P. Landrock and 0. Manz
Nagoya Math. J.
Vol. 125 (1992), 33- 51
S YM M ETRIC FORM S , ID EM P OTEN TS
AN D
IN VOLU TARY AN TΠ S OM ORP H IS M S
PETER LANDROCK AND OLAF MANZ
Introduction
Let G be a finite group, zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
F a field and M an irreducible F[G ]- module.
By we denote the F - linear involutary antiautomorphism of F[G], induced by inversion on group elements. Suppose t h at char (F) Φ 2. We then
show th at M carries a non- singular G - invariant symmetric bilinear form
with values in F if and only if there exists a A- invariant idempotent
e e F[G] which generates the projective cover of M. This extends earlier
results of W. Willems [Wi]. The assertion is not true if c h a r ( F ) = 2.
We even consider this question in th e class of those finite- dimensional
algebras which admit an .F- linear involutary antiautomorphism r and
which are symmetric with respect to a r- invariant symmetric functional.
Besides group algebras, also involutary semi- simple F - algebras belong to
t h at class.
Λ
In th e final part of this paper, we let G be represented irreducibly
and orthogonally on a real vector space M. We then show t h at there
is a relationship between G - orbits on the un it sphere of M and idempotents e e R[G ] such t h at M ^ R[G ]e and e = e. This has some connection to a problem in Coding Theory, namely to find G - orbits on the
un it sphere whose minimal Euclidian distance is considerably large.
§ 1.
Involutary and symmetric algebras
Let A be a finite- dimensional F - algebra over a field F. We set A =
A/ J(A), where J(A) denotes th e Jacobson Radical of A. ΐn th e following,
each A- module should be understood as a finitely generated A- Ze/ ΐ- module.
1.1 LEMMA. Let e, f e A be primitive d
ί empotents such that ef Φ
Then the following assertions hold.
0.
Received Ju ly 9, 1990.
33
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
34
PETER LAN D ROCK AN D OLAF M AN Z
(a) zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
The map Ae —> Af, ae^- ^ae f, is an A- module isomorphism.
(b) fAe = fAf (as F- vector spaces).
(c) fAe is a local algebra sί omorphic to fAf, via the algebra- isomorphism fae H- > foe • / (a e A).
Proof, (a) Since ~ef Φ
0 and both Ae and A/ are irreducible, the
map
Ae- +Af,
ae- >ae- f,
is an isomorphism. Consequently, Ae ^ Af via αβ *- >ae- f (cf. [H B; VII,
11.6]).
(b) I t follows from (a) th at
fAf = Hom A(A/ , A/ ) ^ Hom A(A/ , Ae) = / Ae
(as F- vector spaces).
(c) By (a), the map fae *- +fae- f (a e A) is a vector space monomorphism between fAe and fAf, and (b) implies th at it even is an isomorphism. The assertion now follows from
(fae)(fbe)f = (fae)f. (fbe)f
(a, be A).
D
If A admits an F - linear involutary antiautomorphism τ , we call (A, τ )
an involutary F- algebra. Observe th at τ leaves J(A) in varian t and th us
τ induces an involutary antiautomorphism on A. Let V be an A- module
over an involutary F - algebra (A, τ ) . An F - bilinear form
< , >: Vx
V- >F
is called a τ - form if it is non- degenerate and if
(av, w} = yζ , aτ w}
for all ae A, v, w e V
(i.e. th e adjoint mapping of a with respect to <( , ) is given by aT ).
1.2. LEMMA. Let (A, τ) be an involutary F- algebra and M an irreducible A- module.
(a) / / there exists a primitive idempotent f e A such that M = Af and
f f φ 0, then there even exists a primitive idempotent e e A such that M =
Ae and eτ = e. Moreover, e can be chosen an the 1- element in fAfτ .
(b) Let M carry a τ - form < , ) . If M contains a non- sί otropίc vector
x, then there exists an idempotent feA
which satisfies M— Af, fτ f φ θ
and fx = x.
τ
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
SYMMETRIC FORMS, IDEMPOTENTS
35
Prcof. (a) By Lemma 1.1 (c), th e mapping
fAΓ- + fAf, faf*~faf*.f,
is an algebra- isomorphism. Let eefAfτ be th e preimage of fefAf.
Then
τ
e is a primitive idempotent and is th e 1- element of fAf\ Since fAf is
τ
r- invariant, we also have e = e. F inally
ae ι- » αe • / = af (a e A)
yields Ae = A/ , hence M ~ Ae.
(b) We consider th e map A —> M, α (- > αx. Then there exists a primitive idempotent / e A such th at A(ϊ — f) is its kernel. Consequently,
M = A/ and / x = x. Since
0φ
<x, x> = </x, fx) = </</*, x> =
f / ^ 0 follows.
•
We denote by P(V) th e projective cover of an A- module V.
1.3 TH EOREM. L et (A, τ) 6β an inυ olutary F- algebra. Suppose that
the irreducible A- module M carries a symmetric τ - form < , >. If c h ar( F )
Φ 2, then there exists a primitive idempotent e e A swcΛ that eτ = e and
P(M) s Ae.
Proof. Since char (F ) ^ 2, th e symmetric form < , > is n ot symplectic.
Therefore, Lemma 1.2(b) applies and Lemma 1.2(a) yields an idempotent
e e A such th at eτ = e and M ^ Ae. H ence P(M) = Ae.
Π
We shall see in Example 3.1 th at th e hypothesis c h a r ( F ) Φ 2 is n ot
superfluous.
