Solution of Polynomials by Real Radicals

Author(s): I. M. Isaacs
Source: The American Mathematical Monthly, Vol. 92, No. 8 (Oct., 1985), pp. 571-575
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2323164 .
Accessed: 15/11/2013 11:40
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 [email protected].
.
Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to
The American Mathematical Monthly.
http://www.jstor.org
This content downloaded from 168.176.5.118 on Fri, 15 Nov 2013 11:40:50 AM
All use subject to JSTOR Terms and Conditions
1985] NOTES 571
Acknowledgement. The author wishes to thank N. Linial for his helpful remarks during an early presentation of
this subject in Jerusalem, 1982.
References
1. R. L. Graham and H. 0. Pollak, On the addressing problem for loop switching, Bell System Tech. J., 50
(1971), 2495-2519; see also "On embedding graphs in squashed cubes", in Graph Theory and its Applications, Y.
Alavi et al. ed., Lecture Notes in Math., Springer-Verlag, vol. 303 (1972) 99-110.
2. J. Kasem, Ph.D. Thesis, Hebrew University, Jerusalem, 1984.
3. G. W. Peck, New proof of a theorem of Graham and Pollak, Discrete Math., 49 (1984) 327-328.
4. B. Reznick, P. Tiwari, and D. B. West, Decomposition of product graphs into complete bipartite subgraphs,
U. of Illinois, Urbana, 1984.
5. H. Tverberg, On the decomposition of
K,,
into complete bipartite graphs, J. Graph Theory, 6 (1982) 493-494.
6. J. Zaks, Bounds of neighborly families of convex polytopes, Geom. Dedicata, 8 (1979) 279-296.
NOTES
EDITED BY SABRA S. ANDERSON, SHELDON AXLER, AND J. ARTHUR SEEBACH, JR.
For instructions about submitting Notes for publication in this department see the inside front cover.
SOLUTION OF POLYNOMIALS BY REAL RADICALS*
I. M. ISAACS
Department of Mathematics, University of Wisconsin, Madison, WI 53706
Let F be a field and let f E F[x] be a polynomial. Recall that f is said to be solvable by
radicals if f splits over some extension field E D F for which there exists a chain of fields
F=
Fo
cF,
c
...
c
Fk
=
E
such that
J,
= -
1[i
] for 1 < i < k, where
a,
has some power lying in
J,
1 Such an extension
field E is called a repeated radical extension of F.
By the work of Galois, we know that (in characteristic zero) every polynomial of degree < 4 is
solvable by radicals, and that over suitable fields (the rational numbers Q, for instance), there
exist, for every integer n > 5, some irreducible polynomials which are solvable by radicals and
some which are not.
Let us now limit the discussion to the case F= G. What is the situation with respect to
solvability by radicals if we require that we work entirely inside the real numbers R? We shall say
that a Eli R is a real radical element if it lies in some repeated radical extension E of a with
E c R. In other words, a is constructible from the rationals by some combination of multiplica-
tions, additions and extractions of real nth roots.
A polynomial f
E QG[x] is solvable by real radicals if all of its (complex) roots are real radical
elements. As we shall see, a polynomial which is solvable by radicals is rarely solvable by real
radicals even given that all of its roots are real. More surprisingly, we have the following:
THEOREM. Suppose f E Q[x] is an irreducible polynomial which splits over R. (In other words,
its roots are all real.) Iff has any root which is a real radical element, then the degree off is a power
of 2 and the Galois group off over C is a 2-group.
The fact that irreducible cubics are never solvable by real radicals is classical. This phenome-
non is sometimes referred to as the "casus irreducibilis". (See the discussion of this in [2].)
*Research partially supported by an N.S.F. grant.
This content downloaded from 168.176.5.118 on Fri, 15 Nov 2013 11:40:50 AM
All use subject to JSTOR Terms and Conditions
572 I. M. ISAACS [October
A striking consequence of the theorem is the following:
COROLLARY. Let f be as in the theorem. Then all of the roots of f lie in a repeated square-root
extension of Q and thus they are constructible with compass and straightedge.
Proof. The Galois group, because it is a 2-group, necessarily has a chain of subgroups, each of
index 2 in the preceding, which begins with the whole group and ends with the trivial subgroup.
By the fundamental theorem of Galois theory, the corresponding fields form a tower of quadratic
(and hence square-root) extensions, starting with C and ending with a splitting field for f. M
The key step in the proof of the theorem is the following result in which we use the notation
IL:
Fl to denote the
degree
of the field extension F c L.
PROPOSITION. Let F c E be subfields of R and assume that E = F[a] with all c F. If
F c L c E with L Galois over F, then IL: Fl < 2.
