A COMBINATORIAL METHOD FOR COMPUTING
STEENROD SQUARES
Roc´ıo Gonz´alez-D´ıaz and Ped o Real
Dp o. de Ma em´a ica Aplicada I
Facul ad de In o m´a ica y Es ad´ıs ica
Uni e sidad de Se illa
ogo[email p o ec ed], [email p o ec ed]
Abs ac
We p esen he e a combina o ial me hod o compu ing cup-ip oduc s and S een od
squa es o a simplicial se X. This me hod is essen ially based on he de e mina ion
o explici o mulae o he componen mo phisms o a highe diagonal app oxima-
ion (i.e., a amily o mo phisms measu ing he lack o commu a i i y o he cup
p oduc on he cochain le el) in e ms o ace ope a o s o X. A gene aliza ion o
his me hod o S een od educed powe s is ske ched.
Bo h au ho s a e pa ially suppo ed by he PAICYT esea ch p ojec FQM-0143 om Jun a de Andaluc´ıa and he
DGESIC esea ch p ojec PB97-1025-C02-02 om Educa ion and Science Minis y (Spain).
1
1 In oduc ion
Cohomology ope a ions a e algeb aic ope a ions on he cohomology g oups o spa-
ces commu ing wi h he homomo phisms induced by con inuous mappings. This
machine y is use ul when he g aded ec o space s uc u e and he cup p oduc
on cohomology ail o dis inguish wo spaces by hei cohomology. S een od squa-
es [20] cons i u e an ex emely impo an class o cohomology ope a ions no only
in Algeb aic Topology bu also in he a ea o simplicial me hods in Homological
Algeb a (cohomology o g oups, Hochschild cohomology o algeb as, ...).
We he e s udy in de ail he unde lying combina o ial s uc u es o he de ini ion
o hese cohomology ope a ions. Mo e conc e ely, we will mo e in he amewo k
o Simplicial Topology [12] in which he basic objec s a e simplicial se s, ha is,
g aded se s endowed wi h ace (∂i) and degene acy (si) ope a o s, sa is ying se e al
commu a i i y ela ions. Roughly speaking, a simplicial se can be conside ed as
an algeb aic gene aliza ion o he s uc u e o a iangula ed polyhed on al hough
he o me ea u es a mo e igid combina o ial s uc u e han he la e . In his
con ex , we de elop a sui able se ing in which all S een od cohomology ope a ions
can be s udied simul aneously.
I is well-known ha he e a e se e al me hods o cons uc ing S een od squa-
es. One o hem consis s o making hese ope a ions using he cohomology o
Eilenbe g-Mac Lane spaces (see, o ins ance, [[12]; p. 107]). Ano he me hod is
de e mined by he cons uc ion o a amily o mo phisms {Di}measu ing he lack o
commu a i i y o he cup p oduc on he cochain le el [20]. This sequence o mo p-
hisms is called highe diagonal app oxima ion. I s exis ence is always gua an eed
by he acyclic models me hod [5]. In his way, i is possible o de i e a ecu si e
p ocedu e o ob ain he explici o mula o any Di(see [4], [[13]; Sec . 7]).
In his pape , we p esen an al e na i e me hod o ob aining he explici o mula
o a highe diagonal app oxima ion. In [15], he o mula o Diis es ablished
in e ms o he componen mo phisms o a gi en Eilenbe g-Zilbe con ac ion (a
special homo opy equi alence) om CN
∗(X×X) on o CN
∗(X)⊗CN
∗(X), whe e CN
∗(X)
deno es he no malized chain complex o a gi en simplicial se X.
Now, we ha e he ollowing p oblems. On one hand, he componen mo phisms
o he con ac ion abo e a e de ined in e ms o ace and degene acy ope a o s o
he simplicial se X. On he o he hand, he o mula o a mo phism Dialways
in ol es he use o he homo opy ope a o o he Eilenbe g-Zilbe con ac ion and
he explici o mula o his las mo phism is de e mined by shu les (a special ype
o pe mu a ion) o degene acy ope a o s. In consequence, i we y o exp ess in
2
his way he mo phisms Diin e ms o ace and degene acy ope a o s o X, he
numbe o summands appea ing in he o mula o a mo phism Die alua ed o e
an elemen o deg ee nis, in gene al, a leas 2n. The e o e, an algo i hm ha
would be designed s a ing om hese o mulae would be oo slow o p ac ical
implemen a ion.
Because o his, he idea o simpli ying hese o mulae a ises in a na u al way.
This simpli ica ion o no maliza ion is based on he ac ha any composi ion o
ace and degene acy ope a o s o he simplicial se Xcan be pu in a “canonical”
way. Tha is, a composi ion o his ype can be exp essed in he unique o m:
sj · · · sj1∂i1· · · ∂is,
whe e j >· · · > j1≥0 and is>· · · > i1≥0.
Mo eo e , aking in o accoun ha he image o a mo phism Dilies in
CN
∗(X)⊗CN
∗(X), hose summands o he simpli ied o mula o Diwi h a ac o
ha ing a degene acy ope a o in i s exp ession, can be elimina ed. The eason is
ha his ac o applied o an elemen o he simplicial se Xis ze o in he no -
malized chain complex associa ed o X. In his way, we he e ob ain mo e simple
o mula o he mo phism Di.
