalgebra transversal are equivalent tasks. One is phrased as a map; the other as a closed subset. Moving between them is useful: a formula can suggest the subset, while a closure failure can quickly rule out a proposed representative system.

Contents · Previous chapter · Next chapter

Consider reduction q:C4→C2. A section must send the generator 1 of C2 to an odd residue of C4, so its possible images are 1 or 3. A group homomorphism must also respect 1+1=0 in C2, requiring twice that chosen image to be zero in C4.

Both choices fail. Twice 1 is 2 modulo 4, and twice 3 is 6≡2 modulo 4. Neither is zero. Hence no homomorphic section exists. This is an exhaustive obstruction based on the generator relation, not simply a failure of the most obvious choice.

As a set map, a section certainly exists: choose 0 for the even class and 1 for the odd class. It returns the correct class under q. But the chosen odd representative cannot have the required order two. The group operations make the representative-selection problem stricter than ordinary set selection.

The same issue can be seen through subgroups. The only two-element subgroup of C4 is {0,2}, and both of its elements map to zero in C2. It does not meet the odd fibre. There is no subgroup transversal, agreeing with the generator calculation.

This example defeats the claim that every surjective homomorphism can be reversed by another homomorphism. Surjectivity gives preimages. It does not guarantee that preimages can be chosen consistently with every operation. The distinction remains even for a quotient with only two classes and a source with only four elements.

Contents · Previous chapter · Next chapter

Let N=dm with positive integers d,m, and let q:C_N→C_d be reduction modulo d. A homomorphic section is determined by k=s(1). Its section condition is k≡1 modulo d. Its well-definedness condition is dk≡0 modulo N, which is equivalent to m dividing k.

Write k=mt. We then need mt≡1 modulo d. Such a t exists exactly when gcd(m,d)=1. If the greatest common divisor exceeds one, every mt shares that divisor modulo d and cannot be congruent to one. If the greatest common divisor is one, Bezout's identity supplies a solution.

The section is then s(b)=kb modulo N. This formula respects addition and the cyclic relation, and q(s(b))=b. The solution t is unique modulo d, so k is unique modulo N. Therefore this particular cyclic reduction has a unique homomorphic section whenever it has one.

For C12→C4, m=3 and d=4 are coprime. The inverse of 3 modulo 4 is 3, giving k=9. For C4→C2, m=d=2 are not coprime, and the criterion predicts the obstruction already calculated directly.

The criterion concerns this specific family of additive cyclic quotient maps. It is not a general condition for sections in all algebras, nor does it say all groups of a given size behave the same way. Its value is that every hypothesis and arithmetic step can be inspected and transferred to a new modulus.

Contents · Previous chapter · Next chapter

Given a homomorphic section s of q, define p=s∘q:A→A. Both factors preserve operations, so p is an endomorphism: a homomorphism from the algebra to itself. Applying it twice gives p²=s∘q∘s∘q=s∘id_B∘q=p. Such an endomorphism is called idempotent.

Idempotence means one replacement step is enough. Once an element has been moved to its selected representative, applying the same replacement again leaves it there. It does not mean p is the identity on every original element. Elements outside the selected copy can move on the first application.

For C12→C4 with s(b)=9b, the resulting p is p(a)=9a modulo 12. Its outputs lie in {0,3,6,9}. For example p(1)=9, p(5)=9 and p(9)=9: these elements share the same quotient class and are replaced by its selected representative.

Check the algebraic idempotence numerically: 9²=81≡9 modulo 12, so applying the multiplier twice equals applying it once. The formula also shows addition preservation. Unlike a general representative selector, this stable replacement commutes with the group's operations.

This construction supplies a useful certificate. Instead of storing a separate quotient and representative list, one can specify an idempotent endomorphism and identify its fixed image. But idempotence alone is not enough: a stable arbitrary function may still fail operation preservation, as a later counterexample will show.

Contents · Previous chapter · Next chapter

Now start with an idempotent endomorphism p:A→A. Its image S=p(A) is a subalgebra: applying an operation to image elements equals applying p to the corresponding operation on preimages. Constants are preserved. This image-closure statement holds for any endomorphism.

