Annales de l’institut Fourier (AIF)

Amerik, Ekatarina

Rovinsky, Marat

Van De Ven, Antonius

A boundedness theorem for morphisms between threefolds

Brisebarre, Nicolas

Habsieger, Laurent

Sur les fonctions entières à double pas récurrent

Courcelle, Bruno

Olive, Frédéric

Une axiomatisation au premier ordre des arrangements de pseudodroites euclidiennes

Brinkmeier, Michael

ANNALES DE L’INSTITUT FOURIER
MICHAEL BRINKMEIER
The Milgram non-operad
Annales de l’institut Fourier, tome 49, no 5 (1999), p. 1427-1438
<http://www.numdam.org/item?id=AIF_1999__49_5_1427_0>
© Annales de l’institut Fourier, 1999, tous droits réservés.
L’accès aux archives de la revue « Annales de l’institut Fourier »
(http://annalif.ujf-grenoble.fr/) implique l’accord avec les conditions gé-
nérales d’utilisation (http://www.numdam.org/conditions). Toute utilisa-
tion commerciale ou impression systématique est constitutive d’une in-
fraction pénale. Toute copie ou impression de ce fichier doit conte-
nir la présente mention de copyright.
Article numérisé dans le cadre du programme
Numérisation de documents anciens mathématiques
http://www.numdam.org/
Ann. Inst. Fourier, Grenoble
49, 5 (1999), 1427-1438
THE MILGRAM NON-OPERAD
by Michael BRINKMEIER
Introduction.
R.J. Milgram introduced in [10] geometric models J71 for the iterated
loop-space operads ^E^. Later J.M. Boardman, R.M.Vogt and P.May
proved that n-fold loop spaces are closely related to En operads in general,
and thus to the little cube operads (see [5], [6], [8]). Hence the question
arises if the operad structure of^E71 translates to the geometric model J71.
C. Balteanu, Z. Fiedorowicz, R. Schwanzl and R.M. Vogt construct
in [1] an operad A^n, which codifies n-fold monoidal categories, a categorial
analog of n-fold loop spaces. They observe that an equivalent preoperad is
embedded in M.ni whose free space is of the same homotopy type as J^X.
For n = 2 the spaces are even homeomorphic (see [I], §§ 13.12-13.14]).
Due to the underlying polytopes, the permutohedra, the Milgram-
construction J71 is of some importance for the examination of coherent
homotopy-commutativity. Similar to the associahedra introduced by
Stasheff in [II], Williams uses the permutohedra and the Milgram-
const ruction in [12] to define his notion of C^-spaces, which is used in
several subsequent papers, and is occuring in papers of McGibbon and
Hemmi (see for example [9] and [7]).
In fact there exists an operad structure with the permutohedra as
underlying spaces (this was pointed out to me by Clemens Berger and Zig
Fiedorowicz), simply by using the convex extension of the permutation
operad. But since the permutohedra are contractible this is an Eoo operad
in the sense of Boardman and Vogt (i.e. the symmetric group action does
-Keywords : En -operad — Permutohedron — Free n-fold loop space.
Math. classification: 06A07 - 20B30 - 55P35.
1428 MICHAEL BRINKMEIER
not need to be free). Therefore an algebra of this operad is homotopy
equivalent to an infinite loop space and hence its associated monad can not
be of the same homotopy type as the Milgram construction J71 which is
just an n-fold loop space.
C. Berger conjectured in [3] and [4] an En operad structure of the
permutohedra, whose associated monad is the Milgram construction Jn.
I will show that Berger's construction does not work. The first obser-
vation is that the would-be operad bears a structure far too strong, namely
that of strictly abelian monoids. This collapse of structure is then used to
show that the suggested structure does not define an operad at all. In fact
the proof shows that the multiplication defined by Berger does not respect
the degeneration conditions.
Nonetheless the construction defines preoperads J^\ which are
homotopy equivalent to the little n-cubes. In particular, the k-th
space J^ is Sfc-equivariantly homotopy equivalent to the real configuration
space ^(R71,^). Furthermore the "non-monad" associated to the pre—
operad J^ is an alternative description of the Milgram construction Jn.
I would like to thank Zig Fiedorowicz and especially Clemens Berger
for some helpful and enlightening discussions, and James McClure for the
discovery of some mistakes and errors in an earlier version of this paper.
I thank the Deutsche Forschungsgemeinschaft for support during the
preparation of this paper.
1. The permutohedra and Berger ̂ s construction.
Let A be the category of finite sets n = { 1 , . . . , n} and injective maps.
Each injective map ( p : n —^ m has an unique decomposition of the form
(p = (^lnc o (^, such that y?1110: n —>• m is increasing and ^ : n —> n is a
permutation.
DEFINITION 1.1. — The permutation preoperad S : A015 —> Top is the
functor with S(n) = Eyi and </?* := S((^) : Syn —^ ^n given by a ̂  {a o ̂
for (p e A(n,m).
The multiplications of the permutation operad S are given by
7 ,̂..,z, '• Sn X S,, X . . . X S,, —— ̂ +...+,,
(<7;Ti,...,Tn) 1——>a(z i , . . . ,Zn)o(n C • • • © 7-n)
THE MILGRAM NON-OPERAD 1429
where c r ( % i , . . . ,^n) permutes the blocks {see [4], 1.15 (a)) and — (B — :
Sn x Sy^ —» Sn+m i-s the canonical product of permutations.
The product S^ 9 • • • C S^ C S .̂..-^ wiJJ be denoted with
S(%i, . . . ,^) .
For more details about (pre)operads in general the reader is referred
to [3], [4] or [8].
In contrary to Clemens Berger I will use the left action of the
symmetric group on IR"", which seems to be the more common description.
DEFINITION 1.2. — For n > 1 the symmetric group S^ acts on R71
from the left by permuting the coordinates in the following way:
d{x^ . . . ,Xn) =(a^-i(i), . . . ,^-i(^)).
The n-th permutohedron Pn C M71 is the convex hull of the orbit of
(1,2, . . . , n) € y1 under this operation.
The convex hull of the orbit of S ( % i , . . . , ir) C S^ will be denoted
with P(^i , . . . , ir) C Pn' The point a(l , . . . , n) € Pn will be denoted with a.
Remark 1.3. — The notation of a for the point a( l , . . . ,n) =
((T"'1^),... ̂ ^(n)) seems somewhat confusing. But since we will extend
the permutation operad to the permutohedra, we can calculate the vertices
right from the permutations, without applying it to the R71.
The geometric and simplicial properties of these polygons were
examined in [10], [12] and [2]. Here I will give only a rough sketch of
the few details I will use.
PI consists of only one point, ?2 is homeomorphic to the unit interval
in R and ?3 to the hexagon in R2. In general Pn is a (n — l)-dimensional
polytope.
Obviously there exists a left S^-action on Pn. The vertices are
mapped to vertices, and for each a 6 Syi the map Pn —^ Pn with x ^—> ax
is a homeomorphism. But unfortunately this action is not free, since the
barycenter of each permutohedron is a fixed point.
An arbitrary point x € Pn will be denoted by a linear combination of
permutations
x = V^ taO- with y^ to- = 1.
<r€Sn o-CEn
1430 MICHAEL BRINKMEIER
If x = ^ s^o- and y = ^ t^r are points of Pn and Pyyi respectively, the
creE^ rGE^
point a; 9 ̂  C Pn+m is given by
^ © y = ̂  ^ SatrO- C T.
o-eE^ TeEm
Milgram defined in [10] maps Ik : Pk x Pn-fc -^ ^n given by (a:, y) ̂  x C ?/
mapping the product of two permutohedra into certain faces of a higher
dimensional permutohedron. More general the codimension (r — l)-faces
of Pn are in one-to-one correspondence with the ordered partitions of
{ ! , . . . , n} of type ( % i , . . . ,ir) with %i + • • • + ir = n,ik ^ 1.
Remark 1.4. — Each partition of type ( z i , . . . , ir) can be interpreted
as a permutation of i\ + • - • + ir elements. If the classes are given by
VI? • • • ^ J i i } ^ U^i+15 • • • ̂ ii^}^ • • ' ? Un+---+Zr-i+l^ • • • ^ii-}----\-ir}
with jk < 3t for im < k < £ < Zm+i; then the corresponding permutation is
given by k i—> jk for 1 <^ k <: %i + • • • + ir.
The converse does not hold, since the same permutation can be
associated to different partitions. For example the identity in Ss corresponds
to the partion {1}, {2,3} and to {1, 2}, {3}.
The vertices of the codimension (r — l)-face, corresponding to a
partition of type ( z i , . . . ,z^) , with associated permutation a € S^, are
given by the coset aS(%i, . . . ,z^) . In addition there is a homeomorphism
la : P^ x • • • x P^ -^ c rP(zi , . . . , ir) C Pn with
(.ri,...,^) i—> a(rci C • • • ©a^).
Using the (right) weak Bruhat order on the symmetric groups, the 1-skeleton
of these faces can be oriented.
DEFINITION 1.5. — The inversion index inv(a) of a permutation
a G Sn is the number of ordered pairs (z,j), 1 < i < j < n, whose orders
are inverted by a, i.e. a{i) > cr(j').
The (right) weak Bruhat order ofY^n is the partial order generated by
a < r, ifr is the composition of a and a transposition of two subsequent
numbers, that is r = a o (z, i + 1), and inv(cr) < inv(r).
Remark 1.6. — Since we use the left action of Syi on R71, we have to
use the right weak Bruhat order, instead of the left weak Bruhat order, as
Clemens Berger did.
THE MILGRAM NON-OPERAD
Example 1.7. — Es is given by the poset
(123)
(213) (132)
1 1
(231) (312)
(321)
Here and in the following the permutation
1431
( 1 2 • • • n \
\a(l) cr(2) ... a(n))
will be denoted with (cr(l) ,a(2), . . . ,cr(n)) (the commas will be left, if
unnecessary).
Remark 1.8. — If the vertices of the poset {i.e. the permutations)
are interpreted as points in M3, it does not seem to describe the border
of ?3 - the points (2,3,1) and (3,1,2) have to be exchanged. But since we
use the left action of S^ on R72, the permutation (231) corresponds to the
point (3,1,2) and the permutation (312) to the point (2,3,1). In fact the
correspondence holds for all vertices of Ps and the poset Ss.
Applying this partial order to the permutohedra the edges will be
oriented. The face corresponding to a certain partition of type ( z i , . . . , ir)
and associated permutation a has exactly one initial vertex, given by a and
a unique terminal vertex, given by the permutation that turns the classes
of the partition "upside-down", i. e.
t\—> k^ with £' = 2(?i + • • • +^) +^+i -^+ 1 if n < £ ^ ^+1.
DEFINITION 1.9. — An interval [ai.aa] in En with the weak Bruhat
order is called admissible^ if it is the vertex set of some coset crS(zi , . . . , ir).
Hence a\ is the initial and a^ the terminal vertex of a face of Pn.
The geodesic of an arbitrary interval [o-i, 02] of the weak Bruhat order
in Syi, is the average of all oriented edge-paths between a\ and 02 in Pn.
The barycenter of [cri, (T^\ is the barycenter of its geodesic.
1432 MICHAEL BRINKMEIER
(123)
(321)
Figure 1. The geodesic and barycenter of PS
The barycenter of an admissible interval coincides with the barycenter
of the corresponding cell. For example the geodesic of [(123), (213)] C ^3
is the corresponding edge of PS, and its barycenter the barycenter of
the interval. The geodesic of [(123), (321)] C 1:3 is the line between
(123) and (321) and its barycenter is the point B. The geodesic of the
non-admissible intervall [(123), (231)] C Ss consists of the two edges
[(123), (213)] and [(213), (231)] of Pa. Hence its barycenter is the point
(213) e Pa (see Fig. 1).
To define an operad-multiplication on the permutohedra we have to
define maps Pr x P^ x • • • x P^ —> P^+..._^ which satisfy the associativity
conditions. But the intention to formulate an operad, whose associated
monad is the Milgram construction gives certain additional restrictions to
the choice of the multiplication.
The operad structure has to extend the permutation operad S.
This can be done very easily by mapping the vertices of the product
Pr x P^ x • • • x P^ to the corresponding vertices of P^+...+^, given by S.
But the extension of this map can be done in two different ways.
The first possibility is the convex extension of the permutation operad,
by mapping Pr x P^ x • • • x P^ to the convex hull of the image vertices
(see Fig. 2). This construction does in fact define an operad whose n-th
space is Pyi. It is £'oo m the sense ofBoardman and Vogt, i.e. its underlying
spaces are contractible but the actions of the symmetric groups does not
need to be free. Thus its algebras are homotopy equivalent to infinte loop
spaces (cmp. [6], §VI.3). But since the Milgram construction is only a
model for n-fold loop spaces the associated monad of the convex extension
is of the wrong homotopy type.
THE MILGRAM NON-OPERAD 1433
(123)(21);(1),(12) (21);(1),(21)
(213) ̂  / \ \. (132)
(^l)^ / ^ (312)
V.__________I
(12);(1),(12) (12);(1),(21)
?2 X PI X ?2
Figure 2. The convex multiplication
Berger tried to get the correct operad by application of two changes
to the convex extension of the permutation operad. First he deformed the
multiplication such that it respects the relations on the borders of the
permutohedra, given in Milgrams construction. But this does not affect the
homotopy type of the spaces. In a second step he made the action of the
symmetric group free, which would give the correct homotopy type.
In the Milgram construction the border of Pr x P^ x • • • x P^ has to
be mapped to the border of P^4-...+^. Berger did this by using a cubical
extension of the symmetric operad instead of the convex one. He denned
the operad-multiplication 7^^...^ : Pn x Pii x - • • x P^ —^ P^+...+p^ as
the affine extension of the permutation operad, such that the bary center
of any interval in E^-}-...-}-^ is mapped to the barycenter of its geodesies in
P^+...4-^ (see Fig. 3). Hence for n = 2 and i\ + ^2 = 3 the squares A, jE?, C
and P of ?2 x Pi x ?2, resp. Pa x ?2 x Pi are mapped to the corresponding
segments of PS.
The second step, in which the symmetric group action is made free, is
done in the definition of the would-be operad. For each permutation in S^
a copy of P^ is added and an appropriate quotient of the space P^ x E^ is
taken to be the n-th space of the new Ek operad.
Here I will only describe the suggested construction for the operad J^
corresponding to J^ ^6. 2-fold loop spaces.
DEFINITION 1.10 (cmp. [4], § 2.12). — Let J^ be the quotient space
of Pn x Ey^ under the relation
(rx\ a) ~ (x', ra)
for any partition r € Sn of type (^i,... ,Zr)^i 4- • • • + ir = n, where
x e P(zi , . . . ,%y) and ff € Sn.
1434 MICHAEL BRINKMEIER
(123)(21);(1),(12) (21);(1),(21)
^P
72;1,2
B
A
C
D
(12);(1),(12) (12);(1),(21)
P2 X PI X ?2
(213)
(231)
(132)
(312)
(123)
(213)
(231)
(21);(12),(1) (21);(21),(1)
(132) p
^2-2.1
(312)
(12);(12),(1) (12);(21),(1)
?2 X ?2 X P\
Figure 3. The cubical multiplication
The action of y? C A(n, n) is induced by
^ : Pm X ̂ Pn X S^
^ ̂  (^r^a^).
Remark 1.11. — Since I use the left instead of the right action, the
relation given by Berger has to be changed slightly.
Following [4], § 2.14, the A structure and the maps 7^...^ x 7^ ^
fsy\
induce an E^-operad structure on Jn .
2. The commutativity ofBerger^s construction.
In the following we assume that Berger's construction defines an
operad. If (X, *) is a J^-algebra, there exist maps Fn : Pn x Sn x X71 —> X.
THE MILGRAM NON-OPERAD 1435
These fulfil certain conditions, induced by the operad structure and the
relations on J^. Used here are
1) the associativity condition
F,(5;^=(^;^,...,^),...,F,J^;^,...,^))
= F^+...+^ (7n;^l,...,^n(ts^l^ • • ̂  ̂ n)\X\, . . . a^J;
2) the degeneration relations on the associated monad, induced by
the two maps 1-^2, F^(s',x^) = F^x) = x = i^O?;*, x);
3) and the S^-equivariance
Fn(sa',X^, . . . , Xn) = Fn(5;^-i(i), . . . , ̂ -i(^)).
Remark 2.1. — Since only the identities of S^ are used, in the future
this coordinate will be dropped.
There is a multiplication on X, given by xy = F^((12),x,y). The
associativity of the permutation operad shows the associativity of this
multiplication. Since the degeneration relations hold * is a unit. Therefore
X is a associative monoid with strict unit.
Obviously the map ( f ) : I —^ P^ with t ^ (12)(1 - t) + (21)t is a
homeomorphism. Thus there exists a homotopy ht: X x X —> X with
h^y) = F^((12)(t - 1) + (21)^,2/),
running from ho(x, y) == xy to
hi(x^y) = F2((21);^,2/)) = ̂ ((12);^) = yx.
Via Fs and the degenerations one gets even stricter conditions for the
commuting homotopy ht. The maps 7^2 and 7^3 ^ are homeomorphisms.
Therefore there exist 5^, ti € P2^ = 1,2 for each r e PS such that
7f:2,l(^2,(l)) =y=7^1,2(5l;(l)^l).
Hence for x, y e X one gets
F2(s^x,y) = F3(r;*,a;,2/) = F^(t^x,y).
In the first case the homotopy ht is mapped to the edges [(123), (132)] and
[(231), (321)] (cmp. Fig. 4). In the second case ht is mapped to the geodesies
of the intervals [(123), (312)] and [(213), (321)], such that the center point of
the homotopy h^ (x,y) is mapped to (132) and (231). Thus the homotopy
needs to be equal to xy on its first half and equal to yx on the second
half (in the first case the edge [(213), (231)] is mapped to xy and the edge
[(132), (312)] to yx). Thus (X, *) must be an abelian monoid.
1436 MICHAEL BRINKMEIER
(123)(21);(1),(12) (21);(1),(21)
^1,2
(12);(1),(12) (12);(1),(21) (321)
(123)
(213)
(231)
(21);(12),(1) (21);(21),(1)
(132)
(312)
^2-2,1
} ht(x,y}
(321) (12);(12),(1) (12);(21),(1)
Figure 4. The commuting homotopy
Remark 2.2. — Since the permutation coordinate wasn't used, the
freeness of the symmetric group action is not involved in the failure of the
suggested construction. In fact the cubical extension of the multiplication
does not fulfil the needed degeneration properties.
Now let X be a 2-connected CW-complex with non-degenerate base
point * (i.e. the inclusion * ^-> X is a closed cofibration). Then the two-fold
Moore loop space Y: = fl^X of X is a connected CW-complex. The
canonical evaluation map e: S^^X -^ X induces a homomorphism of
monoids Q^e: ̂ T^Y -> Y.
If J^ is an E^ operad, whose associated monad is the Milgram
construction, there exists a homomorphism of monoids ̂ : J^Y —> ̂ I^V
(the map is given in [10], §5.2). Therefore we get a homomorphism
( p : = ^eo-^J^y -» y. From the construction of -0 in [10] one
can see that the diagram
j(2)y _^ ̂ .Y ^^ Y
THE MILGRAM NON-OPERAD 1437
with the inclusion i: Y ^—>- f^Z^y given by
( i ( y ) ) ( s , t ) = y ^ s / \ t ,
commutes. Therefore the homomorphism (p = ^^-'0 is an extension of the
identity and hence surjective.
Since J^Y is the free J^-algebra of the space V, it is an abelian
monoid. The surjectivity of (p now shows that Y = fl]^X is strictly
commutative, too. But obviously this is wrong. Therefore J^ can not be
an operad whose associated monad is the Milgram construction.
BIBLIOGRAPHY
[1] C. BALTEANU, Z. FIEDOROWICZ, R. SCHWANZL, R. VOGT, Iterated monoidal
categories, preprint 98-035, Universitat Bielefeld, 1998.
[2] H.-J. BAUES, Geometry of loop spaces and the cobar-construction, Mem. Amer.
Math. Soc., 230 (1980).
[3] C. BERGER, Operades cellulaires et espaces de lacets iteres, Ann. Inst. Fourier,
46 (1996), 1125-1157.
[4] C. BERGER, Combinatorial models for real configuration spaces and en-operads,
Cont. Math., 202 (1997), 37-52.
[5] J.M. BOARDMAN, R.M. VOGT, Homotopy-every thing /i-spaces, Bull. Amer.
Math. Soc., 74 (1968), 1117-1122.
[6] J.M. BOARDMAN, R.M. VOGT, Homotopy invariant algebraic structures on
topological spaces, Lecture Notes in Math. 347 (Springer, Berlin), 1973.
[7] Y. HEMMI, Higher homotopy commutativity of H-spaces and the mod p torus
theorem, Pac. J. Math., 149 (1991), 95-111.
[8] J.P. MAY, The geometry of iterated loop spaces, Lecture Notes in Math. 271
(Springer, Berlin), 1972.
[9] C.A. McGlBBON, Higher forms of homotopy commutativity and finite loop
spaces, Math. Zeitschrift, 201 (1989), 363-374.
[10] R.J. MILGRAM, Iterated loop spaces, Ann. of Math., 84 (1966), 386-403.
1438 MICHAEL BRINKMEIER
[11] J.D. STASHEFF, Homotopy associativity of /i-spaces, I, Trans. Amer. Math. Soc.,
108 (1963), 275-292.
[12] F.D.WILLIAMS, Higher homotopy-commutativity, Trans. Amer. Math. Soc., 139
(1969), 191-206.
Manuscrit recu Ie 21 octobre 1998,
accepte Ie 19 mars 1999.
Michael BRINKMEIER,
Universitat Osnabruck
Fachbereich Mathematik/Informatik
49069 Osnabriick (Germany).
mbrinkme@ mat hemat ik. uni—osnabrueck. de


The Milgram non-operad

Balibrea, Francisco

Peña, Jose Salvador Cánovas

López, Víctor Jiménez

Commutativity and non-commutativity of topological sequence entropy

Merdy, Christian Le

Finite rank approximation and semidiscreteness for linear operators

Allouche, Jean-Paul

Peyrière, Jacques

Wen, Zhi-Xiong

Wen, Zhi-Ying

Hankel determinants of the Thue-Morse sequence

Huebschmann, Johannes

Lie-Rinehart algebras, Gerstenhaber algebras and Batalin-Vilkovisky algebras

Iwatsuka, Akira

Tamura, Hideo

Asymptotic distribution of negative eigenvalues for two dimensional Pauli operators with nonconstant magnetic fields

Alexander, Herbert

Hulls of subsets of the torus in ${\mathbb {C}}^2$

Not a member yet