In his way, he classical de ini ion o S een od squa es:
Sqi(c)(x) = (µ(< c ⊗c, Dj−i(x)>), i ≤j
0, i > j (1)
whe e c∈Hom(CN
j(X),Z2), x∈CN
i+j(X) and µis he homomo phism induced by
he mul iplica ion in Z2, is complemen ed by a manageable combina o ial o mula-
ion o he highe diagonal app oxima ion. As we will ema k la e , his desc ip ion
can be conside ed as a di ec ansla ion o he mos ancien de ini ion o S een od
squa es (see [18]) o he gene al se ing o he Simplicial Topology. We gi e a mo e
de ailed explana ion in he hi d sec ion.
S ill, we hink ha his combina o ial machine y could be subs an ially imp o ed
in he u u e by exploi ing he well-known p ope ies o S een od squa es [1, 20] and
ad anced echniques o calcula ing cocycles (see, o example, [9, 10]). In his way,
a “ easonably e icien ” algo i hm compu ing he cohomology algeb a o se e al
impo an simplicial se s could be de i ed.
Finally, we s a an analogous s udy o ha gi en in [15] o S een od edu-
ced powe s. Mo e p ecisely, we p o ide a simplicial desc ip ion o hese ope a-
ions in e ms o he componen mo phisms o an Eilenbe g-Zilbe con ac ion om
3
CN
∗(X×p imes
· · · ×X) o CN
?(X)⊗p imes
· · · ⊗CN
∗(X), and a 0-sequence {γi}i≥0in he
symme ic g oup Gp.
He e is a summa y o he p esen pape . Sec ion 2 is dedica ed o no a ion,
e minology and a p esen a ion o he p oblem. In Sec ion 3, we show an explici
combina o ial de ini ion o cup-ip oduc s and, consequen ly, o S een od squa es.
In Sec ion 4, we s udy S een od educed powe s om his poin o iew. Finally, a
p oo o he main heo em enuncia ed in Sec ion 3 is gi en in Sec ion 5.
We a e g a e ul o P o . Tom´as Recio o his help ul sugges ions o making
he exposi ion mo e eadable and, hope ully, mo e in o ma i e. We also wish o
acknowledge ou deb o he e e ees o hei many aluable indica ions.
2 The highe homo opy commu a i i y o he
Alexande -Whi ney ope a o
The aim o his sec ion is o gi e some p elimina ies and a b ie accoun o he wo k
done in [15] in o de o acili a e he unde s anding o he es o he pape . Mos
o he ma e ial gi en in his sec ion can be ound in [12], [11], [17] and [21].
Asimplicial se Xis a sequence o se s X0, X1, . . ., oge he wi h ace ope a o s
∂i:Xn→Xn−1and degene acy ope a o s si:Xn→Xn+1 (i= 0,1, . . . , n), which
sa is y he ollowing “simplicial” iden i ies:
(s1) ∂i∂j=∂j−1∂i,i i < j;
(s2) sisj=sj+1si,i i≤j;
(s3) ∂isj=sj−1∂i,i i < j,
(s4) ∂isj=sj∂i−1,i i > j + 1,
(s5) ∂jsj= 1X=∂j+1sj.
The elemen s o Xna e called n-simplices. A simplex xis degene a ed i x=siy
o some simplex yand degene acy ope a o si; o he wise, xis non degene a ed.
The ollowing elemen a y lemma will be essen ial in he p oo o he main
heo em o his pape .
4
Lemma 2.1 [12] Any composi ion µ:Xm→Xno ace and degene acy ope a o s
o a simplicial se Xcan be pu in a unique “canonical” o m:
sj · · · sj1∂i1· · · ∂is,
whe e n>j >· · · > j1≥0,m≥is>· · · > i1≥0and n− +s=m.
Le Rbe a ing which is commu a i e wi h uni . A chain ( esp. cochain)complex
M={Mn, dn}( esp. C={Cn, δn}) is a g aded (o e he in ege s) R-module
oge he wi h a R-module map do deg ee −1 ( esp. o deg ee +1), called he
di e en ial, such ha dndn+1 = 0 ( esp. δn+1 δn= 0). An elemen co Cnis called
n-cochain. The homology H∗(M) ( esp. he cohomology H∗(C) is he amily o
modules
Hn(M) = Ke dn/Im dn+1 ( esp. Hn(C) = Ke δn/Im δn−1).
I Mis a chain complex o e Rand Gis an R-module, he e is a cochain
complex Hom(M, G) = {Hom(Mn, G), δn}, whe e, i c∈Hom(Mn, G), hen
δnc∈Hom(Mn+1, G) is de ined by
(δnc)(x) = c(dn+1(x)), x ∈Mn+1.
We also w i e < c, x > ins ead o c(x) and se < c, x >= 0 i he deg ee o he
cochain cis no equal o he deg ee o he elemen x. In his no a ion,
< δnc, x > =< c, dn+1(x)> .
Now, gi en a simplicial se X, le C∗(X) deno es he chain complex {Cn(X), dn},
in which Cn(X) is he ee R-module gene a ed by Xnand dn:Cn(X)→Cn−1(X)
is de ined by dn=
n
X
i=0
(−1)i∂i. Le us deno e by s(C∗(X)) he g aded
R-module gene a ed by all he degene a ed simplices. In C∗(X), we ha e ha
dn(s(Cn−1(X))) ⊂s(Cn−2(X)) and, hen CN
∗(X) = {Cn(X)/s(Cn−1(X)), dn}is a
chain complex called he no malized chain complex associa ed o X. Gi en a
R-module G, le us de ine by C∗(X;G) he cochain complex associa ed o CN
∗(X).
