6

Let $S^l \vee S^k$ wedge sum of two spheres $S^l, S^k$.

How can I simplify the calculation of homotopy groups $\pi_n(S^l \vee S^k)$?

I know that $\pi_n(S^l \times S^k) = \pi_n(S^l) \oplus \pi_n(S^k)$,

as well $H_n(S^l \vee S^k) = \pi_n(S^l) \oplus H_n(S^k)$ holds, therefore we have $\pi_n ^{ab}(S^l \times S^k) = \pi_n ^{ab}(S^l) \oplus \pi_n ^{ab}(S^k)$,

but do we have similar reduction rules for $\pi_n(S^l \vee S^k)$? If, yes, why, but I suppose not generally.

What other strategy can I use to simplify the calculation?

Sumanta
  • 9,534
user267839
  • 7,293
  • 4
    In general $H_n$ is NOT the abelianisation of $\pi_n$ when $n\geq 2$, for the very good reason that $\pi_n$ is already abelian for $n\geq 2$. – Arnaud Mortier Apr 10 '18 at 10:12
  • The homotopy groups of spheres alone are already extremely hard to compute (to this day we do not know a general formula for them). Do you want something that depends on these? – Najib Idrissi Apr 10 '18 at 14:49

2 Answers2

5

First, if $k,l=1$, then $S^1\vee S^1$ is the Eilenberg Mac Lane space $K(\mathbb{Z}\ast\mathbb{Z},1)$ amd $S^1\times S^1$ is the Eielenberg Mac Lane space $K(\mathbb{Z}\oplus\mathbb{Z},1)$. It follows that $\pi_1(S^1\times S^1)$ is the abelianisation of $\pi_1(S^1\vee S^1)$, and neither the wedge nor product have any other homotopy groups.

If one of $k,l$ is equal to $1$ (say, $k$), and the other is greater than one, then you can use cover space theory to calculate the homotopy groups of $S^1\vee S^l$. I believe its universal cover may be the infinite wedge $\bigvee^\infty_{i=1} S^l$, but i may not be remembering that correctly. In any case $\pi_*(S^1\times S^l)\cong\pi_*(S^1)\oplus\pi_*(S^l)$ is $\pi_*S^l$ with an extra copy of $\mathbb{Z}$ in degree $1$, whilst $\pi_*(S^1\vee S^l)$ is exceptionally complicated.

Therefore, in the following, we assume that $k,l$ are both $>1$.

Let $j:S^k\vee S^l\hookrightarrow S^k\times S^l$ be the inclusion, and let $F_j$ be its homotopy fiber. Then, after looping, there is a homotopy fibration

$$\Omega F_j\xrightarrow{\Omega i} \Omega(S^k\vee S^l)\xrightarrow{\Omega j} \Omega(S^k\times S^l),$$

where $i:F_j\rightarrow S^k\vee S^l$ is the fibre inclusion.

The projections onto each factor induce a homeomorphism

$\Psi:\Omega(S^k\times S^l)=Map_*(S^1,S^k\times S^l)\xrightarrow{\cong}Map_*(S^1,S^k)\times Map_*(S^1,S^l)=\Omega S^k\times \Omega S^l,$

sending $\omega\mapsto (pr_1\circ\omega,pr_2\circ \omega)$. Composing with $\Omega j$, this map acts as

$\Psi\circ \Omega j(\omega)=\Theta(j\circ \omega)=(pr_1,circ j\circ \omega,pr_2\circ j\circ \omega=(q_1\circ \omega,q_2\circ \omega),$

where $q_1:S^k\vee S^l\rightarrow S^k$, $q_2:S^k\vee S^l\rightarrow S^l$ are the pinch maps. Thus we are left with the homotopy fibration sequence

$$\Omega F_j\rightarrow \Omega(S^k\vee S^l)\xrightarrow{(\Omega q_1,\Omega q_2)} \Omega S^k\times \Omega S^l.$$

