Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Say you're given a ordered set of $n$ relatively prime elements, $a_1,\dots,a_n$ in a principal ideal domain $D$. If I relabel these elements $a_{11},\dots,a_{1n}$ in the same order, is it possible to find some remaining $a_{kj}$ in $D$ such that $(a_{kj})$ is an invertible $n\times n$ matrix over $D$?

share|improve this question
    
In a more general setting, this became known as Serre's problem, and later the Quillen-Suslin Theorem. –  Gerry Myerson Oct 28 '11 at 22:23
    
@Gerry: Can you indicate how this is related to Quillen-Suslin? –  Soarer Oct 29 '11 at 4:38
    
@Soarer, sorry, it's a bit out of my depth. Some statements of the problem/theorem don't look anything like Aria's qiestion, but if you type Quillen Suslin matrix into a search engine you will probably find some references that bring out the relation. –  Gerry Myerson Oct 29 '11 at 8:17
    
@Soarer, Here's a reference: Moshe Roitman, Completing unimodular rows to invertible matrices, J. Algebra 49 (1977), no. 1, 206–211, MR0453779 (56 #12033). From the review: Quillen's solution of Serre's problem implies that if $R$ is a commutative ring, then any unimodular row over $R[x]$ which contains a unitary polynomial is completable to an invertible matrix. The author shows.... Other related results on this completion problem are discussed. –  Gerry Myerson Oct 31 '11 at 6:11

2 Answers 2

We can prove this by induction. For $n=1$, the set $\{a_1\}$ is relatively prime if and only if $a_1$ is a unit, and thus $(a_{11})$ is invertible.

So assume that we can extend any set of $n$ relatively prime elements to an invertible matrix, and let $n+1$ relatively prime elements $a_1,\dotsc,a_{n+1}$ be given. Let $g=\gcd(a_1,\dotsc,a_n)$. Then we can write $a_i=gb_i$ for $1\le i\le n$, with the $b_i$ relatively prime, and we can extend the $b_i$ to an invertible $n\times n$ matrix $B$. If we multiply the first row of $B$ by $g$, we can place it into the upper left $n\times n$ block of the matrix $A$ to be constructed, and we'll have $a_i=gb_i$ in the right place for $1\le i\le n$. Now place a multiple of the first row of $B$ into the $n+1$-th row of $A$, with the multiplier $r$ yet to be determined, and fill the $n+1$-th column with zeros except for the first and last elements $a_{n+1}$ and $s$, with $s$ yet to be determined. Now the determinant of the upper left $n\times n$ block is $g\det B$, and the determinant of the lower left $n\times n$ block is $(-1)^{n+1}r\det B$, so by Laplace expansion the determinant of the entire matrix $A$ is

$$\det A=sg\det B+ a_{n+1}r\det B=(sg+ ra_{n+1})\det B\;.$$

Since $a_1,\dotsc,a_{n+1}$ are relatively prime, $g$ and $a_{n+1}$ are relatively prime. Thus we can choose $r$ and $s$ such that the expression in parentheses is $1$ and $\det A = \det B$. But a matrix over a commutative ring is invertible if and only if its determinant is invertible (see Do these matrix rings have non-zero elements that are neither units nor zero divisors? and necessary and sufficient condition for trivial kernel of a matrix over a commutative ring). Thus, since $B$ is invertible, $A$ is invertible.

share|improve this answer

There is much written on this and related problems, e.g. it is the special case $\rm\:k=1\:$ of below.

MR 82k:15013
Gustafson, William H.; Moore, Marion E.; Reiner, Irving
Matrix completions over Dedekind rings.
Linear and Multilinear Algebra 10 (1981), no. 2, 141--144.

Let d be any element of the ideal generated by the k by k minors of a k by n matrix M over a Dedekind domain R . The authors show that M can be "completed" to form the top k rows of some invertible n by n matrix of determinant d .

In fact, their proof works if R is any Prufer domain whose finitely generated ideals can be generated by $\ 1\ 1/2\ $ elements; merely precede their lemma by the proof of the formula $\rm\ A \oplus B \cong R \oplus AB\ $ given in Theorem 4.1 of R. C. Heitmann and the reviewer [Rocky Mountain J. Math. 5 (1975), 361--373; MR 52 #3141].

Reviewed by Lawrence S. Levy 91k:15028

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.