In his way, we de ine he homology and cohomology o Xwi h coe icien s in a
5

R-module Gby H∗(X;G) = H∗(CN
∗(X)⊗G) and H∗(X;G) = H∗(C∗(X;G)), es-
pec i ely.
Eilenbe g and Mac Lane de ined in [6] a con ac ion o chain complexes om N
on o M, as a iple ( , g, φ) in which :N→M(p ojec ion) and g:M→N
(inclusion) a e chain maps, and φ:N→N(homo opy ope a o ) is a map o
R-module aising deg ee by 1. Mo eo e , i is equi ed ha
(c1) g = 1M,(c2) φd +dφ +g = 1N,
(c3) φg = 0,(c4) φ = 0,(c5) φφ = 0.
Hence, his de ini ion implies ha he “big” complex Nis homology equi alen
o he “small” complex Min a s ong way. In ac , a con ac ion is a special
homo opy equi alence be ween chain complexes.
I we ha e wo con ac ions ( i, gi, φi) om Ni o Mi, wi h i= 1,2 hen, he
ollowing con ac ions can be cons uc ed (see [6]):
•The enso p oduc con ac ion ( 1⊗ 2, g1⊗g2, φ1⊗g2 2+ 1N1⊗φ2) om
N1⊗N2 o M1⊗M2.
•I N2=M1, he composi ion con ac ion ( 2 1, g1g2, φ1+g1φ2 1) om N1 o
M2.
I pand qa e non-nega i e in ege s, a (p, q)-shu le (α, β) is a pa i ion o he
se {0,1, . . . , p +q−1}o in ege s in o wo disjoin subse s, α1<· · · < αpand
β1<· · · < βq, o pand qin ege s, espec i ely. The signa u e o he shu le (α, β)
is de ined by sig(α, β) =
p
X
i=1
αi−(i−1).
A e hese p elimina ies, we a e able o desc ibe a e y impo an homo opy
equi alence in Algeb aic Topology. This con ac ion ells us ha he associa ed
chain complex CN
∗(X×Y) educes o he enso p oduc o chain complexes CN
∗(X)
and CN
∗(Y).
An Eilenbe g-Zilbe con ac ion [8] om CN
∗(X×Y) o CN
∗(X)⊗CN
∗(Y), whe e
Xand Ya e gi en simplicial se s, is de ined by he iple (AW, EML, SHI) whe e:
6
•The Alexande -Whi ney ope a o AW :CN
∗(X×Y)−→ CN
∗(X)⊗CN
∗(Y) is
de ined by:
AW(am×bm) =
m
X
i=0
∂i+1 · · · ∂mam⊗∂0· · · ∂i−1bm.
I X=Y,AW can be conside ed as a “simplicial app oxima ion” o he dia-
gonal and his ope a o p o ides a me hod o cons uc ing he cup p oduc in
cohomology. I we in e change he ac o s amand bmin he o mula, we ob ain
a di e en app oxima ion. Compa ison o hese wo di e en app oxima ions
o he diagonal leads o he S een od squa es.
•The Eilenbe g-Mac Lane ope a o EML :CN
∗(X)⊗CN
∗(Y)−→ CN
∗(X×Y)
is de ined by:
EML(ap⊗bq) = X
(α,β)∈{(p,q)−shu les}
(−1)sig(α,β)sβq· · · sβ1ap×sαp· · · sα1bq.
This ope a o can be seen as a p ocess o “ iangula ion” in he ca esian
p oduc X×Y.
•And he Shih ope a o SHI :CN
∗(X×Y)−→ CN
∗+1(X×Y) is de ined by:
SHI(a0×b0) = 0;
SHI(am×bm) = X(−1)¯m+sig(α,β)+1 sβq+ ¯m· · · sβ1+ ¯ms¯m−1∂m−q+1 · · · ∂mam
×sαp+1+ ¯m· · · sα1+ ¯m∂¯m· · · ∂m−q−1bm;
whe e ¯m=m−p−q,sig(α, β) =
p+1
X
i=1
αi−(i−1), and he las sum is
aken o e all he indices 0 ≤q≤m−1, 0 ≤p≤m−q−1 and
(α, β)∈ {(p+ 1, q)-shu les}.
A ecu si e o mula o he SHI ope a o has al eady been gi en by Eilenbe g
and Mac Lane in [7]. The explici o mula gi en he e was s a ed by Rubio in [16]
and p o ed by Mo ace in he appendix o [14]. In con as o he deep s udies ound
in he li e a u e on he AW and EML ope a o s, i u ns o be qui e su p ising he
lack o in e es shown up o now in he s udy o he homo opy ope a o in ol ed
in an Eilenbe g-Zilbe con ac ion, no only om he poin o iew o ge ing i s
explici o mula bu also o ob aining algeb aic p ese a ion esul s o his ope a o
wi h ega d o he unde lying coalgeb a s uc u e on CN
∗(X×X).