Idempotence adds that every image element is fixed. If y=p(x), then p(y)=p²(x)=p(x)=y. Conversely, any fixed element y=p(y) belongs to the image. Thus the image equals the fixed-point set for this particular map.

Regard p as a map r:A→S and let j:S→A be inclusion. Their composite r∘j is the identity on S, because p fixes S. The smaller algebra S is called a retract of A, with r a retraction and j its section. The other composite j∘r is the original p.

The kernel congruence of p identifies exactly those elements receiving the same representative. The first isomorphism theorem gives A/ker(p) isomorphic to S. This explains how an idempotent endomorphism simultaneously describes a quotient and an embedded copy of that quotient's image.

Do not reverse the final statement too broadly. Every homomorphism has an image and a kernel quotient, but its image need not be fixed by the map when domain and codomain coincide. The equation p²=p is what turns the image into a retract with the displayed inclusion and makes one-step replacement stable.

Contents · Previous chapter · Next chapter

Every endomorphism of C12 has the form p_k(a)=ka modulo 12, determined by the image k of the generator. It is idempotent precisely when k²≡k modulo 12. Testing the twelve residues gives k=0,1,4,9. These four maps have different images and kernels.

For k=0, every input goes to zero; the image is the trivial subgroup and the kernel is all of C12. For k=1, the map is the identity; its image is all of C12 and its kernel is {0}. These are the two extreme stable replacements.

For k=4, the image is {0,4,8} and the kernel is {0,3,6,9}. For k=9, the image is {0,3,6,9} and the kernel is {0,4,8}. Direct multiplication verifies these lists. Both proper images are fixed pointwise by their corresponding multipliers.

One can reduce the idempotence test modulo 3 and modulo 4: k must be either 0 or 1 in each of these rings. The four combinations produce the four listed residues modulo 12. The finite enumeration is already a complete proof here; this second viewpoint explains the pattern behind the list.

These maps are endomorphisms of the additive group. A proper one is not a unital ring endomorphism of Z/12Z, because it sends the designated multiplicative identity 1 to 0,4 or 9 instead of 1. Changing the signature changes which replacement maps are permitted, even when the same multiplication-by-k formula is written down.

The corresponding ring quotient makes the boundary especially sharp. Reduction from Z/12Z to Z/4Z is a unital ring homomorphism, but it has no unital ring section. Such a section would have to send the target's 1 to the source's 1. Since four copies of 1 add to zero in the target, addition preservation would require four copies of the source's 1 to equal zero. They instead give the nonzero residue 4 modulo 12. The additive section sending 1 to 9 therefore cannot repair the ring problem. Its failure is already forced by the constant and addition, before checking multiplication. This example shows why adding an operation or constant can remove a section even though the quotient function and its fibres remain exactly the same.

Contents · Previous chapter · Next chapter

In the successful C12 example, q∘s is the identity on C4, while s∘q=p9 is not the identity on C12. A section provides a one-sided inverse. It embeds a copy of the smaller target without recovering every original element of the larger source.

The quotient's kernel subgroup is K={0,4,8}, and the selected copy is S={0,3,6,9}. Every a in C12 can be written as (a−p9(a))+p9(a). The first part lies in K because q(a−p9(a))=0, and the second lies in S.

The decomposition is unique. If k+s=k'+s', then k−k'=s'−s lies in K∩S={0}. Therefore k=k' and s=s'. In this abelian-group example, the two components combine by ordinary addition and give a direct-product description of C12.

For a concrete value, a=5 has p9(5)=45≡9, so the kernel component is 5−9≡8. The decomposition is 5=8+9 modulo 12. The quotient alone records the S component through its class; the kernel component carries information that q discarded.

This direct-product conclusion uses the group being abelian. For general groups, a split quotient can involve a nontrivial conjugation action and a semidirect product. For arbitrary universal algebras, no group-style subtraction or product decomposition is automatically available. The section and idempotent-endomorphism statements are broader than this particular decomposition formula.

Contents · Previous chapter · Next chapter

Let V=F3² with coordinatewise vector operations, and let q:V→F3 project onto the first coordinate. A linear section must send 1 to a vector whose first coordinate is 1. The possibilities are (1,0),(1,1),(1,2). Each choice extends linearly.