Now the pinch maps $q_1,q_2$ individually have sections, namely the inclusions $i_1:S^k\hookrightarrow S^k\vee S^l$, $i_2:S^l\hookrightarrow S^k\vee S^l$, and with these we use the monoid structure on $\Omega (S^k\vee S^l)$ to define a map $\Omega i_1\circ pr_1+\Omega i_2\circ pr_2:\Omega S^k\times\Omega S^l\rightarrow \Omega(S^k\vee S^l)$, $(\omega_1,\omega_2)\mapsto\Omega i_1(\omega_1)+\Omega i_2(\omega_2)$. Observe that

$(\Omega q_1,\Omega q_2)\circ(\Omega i_1\circ pr_1+\Omega i_2\circ pr_2)(\omega_1,\omega_2)=(\Omega q_1,\Omega q_2)(\Omega i_1(\omega_1)+\Omega i_2(\omega_2))=(\Omega(q_1)(\Omega i_1(\omega_1)+\Omega i_2(\omega_2),\Omega q_2(\Omega i_1(\omega_1)+\Omega i_2(\omega_2))=(\Omega(q_1\circ i_1)(\omega_1)+\Omega(q_1\circ i_2)(\omega_2),\Omega(q_2\circ i_1)(\omega_1)+\Omega(q_2\circ i_2)(\omega_2))=(\Omega(id_{S^1})(\omega_1)+\Omega(\ast)(\omega_2),\Omega(\ast)(\omega_1)+\Omega(id_{S^l})(\omega_2))=(\omega_1+\ast,\ast+\omega_2)\simeq (\omega_1,\omega_2).$

The point is that the map we have just produced is a homotopy section of $\Omega j$. Using again the monoid structure we define a map

$\Theta:\Omega S^k\times\Omega S^l\times\Omega F_j\rightarrow \Omega(S^k\vee S^l)$

by $\Theta=\Omega i_1\circ pr_1+\Omega i_2\circ pr_2+\Omega i\circ pr_3$.

Turning to study the long exact sequence of homotopy groups of the homotopy fibration $F_i\xrightarrow{i}S^k\vee S^l\xrightarrow{j} S^k\times S^l$ we find that the map $\Theta$ is a weak homotopy equivalence. Thus

$$\pi_r(S^k\vee S^l)\cong \pi_{r-1}(\Omega(S^k\vee S^l))\cong \pi_{r-1}(\Omega S^k\times\Omega S^l\times\Omega F_j)\cong \pi_r(S^k\times S^l)\oplus\pi_r(F_i)$$

so for $k,l>1$, the homotopy module $\pi_*(S^k\times S^l)$ is always a direct summand of $\pi_*(S^k\vee S^l)$. Sadly this doesn't really make the task of computing $\pi_*(S^k\vee S^l)$ any easier, since we still need to find $\pi_*(F_i)$. What is this space?

Well, we know from homology that

$H_*(S^k\vee S^l)=\begin{cases}H_*(S^k\vee S^l)&*<k+l\\\mathbb{Z}& *=k+l\\0&\text{otherwise}\end{cases}$

so since both spaces are simply connected we find that the map $j$ is a $(k+l-1)$-equivalence. In fact the attaching map for the top cell of $S^k\vee S^l$ is the Whitehead product $[i_1,i_2]:S^{k+l-1}\rightarrow S^k\vee S^l$. In any case, the $F_i$ must be $(k+l-1)$-connected, and its not difficult to argue that $\pi_{k+l-1}F_i\cong\pi_{k+l-1}S^{k+l-1}\cong\mathbb{Z}$.

Unfortunately there is no short answer to what $F_i$ is, so I'll just write down

$F_i\simeq (\Omega S^k)\ast(\Omega S^l)\simeq \Sigma(\Omega S^k\wedge \Omega S^l),$

and refer you to Jeff Strom's book "Modern Classical Homotopy Theory" for more details.