It is well- known t h at an idempotent d = d + J(A) can be lifted to an
idempotent e e A which is a polynomial in d with integer coefficients.
This observation applies to our question about τ- invariant idempotents as
follows.
1.4 PROPOSITION . Suppose that (A, τ ) is an involutary F- algebra. If
d = d + J(A) is a τ - invariant idempotent in A, then there exists a τ nί varίant idempotent e e A such that e — d.
Proof. We may assume t h at d is τ- invariant. Otherwise namely d
τ
τ
2
can be replaced by dd% because dd = dd — d = d. Arguing by induc2
tion, we may as well assume th at J( A) = 0. We set e = 3d2 — 2d\
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
36
PETER LAN D ROCK AN D OLAF M AN Z
Then e2 = zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
e + (d 2 - d)\ 2d + l)(2d - 3) = β, because d2 - de J(A). Since
Γ
e = e and e = d, th e assertion follows.
•
It is clear t h at in Proposition 1.4, J(A) can be replaced by any τin varian t nilpotent ideal / .
A finite- dimensional F - algebra is called symmetric if there exists a
functional ω e H om F (A, F) which satisfies φ ( ab) = ^>(6α) for all a, be A
and which does not contain any non- zero right- (or left- ) ideal of A in
its kernel. We call φ a symmetric functional for A. (Equivalently, A can
be characterized by a non- degenerate symmetric associative F - bilinear
form ( , ): A X A ~- > F. Observe t h at then (α, b) = φ ( ab). But we prefer
to work with the functional .ψ )
Let (A, τ) be an involutary F - algebra which is symmetric with respect
to ψ e H om F (A, F). We call (A, τ , φ ) a. symmetric involutary algebra if
φ ( a τ ) = φ ( a)
for all a e A.
1.5 LEMMA. Let (A, τ, φ ) be a symmetric involutary
e e A is an d
ί empotent satisfying eτ = e, then
(υ , w} = φ ( vw τ ),
F- algebra. If
υ , w e Ae,
defines a symmetric τ - form on Ae.
Proof. Suppose t h at 0 = (v0, w} = φ {\ι w τ ) for all w = ae e Ae. Since
eT = e, we obtain 0 = φ ( v ϋ eaτ ) — φ ( v Qaτ ) for all ae A, and £> contains the
right- ideal v0A in its kernel. Thus vQ = 0 and < , > is non- degenerate.
Since φ is τ- invariant, we have
(v, w} = <p((vwT ) τ ) = 9?(α;ι;Γ) = <w;, u>
and < , > is symmetric.
Let ae A.
Then
<αu, w} = <p(avwτ ) = φ ( vw ta) = φ ( v(a T w) 7) — <u, aτ uf)
and < , > is a τ- form.
•
The following is inspired by [CR; 9.17], where the element z is defined
for semi- simple algebras over a splitting field.
1.6 L E M M A. Let A be a symmetric algebra with respect to ψ e
H o m F ( A, F), and let M be an irreducible A- module with character β e
H o m F ( A, F). For dual bases {α lr
, un} and {bl9 *• - ,&«} of A (i.e. φ {a tb^
= δij),
we set
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
SYMMETRIC FORMS, IDEMPOTENTS
z = Σ βiaΰ bi
We then have:
(a) β( a) = φ ( za) for all
(b) z e Z(A) Π soc (A).
(c)
37
6A .
aeA.
Let e e A be a primitive idempotent such that M = Ae.
Then the
map Ae - > soc (Ae), ae >- > aez, is an A- module isomorphism.
(d) Suppose t h at (A, τ , φ ) is a symmetric involutary F - algebra and
T
x
assume t h at β is r- invariant (i.e. β( a ) — β( a) for all a e A). Then z = z.
Proof, (a) Observe t h at φ ( za 3) = Σ?= i β( )βι Φψia
,/ = 1,
3) = β ( a ά ) for
, n. Since {α 1?
, α j is an F- basis of A, the assertion follows,
(b) Let c e J(A). By (a), φ ( zc) = β( c) = 0 and φ contains the rightideal zJ(A) in its kernel. Thus z e ann (J(A)) = soc (A). If α, 6 e A, then
Therefore,
again (a) shows φ ( a zb) — <p(zb a) = /3(6α) = β( ab) = φ ( zab).
(α^ — 2α)A < ker(9) and az = a α for all α e A.
(c) Let ε e A be any lift of the Wedderburn idempotent ε e A, corresponding to M. F or each α e A, we th us obtain
•
(ψ za)
= j9(α) = j8(αe) = ^(^α ε) =
φ ( ε za) .
This implies t h at ε z = z and β2 ^ 0. Since soc (Ae) is irreducible and
z e Z(A) Π soc (A), the map Ae - » soc (Ae), αβ ι- > aez = aze, is an isomorphism.
(d) Since φ is r- invariant, we have φ ( a\ b)) — φ ( (a,ibj) τ ) — φ ( a ί bj) — δ i}
and {αί,
, α^}, {bl, - - , bT n) are dual bases of A as well. Since also β is
assumed to be r- invariant, we obtain
We now apply (a) to both z and z\
for all α e A, i.e. z = zτ .
Γ
Consequently, (ψ za) — β( a) = ^(2 α)
Π
If (A, τ ) is an involutary F - algebra and M an A- module, then M * =
H om F (M, F) becomes an A- left- module by
(aa)(m) = a(aτ m),
a e A, m e M, a e M *.
The module M * is called the d^αZ module of M (with respect to τ). It is
easy to see t h at M is self- dual (i.e. M * = M ) if and only if M carries a
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
38
PETER LANDROCK AN D OLAF MAN Z
τ- form (cf. [H B; VII, 8.10]).