For c in F3, define s_c(x)=(x,cx). Then q(s_c(x))=x, and linearity follows from coordinatewise addition and scalar multiplication. These are three different homomorphic sections of the same q. Thus uniqueness in the cyclic criterion was a feature of that family, not a general property of sections.

Their images are three different lines through the origin: the horizontal line and the two nonvertical sloping lines. Each meets every fibre of q once. The vertical line is the kernel and cannot be the image of a section, because q sends all of it to zero.

The corresponding idempotent endomorphisms are p_c(x,y)=(x,cx). They ignore the original second coordinate and replace it by the selected multiple of the first. Applying the map twice changes nothing because the first coordinate remains x. Their shared kernel is the vertical line.

Take input (2,1). The three stable representatives are (2,0),(2,2),(2,1) for c=0,1,2 respectively. Each gives the same quotient value 2, but they are different source elements. A quotient and its kernel do not necessarily specify a unique embedded copy or a unique operation-preserving representative choice.

Contents · Previous chapter · Next chapter

Return to C4→C2 and choose the ordinary set representatives 0 and 1. The replacement function p:C4→C4 sends 0 and 2 to 0, and 1 and 3 to 1. It is idempotent: its two possible outputs are fixed by another application.

Yet it does not preserve addition. We have p(1+1)=p(2)=0, while p(1)+p(1)=1+1=2 in C4. Therefore p is not an endomorphism of the additive group, despite satisfying p²=p as a function.

Its image {0,1} is not a subgroup, since 1+1 leaves that image. This gives another way to see the failure. The general proof that the image of an endomorphism is a subalgebra cannot be applied until operation preservation has been established.

The distinction is useful when a mathematical model describes normalisation or representative selection. Running a procedure twice and seeing no further change demonstrates stability on the tested inputs. It does not show that normalising before an operation equals normalising after it, or that the output set supports the intended algebra.

For a finite proposed normaliser, test both properties independently: p(p(a))=p(a) for every input, and p(m(a,b))=m(p(a),p(b)) for every operation tuple, plus constants. In an infinite algebra, supply general proofs. Stable output is one obligation; compatibility with the operations is another.

Contents · Previous chapter · Next chapter

Suppose p and r are idempotent endomorphisms of the same algebra and they commute: p∘r=r∘p. Their composite is then idempotent, because (pr)²=prpr=pprr=pr. It remains an endomorphism by composition. Commutativity is the step that allows the middle factors to be exchanged.

Without that condition, composition can fail idempotence. In F3², let p(x,y)=(x,0) and r(x,y)=(y,y). Both maps are linear and idempotent. The first projects onto the horizontal line; the second fixes the diagonal line and replaces the first coordinate by the second.

Their composite p∘r sends (x,y) to (y,0). Applying it again sends (y,0) to (0,0). At input (0,1), one application gives (1,0) and two applications give (0,0). Hence the composite is not idempotent, even though both individual maps are.

The opposite composite r∘p is the zero map, so the maps do not commute. This exact calculation identifies the failed hypothesis. It prevents an appealing but false rule that any succession of stable algebraic simplifications must itself be a stable simplification.

When combining retractions, state the intended composite order and test it. If a commuting argument is unavailable, compute the composite's square directly or prove another condition that ensures idempotence. A property of each stage separately does not automatically survive their composition in the form desired.

Contents · Previous chapter · Next chapter

Begin by checking q itself: domain, codomain, surjectivity and preservation of the declared operations. List its fibres or a formula for them. A candidate section must choose exactly one element in each fibre, so test q(s(b))=b before considering any stronger claim.

Next verify the operations for s. In a finite additive example, check zero and all sums, with negation following from the group reasoning or checked explicitly. In a cyclic example, reduce the work to the generator image and its defining relation, but explain why that reduction covers all elements.

If the candidate fails, distinguish failure of that choice from impossibility of every choice. The usual representatives fail for C12→C4, yet a different section exists. For C4→C2, exhausting the two possible generator images proves that no homomorphic section exists.

For a successful section, compute its image and show that q restricts to an isomorphism there. Then form p=sq, check idempotence and identify its fixed points and kernel. These mutually supporting calculations help catch an incorrect representative formula or a mistaken composite order.

