Algebraically Closed Division Rings

Q: Why is it enough to require roots only for polynomials with central coefficients?

Because that hypothesis alone bounds the degrees of irreducible polynomials over the centre by _Z(D)D, and everything else in Baer's proof is field theory about a field with bounded irreducible degrees. Once the conclusion is reached, the Niven–Jacobson theorem supplies roots for all polynomials over D, so the weak hypothesis retroactively implies the strong one.

Q: Are there many such division rings, or essentially one?

Many. There is exactly one for each real-closed field, and real-closed fields are abundant: R, the real algebraic numbers, the real closure of any ordered field, fields of real Puiseux series, ultrapowers of R. What the classification says is that no *new* structure appears at the quaternionic level.

Q: Does the classification cover division rings of infinite dimension over their centres?

No, and Lam is explicit that this case is not understood. Baer's proof begins by bounding the degree of an irreducible polynomial over the centre by _Z(D)D; with infinite dimension there is no bound and no known replacement argument.

Q: Why does the proof need the field to be perfect?

To apply the primitive element theorem. Perfection makes every finite algebraic extension separable, hence simple, which is what lets a hypothetical element outside the largest simple extension generate a larger simple extension and produce the contradiction. In characteristic zero perfection is automatic; the work is entirely in characteristic p.

Q: How does this relate to Frobenius' theorem?

Frobenius' theorem is the last step, not the whole story. It classifies finite-dimensional division algebras over a real-closed field; Baer's contribution is to prove that the centre *must be* real-closed, which is where Artin–Schreier and the perfection argument are needed.

Q: Is right algebraic closure equivalent to left algebraic closure in general?

It is equivalent for noncommutative centrally finite division rings, as a consequence of (16.16): both conditions single out the same class of objects. In general — for centrally infinite division rings — no such equivalence is known, and the two conditions should be kept distinct.

Executive Summary

Call a division ring $D$ right algebraically closed if every nonconstant $f \in D [t]$ has a right root — equivalently, if every polynomial splits completely in $D [t]$ . Which division rings have this property?

For fields the answer is classical and the objects are classified by characteristic and transcendence degree. For noncommutative $D$ that is finite-dimensional over its centre, Baer's theorem $(16.15)$ answers it completely and with a strikingly weak hypothesis: it is enough that every polynomial with coefficients in the centre have a root in $D$ . The centre is then forced to be a real-closed field and $D$ to be the quaternion algebra over it.

Combining with the Niven–Jacobson theorem $(16.14)$ , which supplies the converse, gives the classification $(16.16)$ : the noncommutative centrally finite right algebraically closed division rings are exactly the quaternion algebras over real-closed fields, and these are left algebraically closed as well. The centrally infinite case remains open.

$1$ Isomorphism type per real-closed field

$4$ Dimension over the centre

$2$ Degree of the algebraic closure of the centre

centralCoefficients needed for the hypothesis

Overview

The definition has to choose a side, since roots do. $D$ is right algebraically closed if every nonconstant $f \in D [t]$ has a right root; by the remainder theorem $(16.2)$ this is the same as every $f \in D [t]$ being a product of linear factors. The left-handed notion is defined by mirror image, and one of the payoffs of the classification is that the two coincide in the centrally finite case.

{\begin{matrix} noncommutative, centrally finite, \\ right algebraically closed \end{matrix}} = {\begin{matrix} quaternion algebras over \\ real-closed fields \end{matrix}}

(16.16)

Both classes are also left algebraically closed. The classification is thereby reduced to the classification of real-closed fields.

The proof is field theory, not ring theory. Finite dimension over the centre bounds the degrees of irreducible polynomials over the centre, that bound makes the centre perfect and its algebraic closure finite over it, and the Artin–Schreier theorem converts a finite proper algebraic closure into real-closedness. Frobenius' theorem finishes the job.

Learning Objectives

State the definition of right algebraic closure and its equivalence with complete splitting.
Show that a root in $D$ of an irreducible $f \in R [t]$ bounds $\deg f$ by $\dim_{R} D$ .
Prove that a field whose irreducible polynomials have bounded degree is perfect.
Use the primitive element theorem to conclude that the algebraic closure is a finite extension.
Invoke Artin–Schreier and Frobenius to identify $R$ and $D$ .
State the classification $(16.16)$ and identify what it does not cover.