7
Gi en a simplicial se Xand a posi i e in ege p, we can o m a con-
ac ion ( , g, φ) om CN
∗(X×p imes
· · · ×X) o CN
∗(X)⊗p imes
· · · ⊗CN
∗(X), app o-
p ia ely composing Eilenbe g-Zilbe con ac ions. I p= 2, hen ( , g, φ) is
he con ac ion (AW, EML, SHI). The con ac ion om CN
∗(X×X×X) on o
CN
∗(X)⊗CN
∗(X)⊗CN
∗(X) is de ined by he composi ion o he Eilenbe g-Zilbe
con ac ion om CN
∗(X×X×X) o CN
∗(X)⊗CN
∗(X×X) and he enso p oduc
con ac ion o he iden i y mo phism 1CN
∗(X)and he Eilenbe g-Zilbe con ac ion
om CN
∗(X×X) on o CN
∗(X)⊗CN
∗(X). And so on.
F om now on, he con ac ions ob ained in his way will also be called Eilenbe g-
Zilbe con ac ions.
Le pbe a posi i e in ege . Le us de ine se e al chain maps (o mo phisms) we
use in his pape . We will omi in he no a ion o hese mo phism i s dependency
on p. The diagonal map
∆ : CN
∗(X)→CN
∗(X×p imes
· · · ×X)
is de ined by ∆(x) = (x, p imes
. . . , x). The ollowing au omo phisms a e also de ined:
:CN
∗(X×p imes
· · · ×X)→CN
∗(X×p imes
· · · ×X),
such ha (x1, x2, . . . , xp) = (x2, . . . , xp, x1) and
T:CN
∗(X)⊗p imes
· · · ⊗CN
∗(X)→CN
∗(X)⊗p imes
· · · ⊗CN
∗(X)
de ined by T(x1⊗x2⊗ · · · ⊗ xp) = (−1)|x1|(|x2|+···+|xp|)x2⊗ · · · ⊗ xp⊗x1.
We now ou line he p oblem conce ning S een od squa es in which we a e in e-
es ed. I is well-known ha he AW ope a o is no commu a i e, ha is, assuming
ha X=Y,AW 6=AW. On he o he hand, his ope a o de e mines he
cup p oduc in cohomology. I Gis a ing, gi en wo cochains c∈Ci(X;G) and
c0∈Cj(X;G), and x∈CN
i+j(X), he cup-p oduc o cand c0is de ined by:
c^c0(x) = µ(< c ⊗c0, AW ∆(x)>)
=µ(< c, ∂i+1 · · · ∂i+jx > ⊗< c0, ∂0· · · ∂i−1x >),
whe e µis he homomo phism induced by he mul iplica ion on G. S een od in [19]
de e mined ha he e exis s an in ini e sequence o mo phisms {Di}, called highe
diagonal app oxima ion, which “measu es” his lack o commu a i i y. Mo e p eci-
sely, he e is a sequence o g aded homomo phisms Di:CN
∗(X)→CN
∗(X)⊗CN
∗(X)
o deg ee isuch ha :
8
D0=AW ∆
d⊗Di+1 + (−1)iDi+1 d=T Di+ (−1)i+1Di,
whe e dand d⊗a e he di e en ials o CN
∗(X) and CN
∗(X)⊗CN
∗(X), espec i ely.
Mo eo e , he mo phism Dican be exp essed in he o m Di=hi∆, whe e
hi:CN
∗(X×X)→CN
∗(X)⊗CN
∗(X) is a homomo phism o deg ee i. In he li e-
a u e, he exis ence o he owe o i e a ed Alexande -Whi ney ope a o s {hi}is
gua an eed by he acyclic models me hod (see [5]). This echnique can be conside-
ed as a cons uc i e me hod in he simplicial ca ego y and ecu si e o mulae o
{Di}can be es ablished (see [13]).
In [15], a di e en app oach is p esen ed. The s ong homo opy commu a i i y
o AW is de e mined by making use o he explici Eilenbe g-Zilbe con ac ion
(AW, EML, SHI). Hence, he o mula o a mo phism Diis gi en in e ms o he
mo phisms AW,SHI, he diagonal ∆ and he au omo phism . Mo e p ecisely, he
o mula o hiis AW ( SHI)i, o all i∈N.
Now, he de ini ion o he cohomology ope a ion Sqi:Hj(X;Z2)→
Hj+i(X;Z2) (see (1)) akes he o m:
Sqi(c)(x) = (µ(< c ⊗c, AW ( SHI)j−i(x, x)>), i ≤j
0, i > j (2)
whe e c∈Cj(X;Z2) and x∈CN
i+j(X).
I is ob ious ha he o mulae o he mo phism hican be gi en in e ms o ace
and degene acy ope a o s o X. Ou objec i e in he nex sec ion is o show how o
“simpli y” hese o mulae and o ob ain an explici de ini ion o S een od squa es
only in e ms o ace ope a o s o X.
3 An explici combina o ial desc ip ion o he
cup-ip oduc s
I is clea ha he image o hilies in CN
∗(X)⊗CN
∗(X). The e o e, i we exp ess he
ac o s o he summands o he o mula o Diin a canonical way (see Lemma 2.1),
hose summands o he simpli ied o mula o Diha ing a ac o wi h a degene acy
ope a o in i s exp ession mus be elimina ed.
