In differential geometry, a hypercomplex manifold is a manifold with the tangent bundle equipped with an action by the algebra of quaternions in such a way that the quaternions define integrable almost complex structures.
(with acting as a multiplication by a quaternion , ) is hypercomplex, but not Kähler, hence not hyperkähler either. To see that the Hopf surface is not Kähler, notice that it is diffeomorphic to a product hence its odd cohomology group is odd-dimensional. By Hodge decomposition, odd cohomology of a compact Kähler manifold are always even-dimensional. In fact H. Wakakuwa proved  that on a compact hyperkähler manifold . M. Verbitsky has shown that any compact hypercomplex manifold admitting a Kähler structure is also hyperkähler. 
In 1988, left-invariant hypercomplex structures on some compact Lie groups were constructed by the physicists Ph. Spindel, A. Sevrin, W. Troost, A. Van Proeyen. In 1992, D. Joyce rediscovered this construction, and gave a complete classification of left-invariant hypercomplex structures on compact Lie groups. Here is the complete list.
where denotes an -dimensional compact torus.
It is remarkable that any compact Lie group becomes hypercomplex after it is multiplied by a sufficiently big torus.
Hypercomplex manifolds as such were studied by Charles Boyer in 1988. He also proved that in real dimension 4, the only compact hypercomplex manifolds are the complex torus , the Hopf surface and the K3 surface.
Much earlier (in 1955) M. Obata studied affine connection associated with almost hypercomplex structures (under the former terminology of Charles Ehresmann of almost quaternionic structures). His construction leads to what Edmond Bonan called the Obata connection which is torsion free, if and only if, "two" of the almost complex structures are integrable and in this case the manifold is hypercomplex.
There is a 2-dimensional sphere of quaternions satisfying . Each of these quaternions gives a complex structure on a hypercomplex manifold M. This defines an almost complex structure on the manifold , which is fibered over with fibers identified with . This complex structure is integrable, as follows from Obata theorem (it was first explicitly proved by Kaledin). This complex manifold is called the twistor space of . If M is , then its twistor space is isomorphic to .
- H. Wakakuwa, (1958), "On Riemannian manifolds with homogenious holonomy group Sp(n)", Tôhoku Mathematical Journal 10: 274–303, doi:10.2748/tmj/1178244665.
- Verbitsky, Misha (2005), "Hypercomplex structures on Kaehler manifolds", GAFA 15 (6): 1275–1283, doi:10.1007/s00039-005-0537-4
- Ehresmann, Charles (1947), "Sur la théorie des espaces fibrés", Coll. Top. Alg., Paris.
- Bonan, Edmond (1964), "Tenseur de structure d'une variété presque quaternionienne", C. R. Acad. Sci. Paris 259: 45–48
- Bonan, Edmond (1967), "Sur les G-structures de type quaternionien" (PDF), Cahiers de Topologie et Géométrie Différentielle Catégoriques 9.4: 389–463.
- Kaledin, Dmitry (1996). "Integrability of the twistor space for a hypercomplex manifold". arXiv:alg-geom/9612016 [math.AG].
- Boyer, Charles P. (1988), "A note on hyper-Hermitian four-manifolds", Proc. Amer. Math. Soc. 102 (1): 157–164, doi:10.1090/s0002-9939-1988-0915736-8.
- Joyce, Dominic (1992), "Compact hypercomplex and quaternionic manifolds", J. Differential Geom. 35 (3): 743–761.
- Obata, M. (1955), "Affine connections on manifolds with almost complex, quaternionic or Hermitian structure", Jap. J. Math. 26: 43–79.
- Spindel, Ph.; Sevrin, A.; Troost, W.; Van Proeyen, A. (1988), "Extended supersymmetric -models on group manifolds", Nucl. Phys. B308: 662–698.