For an oriented $d$-manifold $M$, we can ask whether the manifold admits a Spin structure, say, if the transition functions for the tangent bundle, which take values in $SO(d)$, can be lifted to $\operatorname{Spin}(d)$ while preserving the cocycle condition on triple overlaps of coordinate charts.

For an unoriented $d$-manifold $M$, we can ask whether the manifold admits Pin$^+$ or Pin$^−$ structures (that is, lifts of transition functions to either $\operatorname{Pin}^+(d)$ or $\operatorname{Pin}^−(d)$ from $O(d)$. This is analogous to lifting the transition functions to $\operatorname{Spin}(d)$ from $SO(d)$ for the spin manifold).

If the manifold $M$ is orientable, then the conditions for Pin$^+$ or Pin$^−$ structures are the same. So it reduces to the old condition that $M$ admits a Spin structure or not.

Every 2-dimensional manifold admits a Pin$^−$ structure, but not necessarily a Pin$^+$ structure.

Every 3-dimensional manifold admits a Pin$^-$ structure, but not necessarily a Pin$^+$ structure.

Question: For some any other $d$, say $d = 0, 1, 4,$ etc. are there statements like: every $d$-dimensional manifold admits a Pin$^-$ structure? Or, every $d$-dimensional manifold admits a Pin$^+$ structure? Or are there useful conditions like SW classes $w_2+w_1^2=0$ like the case for $d = 2, 3$? What are these conditions in other dimensions?

P.S. The original post on MSE received almost no attention for a week.

  • $\begingroup$ Why does every three-manifold admit a Pin$^-$ structure? I know that every closed three-manifold admits a Pin$^-$ structure. $\endgroup$ Dec 24, 2016 at 0:11

2 Answers 2


The condition for having a $Pin^+$-structure is the vanishing of $w_2$, and for having a $Pin^-$-structure is the vanishing of $w_2 +w_1^2$, for manifolds of any dimension. This is because the Lie group $Pin^\pm(d)$ is a central extension $$\mathbb{Z}/2 \longrightarrow Pin^\pm(d) \longrightarrow O(d)$$ so there are homotopy fibre sequences $$BPin^\pm(d) \overset{p^\pm}\longrightarrow BO(d) \overset{k^\pm}\longrightarrow K(\mathbb{Z}/2,2)$$ for $k^\pm \in H^2(BO(d) ; \mathbb{Z}/2)$ the class classifying the central extension, which convention dictates is $k^+ = w_2$ and $k^- = w_2 + w_1^2$.

If if $\pi : E \to B$ is a $d$-dimensional vector bundle (which need not be the tangent bundle of a manifold) classified by a map $f : B \to BO(d)$, then $f$ lifts along $p^\pm : BPin^\pm(d) \to BO(d)$ if and only if $k^\pm \circ f : B \to K(\mathbb{Z}/2,2)$ is nullhomotopic, i.e. if and only if $f^*k^\pm =0 \in H^2(B;\mathbb{Z}/2)$.

  • $\begingroup$ If Pin+-structure is the vanishing of w2, isn't that the same as the Spin structure is the vanishing of w2? What are the differences of criteria for Pin+-structure and Spin-structure ? $\endgroup$
    – miss-tery
    Dec 23, 2016 at 17:45
  • 4
    $\begingroup$ A Spin structure exists if both $w_1$ and $w_2$ vanish, whereas a $Pin^+$ structure only requires the latter. $\endgroup$ Dec 23, 2016 at 17:56
  • $\begingroup$ I admit I am a bit of an outsider in this discussion, but I am very confused by the following: Why do you write $w_1^2$ instead of just $w_1$? I though these are $\mathbb{Z}_2$ numbers. $\endgroup$ Apr 8, 2020 at 18:41
  • $\begingroup$ @TomasBzdusek: They are not elements of $\mathbb{Z}_2$, they are cohomology classes: $w_1 \in H^1(BO(d); \mathbb{Z}_2)$ and $w_1^2 \in H^2(BO(d); \mathbb{Z}_2)$. $\endgroup$ Jul 5, 2020 at 1:02

For $d \geq 4$, let $M_d = (S^1)^{d-4}\times\mathbb{CP}^2$. As tori are parallelisable, $w(M_d) = w(\mathbb{CP}^2)$, in particular $w_1(M_d) = 0$ and $w_2(M_d) \neq 0$, so $M_d$ does not admit a Spin, Pin$^+$, or Pin$^-$ structure.

Therefore, there is no $d \geq 4$ such that every $d$-dimensional manifold admits a Pin$^+$/Pin$^-$ structure.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.