Tyrone
  • 16,116
  • 1
    Very nice! The universal cover of $S^1\vee S^k$ for $k>1$ has the homotopy type of an infinite bouquet of $S^k$, but not the homeomorphism type. You think of the universal cover as a real line with an $S^k$ attached at each integer point. Also, when you write "the homotopy module .... is always a direct summand", I think you mean $\pi_\ast(S^k\times S^l)$. – Jason DeVito - on hiatus Apr 10 '18 at 18:06
  • Thank you for spotting the error :). I've now corrected it. – Tyrone Apr 10 '18 at 19:07
  • @Tyrone: Hi, thank you for very enlightening answer. One question to your conclusion: Why the simply connectness of both spaces imply that $j$ is a $(k+l-1)$-equivalence? This argument isn't clear to me. – user267839 Apr 11 '18 at 20:27
  • The map $j$ is an isomorphism in homology up to degree $k+l-1$. For simply connected CW complexes this is sufficient to conclude that it is a $(k+l-1)$-equivalence. This is essentially Corollary 4.33 in Hatcher's "Algebraic Topology". – Tyrone Apr 11 '18 at 20:56
  • sorry for picking up this thread again but following point confuses me: In the case $k=1$ you wrote that one can calculate the homotopy groups of $S^1\vee S^l$ by exploit the property of the (universal) cover. Could you explain this step a bit more? I don't understand it. Futhermore I think that $\pi_(S^1\times S^l)\cong\pi_(S^1)\oplus\pi_*(S^l)$ should hold always and not only in the case $k=1$ but I don't see how does this observation concretely help to cope with the calculation problem? – user267839 Apr 27 '19 at 03:04
  • If you want to read about universal covers you should check out section 1.3 of Hatcher's Algebraic Topology. In particular it includes an in depth discussion of howIt is true that for any spaces $X,Y$ you have $\pi_r(X\times Y)\cong \pi_r(X)\oplus\pi_r(Y)$ for any $r\geq 0$. I have written this in the answer. – Tyrone Apr 27 '19 at 08:30
3

This solution Hilton's theorem, from his paper ON THE HOMOTOPY GROUPS OF THE UNION OF SPHERES. It makes the calculation of $\pi_n(S^{r_1+1}\vee \cdots \vee S^{r_m+1})$ for $r_i\geq 1$ easy to do by hand for small $n, m$.

In the Hilton's paper there is Theorem A

$$ \pi_n(T)=\sum_{i=1}^{\infty} \pi_n(S^{q_{i}+1}) $$ where $T=S^{r_1+1}\vee \cdots \vee S^{r_m+1}$, $r_i\geq 1$.

He inductively defines so-called basic generators $p_1,p_2,\cdots$. For each basic generator we get an integer $q_s$, called height.

The basic generators of weight $1$ are $\iota_1,\cdots,\iota_m$ where $\iota_i$ is the generatror of $\pi_{r_i+1}(S^{r_i+1})$ and can be taken to be the generator of a direct summand in $\pi_{r_i+1}(T)$. Order the generators arbitrarily $\iota_1<\cdots<\iota_m$.

Assume the basic generators of weight less then $w$ have been defined and ordered. The basic generators of weight $w$ are $[a,b]$ where $a,b$ are the basic generators of order $u$ and $v$ respectively, $u+v=w$ and $[-,-]$ is the Whitehead product. Furthermore we demand that $a<b$ and if $b=[c,d]$ then $c\leq a$.

Order the basic products of weight $w$ among themselves arbitrarily, and set them to be larger then any basic product of smaller weight.

Any basic product $p_q$ is then some string of basic generators of weight $1$ and some brackets. Define its height as $q_s=\sum_{i=1}^k r_iw_i$ where $w_i$ is the number of occurences of $\iota_i$ in $p_s$.

Let us have a basic product $p_s$ with height $q_s$. There is then an embedding as a direct summand $\pi_n(S^{q_s+1})\hookrightarrow \pi_n(T)$ induced by the map $p_s\in\pi_{q_s+1}(T)$(an easy excercise that $p_s$ is from that group)