1. The question
An algebra has many possible quotients: one can identify elements in ways compatible with its operations. These identifications form a lattice, an ordered structure recording which quotients refine which others.
Can every finite lattice arise as the complete lattice of compatible identifications of some finite algebra? The operations are entirely unrestricted. We may choose their number, their arities and their tables to suit the desired lattice. The only requirement is that the underlying set remain finite.
Family 206 contains two manuscripts dated 24 September 2026. Both claim a negative answer. M2 constructs an obstruction through subgroup lattices of finite groups. M1 goes further, claiming that no algorithm can decide finite representability from a finite lattice table. It also gives finite certificates for positive instances and a mathematically specified, but not numerically evaluated, smallest negative example.
2. Background and essential definitions
A lattice is an order with meets and joins
A lattice is a partially ordered set in which every pair has a greatest lower bound , called the meet, and a least upper bound , called the join. The subsets of a finite set, ordered by inclusion, form a familiar example: meet is intersection and join is union.
A finite nonempty lattice has a least element and a greatest element . It can be specified by a finite order table, so it is a legitimate input to an algorithm.
Congruences record compatible identifications
A finite algebra here is a finite nonempty set equipped with finitely many total operations, each taking finitely many arguments. It need not be a group or a ring. Unary operations, constants and even an empty list of operations are allowed.
A congruence is an equivalence relation preserved by every operation. For a binary operation , compatibility means
This makes the induced operation on equivalence classes well defined. The congruences form a lattice , ordered by inclusion. Its least element identifies only equal elements; its greatest identifies everything. Meets are intersections, while joins are the smallest congruences containing the given relations.
The word full is essential. Finding the desired lattice among some of the congruences is insufficient. A representation requires an isomorphism
with no extra congruences left over.
Group intervals and algorithms
For a subgroup of a finite group , the full subgroup interval is
ordered by inclusion. It contains every intermediate subgroup, not a selected collection.
A decision algorithm must halt on every input and answer correctly. A semidecision procedure for representability need only halt when a representation exists; it may search forever on a negative instance. The distinction is central here. Every individual proposed finite algebra can be checked, but there may be no computable limit on how large a witness must be sought.
3. Development history and previous partial results
Grätzer and Schmidt proved in 1963 that every algebraic lattice is a congruence lattice of an algebra. Finite lattices belong to that class. The representing algebra in their theorem, however, may be infinite. The finite lattice representation problem asks whether a finite carrier can always be used.
Pudlák developed a graph-based proof of the general representation theorem. Pudlák and Tůma later proved that every finite lattice embeds into the partition lattice of a finite set. A partition lattice is the lattice of all equivalence relations on that set. This is a strong finite embedding result, but it does not ensure that exactly the chosen sublattice is preserved by some operations.
Pálfy and Pudlák established a crucial bridge in 1980: the assertion that every finite lattice has a finite congruence representation is equivalent to the assertion that every finite lattice is an interval in a finite group’s subgroup lattice.
This is an equivalence of universal statements. It does not say that each individual lattice has one representation if and only if it has the other. That logical distinction determines what can be concluded from a particular subgroup obstruction.
Positive results cover all finite lattices of width at most two, where width is the largest size of a set of pairwise incomparable elements. DeMeo’s analysis left only one possible exception among lattices with at most seven elements. M1 supplies a representation of that remaining small case, so its claimed minimum obstruction has more than seven elements.
On the group side, Baddeley, Lucchini, Börner and Aschbacher developed reductions using finite simple groups and special interval configurations. Repnitskiĭ and Tůma showed that an infinite, countable locally finite group can represent every finite lattice as an interval. Again, allowing an infinite ambient object changes the problem.
The new papers claim obstructions uniform over every possible finite carrier or finite group, regardless of its size. M1 additionally connects that uniformity problem with Diophantine undecidability.
4. What the new manuscripts claim
M2: a negative existence theorem
M2 asserts that some finite nonempty lattice is not the full congruence lattice of any finite algebra of any finite signature. Its direct construction produces a self-dual finite lattice that is not a full subgroup interval in any finite group. Self-dual means that reversing the order gives an isomorphic lattice.
The Pálfy–Pudlák equivalence then gives the negative congruence-lattice conclusion globally. M2 explicitly does not identify its particular decorated subgroup obstruction with the congruence obstruction point by point.
The self-dual subgroup obstruction also cannot be the entire lattice of intermediate fields of a finite separable field extension. Passing to a finite normal closure would, by Galois correspondence, turn that field lattice into the reverse of a finite subgroup interval; self-duality would remove the reversal.
M1: characterization and undecidability
M1 claims a finite colored-graph characterization of representability. Each graph witness is checkable by a finite procedure, and enumerating all possible witnesses eventually finds a representation whenever one exists.
At the same time, it asserts that no deterministic Turing machine always halts and decides whether a given finite lattice is representable. It separately asserts undecidability of recognition as a full subgroup interval of a finite group.
These statements are compatible. Positive certificates can be searchable without there being a stopping rule for unsuccessful searches.
M1 also specifies a nonrepresentable lattice of minimum cardinality, with . The finite description uses a particular Busy Beaver integer, a maximum associated with halting computations of bounded program size. It does not evaluate that integer, the minimum cardinality or the entries of the lattice’s order table. “Specified” therefore should not be read as “a small diagram explicitly printed and computed.”
5. Why the advance matters
Importance: 5/5, provisional. A negative solution would settle a foundational finite representation problem in universal algebra. The undecidability claim would go further, explaining why no universal recognition procedure can classify all finite instances.