Our next aim is to "lift" symmetric r- forms from zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQ
P(M) to M.
1.7 PROPOSITION . Let (A, τ , φ ) be a symmetric involutary F- algebra.
For a primitive idempotent e e A, suppose that Ae carries a symmetric τ - form
( , >. Then M ^ Ae = soc(Ae) as well carries a symmetric τ - form.
Proof
module
We may assume th at Ae is reducible and consider the subsoc (Ae) 1 = {veAe\ (v, w} = 0 for all w e soc (Ae)}.
Since dim (soc (Ae) 1) = dim (Ae) — dim (soc (Ae)) = dim (J(A)e), we conclude soc(Ae) 1 = J(A)e.
As Ae is assumed to carry a τ- form, Ae is self- dual and th us M= M *.
If β denotes th e F - eharaeter of M, it follows t h at β( a τ ) — β( a) for all ae A.
F or dual bases {α£} and {bj} of A, we consider z = ίΣ β( ai)°t a n ( i define
a bilinear form { , ) ; on M by
Since by Lemma 1.6(c), x H- > XZ is an isomorphism from Ae onto soc (Ae),
and since soc (Ae) 1 = J(A)e, the form < , >' is well- defined. Suppose t h at
0 = (x, 3>o>' = (xz, y0} for all x e Ae. Thus xz run s through the whole of
1
soc (Ae) and y0 e soc (Ae) = J(A)e. Therefore, yQ = 0 and < , y is nondegenerate. Since obviously (ax, y}; = (x, aτ y}' for all ae A, itrem ains
to show t h at ( , >' is symmetric. By part (b) and (d) of Lemma 1.6,
T
z = ze Z(A), and therefore
(x>y)' = (xz, y} = (zx, y) = <x, zT^> = <x, 2:,y> = <y2, x> = (y, x}'
for all x,y e Ae.
This completes the proof.
•
We are now able to formulate our main result.
1.8 TH EOREM. Let (A, r, φ ) be a symmetric involutary F- algebra, and
M an irreducible A- module. If char (F) Φ 2, then the following statements
are equivalent.
(1) P(M) carries a symmetric τ - form.
(2) M carries a symmetric τ - form.
(3)
(4)
There exists an idempotent ee A such that M = Ae and eτ = e.
τ
There exists an idempotent ee A such that P(M) ^ Ae and e = e.
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
SYMMETRIC FORMS, IDEMPOTENTS
Proof. (1)
(2)^(3):
(3)= >(4):
(4)= φ (l):
39
z=> (2): Proposition 1.7.
Theorem 1.3.
Proposition 1.4.
Lemma 1.5.
Π
Recall t h at th e symmetry of A is not needed for (2) => (3) and (3) Φ
(4), and t h at char (F) φ 2 is only relevant for (2) => (3).
§ 2. SomezyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
applications
As our main application, we consider the group algebra F[G] of a
finite group G over the field F. Then
= Σ a8g~ x
α = Σ agg^ά
geσ
gβG
is an F - linear involutary antiisomorphism of F[G]. Moreover, λ { e
HomF(F[G], F) defined by λ ^ a) = ax is a symmetric functional on F[G].
A
Since X0) = λ ^ a) for all aeF[G], (F[G], , λ t) is a symmetric involutary
F - algebra.
Let V be an F[G ]- module and < , ) a A- form on V. Then
(gυ , gw} = (υ , w}
for all υ , w e V, g e G.
Thus th e Λ- forms on V" are just the G - invariant non- degenerate i^- bilinear
forms on V.
2.1,
COROLLARY.
Theorem 1.8 Λo/rfs for (A, r, ^) = CF[G], A, ^ ) .
The Corollary extends earlier results of W. Willems. H e showed in
his (unpublished) dissertation [Wi; 2.19] t h at F[G]e (for a primitive idempotent e) carries a symmetric G - invariant non- degenerate F - bilinear form
if and only if there exists deF[G] such t h at d = d and F[G]e ^ F[G]d.
Observe t h at it is easy to see th at (F[G]e)* ^ F[G]e (cf. [OT; Lemma
1]) and hence e = e implies the existence of a G - invariant non- degenerate
F - bilinear form on F[G]e.
We generalize the approach above. Let H < G be a subgroup of G
such th at c h a r ( F ) ||iff|. Then / = / * = (1/ |# |) Σ ΛG ITΛ is an idempotent
of ^ [G ], and F[G]f is a transitive permutation module. Its endomorphismring
EndFίG1 (F[G]f)
^ nΛ ti fF[G\ f
= : hF(H, G) = 6
is called Hecke algebra. Observe th at H = 1 implies t h at 6 =
JF [G ].
We
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
40
PETER LAN D ROCK AN D OLAF M AN Z
choose representativeszyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
xt == 1, rc2,
, xt for (£Γ, i7)- double cosets in G and
Xi
set in d fo ) = \ H: HC) H\ . Then 23 = [fxj\ i = 1,
, ί} is an F- basis for
b.
Let a — 2ΐ= i QiifXif) e b (α* e F ) . Then b is a symmetric algebra with
respect to ^ eH o m F ( b, F ) , defined by φ ( a) = α 1? Also, {i n d ^ ) - / ^ 1 / ^ =
1,
, £} is a dual basis of 93, whence α^ = φ ( a n
Λ d(x jj- fxj^), j = 1,
, ί.
(This paragraph is a special case of [CR; 11.30 (i) and (Hi)].)
