Two algebraic generators can produce an infinite-dimensional division ring

← All digests

Family 201 · Algebra
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-21

1. The question

For ordinary fields, adjoining finitely many algebraic elements produces a finite-dimensional extension. If $x$ and $y$ each satisfy polynomial equations, their powers can be reduced, and commutativity lets every expression be rearranged into a bounded list of monomials.

What happens when multiplication is not commutative, but division by every nonzero element remains possible? Kurosh’s division-ring problem asks whether algebraicity over the center still forces every finitely generated division subring to be finite-dimensional over that center.

Family 201 claims a negative answer. It constructs a countable characteristic-zero division ring generated by just two elements as an algebra over its center. Every element satisfies a polynomial over that center, but the whole ring has infinite dimension over it.

2. Background and essential definitions

Division rings and their centers

A division ring is a ring in which every nonzero element has a multiplicative inverse. Unlike a field, its multiplication need not commute. The familiar quaternions are a finite-dimensional example over the real numbers.

The center $F=Z(D)$ consists of elements commuting with every element of $D$. It is an ordinary commutative field. The division ring $D$ is algebraic over $F$ if every $a\in D$ satisfies some nonzero polynomial

$$c_0+c_1a+\cdots+c_na^n=0, \qquad c_i\in F.$$

The polynomial and its degree may depend on $a$. Algebraicity does not include a common degree bound.

For a subset $S\subseteq D$, write $F[S]$ for the algebra generated by $F$ and $S$: finite sums of finite products. Write $F(S)$ for the division subring generated by them, allowing inverses as well. Algebra generation is the stronger finiteness requirement.

The ring is locally finite over its center if

$$[F(S):F]<\infty \quad\text{for every finite subset }S\subseteq D.$$

Here brackets denote vector-space dimension. The word “locally” refers to finite sets of generators, not to a topology.

Why the field argument breaks

In a commutative algebra, polynomial relations for $x$ and $y$ reduce every monomial to $x^iy^j$ with bounded exponents. In a noncommutative algebra, words such as $xyx^2yxy$ cannot generally be rearranged into that form. Bounding individual powers does not immediately bound alternating words.

The hypothesis that every element is algebraic, including all these words and their sums, is much stronger than algebraicity of the generators alone. The difficulty is imposing that hypothesis without eliminating all the remaining infinite-dimensional freedom.

3. Development history and previous partial results

Kurosh formulated the question in 1941 in connection with Burnside-type finiteness problems for groups. His original division-ring formulation asks whether algebraicity and finite algebra generation over a central subfield force finite dimension. The chosen central subfield need not be the full center.

A broader question for arbitrary associative algebras has a different answer and mechanism. Golod’s construction, using Golod–Shafarevich estimates, provides finitely generated infinite-dimensional nil algebras. Their unitizations are algebraic but contain nonzero nilpotents. A division ring cannot contain a nonzero nilpotent, so those counterexamples do not settle Kurosh’s division-ring problem.

Positive results show why degree restrictions matter. Jacobson’s bounded-degree theorem leads to a polynomial identity, and Kaplansky’s theorem makes algebraic polynomial-identity algebras locally finite. A polynomial identity is one noncommutative polynomial vanishing under every substitution, unlike an element-specific algebraic equation.

Bell, Drensky, and Sharifi proved a stronger bounded-degree result in a related setting: if elements satisfy polynomial relations of uniformly bounded degree $d$ with coefficients on the left in a possibly noncentral subfield, the division ring has dimension at most $d^2$ over its center. The new construction deliberately allows unbounded element-dependent degrees.

Amitsur proved that finite algebra generation over an uncountable central field forces algebraicity. That is a theorem about obtaining algebraicity, not a conclusion that the division ring is locally finite. Countability serves another purpose in the present construction: it lets every polynomial expression be put into a schedule.

Recent work of Goodearl and Letzter studies failure of finite algebra generation for classes of transcendental skew fields, explicitly distinguishing that question from the algebraic Kurosh problem. The claimed example here has the opposite combination: finite algebra generation together with algebraicity, yet infinite central dimension.

The technical background includes cyclic and symbol division algebras, trace identities for noncommuting words, and the interpretation of inner derivations as infinitesimal conjugations. The new manuscript combines these tools in a lifting theorem designed to impose one relation while preserving room for further enlargement.

4. What the new manuscripts claim

The manuscript, dated 23 September 2026, claims a countable division ring $D$ of characteristic zero, with actual center $F=Z(D)$, and elements $x,y\in D$ satisfying

$$D=F[x,y]=F(x,y),\qquad [D:F]=\infty,$$

while every element of $D$ is algebraic over $F$.

The same pair $\{x,y\}$ therefore witnesses failure of local finiteness. The theorem asserts algebra generation, not only division generation, and infinite dimension over the actual center, not merely over a smaller field chosen during the construction.

The center is not prescribed in advance. The proof first builds an algebraic division ring over a countable central coefficient field, then separately shows that enlarging to the full center does not make its dimension finite.

The principal intermediate claim is a lifting theorem for formal series in two noncommuting variables over a finite-dimensional central division algebra. It makes one prescribed series algebraic over a larger countable central field while increasing the division algebra’s degree, preserving all earlier polynomial identities, restoring generation, and preserving two independent deformation directions.

The successive division algebras are not asserted to embed into one another. The final construction uses an ultraproduct to retain the compatible identities rather than treating them as an increasing union of division rings.

5. Why the advance matters

Provisional importance: 5/5. If correct, this resolves a central noncommutative finiteness problem negatively in a particularly strong form: countable, characteristic zero, and generated by two elements as an algebra over the center.