Definitions

Definition§16Algebraic closure conditions for a division ring

A division ring $D$ is right algebraically closed if every nonconstant $f \in D [t]$ has a right root in $D$ . Since a right root $c$ yields $f = g \cdot (t - c)$ by $(16.2)$ , an induction on degree shows this is equivalent to: every $f \in D [t]$ is a product of linear factors in $D [t]$ . Left algebraically closed is the mirror condition, with the variable substituted on the left.

$R = Z (D)$: The centre, a field. Baer's theorem hypothesises roots only for polynomials in $R [t]$ .
Centrally finite: $\dim_{R} D < \infty$ . The dimension is then a perfect square $m^{2}$ , and $m$ is the degree of $D$ .
$R (α)$: For $α \in D$ algebraic over $R$ , the commutative subfield $R [α] \subseteq D$ , of dimension $\deg (min. poly. of α)$ over $R$ .
Simple extension: A field extension generated by one element. The primitive element theorem makes every finite separable extension simple.
Artin–Schreier theorem: If $C$ is algebraically closed and $F \subseteq C$ satisfies $1 < \dim_{F} C < \infty$ , then $F$ is real-closed and $C = F (\sqrt{- 1})$ ; in particular $\dim_{F} C = 2$ .
Frobenius' theorem, real-closed form: A division algebra finite-dimensional over a real-closed field $R$ with centre $R$ is either $R$ itself or the quaternion algebra over $R$ .

Following Lam, $R$ denotes the centre of $D$ throughout this page, and it turns out to be a real-closed field.

Core Concepts

Why finite dimension bounds degrees

If $f \in R [t]$ is irreducible and has a root $α \in D$ , then $R [α]$ is a commutative domain finite-dimensional over $R$ , hence a field, and $\dim_{R} R (α) = \deg f$ . Since $R (α) \subseteq D$ ,

\deg f = \dim_{R} R (α) \leq \dim_{R} D = n .

(16.15a)

Every irreducible polynomial over $R$ has degree at most $n$ . The hypothesis is used exactly here and nowhere else.

A field whose irreducible polynomials have bounded degree is severely constrained. It has no proper algebraic extension of large degree, so its algebraic closure is finite over it — and by Artin–Schreier a field with a proper finite algebraic closure is real-closed, with the closure of degree exactly $2$ .

The two field-theoretic steps

PerfectionIn characteristic

p

, the tower

R (a^{1 / p}) \subseteq R (a^{1 / p^{2}}) \subseteq \dots

has bounded dimension, so it stabilises; raising the stabilisation identity to a

p

-th power shows

a

already has a

p

-th root in

R

Finiteness of the closurePerfection makes every algebraic extension separable, so the primitive element theorem turns any strictly larger finite extension into a simple extension of larger degree — contradicting maximality of the largest simple extension.

Artin–Schreier

1 < \dim_{R} C < \infty

forces

R

real-closed and

C = R (\sqrt{- 1})

of degree

2

FrobeniusA noncommutative division algebra with real-closed centre and finite dimension over it is the quaternion algebra.

Key Results

Theorem(16.15)Baer

Let $D$ be a noncommutative division ring with centre $R$ , of finite dimension over $R$ , and suppose every polynomial in $R [t]$ has a root in $D$ — as is the case, in particular, if $D$ is right algebraically closed. Then $R$ is a real-closed field and $D$ is the division ring of quaternions over $R$ .

Proof

Write $n = \dim_{R} D < \infty$ and fix an algebraic closure $C$ of $R$ .

Step 1: degrees are bounded. Every irreducible $f \in R [t]$ has a root $α \in D$ by hypothesis, and $R (α) \subseteq D$ is a field with $\dim_{R} R (α) = \deg f$ . Hence $\deg f \leq n$ , and every simple algebraic extension of $R$ inside $C$ has dimension at most $n$ .

