Why can't I just take the product (t-c_1)…(t-c_n)?

Because evaluation is not multiplicative over a division ring. Only the rightmost factor is guaranteed to contribute a root: in H, (t-j)(t-i) has i as its only root and does not vanish at j. The recursion replaces each node by a conjugate corrected for the factors already present.

The Bray–Whaples Theorem | KEVOS® Mathematics

Q: Does the theorem require the nodes to be algebraic over the centre?

No. Unlike the vanishing-polynomial theory for conjugacy classes, Bray–Whaples needs no algebraicity: it uses only the conjugation rule and the class bound, both of which hold for arbitrary elements of an arbitrary division ring.

Q: Is the polynomial g independent of the order in which the nodes are processed?

Yes — uniqueness says so. The intermediate factors t - d_i do depend on the order, so two different orderings give visibly different factorisations of the same g. That is the same non-uniqueness of factorisation seen in Wedderburn's theorem.

Q: How does this interact with the Gordon–Motzkin dichotomy?

It shows the finite side of the dichotomy is always realised. (16.12) says a degree-n polynomial has at most n roots or infinitely many; Bray–Whaples constructs, for each n, a degree-n polynomial with exactly n roots, provided D has at least n conjugacy classes.

Q: Is the Vandermonde criterion an equivalence?

No. Pairwise nonconjugacy implies invertibility, but not conversely. Exercise 16.5 shows V(a,b,c) is invertible whenever a, b, c do not all lie in a single conjugacy class, and gives an explicit criterion for failure when they do. The general theory is Lam's Vandermonde theory over division rings.

Q: What is the analogue for skew polynomial rings?

Replace conjugacy by -conjugacy, c (d)cd^-1, and the Vandermonde matrix by the Moore matrix built from -twisted powers. Pairwise non--conjugate nodes then give an invertible Moore matrix; this is the algebraic core of Gabidulin and skew cyclic code constructions.

Executive Summary

Over a field, $n$ distinct points determine a unique monic polynomial of degree $n$ vanishing on them, namely $(t - c_{1}) \dots (t - c_{n})$ . Over a division ring the naive product fails — its roots are not the $c_{i}$ — but the conclusion survives under the right hypothesis. The Bray–Whaples theorem $(16.13)$ : if $c_{1}, \dots, c_{n}$ are pairwise nonconjugate, there is exactly one monic $g \in D [t]$ of degree $n$ with $g (c_{1}) = \dots = g (c_{n}) = 0$ .

The polynomial produced is extremal in two ways. It has no other roots, so it realises the maximum finite root count permitted by the Gordon–Motzkin dichotomy; and every polynomial vanishing on the $c_{i}$ is a left multiple of it, so it generates the vanishing left ideal. The same hypothesis makes the noncommutative Vandermonde matrix at those nodes invertible, which is the interpolation statement in matrix form.

$1$ Monic polynomial of degree

n

$n$ Roots, and no more

$D [t] g$ Vanishing left ideal

$\det \neq 0$ Vandermonde at the nodes

Overview

Two hypotheses are conceivable when prescribing roots in a division ring: that the $c_{i}$ be distinct, or that they be pairwise nonconjugate. Distinctness is too weak — the class bound $(16.4)$ says a degree- $n$ polynomial can only reach $n$ conjugacy classes, so $n$ distinct roots inside a single class already force infinitely many roots and destroy uniqueness. Nonconjugacy is exactly the right strength.

c_{1}, \dots, c_{n} pairwise nonconjugate ⟹ \exists! g \in D [t] monic, \deg g = n, g (c_{i}) = 0 (1 \leq i \leq n) .

(16.13)

Uniqueness is the delicate half; existence follows from a one-step recursion.

The construction is a noncommutative Newton recursion: having built the monic $f$ of degree $n - 1$ vanishing on $c_{2}, \dots, c_{n}$ , one multiplies on the left by a single linear factor $t - d$ , and the requirement $g (c_{1}) = 0$ determines $d$ uniquely as a conjugate of $c_{1}$ . The whole theorem is that recursion, plus a check that no extra roots have been introduced.