The contrast with fields is precise. Algebraicity controls each element separately, but in a division ring that local control would not force a finite common vector-space description, even with only two algebra generators.

The actual-center conclusion is especially important. A division algebra can be infinite-dimensional over a small central subfield for elementary reasons while remaining finite-dimensional over its center. The manuscript explicitly excludes that explanation, so the result addresses the intrinsic noncommutative structure.

6. The main idea

Impose one algebraic relation without exhausting freedom

The construction proceeds through finite-dimensional central division algebras with increasingly large degrees. Formal series in two noncommuting variables record possible deformations of the chosen generators.

At each stage, the constant terms generate the current algebra. Two linear coefficients remain independent after quotienting out changes obtainable by conjugation. Infinitesimal conjugation changes an element $a$ by a commutator $[T,a]=Ta-aT$; directions of this form do not represent genuinely new generator freedom.

Given one expression $w$, the lifting theorem changes the formal input so that $w$ satisfies an algebraic equation, while keeping both independent directions and increasing the coefficient algebra’s degree. Losing either generation or the available deformation freedom could prevent the next stage from working.

Use matrix responses to solve the formal equations

After extending scalars to split the coefficient division algebra into matrices, the proof studies generic block inputs. Changing a diagonal block gives one type of eigenvalue response; adding an arrow to another block and a return arrow gives a paired response.

The manuscript identifies relations among these responses and transfers each first unsolved word coefficient into the same space. Repeated blocks and auxiliary comparison blocks distinguish local relations from common relations across blocks. Trace identities impose cyclic sums, reflecting the fact that traces do not change when a product is cyclically rotated.

An eleven-block arrangement with two kinds of connections supplies corrections whose obstructions are exactly those cyclic relations. The formal equation is solved one word length at a time through finite affine linear systems.

These matrix calculations must then be realized inside a division algebra. A cyclic-algebra construction supplies that realization, while a translation of the formal input restores generation and the two independent directions. Finally, only countably many scalar coefficients are retained, returning to a countable center for the next step.

This description separates the roles: spectral calculations impose the relation, division realization preserves the ring type, and translation restores the hypotheses required to iterate.

Schedule all expressions and pass to a limit

Every coefficient field at a finite stage is countable, so its polynomial expressions in two noncommuting generators can be enumerated. A diagonal schedule ensures that each expression with coefficients from any stage eventually receives a nonzero algebraic relation. Later stages preserve it.

The fields embed successively, but the division algebras need not. A nonprincipal ultraproduct supplies a common ambient division ring. Informally, it identifies sequences that agree on an ultrafilter-large set of stages, retaining every identity that holds from some point onward.

Let $k_\infty$ be the union of the central fields and form the two-generated subalgebra $R=k_\infty[x,y]$ inside that ultraproduct. Every element of $R$ is algebraic by the schedule.

This already makes $R$ a division ring. If $r\ne0$ has a least-degree relation, its constant coefficient is nonzero. Rearranging gives

$$r^{-1}=-c_0^{-1}(c_1+c_2r+\cdots+c_nr^{n-1})\in R.$$

Thus adjoining inverses adds nothing. Countability follows from the countable coefficient field and finite algebra generation.

Prove that the full center cannot absorb the growth

At the finite stages, generator words span division algebras of unbounded dimension. For each fixed word-length bound, an ultrafilter selects a fixed finite list that remains linearly independent at sufficiently many stages. These lists show $[R:k_\infty]=\infty$.

Now suppose $R$ were finite-dimensional over its full center $F$. Choose a finite $F$-basis and record its multiplication table and the coordinates of $x,y$. Only finitely many elements of $F$ occur in those data. Since every element of $F$ is algebraic over $k_\infty$, they generate a finite field extension $E/k_\infty$.

The $E$-span of the chosen basis is closed under multiplication and contains both generators. It would contain all of $R$, making $R$ finite-dimensional over $k_\infty$, a contradiction. This separate descent argument supplies the actual-center conclusion.

7. Scope, limitations, and open questions

The annihilating polynomial is allowed to vary with the element, with no uniform degree bound. This is compatible with the known positive bounded-degree theorems.

The construction does not prescribe the center or give a small explicit multiplication table for the infinite-dimensional ring. Its ultraproduct is an existence device; the selected two-generated subalgebra is countable even though the ambient ultraproduct can be much larger.

The result is not a claim about commutative fields, where finite algebra generation by algebraic elements still gives finite dimension. Nor does it rely on nilpotents, which are excluded in a division ring.

Determining what additional structural or effective conditions restore local finiteness remains a separate question. The claimed example shows that algebraicity, characteristic zero, and two algebra generators alone would not suffice.

8. Sources and further reading

Kurosh’s Ringtheoretische Probleme, die mit dem Burnsideschen Problem über periodische Gruppen in Zusammenhang stehen (1941) formulates the problem. Jacobson’s Le problème de Kurosch (Bourbaki, 1952) describes early positive theory, while Golod’s 1964 nil-algebra construction addresses the broader algebra question.

Bell, Drensky, and Sharifi’s bounded-degree theorem gives a useful contrast. Amitsur’s Algebras over infinite fields (1956) concerns algebraicity under finite generation, and Goodearl–Letzter’s skew-field work treats a distinct transcendental setting.

Gille and Szamuely’s Central Simple Algebras and Galois Cohomology explains cyclic division algebras. Keisler’s The ultraproduct construction (2010) supplies background on the limiting device used in the manuscript below.

Source manuscripts

• M1. OpenAI, A Counterexample to Kurosh’s Division-Ring Problem (September 23, 2026).

Sources consulted on 21 October 2026 at repository revision 1581d4c. Download the manuscript citations.