**Step 2: $R$ is perfect.** If $char R = 0$ there is nothing to prove. Let $char R = p$ and $a \in R$ ; write $a^{1 / p^{m}}$ for the unique $p^{m}$ -th root of $a$ in $C$ . The chain $R (a^{1 / p}) \subseteq R (a^{1 / p^{2}}) \subseteq \dots$ consists of simple extensions of $R$ , so by Step 1 their dimensions are bounded and the chain stabilises: $a^{1 / p^{m + 1}} \in R (a^{1 / p^{m}})$ for some $m$ . Raising to the $p^{m}$ -th power and using that the Frobenius map is a ring homomorphism gives $a^{1 / p} \in R^{p^{m}} (a) \subseteq R$ . So every element of $R$ is a $p$ -th power and $R$ is perfect.

**Step 3: $\dim_{R} C < \infty$ .** Among the simple extensions of $R$ inside $C$ choose $K$ of largest $R$ -dimension; it exists by Step 1. If some $β \in C ∖ K$ existed, then $K (β)$ would be a finite extension of $R$ , separable because $R$ is perfect, hence simple by the primitive element theorem — and of dimension strictly greater than $\dim_{R} K$ , contradicting the choice of $K$ . Therefore $K = C$ and $\dim_{R} C < \infty$ .

**Step 4: $R \neq C$ .** If $R$ were algebraically closed then every $α \in D$ would satisfy $\dim_{R} R (α) = 1$ , forcing $D = R$ and contradicting noncommutativity.

Step 5: conclude. Now $1 < \dim_{R} C < \infty$ , so the Artin–Schreier theorem makes $R$ a real-closed field with $C = R (\sqrt{- 1})$ . By Frobenius' theorem in its real-closed form, a division algebra of finite dimension over the real-closed field $R$ with centre $R$ is $R$ or the quaternion algebra over $R$ ; noncommutativity selects the latter.

Theorem(16.16)Classification

The noncommutative centrally finite division rings which are right algebraically closed are precisely the division rings of quaternions over real-closed fields. All of these are left algebraically closed as well.

Proof

One inclusion is $(16.15)$ : right algebraic closure certainly gives roots for central polynomials, so such a $D$ is a quaternion algebra over a real-closed field. The other is the Niven–Jacobson theorem $(16.14)$ : quaternions over a real-closed field are right and left algebraically closed. The two statements together give both the classification and the left–right symmetry.

Corollary—Central roots suffice

For a noncommutative centrally finite division ring $D$ , the following are equivalent: (i) every $f \in Z (D) [t]$ has a root in $D$ ; (ii) $D$ is right algebraically closed; (iii) $D$ is left algebraically closed; (iv) $D$ is the quaternion algebra over a real-closed field. In particular the apparently much weaker condition (i) implies the others.

Remark—The commutative and the centrally infinite cases

For fields the classification is classical: an algebraically closed field is determined up to isomorphism by its characteristic and its transcendence degree over the prime field, which for uncountable fields equals the cardinality. For centrally infinite division rings nothing comparable is known; the structure of algebraically closed division rings of infinite dimension over the centre is not understood.

Proof Techniques and Method

How these proofs work, and which move to reuse.

Move 1

Convert an embedding into a degree bound

A root of an irreducible polynomial generates a subfield, and a subfield of a centrally finite division ring has bounded dimension. This is the standard way to transfer a finiteness hypothesis on $D$ into a finiteness hypothesis on $R$ .

Move 2

Stabilise a bounded chain

An increasing chain of subfields with bounded dimension must stabilise; the stabilisation identity, pushed through Frobenius, yields perfection. Bounded chains are the workhorse of the whole proof.

Move 3

Maximise a simple extension

Choosing the largest simple extension and contradicting its maximality with the primitive element theorem is how bounded degree becomes finite algebraic closure.

The proof imports two heavy external theorems. Artin–Schreier — a field with a proper algebraic closure of finite degree is real-closed — is genuinely deep; the characteristic-zero case is comparatively easy, and the difficulty lies in ruling out positive characteristic. Frobenius' theorem in the real-closed form is elementary by comparison.