9
P oposi ion 4.1 Le pbe p ime wi h p≥2and ia non-nega i e in ege , hen
φdγi· · · φγ1φ= (−1)i−1φγi· · · φγ1dφ +
i−1
X
k=1
(−1)i−kφγi· · · φγk+2φγφγk−1· · · φγ1φ.
P oo
We p o e he p oposi ion by induc ion on he pa ame e i.
•I i= 1 hen, φdγ1φ=φγ1dφ.
•I i= 2 hen,
φdγ2φγ1φ=φγ2dφγ1φ=φγ2(g −1−φd)γ1φ
=−φγφ −φγ2φγ1dφ.
•In gene al,
φdγiφγi−1· · · φγ1φ=φγidφγi−1· · · φγ1φ
=φγi(g −1−φd)γi−1· · · φγ1φ
=−φγφγi−2· · · φγ1φ−φγiφdγi−1· · · φγ1φ
(by induc ion assump ion)
=−φγφγi−2· · · φγ1φ−(−1)i−2φγiφγi−1· · · φγ1dφ
−
i−2
X
k=1
(−1)i−1−kφγiφγi−1· · · φγk+2φγφγk−1· · · φγ1φ
= (−1)i−1φγi· · · φγ1dφ
+
i−1
X
k=1
(−1)i−kφγi· · · φγk+2φγφγk−1· · · φγ1φ.
2
Le Γi(k) = γiφγi−1· · · φγk+2φγφγk−1φ· · · φγ1φ∆. We can p o e
P oposi ion 4.2 Le pbe p ime and ia non-nega i e in ege , hen
γiφγi−1· · · φγ2φ∆ = γi−1φγi−2· · · φγ2φγ1φ∆ +
i−1
X
k=1
(−1)k+1Γi(k).
16

P oo
Using γk+1 =γk+ (−1)k+1γ, o all 1 < k < i, we ha e
γiφγi−1· · · φγk+2φγk+1φγk−1· · · φγ1φ∆
= γiφγi−1· · · φγk+2φγkφγk−1· · · φγ1φ∆+(−1)k+1Γi(k).
I we use his ac successi ely, we ob ain he ollowing iden i y:
γiφγi−1· · · φγ2φ∆ = γi−1φγi−2· · · φγ2φγ1φ∆ +
i−1
X
k=1
(−1)k+1Γi(k)
2
The main esul o his sec ion is he ollowing one.
Theo em 4.3 Le pbe a p ime numbe , p≥2, and ia non-nega i e in ege . Then,
he e exis s a sequence o mo phisms {Di}de ined by
Di= γiφγi−1· · · φγ1φ∆,
e i ying ha
dDi+ (−1)i+1Did=αiDi−1,
whe e α2j−1=T−1and α2j= 1 + T+T2+· · · +Tp−1.
P oo
Le us begin wi h he i s e m o he iden i y:
dDi+ (−1)i+1Did=d γiφγi−1· · · φγ1φ∆+(−1)i+1 γiφγi−1· · · φγ1φ∆d
= γidφγi−1· · · φγ1φ∆
+(−1)i+1 γiφγi−1· · · φγ1(g −1−dφ)∆
= γidφγi−1· · · φγ1φ∆
+(−1)i γiφγi−1· · · φγ2φ∆+(−1)i γiφγi−1· · · φγ1dφ∆
(by P oposi ion 4.1, we ge )
= γidφγi−1· · · φγ1φ∆+(−1)i γiφγi−1· · · φγ2φ∆
+(−1)i(−1)i−2 γiφdγi−1· · · φγ1φ∆
+
i−2
X
k=1
(−1)i(−1)k γiφγi−1· · · φγk+2φγφγk−1φ· · · φγ1φ∆
17
= γi(dφ +φd)γi−1· · · φγ1φ∆
+(−1)i γiφγi−1· · · φγ2φ∆ +
i−2
X
k=1
(−1)i−kΓi(k)
= γi(g −1)γi−1· · · φγ1φ∆
+(−1)i γiφγi−1· · · φγ2φ∆ +
i−2
X
k=1
(−1)i−kΓi(k)
(le β2j=T+T2+· · · +Tp−1and β2j−1=T, hen we ha e)
=βi γi−1φγi−2· · · φγ1φ∆
+(−1)i γiφγi−1· · · φγ2φ∆ +
i−1
X
k=1
(−1)i−kΓi(k)
(P oposi ion 4.2 implies)
=βi γi−1φγi−2· · · φγ1φ∆+(−1)i γi−1φγi−2· · · φγ1φ∆
+
i−1
X
k=1
(−1)i+k+1Γi(k) +
i−1
X
k=1
(−1)i−kΓi(k)
=αi γi−1φγi−2· · · φγ1φ∆.
2
5 P oo o he main heo em
The p oo consis s in inding ou he ac o s o he o mula (w i en in he canonical
way) ha a e degene a ed and in elimina ing he summands ha ing hese ac o s.
Fi s o all, no ice ha using he commu a i i y p ope ies o he ope a o s o a
simplicial se (essen ially, (s3)), i is easy o see ha a ac o o he o mula whose
exp ession begins (on he le ) by
∂j1· · · ∂j sk· · · (6)
such ha 0 ≤j1<· · · < j < k, is degene a ed in i s simpli ied o m.
Ha ing said ha , le us begin wi h he p oo o he heo em.
