Motivation
The Chinese remainder theorem says that residues modulo comaximal ideals can be chosen independently. Without comaximality, some information is shared. Two prescribed residues can still come from a single ring element, but they must agree on that shared information. The useful question is therefore more precise than whether the usual theorem applies: which pairs can actually be reconstructed?
For polynomials in two variables, prescribing a residue modulo specifies a polynomial on the vertical axis; prescribing one modulo specifies a polynomial on the horizontal axis. Both prescriptions include a value at the origin. That value must agree. We will prove that it is the only obstruction for these two axes, while three intersecting lines can impose a further obstruction that checking each pair misses.
Use the elementary definitions of ideals, ideal sums, ring quotients and kernels from the preceding ring notes. Throughout, is a commutative ring with identity , and are ideals. Ideals need not be proper. A quotient by the whole ring is the zero ring, which is permitted as a quotient here and is never called a field or a domain.
Definitions
Definition 3.2.1
(Congruence and shared information)
Write when . Since , there is a natural unital homomorphism
There is likewise a map from . A pair is compatible when , equivalently when .
These maps are well defined: changing by an element of changes it by an element of . Notice the direction. Enlarging an ideal forgets more information, so a map goes from the quotient by the smaller ideal to the quotient by the larger one. No inclusion between and is needed.
Let be the set of compatible pairs inside . It is a subring under coordinatewise operations. For example, if two pairs have the same respective images in the common quotient, their sum has image in both coordinates, and their product has image in both. The identity pair belongs to , as do additive inverses. Thus compatibility is preserved by both operations, rather than being only a test on sets.
Theorems
Theorem 3.2.2
(Exact reconstruction for two arbitrary ideals)
The map
has kernel and image exactly . Consequently it induces a ring isomorphism
If realizes a compatible pair, its complete set of representatives in is .
Theorem 3.2.3
(Independence is equivalent to comaximality)
Every pair of residues is compatible, or equivalently is surjective onto the whole product, exactly when . Under that condition , recovering the usual Chinese remainder theorem. Without the condition, the isomorphism with still uses ; it cannot in general be replaced by .
Proof sketch or proof idea
Proof
(Kernel, image and the comaximal case)
We prove both statements completely. The map preserves the ring operations coordinatewise. An element maps to zero precisely when it belongs to both ideals, proving the kernel assertion. Every realized pair is compatible because the same element has the same image in the common quotient.
Conversely, suppose . By definition of ideal sum, choose and with . Set
The first expression gives ; the second gives . This proves the image assertion constructively. The first isomorphism theorem gives the displayed isomorphism. If is another representative, lies in both ideals. Conversely adding any element of their intersection changes neither residue. That proves the final assertion and also specifies exactly what uniqueness means.
If , the common quotient is the zero ring, so every pair has equal images there. Conversely, if is onto, realize the particular pair by some . Then and , so and . To identify the intersection in this case, write with . For ,
The reverse containment always holds because an ideal absorbs arbitrary ring multiples and sums. This completes the proof.
Proof X-ray
Compatibility supplies an element, not just a necessary test
Subtracting the two required congruences immediately shows that must lie in . Sufficiency needs an additional step: split that difference as and use . The signs matter. The construction then separates existence from ambiguity: the difference split constructs one element, while the kernel calculation describes every possible change to it.
Counting the obstruction when the quotients are finite
There is a useful additive map
It is well defined and additive, is onto since , and has kernel . Each fiber is a translate of : two pairs have the same image exactly when their difference is in the kernel. If and are finite, this counts the compatible pairs and gives
The common quotient measures how many equally sized compatibility classes the whole product has. This counting argument uses additive groups. In general is not a ring homomorphism: with and , the pairs and have product zero, but their images under have product .
Reading inverses through the two coordinates
The isomorphism also tells us how to do ring arithmetic after reconstruction. In particular, a compatible pair is a unit in if and only if each coordinate is a unit in its quotient ring. Here the zero ring's sole element is its identity and is a unit, since its product with itself equals that identity. Necessity follows by projecting an inverse to either coordinate. For sufficiency, suppose the two coordinates have inverses. Their images in are inverses of the same element, because both quotient maps preserve multiplication and identity. Inverses are unique: if in a commutative ring, multiplying by gives . Hence the two inverse coordinates have equal images and form a compatible pair themselves. Their coordinatewise product with the original pair is the identity of .
It follows that is invertible precisely when both and are invertible. This does not require the ideals to be comaximal. It gives a practical method: find the two inverse residues, check their compatibility, and reconstruct them by the same difference-splitting argument. Existence of an inverse in the larger ring is a different question; quotienting may create inverses.
For example, the integer reconstruction below gives modulo . Its two coordinates are modulo and modulo . Both have inverse in their respective quotients, since and leave remainder one. These inverse residues are compatible and reconstruct the class of modulo . Indeed . The integer itself is not a unit of ; the statement concerns its class in the quotient.