The unrestricted signature makes the negative result especially strong. Failure cannot be repaired by adding a clever operation, increasing its arity, or changing from groups to some other finite algebraic system. The obstruction concerns every finite algebra on every finite carrier.
The decision-theoretic conclusion also gives a structural consequence: there can be no computable bound, depending only on the size of a representable lattice, on the size of the graph witness in M1’s characterization. If such a bound existed, a finite exhaustive search would decide the problem.
The significance is theoretical, not a claim that ordinary small examples are impossible to handle. Indeed, the paper’s small-lattice result points in the opposite direction: substantial finite ranges can have uniformly positive answers while the general recognition problem remains undecidable.
6. The main idea
Replace operations by a colored graph
M1 assigns a lattice element to each pair of vertices in a finite complete graph. Think of the color as the least identification needed to merge and . Diagonal colors are zero, and the labels satisfy a lattice-valued triangle inequality:
For each lattice element , the pairs with form an equivalence relation. Additional separation conditions ensure that different lattice elements give different relations.
A self-map of the vertices is admissible if it only decreases labels. Give the vertex set all such maps as unary operations. A path condition then ensures that every congruence of this algebra is one of the intended relations. That last condition prevents extra congruences from appearing and turns an embedding into a full representation.
For a fixed number of vertices, all colorings, maps and paths can be checked finitely. The unbounded witness size is where algorithmic difficulty remains.
Force incompatible finite-group configurations
Both manuscripts use carefully decorated finite lattices to impose constraints on any hypothetical subgroup representation. Special intervals test normal subgroups and force nonabelian simple-group components to occur along selected chains.
The challenge is simultaneous compatibility. Recognizing the abstract isomorphism type of a simple group at each step does not control how the actual groups embed into one another. The constructions retain those embeddings and impose tests relative to earlier groups along the chain.
M2 derives an absolute upper bound on the length of certain tested chains with different endpoint labels. It uses finite simple-group classification, restrictions of representations, field parameters and growth estimates. The bound is independent of the size of a hypothetical ambient group.
It then builds a finite lattice whose inserted chains exceed that bound while preserving selected Boolean meets and joins. In a supposed representation, the surviving subgroup data describe homomorphisms extending a common action. The Boolean faces force these extensions to agree locally, while another part of the lattice forbids a common extension. The final gluing argument forces exactly that forbidden extension.
M1 develops a separate obstruction using affine charts and protected chains. Its undecidability argument needs further quantitative constructions beyond the existence of one negative lattice.
Turn a hypothetical decider into bounded arithmetic
M1 encodes positive integer solutions of arithmetic circuits in weighted subgroup templates. Sizes of permutation sets represent numbers; equal totals express addition, and square-size constructions express multiplication.
Two bounds work together. The number of lattice elements is bounded computably by the circuit’s syntax, independently of the sizes of its numerical solution. Conversely, any finite representation gives a fresh numerical reading whose values are computably bounded in the size of that representation.
If a representability decider existed, one could list all representable lattice tables up to the relevant size and search until a witness had been found for each. This would give a computable witness bound, and hence a computable bound on a replacement arithmetic solution whenever any solution existed.
Searching the resulting finite box would decide the Diophantine problem, contradicting its known undecidability. The argument does not need to construct an unknown-solution template directly from the circuit. It uses the bounded collection of possible lattice tables and the hypothetical decider to obtain the replacement bound.
7. Scope, limitations, and open questions
The algebras and lattices are nonempty and finite. Operations are total and finitary. Allowing infinitely many finitary operations on a finite carrier would not add representations: for each unwanted equivalence relation, one could retain a single operation witnessing its incompatibility, leaving a finite list.
The negative conclusion concerns full congruence lattices and full subgroup intervals. It does not contradict the positive embedding theorems or the general representation theorem with infinite carriers.
Undecidability is a statement about one algorithm handling every finite input. It does not say that each individual lattice is unclassifiable, nor that useful restricted classes lack algorithms.
The minimum obstruction’s numerical size remains unevaluated in M1. Producing a manageable explicit order table, determining sharp small-size thresholds, and finding tractable structural subclasses are distinct questions from the claimed existence and undecidability theorems.
8. Sources and further reading
Grätzer and Schmidt, Characterizations of congruence lattices of abstract algebras, Acta Scientiarum Mathematicarum 24 (1963), 34–59, gives the general representation theorem. Pálfy and Pudlák, Congruence lattices of finite algebras and intervals in subgroup lattices of finite groups, Algebra Universalis 11 (1980), 22–27, supplies the global equivalence.
Pudlák and Tůma, Every finite lattice can be embedded in a finite partition lattice, Algebra Universalis 10 (1980), 74–95, is the finite embedding theorem. DeMeo’s Congruence Lattices of Finite Algebras, doctoral thesis (2012), provides small-lattice results and the unary-closure viewpoint.
For the graphical predecessor, see Kenney, Graphical algebras—a new approach to congruence lattices, Algebra Universalis 64, 313–338. Matiyasevich’s Diophantine representation of enumerable predicates, Mathematics of the USSR-Izvestiya 5 (1971), 1–28, is the arithmetic undecidability input.
Source manuscripts
- M1. OpenAI, Finite congruence lattices: characterization and undecidability (September 24, 2026).
- M2. OpenAI, A negative solution to the finite lattice representation problem (September 24, 2026).
Sources consulted on 21 October 2026 at repository revision 1581d4c. Download the manuscript citations.