We define τ to be the restriction of A on b. Since / is A- invariant,
τ is an involutary antiisomorphism on b. N ote t h at aτ = U
Σ \ «*(/#*" V)We now expand aτ in terms of S3, say α τ = X^ = 1 &*(/#*/), ^ e F . Since
indix^'fx^f
= f = fxj, it follows from the previous paragraph t h at ^(α Γ)
= 6i = 9>(Σί 6^ (/ x,/ ) • /) = ^ ( 2 , α, (fxΐ ι f)./ ) = α! = 9>(α) for α e b. Consequently, (b, r, 9) is a symmetric involutary F - algebra.
2.2. COROLLARY. Lei h be a Hecke algebra over F and τ the involutary
antiisomorphism induced by A. Then Theorem 1.8 holds for (b,τ fφ ) .
Theorem 1.3 clearly can be applied to any
no m atter whether A is symmetric or not.
symmetric with respect to φ , ψ e H om F (A, F ) .
(A, τ , φ ) is a symmetric involutary F - algebra,
involutary F - algebra (A, r),
Suppose however th at A is
It might then happen t h at
but (A, r, )ψ is not.
2.3. EXAM PLE. Let q be an odd prime power. Set A = GF(q2) and
consider A as an algebra over F = GF(q). Let τ be the F robenius automorphism of A over F . Then τ is an F - linear involutary (anti- ) isomorphism of A. F or ae A, we consider
φ ( a)
= trA/ F(a)
= aτ + a
an d
ψ(α ) = aτ — a.
Then φ ( A) = F , and ψ(α) = 0 if and only if aeF.
Thus both φ and ψ
are symmetric functionals for A. H owever φ ( a τ ) = ^(α ), but ψ(α τ) = — ψ(α)
for all ae A.
The situation of Example 2.3 is typical. N amely let (A, τ ) be an
involutary F - algebra with center Z = Z(A). Suppose t h at Z is a field and
consider the subfield i£" of Z, consisting of all r- invariant elements. Then
Z\ K is a separable field extension of degree at most 2 (see [Al; X, Thm.
10]). We do not exploit this further on.
If φ is a symmetric functional for A, th en it is easy to see t h at any
other symmetric functional ψ is given by
(ψ a) = φ ( za)
for all ae A,
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
SYM M ETRIC F ORM S, I D EM P OTEN TS
41
wherezyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
z is a central element of A. The following fact as well is easy
and will be needed later on.
2.4 LEMMA. L et A be a symmetric F- algebra with respect to φ e
H om ^A, F). If x is an inυ ertible element in Z(A), then φ x e H om F (A, F)
defined by φ x(a) = φ ( xa) 9 ae A, as well is a symmetric functional for A.
Proof. Since x e Z(A), we have
φ x(ab) —φ ( xab) = φ ( bxa) = φ ( xbά)
= φ x{ba)
If φ x h as t h e right ideal cA in its kernel, then φ ( xcA)
Since x is invertible, it follows t h at c = 0.
for all a, be A.
= 0, whence xc = 0.
•
We now consider a semi- simple algebra S. Recall t h at S then is
symmetric (cf. [CR; 9.8]). By Wedderburn's Theorem,
n
S = 0 Mat m .(Z) ί)
with finite- dimensional skew- fields Dt.
i= l
It M is an irreducible S- module, it belongs to a unique Wedderburn
component of S. Thus in view of an application of § 1, we may assume
th at S = M at w(D ) is simple.
2.5. PROPOSITION . L et D be a finite- dimensional skew- field over F
and assume that (D, ή) is an involutary F- algebra. Then there exists X e
Hom F (Z), F) such that (/ ), r, X) is a symmetric involutary F- algebra.
Proof. Set Z = Z(D).
Case 1: Suppose t h at η induces th e identity on Z. We then consider
D as a Z- algebra and pick a symmetric functional φ e Hom^(Z), Z). Let L
be a splitting field for D. Since D is centrally simple over Z, it follows
th at L ®z D = M at w(L) for some n e N . Since both φ and η are Z- linear,
we can define φ , r) e H om L (M at n (L), L) by a = id L (x) y> and ή = id L (x) 57. Then
37 is an involutary antiisomorphism on Mat7Z(L) and ψ satisfies
iψ iXiMyij))
-
9((Jΰ)(^;))
for
all ( Xij), (ytj) e M at n (L).
It follows t h at φ is (up to some F - scalar factor; th e trace on M at n (L).
Since (x3ί ) >-> (JC^)' is an L- algebra automorphism on M at n (L), an elementary
version of th e Skolem- N oether Theorem implies t h at there exists an invertible matrix (ckl) e Matn(L) such t h at
( Xijy
= (ckl)'\ xjt)(ckl)
for
all ( Xij) e M at n (L).
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
42
PETER LANDROCK AN D OLAF M AN Z
In particular, zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
^(xtjy) = φ ( (Xij)) for all (xtj) e M at n (L). Consequently, we
have for all ae L and d e D
a (g> φ ( d) = φ ( a <g) d) = $((α <g) d)^) = £>(α <g) d*) = a
i.e.
Let
/i e H OΠ IF(Z, F ) be any non- zero functional.
HomF(D, F).
Since X Φ
We set X = µ φ e
0, th e skew- field Z) is a symmetric F~algebra with
v
respect to X. As X(d ) = X(d) for all de D, th e assertion of th e Proposition
holds in th e first case.
Case 2.