Learning Objectives

Explain why pairwise nonconjugacy, not distinctness, is the correct hypothesis.
Carry out the Newton recursion $g = (t - d) f$ with $d = f (c_{1}) c_{1} f (c_{1})^{- 1}$ .
Prove part (b): every polynomial vanishing on the nodes is a left multiple of $g$ .
Prove part (a): $g$ has no roots outside the prescribed set, using the auxiliary quadratic.
Deduce invertibility of the Vandermonde matrix and unique interpolation of arbitrary values.
Compute a Bray–Whaples polynomial explicitly over the real quaternions.

Definitions

Pairwise nonconjugate: $c_{i} \neq d c_{j} d^{- 1}$ for all $i \neq j$ and all $d \in D^{*}$ . Distinct central elements are automatically pairwise nonconjugate; distinct elements of one class never are.
$g$ monic of degree $n$: $g (t) = t^{n} + b_{n - 1} t^{n - 1} + \dots + b_{0}$ with $b_{i} \in D$ ; monicity is essential, since a left scalar multiple of $g$ vanishes on the same set.
Vandermonde matrix $V (c_{0}, \dots, c_{n})$: The $(n + 1) \times (n + 1)$ matrix over $D$ whose entry in row $i$ and column $j$ is $c_{j}^{i}$ , rows indexed by powers $0$ to $n$ .
Interpolation problem: Given nodes $c_{0}, \dots, c_{n}$ and values $d_{0}, \dots, d_{n}$ in $D$ , find $f \in D [t]$ of degree at most $n$ with $f (c_{i}) = d_{i}$ for all $i$ .
Vanishing left ideal: ${h \in D [t] : h (c_{i}) = 0 for all i}$ , a left ideal because $h (c) = 0$ implies $(q h) (c) = 0$ .

$D$ is any division ring; no algebraicity, finiteness or chain condition is assumed, and the $c_{i}$ need not be algebraic over the centre.

Core Concepts

Why the naive product fails

Over a field the answer is $\prod_{i} (t - c_{i})$ . Over a division ring that product generally does not vanish at the $c_{i}$ : only the rightmost factor supplies a root. In $ℍ$ , $(t - j) (t - i)$ has $i$ as its unique root, and $j$ is not a root at all. The correct polynomial must be assembled so that each new factor is corrected for the values already accumulated.

The correction, in one line

Suppose $f$ is monic of degree $n - 1$ and vanishes on $c_{2}, \dots, c_{n}$ , and we want $g = (t - d) f$ to vanish additionally at $c_{1}$ . Since $c_{1}$ is not conjugate to any $c_{i}$ with $i \geq 2$ , it is not a root of $f$ , so $α : = f (c_{1}) \neq 0$ . The conjugation rule $(16.3)$ gives $g (c_{1}) = (α c_{1} α^{- 1} - d) α$ , so

g (c_{1}) = 0 ⟺ d = f (c_{1}) c_{1} f (c_{1})^{- 1} .

(16.13a)

The new linear factor is not $t - c_{1}$ but $t$ minus a specific conjugate of $c_{1}$ , determined by the accumulated polynomial.

That single equivalence gives existence and uniqueness at each step, and hence for the whole construction. Everything remaining in the proof is the verification that the resulting $g$ acquires no unwanted roots.

Key Results

Theorem(16.13)Bray–Whaples

Let $D$ be a division ring and let $c_{1}, \dots, c_{n} \in D$ be pairwise nonconjugate. Then there is a unique monic $g \in D [t]$ of degree $n$ with $g (c_{1}) = \dots = g (c_{n}) = 0$ . Moreover:

$c_{1}, \dots, c_{n}$ are all the roots of $g$ in $D$ ;
if $h \in D [t]$ satisfies $h (c_{i}) = 0$ for $1 \leq i \leq n$ , then $h \in D [t] \cdot g$ .