Worked examples
Worked example 3.2.4
(Two integer residues with shared information)
Require and . The sum ideal is , and , so the pair is compatible. A suitable split is , giving , or modulo .
Here , because common multiples of and are exactly multiples of their least common multiple. Thus every solution is for . The product ideal is instead , so reduction modulo the product would retain unnecessary distinctions between solutions. There are compatible pairs among the pairs in the full product. A pair such as is impossible because its entries disagree modulo six.
Concept lensAlgorithmic
Bézout produces the difference split
For integer moduli with , the extended Euclidean algorithm returns with . If , put ; then splits the difference into a multiple of and a multiple of , and . In the example above, and , so the split is and , exactly as found by hand. The algorithm either finds a split or shows, through , that none exists.
Worked example 3.2.5
(Gluing polynomials on two axes)
Let for a field , and . A residue modulo is a polynomial , and a residue modulo is a polynomial . The common quotient is , so compatibility is exactly . If their common constant is , then
realizes both residues by substitution at and .
The ambiguity is . For a direct proof, write an element of the intersection as . Substitution at gives in the domain , so . Grouping the powers of shows that divides . Hence every common multiple lies in ; the opposite inclusion is immediate. All reconstructions are therefore . The term vanishes on both axes and records exactly the information the prescriptions cannot see.
Worked example 3.2.6
(Agreement can include coefficients, not just values)
In , take and . Here and . Compatibility requires that the residue modulo , when reduced modulo , equal the prescribed residue modulo .
Thus modulo and modulo are compatible, with all reconstructions
Replacing the first prescription by makes them incompatible, even though both still have value one at zero. Their linear coefficients disagree. When ideals are nested, the finer residue already determines the coarser one; the two entries are especially far from independent.
Why checking three pairs can fail
Now prescribe modulo , modulo , and modulo a third ideal . First solve the first two conditions, obtaining . Every possible first-stage solution is with . It can also satisfy the third condition exactly when
Indeed, writing gives ; conversely such a solution gives the same membership after subtraction. Choosing a different first-stage solution changes by an element of , so it does not change this test. If the test succeeds, the final ambiguity is .
Counterexample mode
Pairwise compatibility does not suffice for arbitrary ring ideals
In take , and . Prescribe residues modulo , modulo , and modulo . Each pair is compatible: every pairwise sum ideal is , and each difference has zero constant term.
However, the first two conditions force . Modulo , substitute , obtaining . It cannot equal in , since no polynomial multiplied by equals . Thus there is no simultaneous solution. The missing test is , which fails by the same substitution. For three lines, agreement at their common point does not capture every restriction on their polynomial prescriptions.
Common mistakes
Common mistake
(Compatibility is tested in )
Compatibility concerns equality in , not equality of chosen representatives in . Changing representatives does not affect it. Uniqueness concerns a class modulo , not a unique element of . Replacing the intersection by the product requires a justification; comaximality supplies one, while nested ideals often disprove the replacement.
Common mistake
(Two-ideal tests do not extend pairwise)
For several ideals, do not carry over a two-ideal test without checking it. After combining two conditions, their new kernel is their intersection. The next common quotient uses that intersection plus the next ideal. Checking only the original pairwise sums can lose this extra information. Finally, a map used for additive counting need not preserve multiplication, as demonstrates.
Summary
Two ring residues combine exactly when they agree modulo the sum ideal. Their reconstructions form one coset of the intersection ideal. Comaximality makes the shared quotient zero and the choices independent. Successive reconstruction preserves these principles, but introduces intersections that pairwise checks alone may miss.
Exercises
Quick check
(Determine all integers that are two modulo eight and six modulo twelve. Count the compatible residue pairs for these two moduli and distinguish the intersection ideal from the product ideal.)
Solution · Answer
The difference lies in , so solutions exist. Writing , the second condition becomes , hence . Thus . The intersection is , while the product is . The counting formula gives compatible pairs.
Quick check
(In , prescribe modulo and modulo . Find every reconstruction. Explain what changes if the second constant term becomes four.)
Solution · Answer
The constants agree, so all reconstructions are . These have both required restrictions, and every difference between two reconstructions lies in . Changing three to four in the second prescription produces unequal images in , so no reconstruction exists.
Quick check
(For , and , prove that using an explicit map. Reconstruct the pair modulo and modulo .)
Solution · Answer
Send to . This is a well-defined ring homomorphism. It is injective because its second coordinate is the input, and every compatible pair has its first coordinate forced by the second, proving surjectivity onto . The specified pair has all reconstructions .
Quick check
(For the three line ideals in the counterexample, replace the last residue by . Find one solution and prove that all solutions are .)
Solution · Answer
Take : it is zero modulo and , and becomes after setting . The ambiguity is the intersection of all three ideals. An element in the first two is ; membership in then gives . Since is a domain, . Division of as a polynomial in by the monic polynomial shows that is divisible by , equivalently by . Thus the intersection is , with the reverse containment immediate. Adding exactly these elements gives all solutions, over every field .