Burrow 1973
Burrow 1973
Burrow 1973
Author(s): M. D. Burrow
Source: The American Mathematical Monthly, Vol. 80, No. 10 (Dec., 1973), pp. 1129-1131
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2318550 .
Accessed: 15/02/2015 02:21
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to
The American Mathematical Monthly.
http://www.jstor.org
MaterialforthisDepartment of Mathematics,
shouldbe sentto David Roselle,Department
BatonRouge,LA 70803.
LouisianaState University,
ao0+a,Th+ *+anTno=O.
1129
< degg(x) < n and so the theoremholds. Suppose now that g(T)V # 0. Let
U = {x: g(T)o = 0}. Then U is a propersubspace of V and dimU = r < n. By
theinductionalassumptionthereis a polynomialml(x) of degreeless thanor equal
to r such that m1(T)U = 0. Moreover,TU c U, since g(T) commuteswith T,
so that U is an invariantsubspace of V. Now considerthe quotientspace V/U.
We have dimV/U = n - r < n and so, by the inductionalassumptionagain, there
is a polynomialm2(x) of degree < n - r such that m2(T)V/U = 0. This means
that m2(T)V c U, so that m1(T)m2(T)V = 0.
Writingh(x) = m1(x)m2(x) we have h(T)V = 0 so that if m(x) is the minimal
polynomial
degm(x) < degh(x) = degml(x) + degm2(x) ? r + n - r = n.
Thus the theoremholds in all cases and the proofis complete.
NOTE: If ml(x) and m2(x) are the minimalpolynomialsof T restricted
to U
then h(x) = m1(x)m2(x)coincides with the minimalpoly-
and V/U respectively,
nomial of T.
THEOREM 2. Let V be a vectorspace of dimensionn over a field F, and let T
be a linear transformation on V. If the minimal polynomial m(x) = (p(x))s,
wherep(x) is irreducibleand of degreer, thenr dividesn.
Proof. Let B = {a1, x2, c* n} be a basis of V. We assertthatthereis one of
thesevectors,which,withno loss of generality, may be taken to be ac,, such that
the set G = {C,XTaj, ..., T"-'oc,} is linearlyindependent.
Suppose the statementis false; then for everya in V, p(T)Jm?= 0 for some
je < s. This is so because the annihilatingpolynomialof minimaldegree (given
here by the assumed dependence)of a vectordividesany otherannihilatingpoly-
nomial,and in particular,then,divides(p(x))s. Thus p(T)s- 1V = 0, a contradiction.
Let U be the subspace of V generatedby the set G. Then dimU equals rs. If
U = V, thenrs = n so thatr dividesn and we are finished.To completetheproof
we use an inductionon n. Suppose that U # V. First note that TU c U, since
foreveryj < rs - 1, T(T(x 1) = T!+ 1x1is a basiselementof U, whereas(p(T))sa1 = 0
gives Tcal in termsof the basis elements.We go now to the quotientspace V/U.
Since TU c U we can considerthetransformation T1 inducedon V/U by T. Since
(p(Ti))s also annihilatesV/U, the minimalpolynomialof T1 is (p(x))t for some
t ? s. Now dimV/U = n - rs < n so that the inductionalhypothesismakes r
dividen - rs and thisimpliesthatr dividesn, completingtheproof.
COROLLARY. Let V be a vectorspace of dimensionn overa fieldF and let T be
a linear transformation on V. If the minimalpolynomialm(x) = (p(x))s, where
p(x) is irreducibleand of degree r, thendeterminant(xI- T) = (p(x))ntr.
Proof. Assumethatthestatementis trueforall spaces of dimensionless thann.
But then
det(xI - T) = det(xI - T1)det(xI - T2) = (p(x))nlr.
whereeach pi(x) is irreducibleof degree ri. Let Vi be the null space of (pj(x))5i
and let ni be the dimensionof Vi, thenfor i = 1,2, ', k we have that ri dividesni
and
f(x) = (pi(X))n,Ir, lI. (P (X))nk/rk
This is immediatefromTheorem2 and its corollary,since V is the directsum
of the Viand each (pi(x))si is the minimalpolynomialof the transformation Ti in-
duced on Vi by T. By Theorem 1 we have risi < ni so that si < nJIriand hence
m(T)V = 0 implies thatf(T)V = 0. Thus the Cayley-Hamiltontheoremfollows
as a consequence.
Note thatthe vectoro1 of Theorem2 is annihilatedby no polynomialof degree
< rs, and in fact that its order(i.e., annihilatingpolynomialof minimaldegree)
is the minimalpolynomialm(x). For the generalcase, in each Vi thereis a vector
ci whose order is the minimalpolynomial(pj(x))si and the vectors(xi given by
i= ( + + (x, for j = 1,2, ..,k, have orders flHj=I(pm(x))si: Thus a-=k
has order m(x), the minimalpolynomialof T.
H. G. JACOB,University
of Massachusetts