Fo n= 0, we ob ain he explici o mula o he Alexande -Whi ney ope a o .
Le us assume ha he o mula is ue o k≤nso, le us p o e ha he o mula
is ue o he case n+ 1.
18
Le us conside ha nis e en (when nis odd, he p oo is simila ). In his case,
by induc ion assump ion, he o mula o e an elemen o deg ee mis as ollows:
AW ( SHI)n+1 =AW ( SHI)n( SHI)
=
m+1
X
in=n
· · ·
i1−1
X
i0=0 X(−1)A(n)+B(n,m+1,¯ı)+C(n,m+1,¯ı)+D(n,¯ı)(−1)¯m+sig(α,β)+1
∂i0+1 · · · · ∂m+1sαp+1+ ¯m· · · sα1+ ¯m∂¯m· · · ∂m−q−1
⊗∂0· · · · ∂in−1sβq+ ¯m· · · sβ1+ ¯ms¯m−1∂m−q+1 · · · ∂m
whe e ¯ı = (i0, i1, . . . , in), ¯m=m−p−q,sig(α, β) =
p+1
X
i=1
αi−(i−1), and he
las sum is aken o e all he indices 0 ≤q≤m−1, 0 ≤p≤m−q−1 and
(α, β)∈ {(p+ 1, q)-shu les}.
Le us ecall ha i he o mula is in he no malized o m, he summands which
ha e a degene a ed ac o mus be elimina ed. The ollowing cases a e conside ed:
I in> m −p, we ha e o conside he ollowing cases:
•I in=m−p+ and βq< q −1 + , wi h 1 ≤ ≤p hen,
αp+1 =p+q > · · · > α +1 =q+ > α =q+ −1> βq.
So, he i s ac o o hese summands is:
∂i0+1 · · · · ∂in−1−1∂m−p+ +1 · · · ∂m+1sm· · · sm−p+ −1sα −1· · · sα1∂¯m· · · ∂m−q−1
=∂i0+1 · · · · ∂in−1−1sm−p+ −1sα −1· · · sα1∂¯m· · · ∂m−q−1.
Since in−1−1< in−1 = m−p+ −1, his ac o is degene a ed by (6).
•I in=m−p+ and βq=q−1 + , wi h 1 ≤ ≤p hen,
αp+1 =p+q > · · · > α +1 =q+ > βq=q+ −1.
And hence, in his case, he exp ession o he i s ac o begins in he o m:
∂i0+1 · · · · ∂in−1−1∂m−p+ +1 · · · ∂m+1sm· · · sm−p+ sα + ¯m· · ·
=∂i0+1 · · · · ∂in−1−1sα + ¯m· · · .
19
Now, we ha e o conside wo di e en cases:
–I in−1−1< α + ¯m hen, his ac o is degene a ed.
–I in−1−1≥α + ¯m, le us deno e α =q+ −1−j, whe e 1 ≤j≤q+ −1.
Then,
αp+1 =p+q > · · · > α +1 =q+ ,
and
βq=q+ −1>· · · > βq−j+1 =q+ −j > α =q+ −1−j.
Hence, he second ac o o hese summands is:
∂0· · · · ∂in−2−1∂in−1+1 · · · ∂m−p+ −1sm−p+ −1· · · sm−p+ −jsβq−j+ ¯m· · ·
=∂0· · · · ∂in−2−1sin−1· · · sm−p+ −jsβq−j+ ¯m· · ·
Since in−2−1< in−1, his ac o is degene a ed.
•I in=m−p+ and βq> q−1+ , wi h 1 ≤ ≤p hen in−1< βq+ ¯m. So, he
second ac o o he summands has he o m (6) and hence, hese summands
mus be elimina ed.
•I in=m+1 and βq< p+q hen αp+1 =p+qand since in−1−1< in−1 = m
hen, he i s ac o o he summands is degene a ed.
•I in=m+ 1 and βq=p+q hen
βq=p+q > · · · > βj+1 =p+j+ 1 > αp+1 =p+j
wi h 0 ≤j≤q−1 and he i s ac o o he summands is:
∂i0+1 · · · · ∂in−1−1sm−q+jsαp+ ¯m· · · sα1+ ¯m∂¯m· · · ∂m−q−1.
We ha e o conside wo di e en cases:
–I in−1−1< m −q+j hen, his ac o is degene a ed.
–I in−1−1≥m−q+j hen he second ac o o he summands is:
∂0· · · · ∂in−2−1∂in−1+1 · · · ∂msm· · · sm−q+j+1sβj+ ¯m· · ·
=∂0· · · · ∂in−2−1sin−1· · · sm−q+j+1sβj+ ¯m· · · ,
which is degene a ed due o he ac ha in−2−1< in−1.
20
I in< m −p, hen in−1< βq+ ¯m. So, hese summands ha e he second ac o in
he o m (6) and hence, mus be elimina ed.
I in=m−p, wo cases hold:
•I βq> q −1, hen he second ac o o hese summands is degene a ed as
abo e.
•I βq=q−1 and in−1>¯m−2 hen
αp+1 =p+q > · · · > α1=q > βq=q−1>· · · > β1= 0,
and he second ac o o he enso p oduc is
∂0· · · · ∂in−2−1∂in−1+1 · · · ∂m−p−1sm−p−1· · · sm−p−q−1∂m−q+1 · · · ∂m
=∂0· · · · ∂in−2−1sin−1· · · sm−p−q−1∂m−q+1 · · · ∂m;
and since in−2−1< in−1, his ac o is degene a ed.