Suppose now t h at η is n ot th e identity on Z. Letx e HomF(JD, F)
be any symmetric functional on D.
instead
ψ e H om F (D , F)
<p(dv) = (ψ d).
If φ
and we are done.
If X is n ot ^- invariant, we consider
defined by φ (d) = %(d) + χ(d'), d e D.
We may th us assume t h at X(dη ) — —X{d) for all
and also t h at char (F) Φ
deD,
2.
We consider th e F - linear map η
exists 0 Φ
Clearly,
0, then (Z>, 37, ^) is a symmetric in volutary F - algebra,
Φ
z
η
x e Z such t h at x = —x.
e Ή .om
Since η
(Z, Z).
F
z
Φ
id z , there
By Lemma 2.4, Xx e H o m F φ , F) de-
fined by Xx(d) = X(xd), d e D, is a symmetric functional on Z) as well. N ow
χ^ Φ )
= X(xΦ )
= X(xd) = Xx(d) for all d e D , and
= X(—χiφ )=—X((xdy)
the proof is complete.
•
In order to extend Proposition 2.5 to simple algebras S = M at w(D ),
we use the fact th at any involutary antiisomorphism on S can be written
as an involutary antiisomorphism on D followed by transposition and
conjugation of matrices.
2.6
THEOREM.
L et S be a simple finite- dimensional F- algebra which
admits an F- linear involutary antiisomorphism τ.
a skew- field D, Z = Z(S) = Z(D)Ί
(a)
[Al; X, Thm. 11]
Then τ
Set S = Mat w(Z)) with
τ
S
and K = {ze Z\ z = z).
induces an involutary
antiisomorphism
η
7] on D such that K — {ze Z\ z — z).
(b)
(dυ y
[Al
X, Thm. 10]
= (cuy\ d]d{cu)
2.7
THEOREM.
There exists a non- singular (ckl) e S such that
for all (di3) e S = M at w(D ).
L et S be a finite- dimensional simple F- algebra which
admits an F- linear involutary antiisomorphism τ .
H om F (S, F) such that (S, r, )ψ
Then there exists ψ e
is a symmetric involutary F- algebra.
Proof. Set S = Matm(Z)) for a finite- dimensional skew- field D.
By
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
SYMMETRIC FORMS, IDEMPOTENTS
43
Theorem 2.6, there exist an invertiblezyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFE
(cu) e S and an F - linear involutary
antiisomorphism η on D such t h at
(d 4 i )' -
(ckl)- \ dh)(cu)
for all (dtj) e S.
By Proposition 2.5, there exists X e H om F (D , F) such t h at (D , 37, %) is a
symmetric involutary F - algebra. We set φ = X- tr eΐlom F(S, F). Then φ
is a symmetric functional on S, and for (d*j) e S we have
?((cU r ) = ?>((<*?,)) = Σ «
= Σ *(<***) = φ iidij)) .
This establishes the claim.
•
2.8 COROLLARY. Let (S, τ ) be an involutary F- algebra. If S is semisimple and char (F) Φ 2, then the following assertions are equivalent for
an irreducible S- module M.
(1) M carries a symmetric τ - form.
(2) There exists an idempotent e e S such that M = Se and eτ = e.
Proof. Let 1 = βj +
+ ε t be th e decomposition of 1 e S in to Wedderburn idempotents ε^. Then τ permutes th e εέ. Observe t h at under each
of the conditions (1) and (2), the idempotent ε t corresponding to M is fixed.
Thus th e assertion follows from Theorems 1.8 and 2.7.
•
F or more examples of involutary algebras we refer to [Al; chap. X].
§ 3.zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Absolutely irreducible (7- modules
3.1 EXAM PLES, (a) Let (A, τ ) be an involutary F - algebra and M an
absolutely irreducible A- module. If M carries a symplectic r- form < , ),
then it is very easy to see t h at there does not exist an idempotent e e A
such t h at eT = e and M = Ae:
Suppose there is such an idempotent e. Since < , ) is non- degenerate,
we find ae A such t h at (ae, e) Φ 0. Since M is absolutely irreducible,
eae = µ efor some µ e F. Consequently,
0Φ
(ae, e> = <eταβ, β> = (eae, e) = µ (e,e> = 0,
a contradiction.
(b) Let (A, τ ) be an involutary F - algebra, c h a r ( F ) = 2 and M an
absolutely irreducible A- module with symmetric r- form ( , ) . If dim F M
> 2, counting yields a non- zero isotropic vector in M. Since the isotropic
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
44
PETER LAN D ROCK AN D OLAF M AN Z
vectors in zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
M form a submodule of M, th e form < , > is symplectic. By
(a), there does not exist ee A such t h at eτ — e and M — Ae. Thus th e
assertion of Theorem 1.3 is false for c h a r ( F ) = 2.
(c) N evertheless, if F is not a splitting field, there might exist such
an idempotent. As a trivial example, consider A — F2[CZ] can let M be
its 2- dimensional irreducible module. Clearly, there exists exactly one
primitive idempotent ee A such t h at M^ Ae. H ence eτ — e, and M carries a symmetric A- form, by Lemma 1.5.
F or char (F) = 2, one might have to consider quadratic forms instead
of bilineai ones. F or more results in this direction, we refer to th e
(unpublished) dissertation of W. Willems [Wi].
In the following, we restrict ourselves to group algebras F[G] with
symmetric functional λ x e H om F (F [G ], F) and involutary antiisomorphism
τ = A. Since the A- forms are just the G - invariant ones, we shall speak
henceforth of G- forms.