Proof

Induct on $n$ . For $n = 1$ take $g = t - c_{1}$ ; both extra properties are clear. Let $n \geq 2$ and let $f$ be the monic polynomial of degree $n - 1$ supplied by the inductive hypothesis for $c_{2}, \dots, c_{n}$ , so that the roots of $f$ are exactly $c_{2}, \dots, c_{n}$ and every polynomial vanishing there is a left multiple of $f$ .

Existence and uniqueness. Any monic $g$ of degree $n$ vanishing on all $c_{i}$ vanishes in particular on $c_{2}, \dots, c_{n}$ , hence $g = (t - d) f$ for some $d \in D$ by the inductive hypothesis and a degree count. Since $c_{1}$ is not a root of $f$ , $α : = f (c_{1}) \neq 0$ , and $(16.3)$ gives $g (c_{1}) = (α c_{1} α^{- 1} - d) α$ . So $g (c_{1}) = 0$ holds precisely for $d = α c_{1} α^{- 1}$ , which both proves existence and forces uniqueness.

Property (2). Let $h (c_{i}) = 0$ for all $i$ . Right-divide by the monic $g$ : $h = p g + r$ with $r = 0$ or $\deg r < n$ . Since $g (c_{i}) = 0$ , the product $p g$ vanishes at each $c_{i}$ by $(16.2)$ , so $r$ vanishes at $n$ pairwise nonconjugate points. If $r \neq 0$ , its roots would meet $n$ distinct conjugacy classes while $\deg r \leq n - 1$ , contradicting the class bound $(16.4)$ . Hence $r = 0$ .

Property (1). Let $c$ be a root of $g = (t - d) f$ . If $f (c) = 0$ then $c \in {c_{2}, \dots, c_{n}}$ by induction and we are done, so assume $f (c) \neq 0$ . Then $(16.3)$ makes $c$ conjugate to $d$ , and $d$ is conjugate to $c_{1}$ ; so $c$ lies in the class of $c_{1}$ . Suppose $c \neq c_{1}$ .

Construct the monic quadratic $λ (t) = (t - e) (t - c_{1})$ with $e = (c - c_{1}) c (c - c_{1})^{- 1}$ ; by the case $n = 2$ of the construction, $λ$ vanishes at both $c_{1}$ and $c$ , and by $(16.3)$ every root of $λ$ is conjugate to $c_{1}$ . Write $g = q λ + s$ with $s = 0$ or $\deg s \leq 1$ . Both $c_{1}$ and $c$ are roots of $g$ and of $λ$ , so $s$ has two distinct roots; a nonzero polynomial of degree $\leq 1$ over a division ring has at most one root, so $s = 0$ and $g = q λ$ with $\deg q = n - 2$ .

For $i \geq 2$ we have $λ (c_{i}) \neq 0$ , since the roots of $λ$ all lie in the class of $c_{1}$ . Hence $(16.3)$ makes each $c_{i}$ $(i \geq 2)$ conjugate to a root of $q$ , so $q$ has roots in $n - 1$ distinct conjugacy classes while $\deg q = n - 2$ — impossible by $(16.4)$ . Therefore $c = c_{1}$ .

Corollary—Extremal root sets

For every $n \geq 1$ and every choice of $n$ pairwise nonconjugate elements of $D$ , there is a monic polynomial of degree $n$ whose root set is exactly that finite set. Combined with $(16.12)$ — a degree- $n$ polynomial has at most $n$ roots or infinitely many — this shows the bound $n$ is attained for every $n$ and every division ring with at least $n$ conjugacy classes.

TheoremEx. 16.4Vandermonde invertibility and interpolation

Let $c_{0}, \dots, c_{n} \in D$ be pairwise nonconjugate. Then the Vandermonde matrix $V (c_{0}, \dots, c_{n}) = (c_{j}^{i})_{0 \leq i, j \leq n}$ is invertible over $D$ . Equivalently, for any prescribed values $d_{0}, \dots, d_{n} \in D$ there is a unique $f (t) = a_{n} t^{n} + \dots + a_{1} t + a_{0} \in D [t]$ with $f (c_{i}) = d_{i}$ for $0 \leq i \leq n$ .