Worked Example

The model case, and why it is the only one over $ℝ$

$ℝ$ is real-closed, so $ℍ$ is right and left algebraically closed by $(16.14)$ , and $(16.16)$ says it is the only noncommutative centrally finite example with centre $ℝ$ — as Frobenius' theorem already told us, since $ℝ$ , $ℂ$ and $ℍ$ exhaust the finite-dimensional real division algebras and only $ℍ$ is noncommutative.

A near miss: the rational quaternions

Let $D = (\frac{- 1, - 1}{ℚ}) = ℚ \oplus ℚ i \oplus ℚ j \oplus ℚ k$ , a noncommutative division ring with $Z (D) = ℚ$ and $\dim_{ℚ} D = 4$ . Is it right algebraically closed? Test the central polynomial $f (t) = t^{2} - 2$ . For $q = a + v$ with $v$ purely imaginary,

q^{2} = (a^{2} - | v |^{2}) + 2 a v = 2 ⟹ a v = 0 and a^{2} - | v |^{2} = 2 .

(E.1)

If $v \neq 0$ then $a = 0$ and $- | v |^{2} = 2$ , impossible since $| v |^{2}$ is a sum of squares in $ℚ$ ; so $v = 0$ and $a^{2} = 2$ with $a \in ℚ$ , impossible. Hence $t^{2} - 2$ has no root in $D$ , and by Baer's theorem it could not have — $ℚ$ is not real-closed.

The failure is exactly the failure of $ℚ$ to be real-closed, not anything about the quaternion structure. Replacing $ℚ$ by its real closure $ℝ_{alg} = ℝ \cap \bar{ℚ}$ repairs it: the quaternion algebra over $ℝ_{alg}$ is a countable, right algebraically closed, noncommutative division ring.

Frameworks and Models

The landscape of algebraically closed objects, sorted by how they sit over their centre.

Right algebraically closed division rings — every nonconstant $f \in D [t]$ has a right root
- Commutative
  - Algebraically closed fields
  - Classified by characteristic and transcendence degree over the prime field
  - Examples: $ℂ$ , $\bar{ℚ}$ , $\bar{𝔽_{p}}$
- Noncommutative, centrally finite
  - Quaternion algebras over real-closed fields $(16.16)$
  - Always of dimension $4$ over the centre
  - Also left algebraically closed
- Noncommutative, centrally infinite
  - No classification known
  - Baer's argument fails at the first step: degrees of irreducible polynomials are unbounded

What the hypotheses exclude
Drop this hypothesis	What survives	Why the proof breaks
Noncommutative	$D$ may be any algebraically closed field	Step 4 fails: $R = C$ is allowed
Centrally finite	no classification available	Step 1 fails: $\deg f \leq \dim_{R} D$ becomes vacuous
Roots for central polynomials	arbitrary centrally finite division rings, e.g. rational quaternions	Step 1 has no input at all
Nothing (all hypotheses held)	quaternions over a real-closed field	—

Comparison and Classification

Algebraic closure properties across candidate division rings
	Right alg. closed	Left alg. closed	Centrally finite	Noncommutative
$ℂ$	yes	yes	yes	no
$ℝ$	no	no	yes	no
$ℍ$ over $ℝ$	yes	yes	yes	yes
Quaternions over $ℚ$	no	no	yes	yes
Quaternions over the real algebraic numbers	yes	yes	yes	yes
Fraction ring of the Weyl algebra $A_{1} (ℂ)$	no	no	no	yes

Algebraic closure properties across candidate division rings

Commutative theory versus the noncommutative classification
Question	Algebraically closed fields	Noncommutative centrally finite case
How many isomorphism types?	one per characteristic and transcendence degree	one per real-closed field
Dimension over a distinguished subfield	not applicable	always $4$ over the centre
Number of roots of a degree- $n$ polynomial	exactly $n$ with multiplicity	at most $n$ or infinitely many
Uniqueness of factorisation into linear factors	yes	no; infinitely many factorisations are typical
Is the property left–right symmetric?	vacuously	yes, but it is a theorem $(16.16)$

Relationship Map

\dim_{R} D < \infty

bounded irreducible degrees

R

perfect,

\dim_{R} C < \infty

R

real-closed

D

= quaternions

The Niven–Jacobson Theorem supplies the converse half of the classification and the left-handed statement.
Division Rings Containing an Algebraically Closed Field proves the closely related $(15.9)$ by the same Artin–Schreier plus Frobenius route.
Centrally Finite and Centrally Infinite Division Rings explains the dichotomy in the hypothesis and why the centrally infinite case is out of reach.
Generalised Quaternion Algebras provides the structure of the objects that the classification produces.
The Gordon–Motzkin Theorem explains why algebraic closure here cannot carry the usual root count.