We next slightly sharpen the assertion of 3.1 (a) in case of group
algebras. To do so, we need the following result (see [H B; VII, 8.12]).
3.2 TH EOREM. Let M be an absolutely irreducible self- dual F[G]module. Then to within an F- scalar multiple, there exists only one G- form
on M. If char (F) Φ 2, this form is either symmetric or symplectic,
3.3 COROLLARY. Let M be an absolutely irreducible F[G]- module. If
char (F) Φ 2, then the following asesertίons are equivalent
(1) M carries a symplectic G- form.
(2) ff=
0 for all idempotents feF[G]
which satisfy M ^
F[G]f
Proof. (1) iφ (2): Suppose there exists an / such t h at ffφ
0. Then
Lemma 1.2 (b) implies t h at there also exists an idempotent ee A such t h at
M = F[G]e and e = e. By Lemma 1.5 and Proposition 1.7, the module
M carries a symmetric G- form. Since M is absolutely irreducible, M
cannot carry a symplectic G- form, by Theorem 3.2. This contradicts (1).
(2) => (1): Suppose t h at M carries a symmetric G- form. As char (F)
Φ 2,
Theorem 1.3 yields an idempotent eeF[G] such t h at M ^ F[G]e
and e = e. I n particular, ee = e Φ 0, contradicting (2). By Theorem 3.2,
M carries a symplectic G- form.
•
The next lemma, which we only state under th e conditions needed
later on, is probably well- known.
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
SYMMETRIC FORMS, IDEMPOTENTS
45
3.4 LEMMA.zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Let e be an idempotent in F[G], where c h a r ( F ) ||G |.
Then λ 1(e) = (l/ |G |)dim F (F [G ]e).
Proof. Let L ΞΞ> F be an algebraically
Then
dimF(F[G]e) = άim L (L®FF[G]e)
closed extension field of F.
= dimL (L[G]e).
Let e = /Ί +
+ fs be a decomposition of e into primitive idempotents
ft in L[G ], and / = fx. We denote the character of L[G]f by X, and the
corresponding Wedderburn idempotent by ε e L[G]. Thus
2
and ^(ε) = %(1) / |G|. Since al] primitive idempotents corresponding to ε
ι
are conjugate to / , and since λ x(u~ fu) = λ {(f) (for units u in L[G\ ), we
conclude λ x(f) = (l/ χ(l))λ,(e) = X(1)I\ G\ = (1/ |G |) dim L(L[G ]/ ). Consequently,
Ue) = iUfd
= (l/ |G |)gdim L (L[G ]/ , ) = (1/ |G |) dim L(L[G]β)
= (l/ |G D dim F (F [G ]e).
D
As a disadvantage of Theorem 1.3 we recall th at its proof does not
yield an explicit formula for a r- invariant idempotent in terms of the
given τ- form. U nder certain circumstances, we can do better.
Let M be an F[G ]- module with G- form < , ). F or an element xeM,
we define
geσ
Then cx has the following properties.
(1) λ , (c xa) = (ax, Xs) for all a e F[G].
(N amely just observe th at λ x(cxh) = (hx, x) for all h e G.)
(2) If fe F[G] satisfies fx = x, then also fcx = c^.
(To see this, note th at
χ x( Cχ a) (3)
(ax, x) = (afx, x) = ^ifeα/ ) = ί^/ c^α)
for all
aeF[G],
and therefore fcx — cx — 0.)
If < , > is symmetric, then cx = iΣ geo (gx, x)g and cx = cx.
3.5 Remark. Before we proceed, we recall what Lemma 1.2 says in
our present context. Let M be an irreducible F[G]- module which carries
a G- form and which contains a non- isotropic vector x. Then there exist
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
46
PETER LAN DROCK AN D OLAF M AN Z
primitive idempotents / , zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFED
e e F[G] with t h e following properties:
( i ) M^ F[G]f^ F[G]e.
( ii) fx = x.
(iii) e — e.
(iv) e is t h e 1- element of fF[G]f, hence fF[G]f = eF[G]e.
3.6 PROPOSITION . L et M be an irreducible F[G]- module which carries
a symmetric G- form <( , ) . Suppose that M contains a non- isotropic vector
x> and let e = e be chosen according to Remark 3.5. / /
{veeF[G]e\ ϋ
then e = γ c
x
= v} = Fe,
for some γ e F.
Proof. We choose th e idempotent / as in Remark 3.5. By (ii), fx = x
implies fcx = cx. Since < , ) is symmetric, (iv) yields
cx = cx = cj = cj = fcJefF[G]f
= eF[G\ e,
and cx = βe for some βe F. I t remains to show th at cx Φ
because λ x(cx) = <x, x} Φ 0.
0. This follows,
Π
3.7 TH EOREM. L et F[G] be semi- simple, and suppose that the absolutely irreducible F[G]- module M carries a symmetric G- form < , >. Suppose
that M contains a non- isotropic vector x (, which holds provided that
chΆγ( F)φ 2).
Then
e = (dimFM)l(\ G\ .(x, x}) cx = (dim F M )/ (|G |.< *, x» Σ
<gx, *>g
geo
is an idempotent such that M = F[G]e and e = e.
Proof. Let e = e be chosen as in Remark 3.5. Since F[G] is semisimple and M is absolutely irreducible, we have M = F[G]e and F =
E n d P [ β ] (M ) ^ &ntleF[G]e. I n particular, {ueβF [G ]e|ί} = v} = F e, an d Proposition 3.6 implies t h at e = γ c x for some ^ 6 F. I t remains to determine
the scalar γ . By Lemma 3.4,
and t h e assertion follows.