Proof

Regard $D^{n + 1}$ as a left $D$ -vector space of row vectors and let $Φ (a_{0}, \dots, a_{n}) = (a_{0}, \dots, a_{n}) V$ , whose $j$ -th entry is $\sum_{i} a_{i} c_{j}^{i} = f (c_{j})$ . The map $Φ$ is left $D$ -linear because the $a_{i}$ appear on the left.

$Φ$ is injective: if $f (c_{j}) = 0$ for all $j$ with $\deg f \leq n$ , then $f$ has roots in $n + 1$ distinct conjugacy classes, so $(16.4)$ forces $f = 0$ . An injective endomorphism of a finite-dimensional left vector space over a division ring is bijective, so $Φ$ is onto — which is exactly the interpolation statement — and $V$ is invertible.

RemarkEx. 16.5Nonconjugacy is sufficient, not necessary

For three distinct $a, b, c \in D$ , the matrix $V (a, b, c)$ fails to be invertible exactly when $(b - a) b (b - a)^{- 1} = (c - a) c (c - a)^{- 1}$ ; and $V (a, b, c)$ is invertible whenever $a$ , $b$ , $c$ do not all lie in a single conjugacy class. So invertibility can survive some conjugacy among the nodes — the general theory is developed in Lam's work on Vandermonde matrices over division rings.

Proof Techniques and Method

How these proofs work, and which move to reuse.

Build from the right, correct on the left. The recursion multiplies the existing polynomial on the left by a new linear factor and chooses that factor's root to fix the new interpolation condition. This is Newton's method transplanted: each step touches only one new node.
Turn extra roots into class counting. Both the uniqueness argument and the no-extra-roots argument end by exhibiting a low-degree polynomial with roots in too many conjugacy classes and invoking $(16.4)$ . The class bound is used purely as a counting contradiction.
Manufacture an auxiliary quadratic. To rule out a second root in the class of $c_{1}$ , the proof constructs the unique monic quadratic vanishing at $c_{1}$ and at the intruder, then divides by it. Constructing a small interpolating polynomial to divide by is a reusable device.
Injective implies bijective. The interpolation theorem needs no explicit inverse: linear algebra over a division ring behaves like linear algebra over a field, so injectivity in finite dimension suffices.

Worked Example

Two nodes in the real quaternions

Take $D = ℍ$ , $c_{1} = 1$ and $c_{2} = i$ . They are nonconjugate: $1$ is central while $i$ is not. Start from $f (t) = t - i$ , the monic polynomial for the single node $c_{2}$ . Then $α = f (c_{1}) = 1 - i \neq 0$ and

d = α c_{1} α^{- 1} = (1 - i) \cdot 1 \cdot (1 - i)^{- 1} = 1,

(E.1)

Central nodes are their own corrections — conjugation fixes them.

So $g (t) = (t - 1) (t - i) = t^{2} - (1 + i) t + i$ . Check: $g (1) = 1 - (1 + i) + i = 0$ and $g (i) = i^{2} - (1 + i) i + i = - 1 - i + 1 + i = 0$ . Part (1) of the theorem asserts there are no other roots, which can be confirmed directly: if $c \neq i$ is a root then $(16.3)$ forces $(c - i) c (c - i)^{- 1} = 1$ , hence $c = 1$ .

Three nodes, and a noncentral correction

Now take $c_{1} = 1$ , $c_{2} = i$ , $c_{3} = 0$ — pairwise nonconjugate for the same reason. Build from the right. For the nodes ${i, 0}$ : start with $f_{1} (t) = t$ (node $0$ ), then $α = f_{1} (i) = i$ and $d = i \cdot i \cdot i^{- 1} = i$ , giving

f_{2} (t) = (t - i) t = t^{2} - i t .