Record the signature beside every result. The C12 multipliers are valid additive-group endomorphisms, while most fail preservation of one as ring maps. A correct proof in one algebraic language is not a certificate in a richer language until the extra operations have been checked.

For a small general finite quotient, an exhaustive search can choose one candidate from each fibre and reject a choice as soon as an operation equation fails. If a choice survives every tuple and constant, it certifies a section. If every possible choice is rejected, retain the failed equation for each branch or a shorter common obstruction. The method is finite but can be expensive: the number of raw choices is the product of the fibre sizes. Existence of a search procedure does not promise that a large quotient will be easy to split or rule out.

Contents · Previous chapter · Next chapter

Consider reduction q:C20→C4. Does it admit a homomorphic section? If so, find its generator image, list the selected representatives and calculate the stable representative of 7. Then compare with reduction C16→C4. Try the arithmetic conditions before reading the answers.

For C20→C4, N=20,d=4,m=5. Since gcd(5,4)=1, a section exists. We need k divisible by 5 and congruent to 1 modulo 4. The choice k=5 works, so s(b)=5b modulo 20 selects 0,5,10,15.

The composite p(a)=5a modulo 20 is idempotent because 25≡5 modulo 20. It sends 7 to 35≡15, which lies in the same quotient class since both are 3 modulo 4. Applying p again gives 75≡15, confirming stability at this input.

For C16→C4, m=d=4 are not coprime. A section's generator image would have to be divisible by 4 and congruent to 1 modulo 4, which is impossible. There are set sections, but none preserves the additive-group structure.

Finally change C20→C4 to C20→C5. Now m=4,d=5, and the inverse of 4 modulo 5 is 4. The generator image is k=16, giving a different selected subgroup. This last change tests whether the quotient size, rather than only the source size, remains attached to the criterion throughout your calculation.

Contents · Previous chapter · Next chapter

Does a section undo a quotient? It undoes it on the quotient side: q∘s is the identity there. The reverse composite usually replaces an original element by a selected member of its class. It cannot recover distinctions that q has discarded without additional information.

Why do ordinary representatives fail? They need only belong to the correct classes. Their exact sums upstairs may differ from the chosen representative of the quotient sum. The values agree after applying q, but a homomorphism requires equality before that final reduction.

Does every quotient have some clever homomorphic section? No. The C4→C2 order obstruction rules out every possible choice. Conversely, failure of one convenient choice does not prove impossibility, as C12→C4 demonstrates with its less obvious representative list.

Is every idempotent function an algebraic retraction? Only if it also preserves the operations. The parity representative function on C4 is stable but fails addition. An idempotent endomorphism, with both properties, fixes its subalgebra image and does give a retraction onto it.

Are sections unique? Not generally. The three linear sections of F3²→F3 share the same quotient map and kernel but select different lines. Does a split quotient always yield a direct product? The displayed decomposition is justified for the abelian-group example; broader algebraic settings require their own structural theorem and hypotheses.

Contents · Previous chapter · Next chapter

First draw the two typed round trips q∘s and s∘q, or write them as equations with their domains. Explain why only one must be the identity. Use a pair of elements from the same fibre to show exactly which original distinction is lost after replacement.

Next reproduce the failed usual representatives and successful multiplier for C12→C4. The comparison should name an operation equation, not merely describe one choice as better. Then prove the impossibility for C4→C2 by checking all possible images of the quotient generator.

Derive the general cyclic condition from k≡1 modulo d and dk≡0 modulo N. This makes the greatest-common-divisor criterion a consequence of two understandable requirements. Apply it to the changed-modulus exercise and verify one representative sum directly as an independent arithmetic check.

Finally move between a section, its image subalgebra and its idempotent endomorphism. Use the linear example to remember that several choices can work, and the composition counterexample to remember that stable maps need not compose stably without further conditions. These examples test different claims rather than repeating one formula.

The lasting skill is to distinguish selecting a label from selecting an algebraic copy. A homomorphic section must return the correct class and preserve every operation exactly. When it exists, it reveals a retract inside the original algebra; when it fails, the failed relation explains why representative choice cannot repair the quotient in that stronger sense.

Contents · Previous chapter · Continue to related reading