•
3.8 Remarks, (a) I t should be clear t h at Theorem 3.7 still holds if
we drop t h e hypothesis "semi- simple" and assume instead t h at M h as
defect 0 (i.e. t h e block ideal of F[G] corresponding to M is simple).
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
SYM M ETRIC F ORM S, I D EM P OTEN TS
47
(b) IfzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
M h as positive defect however, th en cx definitely is no candidate for an idempotent. To see this note t h at λ x(cxj) = <jx, x) = 0 for
all j e J(F[G]). Consequently, cx e ann (J(F[G])) = soc (F[G\ ). Thus cx =
fcj eeF[G]e is in th e socle and hence in th e radical of th e block ideal
corresponding to M. Therefore, c2x — 0.
(c) We do not know how to generally proceed if F is not a splitting
field for M. The case F — R however will be treated in t h e next section.
§ 4.zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Real orthogonal representations
Let M be an R[G ]- module and fix a symmetric, positive definite bilinear form ( , ) on M. Then [ , ] defined by
[v, w] = Σ (gυ , gw),
υ,
w e M,
obviously is a symmetric, positive definite G- form.
Theorem 1.3.
I t th us follows from
4.1 COROLLARY. L et M be an irreducible ~R[G]- module. Then there
exists an idempotent e e R[G ] such that M ^ R[G ]β and e = e.
4.2
LEMMA.
L et e be a primitive idempotent in R[G ] with e = e. Then
/ : = {vee~R[G]e\ ϋ = v} = R e .
Proof. By Lemma 1.5, <u, w) — λ x{vw), υ , w e R[G ]β, is a symmetric
G- form on R[G ]β. Moreover, < , ) is positive definite, and it holds t h at
<y, wά) = λ i(vάw)
= <uα, w;>
for all a e eR[G ]e.
Suppose th at dim R / > 2 and recall t h at eR[G ]β = R, C or H , where H
denotes th e quaternion skew- field. If eR[G ]β ^ C, we choose i e eH[G]e
corresponding to th e complex unit in C. I t then follows for all 0 Φ υ e
R[G ]β t h at
0 < (υ i, vίy = (υ ii, u> = (vi2, u) = —(y, u> < 0,
a contradiction.
We may thus assume th at βR[G]β = H . If dim R 7 = 2, then / =
span R<β, x) for some xeeR.[G]e and th e elements of I pairwise commute.
Therefore, I is closed under multiplication and I = C. Consequently, if
2
dim R 7 = 2 or 4, then J contains an element ί such t h at i = — e and we
proceed as in th e last paragraph. We still have to consider th e case
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
48
PETER LAN D ROCK AN D OLAF M AN Z
dim R 7 = 3, sayzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
I = span R <e, x,y}. Let {e, i,j, k} be the canonical R- basis
of eR[G ]e = H . After a suitable basis transformation we may assume
Sin ce x2 = — e(l + µ2), we
t h a t x = i + µ k an d y — j + vk for µ ,>
ι e R.
obtain for 0 Φ
υ e R[G ]e
2
2
0 < (υ x, vx) = (vx , v} = - ( 1 + ^ )<u, u> < 0,
again a contradiction. This completes the proof.
•
Let V be an F[G]- module (for an arbitrary field F) and < , ) a G- form
on V. Then the mapping
(x • - * < > >« >
where <y, ι^>α = <y, αw;> ,
yields an isomorphism between End F [ ί ?] (V) and the F- space BG(V) of all
G - invariant bilinear forms on V (possibly degenerate).
Assume in addition th at V = F[G]e for an idempotent e. The isomorphism eF[G]e ^ E n d ^ / V) is given by a H- > α α , where # α(ι>) = uα. H ence
« •- > < » >α ,
where <ϋ, w;>α = <ϋ, wa),
induces a vector space isomorphism between eF[G]e and BG(V).
following serves as a substitute for Theorem 3.2.
The
4.3 PROPOSITION . Let M be an irreducible TR[G]- module. Then M
has exactly one symmetric G- form (up to R- scαZαr factors).
Proof. By Corollary 4.1, we may assume t h at M = R[G ]e for an
idempotent e = e. Consider the symmetric G- form < , ) on M induced
by λ ι (see Lemma 1.5). Then every other G - invariant bilinear form on M
is given by < , > α for a unique a e eR[G\ e. N ow < , > α is symmetric if
and only if
<u, wa) = (v, υι)
a
= (w, v}a = (w, va} = (wά,
v) = (v, wά}
for all v, w e M. This happens if and only if d = a, and Lemma 4.2 implies t h at a = γe for some γ e R. Consequently, < , ) α = γ( , ), which was
to be proved.
D
Let M be an irreducible R[G ]- module. U sing th e form [ , ] and
Proposition 4.3, any given symmetric G- form < , ) on M may be assumed
to be positive definite. The group G is th en said to be represented
orthogonally on M. It makes sense now to consider the unit sphere
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
SYMMETRIC FORMS, IDEMPOTENTS
49
{xe M\ (x, x} = 1} on M. Also a distance function d( , ) can be introduced in the usual way by
d(x, yf = (x -
y, x -
y) ,
x, y 6 Λf.
4.4 TH EOREM. Let G be represented irreducίbly and orthogonally on
the R- space M with respect to the form < , >.
a) Given xe M with (x, x} = 1,
is an idempotent satisfying both M ^ R[G ]e an d e = e. Here, elements of
M in the same G- orbit lead to G- conjugate d
ί empotents.
b) Conversely, given e = Σ gς G<xgg an idempotent with M = R[G ]β and
e = e, ίΛere, exists xe M with (x, x} = 1 an d
<^x, Λ:> = IGI a^/ dimRAf,
geG.