(E.2)

Check: $f_{2} (0) = 0$ and $f_{2} (i) = i^{2} - i \cdot i = - 1 + 1 = 0$ . Note $f_{2} (j) = j^{2} - i j = - 1 - k \neq 0$ , as part (1) demands. Adjoining the node $c_{1} = 1$ : $α = f_{2} (1) = 1 - i$ , and $d = (1 - i) \cdot 1 \cdot (1 - i)^{- 1} = 1$ , so

g (t) = (t - 1) (t^{2} - i t) = t^{3} - (1 + i) t^{2} + i t .

(E.3)

The unique monic cubic over $ℍ$ vanishing at $0$ , $1$ and $i$ .

Verification: $g (0) = 0$ ; $g (1) = 1 - (1 + i) + i = 0$ ; $g (i) = i^{3} - (1 + i) i^{2} + i \cdot i = - i + (1 + i) - 1 = 0$ . By part (1) these are the only roots, and by part (2) any $h \in ℍ [t]$ vanishing at $0$ , $1$ , $i$ is a left multiple of $g$ .

Process and Workflow

Check the hypothesisVerify the nodes are pairwise nonconjugate. Over a centrally finite

D

this is Dickson's criterion: compare minimal polynomials over

Z (D)

InitialiseSet

f : = t - c_{n}

, the monic polynomial for the last node.

EvaluateFor the next node

c

, compute

α : = f (c)

. It is nonzero precisely because

c

is not conjugate to any node already handled.

Correct and extendSet

d : = α c α^{- 1}

and replace

f

(t - d) f

. The degree rises by one and the new node is now a root.

TerminateAfter

n

steps

f

is the Bray–Whaples polynomial

g

; its root set is exactly the node set and

D [t] g

is the vanishing left ideal.

You want a polynomial with a prescribed finite root set $S$ . Is it possible?

$S$ pairwise nonconjugateYes. A unique monic polynomial of degree

| S |

works, and nothing of smaller degree does.

$S$ has two elements in one classNo. Any polynomial vanishing on

S

vanishes on infinitely many elements of that class by

(16.11)

, so the root set cannot be exactly

S

$S$ is a whole conjugacy class, algebraic over $Z (D)$ Yes, but the generator is the class's minimal polynomial and the vanishing set is a two-sided ideal — see Vanishing Polynomials.

Comparison and Classification

Interpolation: field versus division ring
Item	Field $F$	Division ring $D$
Hypothesis on nodes	distinct	pairwise nonconjugate
Monic vanishing polynomial of degree $n$	$(t - c_{1}) \dots (t - c_{n})$	built by the Newton recursion; the naive product is wrong
Its root set	exactly the nodes	exactly the nodes $(16.13)$ (1)
Vanishing set	the ideal generated by it	the left ideal $D [t] g$ $(16.13)$ (2)
Vandermonde at the nodes	invertible iff nodes distinct	invertible if nodes pairwise nonconjugate; the converse fails
Lagrange basis available	yes	no direct analogue; the recursion replaces it

Which conclusions hold under which hypothesis on the nodes
	Unique monic degree- $n$ vanisher	Root set is exactly the nodes	Vanishing set is $D [t] g$	Vandermonde invertible
Pairwise nonconjugate	yes	yes	yes	yes
Distinct, all in one class	no	no	no	no
Distinct, not all in one class	partial	partial	partial	partial
All central and distinct	yes	yes	yes	yes

Which conclusions hold under which hypothesis on the nodes

Relationship Map

(16.3)

conjugation rule

(16.4)

class bound

(16.13)

Bray–WhaplesVandermonde invertibility

The Gordon–Motzkin Theorem supplies the class bound used three times in the proof, and its dichotomy $(16.12)$ explains why the extremal root sets constructed here are exactly the finite ones.
Vanishing Polynomials handles the complementary case where the prescribed set is a whole conjugacy class; there the generator is central and the vanishing set is two-sided.
Wedderburn's Factorisation Theorem is the opposite construction: it starts from a central polynomial and produces linear factors, rather than starting from points and producing a polynomial.
The Vandermonde theory extends to Ore extensions, where it becomes the theory of Moore matrices used in rank-metric coding.