Design Considerations

Design considerations here means the choices made when modelling a problem with these algebraic structures.

Which closure condition? Demanding roots for all $f \in D [t]$ looks much stronger than demanding them for $f \in Z (D) [t]$ , but for noncommutative centrally finite $D$ the two are equivalent. Choose the weak condition when verifying, the strong one when using.
Which side? For centrally finite division rings the left and right conditions agree, so no choice is needed. Do not carry that expectation into the centrally infinite setting, where nothing is known.
Which base field? The classification reduces the problem to choosing a real-closed field. If a countable model is wanted, use the real algebraic numbers; if a non-archimedean one, use real Puiseux series. The quaternionic layer contributes nothing new.
**When to abandon the analogy with $ℂ$ .** Algebraic closure gives existence of roots and complete splitting, and nothing else: no root count, no unique factorisation, no rigidity of the linear factors. Any modelling that relies on those additional properties needs a commutative setting.

Failure Modes and Common Mistakes

Do not conclude from $(16.16)$ that there is only one such division ring; there is one for every real-closed field, and real-closed fields form a proper class of isomorphism types.
Do not assume the quaternion algebra over an arbitrary formally real field is algebraically closed — real-closedness, not formal reality, is what makes $R (\sqrt{- 1})$ algebraically closed.
Do not use Frobenius' theorem over a base that is merely ordered; the real-closed hypothesis is essential to its finite-dimensional form.
Do not read $(16.15)$ as requiring $D$ to be right algebraically closed; roots for central polynomials suffice, and that is the sharper statement.

Historical Notes and Lessons Learned

1877FrobeniusFrobenius classifies the finite-dimensional associative division algebras over $ℝ$ : only $ℝ$ , $ℂ$ and $ℍ$ occur.
1927Artin and SchreierArtin and Schreier characterise real-closed fields and prove that a field with a proper algebraic closure of finite degree is real-closed with closure of degree two — the theorem that makes the classification possible.
1941–44Niven; Eilenberg–NivenNiven solves polynomial equations over the real quaternions; Eilenberg and Niven give a topological proof that every nonconstant quaternionic polynomial has a root.
mid-centuryBaer's converseBaer proves that the quaternionic examples are the only noncommutative centrally finite ones, reducing the classification to that of real-closed fields.
1965–Root sets understoodGordon and Motzkin describe the shape of root sets over arbitrary division rings, explaining what algebraic closure can and cannot mean once conjugacy classes are infinite.

The lesson is about the cost of a definition. Transplanting algebraically closed to the noncommutative world preserves the existence statement and loses everything quantitative, and the classification shows the surviving notion is extremely rigid: over a real-closed field there is exactly one noncommutative example, and beyond finite dimension the question is still open after decades.

Quick Reference

Definitionevery nonconstant

f \in D [t]

has a right root

Equivalent formevery

f \in D [t]

splits into linear factors

Baer's hypothesis

D

noncommutative,

\dim_{Z (D)} D < \infty

, every

f \in Z (D) [t]

has a root

Baer's conclusion

Z (D)

real-closed and

D

the quaternions over it

Classificationnoncommutative centrally finite right alg. closed

=

quaternions over real-closed fields

Symmetryright and left algebraic closure agree in this class

Dimension

\dim_{R} D = 4

\dim_{R} C = 2

Openthe centrally infinite case

Statements and hypotheses
Reference	Hypotheses	Conclusion
$(16.15)$	$D$ noncommutative, centrally finite, every $f \in Z (D) [t]$ has a root in $D$	$Z (D)$ real-closed; $D$ the quaternion algebra over it
$(16.16)$	$D$ noncommutative and centrally finite	right alg. closed $\Leftrightarrow$ quaternions over a real-closed field $\Leftrightarrow$ left alg. closed
$(16.14)$	$R$ real-closed, $D$ quaternions over $R$	$D$ is right and left algebraically closed
Artin–Schreier	$C$ algebraically closed, $1 < \dim_{F} C < \infty$	$F$ real-closed and $C = F (\sqrt{- 1})$
Frobenius, real-closed form	$D$ finite-dimensional over a real-closed centre $R$	$D = R$ or $D$ the quaternions over $R$

Frequently Asked Questions

Why is it enough to require roots only for polynomials with central coefficients?

