Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

In Shastri's Elements of Differential Topology, p. 225, there is written:

enter image description here

I don't understand this map $R$. Why is $\theta\!\in\!\mathbb{S}^{l-1}$, shouldn't we have $\theta\!\in\!\mathbb{B}^l$? Furthermore, how do I construct such a map myself? I see that the idea is for each $(s,x)\!\in\!\mathbb{I}\!\times\!\mathbb{B}^l$ to construct a ray starting in $(0,\frac{x}{\|x\|})\!\in\!\{0\}\!\times\!\mathbb{S}^{l-1}$ and going through $(s,x)$, i.e. the ray $$\mathbb{I}\!\times\!\mathbb{B}^l\!\times\![0,\infty)\!\rightarrow\!\mathbb{I}\!\times\!\mathbb{B}^l,\;\; (s,x,t)\mapsto \begin{cases} (0,\frac{x}{\|x\|})+t\big((s,x)-(0,\frac{x}{\|x\|})\big)& x\!\neq\!0\\ (s,x) & x\!=\!0 \end{cases},$$ and then for each $(s,x)$ to choose the right $t_0\!\in\![0,\infty)$ where that ray hits $\{1\}\!\times\!\mathbb{B}^l\cup \mathbb{I}\!\times\!\{0\}$, divide that ray by $t_0$ so that the ray becomes a line parametrized by $\mathbb{I}$. I do not see a way to do this.

share|cite|improve this question
$\mathbb{S}^{l-1}$ is the boundary of the disc $\mathbb{D}^l$. You can consider a point in the disc as a pair $(\theta,s)$, with $0\leq s\leq 1$. – Sigur Aug 16 '12 at 0:23
Ok, so $\theta\!=\!\frac{x}{\|x\|}$. But how did he get the formula for $R$? – Leon Aug 16 '12 at 0:34
Considering $r$ as the time, I guess that for $0\leq r\leq 1/2$, it sends the disc $\{r\}\times \mathbb{D}^l$ to a point on the segment $[0,1]\times \{0\}$ at time $2r$, i.e. $\{2r\}\times \{0\}$, and for the other half $1/2\leq r\leq 1$ it sends $\{r\}\times \mathbb{D}^l$ onto the discs $\{1\}\times \mathbb{D}^l_\rho$, where $\rho$ is related to the radius of the disc (starting with the disc with radius 0, i.e. the point $\{1\}\times \{0\}$ and ending with the disc with radius 1, i.e. $\{1\}\times \mathbb{D}^l$). – Sigur Aug 16 '12 at 0:55
Roughly speaking, during the first half you collapse over the vertical segment and during the last half you project over the top concentric (from the center to the boundary) discs. – Sigur Aug 16 '12 at 0:58
@LeonLampret I have been told that in algebraic topology it is common practice not to write down explicit formulas for deformation retracts/homotopies/etc. For example, if you look in Hatcher many deformation retracts are presented as pictures. For me I have to the point that if you show me the picture, I will accept it as my deformation retract.... – user38268 Aug 16 '12 at 0:59
up vote 0 down vote accepted

I've figured it out. Turns out I misunderstood what the text was suggesting.

Claim: $\mathbb{B}^l\!\times\!\mathbb{I}$ strongly deformation retracts onto $\mathbb{B}^l\!\times\!\{1\} \cup \{0\}\!\times\!\mathbb{I}$, and $\mathbb{B}^l\!\times\!\{0\}$ goes to $(0,0)$.