The proofs of the proposition and the theorem make heavy use of the theorem on "natural
irrationalities" (Theorem 29 of [1]). We state this here for the convenience of the reader after a
brief review of the relevant notation. If E and L are subfields of some common field, their
compositum
(E,
L) is the (unique) smallest subfield containing both E and L. It is most simply
defined as the intersection of all subfields which contain the two given fields. For fields F c
E, we
write Gal(E/F) to denote the associated Galois group.
THEOREM (Natural irrationalities). Let F, E, and L be subfields of some field Q and suppose
E D F is a Galois extension. Let K =
(E, L), the compositum. Then K is Galois over L and
Gal(K/L) Gal(E/E n L). In particular,
IK:LI
= E: E n LI. (See Fig. 1.)M
K
L
E
F
FIG. 1
COROLLARY 1. In the above situation, assume IL: Fl
<
oo. Then
IK:
El
=
I L: E n
L1.
Proof. Divide both sides of the equation
IK: LI
=
IE: E n LI into the number
IK:
E n LI. M
A somewhat more interesting corollary permits us to draw another line in the natural
irrationalities lattice diagram.
COROLLARY 2. Let F, L, E and K be as before with E Galois over F and K =
(E, L). Suppose
L c M c K. Then M
=
(L, E (n M).
The point here is that by drawing the line segment down from M, parallel to KE, we define a
new node, identified with the field E n M. The resulting diagram, if interpreted as a lattice
diagram, would imply M = KL, E n M). Corollary 2 asserts that this is a valid inference. Of
course, the diagram is not a proof. (See Fig. 2.)
This content downloaded from 168.176.5.118 on Fri, 15 Nov 2013 11:40:50 AM
All use subject to JSTOR Terms and Conditions
1985] NOTES 573
K
M
E L
F
FIG. 2
Proof of Corollary 2. Let M1 = (L, E n M). It is clear that M1 c M, E n) M1 = E n M and
<M1, E)
= K =
<M, E).
Applying natural irrationalities with M in place of L yields that
1K:
Ml
=
IE:
E n
Ml
and
similarly, 1K:
Mll
=
IE:
E n
Mll.
We conclude that
iK: Ml
=
IK: Mll,
and since
M1
c
M, we have
M1
= M as desired.
U
We will also need the following lemma which is probably well known.
LEMMA. Let F c E be fields with E = F[ a] where a' e F, and suppose F contains a primitive
nth root of unity.
Let IE:
Fl
= m. Then
atm c F.
Proof. Let f E
F[x]
be the minimal polynomial of a over F so that deg(f)= m. Write
an = a E F. Then f divides
Xn
- a in F[x] and so all the roots of f in a splitting field are nth
roots of a. Each root, therefore, has the form a8 where 8 is some nth root of unity. It follows
that the constant term of f is equal to + am, where ? is the product of all the 8's and so is an
nth root of unity. Thus
-
e F and we conclude that aim E F as required. E
Proof of the proposition. Adjoin a complex primitive n th root of unity to each of F and E to
obtain the fields F* and E*, respectively. Note that E* = <F*,
E)
and let K = (F*, L) and
M = E n K so that we have the lattice diagram as shown. (See Fig. 3.)
Since F* is a splitting field for the polynomial
Xn
- 1 over F, we see that F* 2 F is Galois
and in fact Gal(F*/F) is abelian. By Corollary 2, M = (L, F* n
M)
and we observe that L is
Galois over F by hypothesis and F* n M is Galois over F by the fundamental theorem of Galois
theory, since Gal(F*/F) is abelian and so Gal(F*/F* n M) is a normal subgroup.
Since each of the fields L and F* n M is Galois over F, it follows that any F-automorphism
of any field which contains <L,
F* n
M>
leaves each of them (setwise) invariant. Since the
compositum (which is equal to M) is uniquely determined, it follows that it too is invariant under
all these automorphisms and therefore M is normal, and hence Galois over F.
By natural irrationalities applied to the situation
<F*,
M) = K with F* Galois over F, we
conclude that K is Galois over M. We can therefore apply natural irrrationalities to the situation
<K, E)
= E* and we conclude from
Corollary
1
that
IE*: KI
=
IE:
Ml.
Now write m =
IE*: KI. Since E* =
K[a],
an E K and K contains a primitive nth root of
unity, we conclude from the lemma that a'm E K. Since, of course, a'm E E, we see that al' r M.
We claim, in fact, that M =
F[ a'].
To see this, write M1 =
F[am] and note that E = M1[a] and a is a root of the polynomial
x a"' E
M4[x].
It follows that IE:
Mll
< m. However, M1 c M and IE:
Ml
=
IE*: KI
=
and it follows that M1 =
M, as claimed.