Finally, i βq=q−1 and in−1≤¯m−2 hen he o mula (sa e o he signs)
co esponding o AW ( SHI)n+1 is:
¯m−2
X
in−1=n−1
· · ·
i1−1
X
i0=0
m−1
X
q=0
m−q−1
X
p=0
∂i0+1 · · · · ∂in−1−1∂m−p+1 · · · ∂m+1sm· · · sm−p∂¯m· · · ∂m−q−1
⊗∂0· · · · ∂in−2−1∂in−1+1 · · · ∂m−p−1sm−p−1· · · s¯m−1∂m−q+1 · · · ∂m
=
m
X
i0
n+1=n+1
i0
n+1−1
X
i0
n=n
i0
n−1
X
in−1=n−1
· · ·
i1−1
X
i0=0
∂i0+1 · · · · ∂in−1−1∂i0
n+1 · · · ∂i0
n+1−1
⊗∂0· · · · ∂in−2−1∂in−1+1 · · · ∂i0
n−1∂i0
n+1+1 · · · ∂m;
whe e i0
n= ¯m−1 and i0
n+1 =m−q.
Now, le us s udy he signs o he o mulae in his las case. Keeping in mind
ha we a e wo king wi h he exponen o (−1), all he iden i ies a e mod 2.
This p oo is based on he ac ha i and only i in=m−pand βq=q−1,
he summands o AW ( SHI)na e non-degene a ed.
21

We i s e i y he o mula in he case n= 1. The exponen o (−1) associa ed
o each summand is:
¯m+sig(α, β) + 1 = ¯m+
p+1
X
i=1
αi−(i−1) + 1
= ¯m+
p+1
X
i=1
(q−1 + i−(i−1)) + 1 = ¯m+q(p+1)+1
(no e i0= ¯m−1 and i1=m−q)
=i0+ (m+i1)(i0+i1) = A(1) + B(1, m,¯ı) + C(1,¯ı) + D(1, m,¯ı)
whe e ¯ı = (i0, i1), as asse ed.
In gene al, we ha e o p o e ha
A(n) + B(n, m + 1,¯ı) + C(n,¯ı) + D(n, m + 1,¯ı) + ¯m+sig(α, β)+1
=A(n+ 1) + B(n+ 1, m,¯ı) + C(n+ 1,¯ı) + D(n+ 1, m,¯ı).
•I nis e en hen D(n, m+1,¯ı) = 0 and he exponen o (−1) in each summand
is:
A(n) + B(n, m + 1,¯ı) + C(n,¯ı) + ¯m+sig(α, β)+1
=A(n) + B(n, m + 1,¯ı) + C(n−1,¯ı)
+(m+p+in−1)(in−1+· · · +i0) + ¯m+q(p+1)+1
(since i0
n= ¯m−1, i0
n+1 =m−qand hese iden i ies a e mod 2 hen)
=A(n) + B(n, m + 1,¯ı) + C(n−1,¯ı)
+(m+1+i0
n+1 +i0
n+in−1)(in−1+· · · +i0)
+i0
n+ (m+i0
n+1)(i0
n+1 +i0
n)
=A(n) + B(n, m + 1,¯ı) + C(n−1,¯ı) + in−1+· · · +i0
+(i0
n+in−1)(in−1+· · · +i0)+(m+i0
n+1)(in−1+· · · +i0)
+i0
n+ (m+i0
n+1)(i0
n+1 +i0
n)
=A(n) + B(n, m + 1,¯ı) + i0
n+in−1+· · · +i0
+C(n+ 1,¯ı) + D(n+ 1, m,¯ı).
22
We ha e o dis inguish wo cases:
–I n≡0 mod 4 hen A(n) = A(n+ 1) and
A(n) + B(n, m + 1,¯ı) + i0
n+in−1+· · · +i0
=A(n+ 1) +
n−2
2
X
j=0
i2j+1 +i0
n+in−1+· · · +i0
=A(n+ 1) +
n−2
2
X
j=0
i2j+i0
n=A(n+ 1) + B(n+ 1, m,¯ı).
–I n≡2 mod 4 hen A(n) = A(n+ 1) + 1 and
A(n) + B(n, m + 1,¯ı) + i0
n+in−1+· · · +i0
=A(n+1)+1+
n−2
2
X
j=0
i2j+m+p+i0
n+in−1+· · · +i0
=A(n+1)+1+
n−2
2
X
j=0
i2j+m+i0
n+1 +i0
n+ 1
+i0
n+in−1+· · · +i0
=A(n+ 1) +
n−2
2
X
j=0
i2j+1 +i0
n+1 +m=A(n+ 1) + B(n+ 1, m,¯ı).