Proof: Each element of $\mathbb{B}^k$ is written in polar coordinates as $(\theta,r)\in\mathbb{S}^{k-1}\!\!\times\!\mathbb{I}$, since $\frac{\mathbb{S}^{k-1}\!\times\mathbb{I}}{\mathbb{S}^{k-1}\!\times\{0\}}\approx\mathbb{B}^k$. The strong deformation retraction $h$ is obtained by not changing $\theta$ (i.e. $h$ is the same on each slice $\{(\theta,r,s)\!\in\!\mathbb{S}^{k-1}\!\!\times\!\mathbb{I}\!\times\!\mathbb{I};\,\theta\!=\!\text{const.}\}\!\approx\!\mathbb{I}\!\times\!\mathbb{I}$), on the lower half of the cylinder, $\mathbb{B}^l\!\times\![0,\frac{1}{2}]$, the outer line $\{(\theta_0,1)\} \!\times\! [0,\frac{1}{2}]$ is pushed to the middle line $\{(\theta_0,0)\} \!\times\! [0,1]$, on the upper half of the cylinder, $\mathbb{B}^l\!\times\![\frac{1}{2},1]$, the outer line $[\frac{1}{2},1] \!\times\!\{(\theta_0,1)\}$ is pushed to the top radial line$\{\theta_0\}\!\times\mathbb{I}\times\!\{1\}$, where the point $(1,\theta_0,1)$ does not move. Let us write $(r,s)$ instead of $(\theta_0,r,s)$. For $0\!\leq\!s\!\leq\!\frac{1}{2}$ the point $(1,s)$ therefore travels to $(0,2s)$, i.e. we have a straight path $t \mapsto (1,s) + t\big((0,2s)\!-\! (1,s)\big) \!=\!\big(1\!-\!t, s(1\!+\!t)\big)$. For $\frac{1}{2}\!\leq\!s\!\leq\!1$ the point $(1,s)$ travels to $(2(s\!-\!\frac{1}{2}),1)$, i.e. we have a path $t \mapsto (1,s) +t\big((2s\!-\!1,1) \!-\!(1,s)\big) \!=\!\big(1\!+\!2t(s\!-\!1),s\!+\!t(1\!-\!s)\big)$. Let us construct the homotopy for the remaining points as well (those from the whole square $\mathbb{I}\!\times\!\mathbb{I}$). The interval $0\!\leq\!s\!\leq\!\frac{1}{2}\!+\!\frac{1-r}{2} \!=\!\frac{2-r}{2}$ will be mapped at the end of the homotopy to $0\!\leq\!s\!\leq\!1$, hence each point $(r,s)$ travels to $(0,\frac{2}{2-r}s)$, i.e. we have a path $t \mapsto (r,s) +t\big((0,\frac{2}{2-r}s)\!-\! (r,s)\big) \!=\!\big(r\!-\!rt,s\!+\!t(\frac{2}{2-r}s\!-\!s)\big) \!=\!\big(r(1\!-\!t),s\!+\!st\frac{r}{2-r}\big)$. Now we must deal with the remaining points $(r,s)$ from the triangle $\frac{2-r}{2}\!\leq\!s\!\leq\!1$. The line through $(0,1)$ and $(1,\frac{1}{2})$ is $y\!-\!1\!=\!\frac{1/2-1}{1-0}(x\!-\!0)$, i.e. $y\!=\!-\frac{1}{2}x\!+\!1$. The line through $(1,1)$ in $(r,s)$ is $y\!-\!1\!=\!\frac{s-1}{r-1}(x\!-\!1)$. These two lines intersect in $(\frac{2s-2}{r+2s-3},\frac{r+s-2}{r+2s-3})$. The ratio of the distances is $\|-(1,1)\!+\!(r,s)\| \,/\, \|-(1,1)\!+\!(\frac{2s-2}{r+2s-3}, \frac{r+s-2}{r+2s-3})\|\!=\!\ldots\!=\!|r\!+\!2s\!-\!3|\!=\!3\!-\!r\!-\!2s\in[0,1]$. Therefore the point $(r,s)$ is mapped to $\big(1\!-\!(3\!-\!r\!-\!2s),1\big)\!=\! (r\!+\!2s\!-\!2,1)$, i.e. we have a path $t\mapsto (r,s)\!+\!t\big( (r\!+\!2s\!-\!2,1) \!-\!(r,s)\big) \!=\!\big(r\!+\!2t(s\!-\!1),s\!+\!t(1\!-\!s)\big)$. The desired strong deformation retraction is therefore

$$\;\;\;\;\;\;\;h\!:\;\;\mathbb{B}^l\!\times\!\mathbb{I}\! \times\!\mathbb{I}\!\longrightarrow\!\mathbb{B}^l\!\times\!\mathbb{I},\;\;\; (\theta,r,s,t)\mapsto \begin{cases} \displaystyle \big(\theta,r(1\!-\!t),s\!+\!st{\textstyle\frac{r}{2-r}})\big) & \hspace{-8pt} ;\;\; \displaystyle r\!+\!2s\!\leq\!2,\\ \displaystyle \big(\theta,r\!+\!2t(s\!-\!1),s\!+\!t(1\!-\!s)\big) &\hspace{-8pt} ;\;\; \displaystyle r\!+\!2s\!\geq\!2. \;\;\;\;\;\;\;\;\;\blacksquare \end{cases}$$

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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