Write f8
=
am. The situation now is that F c M c
R,
M
=
F[f3], /7 E F and M is Galois
This content downloaded from 168.176.5.118 on Fri, 15 Nov 2013 11:40:50 AM
All use subject to JSTOR Terms and Conditions
574 I. M. ISAACS [October
E*
A ~~E
K
F* ~ w
F
FIG. 3
over F. Let a E Gal(M/F). Then
(C,a)n = (3n)O
=
n
and so B=
f,3
where 8 c M is some nth root of unity. Since M c R, however, 8 = +1 and
thus /3a = +?f for every element a c Gal(M/F). Since M = F[f8], a is determined by its action
on
f#
and it follows that
IGal(M/F)I
< 2.
We now have
IL:FI
<
IM:FI
=
IGal(M/F)I
< 2.M
Proof of the theorem. Suppose f E Q[x] is irreducible and let S c R be a splitting field for
f.
Let G
=
Gal(S/C). Our goal is to show that G is a 2-group. Since deg f divides IS: 1
1
=
IGI, this
will yield that degf is a power of 2, as required.
Suppose that G is not a 2-group and choose an odd prime divisor p of
IGI.
Let N C G be the
subgroup generated by all subgroups P C G with
IPI
= p. Since this collection of subgroups is
permuted by conjugation by elements of G, it follows that N is normal in G. Also, by Cauchy's
theorem, there do exist subgroups P C G of order p, and hence N > 1.
We are given the real radical element a with f (a)
= 0, and so a c S and we write T = Q[a].
Let H
=
Gal(S/T), so that H c G is the subgroup corresponding to T via the fundamental
theorem of Galois theory. We claim that N cannot be contained in H. To see this, let
L
=
Fix(N), which is the field corresponding to N. Then L is normal over
a
since N-<G. Also,
L <
S because N > 1. If N c
H,
then L D T and so a c L. Because f is irreducible and L is
normal over 1, we conclude that f splits over L and this is a contradiction since L < S.
We have established that N ? H and thus P
Z
H for some subgroup P c G of order p. Let
F
=
Fix(P). Since P
Z
H, we have F
;
T and so a
4
F.
By our assumption on a, there exists a tower of fields
it=QoCQ1
...
-CQmCR
with a C Qm and
Qi
=
QJ-y1[y]
for some element
y,
with a power in Q11. Construct a new
tower of fields starting with F by setting
Fo
= F and
F,
=
Fi- [Y i]. Then
Q,
_
F,
_ R for all i
and, in particular, a c
EFm.
Since a
4
Fo,
we can choose an integer r > 0 so that a
4
F,-,
and
a E Fr.
This content downloaded from 168.176.5.118 on Fri, 15 Nov 2013 11:40:50 AM
All use subject to JSTOR Terms and Conditions
1985] NOTES 575
Now
IS: Fl
=
jPI
= p, a prime, and thus F is a maximal subfield of S. Also,
F
cF_1
n S c
F,
n S c S.
Since a 0
FI,
nS and a e Fr nS, we conclude that F= F l nS and FI n S = S so that
S C Fr.
Write E =
(S,
F]_ ). Then
F,-
c c F c F_ Also, S is Galois over 1 and so natural irrational-
ities applies and we conclude that E:
F,l-
=
IS:
Fj = p and E is Galois over
Fr-1
However,
F
-i[yr]
=
Fr
c
R and some
power
of Yr lies in
FI_
Our
proposition
now
applies
and
yields
p =
E: Fr 2. This contradiction
proves
the theorem.
(See Fig. 4.)
E
F
E
S
F
FIG. 4
We close with the observation that solvable polynomials with real roots but which are not
solvable by real radicals seem to abound. For example, for any prime p, the polynomial
f(x)
= -
2px
+
p
has this property. (Note that it is irreducible by the Eisenstein criterion and it has three real roots
since f(0)
>
0 and f(1)
<
0.) It is amusing to solve this polynomial by Cardan's method to see
where nonreal numbers come in.
References
1. E. Artin, Galois Theory, Notre Dame Math. Lectures, 1944.
2. B. L. van der Waerden, Modern Algebra (vol. 1, 2nd edition), Ungar, New York, 1953.
A THEOREM IN COMBINATORICS AND WILSON'S THEOREM
HARAGAURI N. GUPTA
Department of Mathematics and Statistics, University of Regina, Regina, Canada S4S OA 2
The purpose of this note is to state and prove a theorem in combinatorics and deduce from it
the famous Wilson's theorem as a special case.
THEOREM 1. Let B(n, r), 1 < r < n, denote the number of distributions of n distinct objects
among r ordered boxes, none remaining empty. Then,
(i) if n is prime, then n divides B(n, r) for all r > 2,
(ii) if n is prime, then n divides B(n - 1, r) + (-1)r, for all r, 1 < r < n -1.
Proof. We shall not require the explicit formula for B(n, r), but will make do with the
following identities I and II, both of which can be established solely by combinatorial arguments:
This content downloaded from 168.176.5.118 on Fri, 15 Nov 2013 11:40:50 AM
All use subject to JSTOR Terms and Conditions