Applications and Industry Use

Applications here means where this structure is used — inside mathematics and in the engineering and computing disciplines that consume it.

Rank-metric and skew cyclic codes. Gabidulin codes are defined by evaluation of skew polynomials at linearly independent points; the invertibility of the corresponding Moore–Vandermonde matrix is the exact analogue of the theorem above and is what guarantees unique decoding up to the designed radius.
Quaternionic interpolation. Fitting a quaternion-valued polynomial through prescribed attitudes requires exactly the nonconjugacy bookkeeping described here; naive coefficientwise interpolation silently assumes commutativity and fails.
Conjugacy testing in algebras. Verifying the hypothesis is itself a useful computation: for a centrally finite $D$ , nonconjugacy is decided by comparing minimal polynomials over the centre, per Dickson's criterion.
**Structure of $D [t]$ .** The theorem identifies a large family of left ideals of $D [t]$ with an explicit monic generator, which is how one exhibits concrete left ideals in this non-principal-in-the-obvious-sense setting.

Honest summary: outside coding theory the result is used mainly as a lemma. Its practical content is the recursion, which is a genuine algorithm and is what implementations actually run.

Computational Notes

Computational notes cover algorithms, cost and library behaviour rather than manufacturing process.

The recursion is the algorithm. Building $g$ from $n$ nodes costs $n$ iterations; iteration $k$ evaluates a polynomial of degree $k - 1$ at one point and performs one inversion in $D$ .

Total cost: $O (n^{2})$ multiplications in $D$ plus $n$ inversions — the same order as commutative Newton interpolation, with no extra asymptotic penalty for noncommutativity.
Evaluation must use Horner from the right, $f (c) = (\dots (a_{n} c + a_{n - 1}) c + \dots) c + a_{0}$ , which is valid because coefficients stay on the left throughout.
Solving the interpolation problem by inverting $V$ directly costs $O (n^{3})$ operations in $D$ and is numerically and symbolically worse; use the recursion.
Verifying the hypothesis is the expensive part in general: deciding conjugacy in a centrally infinite division ring may be undecidable, whereas for a centrally finite $D$ it reduces to comparing minimal polynomials.

Failure Modes and Common Mistakes

Do not assume the Vandermonde criterion is an equivalence: nonconjugate nodes give invertibility, but invertibility can also occur for some conjugate nodes (Exercise 16.5).
Do not reorder the nodes and expect the same intermediate factors; the $d_{i}$ depend on the order, even though the final $g$ does not.
Do not use the theorem to interpolate values at nodes lying in a common class — no polynomial of any degree can distinguish two conjugate nodes if its coefficients are central, and the general problem is not governed by this theorem.
Do not confuse part (2) with divisibility on the right; $h = p g$ , not $h = g p$ .

Quick Reference

Hypothesis

c_{1}, \dots, c_{n}

pairwise nonconjugate in

D

Existence and uniquenessone monic

g

\deg g = n

g (c_{i}) = 0

Recursion

g = (t - d) f

with

d = f (c) c f (c)^{- 1}

Root setexactly

{c_{1}, \dots, c_{n}}

Vanishing setthe left ideal

D [t] g

Vandermonde

V (c_{0}, \dots, c_{n})

invertible over

D

Interpolationunique

f

of degree

\leq n

with

f (c_{i}) = d_{i}

Cost

O (n^{2})

multiplications,

n

inversions

Statements and hypotheses
Reference	Hypotheses	Conclusion
$(16.13)$	$D$ a division ring, $c_{1}, \dots, c_{n}$ pairwise nonconjugate	Unique monic $g$ of degree $n$ vanishing on the $c_{i}$
$(16.13)$ (1)	Same	The $c_{i}$ are all the roots of $g$ in $D$
$(16.13)$ (2)	Same, $h (c_{i}) = 0$ for all $i$	$h \in D [t] g$
Ex. 16.4	$c_{0}, \dots, c_{n}$ pairwise nonconjugate	$V (c_{0}, \dots, c_{n})$ invertible; unique interpolation of any values
Ex. 16.5	$a, b, c$ distinct, not all in one class	$V (a, b, c)$ invertible

Frequently Asked Questions

Why can't I just take the product

(t - c_{1}) \dots (t - c_{n})

Because evaluation is not multiplicative over a division ring. Only the rightmost factor is guaranteed to contribute a root: in $ℍ$ , $(t - j) (t - i)$ has $i$ as its only root and does not vanish at $j$ . The recursion replaces each node by a conjugate corrected for the factors already present.