Proof, a) Consider first xe M with <#, x> = 1, and choose the
idempotent e = β eR [G ] with M ^ R[G ]β according to Remark 3.5. By
Lemma 4.2, {v e eR[G]e\ ΰ = v} = Re, and Proposition 3.6 yields
e = γc
x
= γ Σ
(gx, x}g
for some γ e R.
gea
The scalar γ again is determined by Lemma 3.4, namely
dim R M / |G | = λ 1(e) = γ λ x{cx) = r(x,x)
= r.
F inally observe th at replacing x by hx (he G) replaces e by heh~\
b) Assume conversely th at e — e is given. Then Lemma 1.5 asserts
t h at
υ , weR[G]e,
defines a symmetric G- form < , ) ' on R[G ]β = M.
In particular,
(geG),
and Lemma 3.4 yields (e, ey — 1. By Proposition 4.3, there is a non- zero
γ e R such th at <ι>, α;) 7 = γ( v, w} for all v, w 6 R[G ]β. Then 1 = <e, β) x =
γ( β 9 e), and ^ > 0, since < , > is positive definite. H ence we may take x
to be V 7 e.
•
F or data transmission via a G aussian channel, it turned out to be
successful to consider the codewords as G - orbits on the un it sphere of
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
50
PETER LAN D ROCK AN D OLAF M AN Z
some Euclidian space R \ The question about reasonable lower bounds
for the minimal Euclidian distance—actually our motivation for this
paper—has only got partial answers. The following result was first proved
by D. Splpian in 1968. (See [BM; chap. 6] for this result and some
background in Coding Theory.)
4.5 COROLLARY (Slepian).zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Let G be represented irreducibly and orthogonally, but non- trίva
ί lly, on the -ίΐ space M with respect to the form
< , >. Let xeM with (x, x> = 1. Then
a) *
Σ eβd(gx,xY
= 2\ G\ .
b) / / 9ΐ denotes any conjugacy class in G and ke% then gΣ em d(gx, x) 2
= 2 |9 ΐ |( l - X(k)lX(ί) \
where X is the character of M.
Proof, By Theorem 4.4, e = dim R M/ |G ] g
Σ ec (gx, x)g is an idempotent
affording M.
a) Since M is not th e trivial module, we have
o = ( Σ zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
< gχ, * > £ ) ( Σ h) = Σ < gχ, * > Σ
gQG
hβG
gGG
hβG
2
and therefore g
Σ eod(gx, x) = g
Σ ea^O- - (gx, x» = 2\ G\ .
b) Since d(gx, x) 2 = 2(1 — (gx, x}) 9 it amounts to show t h at
Σ
<gχ, χ> = mW)IO) •
a n d
P ut S = εΣ evg
observe t h at e = X(ΐ) l\ G\ SΣ GG (gx, x}g~ ι . Thus ^(βSR)
= %(1)/ |G| g
Σ em (gx, x) and it is therefore sufficient to show t h at λ ^ effi) =
ι
(\ m\ l\ G\ )X(k). Let d = dim R En d R [ σ] (M )). Then ε = Z(l)/ (|G |- d) Σ geG t{g)g~
is the Wedderburn idempotent corresponding to M. We decompose ε =
e1 + - - - + es into primitive idempotents et e R[G ], where e = ex and s — X(ϊ) / d.
Then ei — u^eUi for units ut e R[G ]. Since
we obtain
(χ(i)/ d)Λ(e3l) = Λ(εS) = χ(i)/ (|G i d) Σ *(g) = (x(i)ld)(\ m\ \ι G\ )x(k)),
gem
which establishes the claim.
•
ACKN OWLED G EMEN T. This paper was written when the first author
was visiting Mainz and the second autor was visiting Aarhus and H aifa.
We th an k th e Deutsche Forschungsgemeinschaft (DFG), th e Danish Nat-
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883
SYMMETRIC FORMS, IDEMPOTENTS
51
ural Science Research Council and the Technίon (Haifa) for th eir financial
support.
We are also indebted to P. F leischmann, R. Kn orr and A. Juh asz for
helpful discussions.
REF EREN CES
[Al] A. Albert, Structure of algebras, AMS Colloquium Publications, 1939.
[BM] I. F . Blake, R. C. Mullin, The mathematical theory of coding1, Academic Press,
New York, 1975.
[CR] C. Curtis, I. Reiner, Methods of representation theory, John Wiley, New York
1981.
[HΓB] B. H uppert, N. Blackburn, F inite groups I I , Springer- Verlag, Berlin, 1982.
[La] P. Landrock, F inite group algebras and their modules, Cambridge U niversity
Press, Cambridge, 1983.
[OT] T. Okuyama, Y. Tsushima, On a conjecture of P. Landrock, J. Algebra 104
(1986), 203- 208.
[Wi] W. Willems, Metrische Moduln ϋber G ruppenringen, D issertation, Mainz, 1976.
Peter Landrock
Department of Mathematics
University of Aarhns
DK- 8000 Aarhus
Denmark
Olaf Manz
Department of Mathematics (IWR)
University of Heidelberg
D- 6900 Heidelberg
Germany
Current Address:
UCI
Utility Consultants International
D- 6000 Frankfurt 71
Downloaded from https://www.cambridge.org/core. IP address: 181.215.217.119, on 12 Jun 2019 at 12:11:26, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0027763000003883