Because that hypothesis alone bounds the degrees of irreducible polynomials over the centre by $\dim_{Z (D)} D$ , and everything else in Baer's proof is field theory about a field with bounded irreducible degrees. Once the conclusion is reached, the Niven–Jacobson theorem supplies roots for all polynomials over $D$ , so the weak hypothesis retroactively implies the strong one.

Are there many such division rings, or essentially one?

Many. There is exactly one for each real-closed field, and real-closed fields are abundant: $ℝ$ , the real algebraic numbers, the real closure of any ordered field, fields of real Puiseux series, ultrapowers of $ℝ$ . What the classification says is that no new structure appears at the quaternionic level.

Does the classification cover division rings of infinite dimension over their centres?

No, and Lam is explicit that this case is not understood. Baer's proof begins by bounding the degree of an irreducible polynomial over the centre by $\dim_{Z (D)} D$ ; with infinite dimension there is no bound and no known replacement argument.

Why does the proof need the field to be perfect?

To apply the primitive element theorem. Perfection makes every finite algebraic extension separable, hence simple, which is what lets a hypothetical element outside the largest simple extension generate a larger simple extension and produce the contradiction. In characteristic zero perfection is automatic; the work is entirely in characteristic $p$ .

How does this relate to Frobenius' theorem?

Frobenius' theorem is the last step, not the whole story. It classifies finite-dimensional division algebras over a real-closed field; Baer's contribution is to prove that the centre must be real-closed, which is where Artin–Schreier and the perfection argument are needed.

Is right algebraic closure equivalent to left algebraic closure in general?

It is equivalent for noncommutative centrally finite division rings, as a consequence of $(16.16)$ : both conditions single out the same class of objects. In general — for centrally infinite division rings — no such equivalence is known, and the two conditions should be kept distinct.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §16, (16.15)–(16.16) (pp. 269–271), with §15 for the Artin–Schreier input.
N. Jacobson, Basic Algebra II, 2nd edition, W. H. Freeman, 1989, for the Artin–Schreier theory of real-closed fields and the primitive element theorem.
I. Niven, “Equations in quaternions”, American Mathematical Monthly 48 (1941), 654–661.
P. M. Cohn, Skew Fields: Theory of General Division Rings, Encyclopedia of Mathematics and its Applications 57, Cambridge University Press, 1995.
P. K. Draxl, Skew Fields, London Mathematical Society Lecture Note Series 81, Cambridge University Press, 1983.
N. Jacobson, Structure of Rings, American Mathematical Society Colloquium Publications 37, revised edition, 1964.

AI Suggested Questions

Sketch the proof of the Artin–Schreier theorem and identify where positive characteristic is ruled out.
Construct a countable noncommutative right algebraically closed division ring explicitly.
Is there any known example of a centrally infinite right algebraically closed division ring?
How does Baer's theorem interact with Cohn's theory of existentially closed skew fields?
What happens to the classification if one requires only that every central polynomial of odd degree has a root?
Compare Baer's theorem with Lam's $(15.9)$ on division rings containing an algebraically closed field: which hypotheses are stronger?

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Why finite dimension bounds degrees

The two field-theoretic steps

06Key Results

07Proof Techniques and Method

Convert an embedding into a degree bound

Stabilise a bounded chain

Maximise a simple extension

08Worked Example

The model case, and why it is the only one over ℝ

A near miss: the rational quaternions

09Frameworks and Models

10Comparison and Classification

11Relationship Map

12Design Considerations

13Failure Modes and Common Mistakes

14Historical Notes and Lessons Learned

15Quick Reference

16Frequently Asked Questions

17Related KEVOS Topics

18References

19AI Suggested Questions

Continue learning