Does the theorem require the nodes to be algebraic over the centre?

No. Unlike the vanishing-polynomial theory for conjugacy classes, Bray–Whaples needs no algebraicity: it uses only the conjugation rule and the class bound, both of which hold for arbitrary elements of an arbitrary division ring.

Is the polynomial

g

independent of the order in which the nodes are processed?

Yes — uniqueness says so. The intermediate factors $t - d_{i}$ do depend on the order, so two different orderings give visibly different factorisations of the same $g$ . That is the same non-uniqueness of factorisation seen in Wedderburn's theorem.

How does this interact with the Gordon–Motzkin dichotomy?

It shows the finite side of the dichotomy is always realised. $(16.12)$ says a degree- $n$ polynomial has at most $n$ roots or infinitely many; Bray–Whaples constructs, for each $n$ , a degree- $n$ polynomial with exactly $n$ roots, provided $D$ has at least $n$ conjugacy classes.

Is the Vandermonde criterion an equivalence?

No. Pairwise nonconjugacy implies invertibility, but not conversely. Exercise 16.5 shows $V (a, b, c)$ is invertible whenever $a, b, c$ do not all lie in a single conjugacy class, and gives an explicit criterion for failure when they do. The general theory is Lam's Vandermonde theory over division rings.

What is the analogue for skew polynomial rings?

Replace conjugacy by $σ$ -conjugacy, $c \mapsto σ (d) c d^{- 1}$ , and the Vandermonde matrix by the Moore matrix built from $σ$ -twisted powers. Pairwise non- $σ$ -conjugate nodes then give an invertible Moore matrix; this is the algebraic core of Gabidulin and skew cyclic code constructions.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §16, (16.13) and Exercises 16.4–16.5 (pp. 267–272).
U. Bray and G. Whaples, “Polynomials with coefficients from a division ring”, Canadian Journal of Mathematics 35 (1983), 509–515.
T. Y. Lam, “A general theory of Vandermonde matrices”, Expositiones Mathematicae 4 (1986), 193–215.
B. Gordon and T. S. Motzkin, “On the zeros of polynomials over division rings”, Transactions of the American Mathematical Society 116 (1965), 218–226.
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.

AI Suggested Questions

Compute the Bray–Whaples polynomial for three pairwise nonconjugate quaternions none of which is central.
Give an explicit example of three distinct quaternions lying in one conjugacy class whose Vandermonde matrix is nonetheless invertible, or prove none exists.
State and prove the $σ$ -twisted version of Bray–Whaples for $D [t; σ]$ .
How does the recursion behave numerically over the real quaternions when two nodes are nearly conjugate?
Is there a Lagrange-style closed formula for the Bray–Whaples polynomial, and if not, what obstruction prevents it?
Characterise the left ideals of $D [t]$ that arise as vanishing ideals of finite sets of pairwise nonconjugate elements.

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Why the naive product fails

The correction, in one line

06Key Results

07Proof Techniques and Method

08Worked Example

Two nodes in the real quaternions

Three nodes, and a noncentral correction

09Process and Workflow

10Comparison and Classification

11Relationship Map

12Applications and Industry Use

13Computational Notes

14Failure Modes and Common Mistakes

15Quick Reference

16Frequently Asked Questions

17Related KEVOS Topics

18References

19AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Key Results

Proof Techniques and Method

Worked Example

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Computational Notes

Failure Modes and Common Mistakes

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions