symmetric monoidal (∞,1)-category of spectra
An $H$-space is a magma in the homotopy category of topological spaces Ho(Top), or in the homotopy category category $Ho(Top)_*$ of pointed spaces, which has a unit up to homotopy.
An $H$-monoid is a monoid object in Ho(Top), hence an $H$-space is an $H$-monoid if the product of the magma is associative up to homotopy.
An $H$-group is a group object in Ho(Top), so an $H$-monoid is an $H$-group if has also a inverses up to homotopy.
To continue in this pattern, one could say that an H-category is an Ho(Top)-enriched category.
Notice that here the homotopies for units, associativity etc. are only required to exist for an H-space, not required to be equipped with higher coherent homotopies. An $H$-monoid equipped with such higher and coherent homotopies is instead called a strongly homotopy associative space or $A_\infty$-space for short. If it has only higher homotopies up to level $n$, it is called an $A_n$-space.
A better name for an $H$-space would be $H$-unitoid, but it is rarely used. The $H$ stands for Heinz Hopf, and reflects the sad fact that the natural name ‘homotopy group’ was already occupied; Hopf and A. Borel found necessary algebraic conditions for a space to admit an $H$-space structure.
Does ‘unitoid’ generically mean a magma with identity? I can't verify this, but it would be convenient to have such a term. —Toby
Good question. Postnikov in his 1980-s course of homotopy theory, talks about internal unitoids in a fixed category before modifying the notion to H-version; however he does complaint that in the universal algebra the term is not standardized. – Zoran
Thanks, that's something to start with at least. —Toby
There are dual notions of $H$-counitoid (or $H'$-space, or co-H-space), $H$-comonoid (or $H'$-monoid) and $H$-cogroup (or $H'$-group) having co-operations with the usual identities up to homotopy.
The main example of an $H$-group is the loop space $\Omega X$ of a space $X$, which however is naturally even an A-infinity space.
(The main example of an H-cogroup in $Top_*$ is the suspension $S X= S^1\wedge X$ of a pointed topological space $X$. )
An example of an H-space that does not lift to an A-infinity space is the 7-sphere $S^7$. It can’t be delooped because its delooping would have cohomology group a polynomial ring on a generator in degree 8, and this is impossible by mod $p$ Steenrod operations for any odd $p$.
If $K$ is an $H$-group then for any topological space $X$, the set of homotopy classes $[X,K]$ has a natural group structure in the strict sense; analogously if $K'$ is an $H$-cogroup then $[K',X]$ has a group structure. If there is more than one $H$-group structure on a space, then the induced group structures on the set of homotopy classes coincide.
If an $H$-space is equivalent to a deloopable one, then it is a group object in the (∞,1)-category Top. In other words, in that case, the associativity and other axioms hold up to coherent homotopy.
For more details see at loop space.
Given an A-∞ space in the (∞,1)-category ∞Grpd/Top, its image in the corresponding homotopy category of an (∞,1)-category Ho(Top) in an $H$-space. Conversely, refining an H-space to a genuine $A_\infty$-space means lifting the structure of a monoid object in an (∞,1)-category from the homotopy category Ho(Top) to the genuine (∞,1)-category ∞Grpd/Top.
Further discussion of this is also at loop space – Homotopy associative structure
The terminology $H$-space is a definition in a Chapter IV, Section 1 (dedicated to loop spaces) of
Some other papers in the 1950s include
E. H. Spanier, J. H, C, Whitehead, On fibre spaces in which the fibre is contractible, Comment. Math. Helv. 29, 1955, 1–8.
Arthur H. Copeland, On $H$-spaces with two nontrivial homotopy groups, Proc. AMS 8, 1957, 109–129.
M. Sugawara:
F. I. Karpelevič, A. L. Oniščik, Algebra of homologies of space of paths, Dokl. Akad. Nauk. SSSR, (N.S.), 106 (1956), 967–969. MR0081478
The theory of $H$-spaces was widely established in the 1950s and studied by Serre, Postnikov, Spanier, Whitehead, Dold, Eckmann and Hilton and many others.
The deloopable case has more coherent structure which has been discovered few years later in J. Stasheff’s thesis and published in
For a historical account see
The description in terms of groupoid object in an (∞,1)-category is due to
see last remark of section 6.1.2 .
Wikipedia's definition (at time of this writing, and phrased in the language of homotopy theory) is rather a unitoid object in the $(\infty,1)$-category Top.