•I nis odd hen he exponen o (−1) is:
A(n) + B(n, m + 1,¯ı) + C(n,¯ı) + D(n, m + 1,¯ı) + ¯m+sig(α, β)+1
=A(n) + B(n, m + 1,¯ı) + C(n,¯ı)
+(m+1+m+p)(m+p+in−1+· · · +i0) + ¯m+q(p+1)+1
(since i0
n= ¯m−1, i0
n+1 =m−qand hese iden i ies a e mod 2 hen)
=A(n) + B(n, m + 1,¯ı) + C(n,¯ı)
+(i0
n+1 +i0
n)(m+1+i0
n+1 +i0
n+in−1+· · · +i0)
+i0
n+ (m+i0
n+1)(i0
n+1 +i0
n)
=A(n) + B(n, m + 1,¯ı) + C(n,¯ı) + (i0
n+1 +i0
n)(i0
n+in−1+· · · +i0)
+i0
n+1 +i0
n+ (i0
n+1 +i0
n)(m+i0
n+1) + i0
n+ (m+i0
n+1)(i0
n+1 +i0
n)
=A(n) + B(n, m + 1,¯ı) + C(n,¯ı)
+(i0
n+1 +i0
n)(i0
n+in−1+· · · +i0) + i0
n+1
=A(n) + B(n, m + 1,¯ı) + i0
n+1 +C(n+ 1,¯ı) + D(n+ 1, m,¯ı).
23
Since nis odd hen n≡1,3,5,7 mod 8 and, in hese cases, A(n) = A(n+ 1).
Now, we ha e o dis inguish wo cases:
–I n≡1 mod 4 hen
B(n, m + 1,¯ı) + i0
n+1 =
n−1
2
X
j=0
i2j+i0
n+1 =
n+1
2
X
j=0
i2j=B(n+ 1, m,¯ı).
–I n≡3 mod 4 hen
B(n, m + 1,¯ı) + i0
n+1 =
n−3
2
X
j=0
i2j+1 +in+m+1+i0
n+1
=
n−3
2
X
j=0
i2j+1 +i0
n+m+1+m+ 1
=
n−1
2
X
j=0
i2j+1 =B(n+ 1, m,¯ı).
2
Re e ences
[1] J. Adem. The i e a ion o he S een od squa es in Algeb aic Topology. P oc.
Na . Acad. Sci. USA, . 38 (1952), 720-724.
[2] M. And ´e. M´e hode Simpliciale in Alg`eb e Homologique e Alg`eb e Commu a-
i e. Lec u e No es in Ma hema ics, . 32, Sp inge -Ve lag (1967).
[3] J. Dieudonn´e. A his o y o Algeb aic and Di e en ial Topology 1990-1960.
Bi kh¨ause , Bos on (1989).
[4] A. Dold. Ube die S een odschen Kohomologieope a ionen. Annals o Ma h.,
. 73 (1961), 258-294.
[5] S. Eilenbe g and S. Mac Lane. Acyclic models. Am. J. Ma h., . 67 (1953),
282-312.
[6] S. Eilenbe g and S. Mac Lane. On he g oups H(π, n), I. Annals o Ma h.,
. 58 (1953), 55-106 .
[7] S. Eilenbe g and S. Mac Lane. On he g oups H(π, n), II. Annals o Ma h.,
. 60 (1954), 49-139.
24
[8] S. Eilenbe g and J. A. Zilbe , On p oduc s o complexes. Am. J. Ma h., . 75
(1959), 200-204.
[9] T. Ekedahl , J. G abmeie and L. Lambe. Algo i hms o algeb aic compu a ions
wi h applica ions o he cohomology o ini e p-g oups. P ep in o Depa men
o Ma hema ics and Cen e o Inno a i e Compu a ion, Uni e si y o Wales
(1997).
[10] L. Lambe. An algo i hm o calcula ing cocycles. P ep in o Depa men o
Ma hema ics and Cen e o Inno a i e Compu a ion, Uni e si y o Wales
(1997).
[11] S. Mac Lane. Homology. Classics in Ma hema ics. Sp inge -Ve lag, Be lin
(1995). Rep in o he 1975 edi ion.
[12] P. May. Simplicial objec s in Algeb aic Topology. Van Nos and, P ince on
(1967).
[13] P. May. A gene al algeb aic app oach o S een od ope a ions. Lec . No es in
Ma h., Sp inge , . 156 (1970), 153-231.
[14] P. Real. Homological Pe u ba ion Theo y and Associa i i y. P ep in o De-
pa men o Applied Ma hema ic I, Uni e si y o Se ille (1996).
[15] P. Real. On he compu abili y o he S een od squa es. Annali de’ll Uni e si `a
di Fe a a, sezione VII. Scienze Ma ema iche, . XLII (1996), 57-63.
[16] J. Rubio. Homologie e ec i e des espaces de lace s i ´e ´es: un logiciel. Th`ese de
doc o a de l’Ins i u Fou ie , G enoble (1991).
[17] E. H. Spanie . Algeb aic Topology. New Yo k: McG aw-Hill (1966) (Rep in ed
by Sp inge (1981)).
[18] N. E. S een od. P oduc s o cocycles and ex ensions o mappings. Annals o
Ma h., . 48 (1947), 290-320.
[19] N. E. S een od. Reduced powe s o cohomology classes. Ann. o Ma h., . 56
(1952), 47-67.
[20] N. E. S een od and D. B. A. Eps ein. Cohomology ope a ions. Ann. o Ma h.
S udies . 50, P ince on Uni e si y P ess (1962).
[21] C. A. Weibel. An in oduc ion o homological algeb a. Camb idge s udies in
ad anced ma hema ics, 38, Camb idge Uni e si y P ess (1994).
25