I am trying to show the standard $n$-simplex is homeomorphic to the $n$-ball.

Here, the standard $n$-simplex is given by $$\Delta^n=\left\{(x_0,x_1,\cdots,x_n)\in\mathbb{R}^{n+1}:\sum x_i=1,x_i\geq0\right\}$$ and the $n$-ball is given by $$B^n=\{x\in\mathbb{R}^n:||x||\leq 1\}$$

Any help will be appreciated.

  • 1
    $\begingroup$ Are you looking for an idea or the details? $\endgroup$
    – copper.hat
    Oct 24, 2013 at 15:30
  • 1
    $\begingroup$ @copper.hat I am looking for the details, since I have actually thought about this problem for days. Thank you! $\endgroup$
    – YYF
    Oct 24, 2013 at 15:35
  • $\begingroup$ Unfortunately I don't have time for the details now. $\endgroup$
    – copper.hat
    Oct 24, 2013 at 15:43

3 Answers 3


Hint: $\Delta^n$ is convex, so you may may project $\Delta^n$ onto a ball $B^n \supset \Delta^n$ with respect to its barycentric center $c$.

The projection $f$ can be described as follow: First, notice that without loss of generality $B^n$ may be supposed to be centered at $c$; let $r$ denote its radius. For every $p \in \Delta^n \backslash \{c\}$, the ray from $c$ to $p$ meets $\partial \Delta^n$ at only one point $f(p)$. Now, we may define the projection $$g(p)= c+\frac{r}{\|f(p)-c\|} \cdot (p-c).$$

(Another related question: Proof that convex open sets in $\mathbb{R}^n$ are homeomorphic?)

  • $\begingroup$ What do you mean by "project"? I did not get it. Actually, I am self-learning algebraic topology by reading Rotman's book, and it is my first time learning simplices. $\endgroup$
    – YYF
    Oct 24, 2013 at 15:54
  • $\begingroup$ I added a description of the projection. $\endgroup$
    – Seirios
    Oct 24, 2013 at 16:42
  • 1
    $\begingroup$ It is pretty clear on a figure, but you can use the following argument: $\partial \Delta^n$ consists in the intersections of $\Delta^n$ with the hyperplanes $x_i=0$. Because $c$ does not belong to any of these hyperplanes, any ray starting from $c$ can meet $\partial \Delta^n$ at most once, and in fact exactly once since $c \in \Delta^n$. $\endgroup$
    – Seirios
    Dec 6, 2014 at 19:27
  • 1
    $\begingroup$ There is a generalization of this in Bredon's Topoloy and geometry, Proposition 16.4, page 56. The proposition says: ''A compact convex body $C$ in $\mathbf{R}^n$ with nonempty interior is homeomorphic to the closed $n$-ball''. $\endgroup$ Feb 4, 2015 at 20:49
  • 1
    $\begingroup$ Why is $g$ continuous, why is $g$ continuous at $c$ and why is $g^{-1}$ continuous? $\endgroup$ May 17, 2017 at 7:12

So, why are $g$ and $g^{-1}$ continuous in @Seirios answer?

Here are the main facts (all easily verifiable)

  1. The barycenter $c$ has all its coordinates equal to $1/(n+1)$.
  2. The standard simplex $\Delta^n$ is included in the hyperplane $H=\{x\mid\sum_ix_i=1\}$.
  3. If $x_{(1)}$ denotes the smallest coordinate of vector $x$, then the application $x\mapsto x_{(1)}$ is continuous.
  4. The projection $f\colon B[c,r]\cap H\setminus\{c\}\to\partial\Delta^n$ is $$ f(x) = c + \rho(x)(x-c), $$ where $$ \rho(x) = \frac{1}{1-x_{(1)}(n+1)}. $$
  5. The homeomorphism $g\colon\Delta^n\to B[c,r]\cap H$, defined as $$ g(x) = \begin{cases} c &{\rm if\ }x=c,\\ c + \frac{r}{\Vert f(x) - c\Vert}(x-c) &\text{otherwise}, \end{cases} $$ is continuous at $c$ because $$ \frac{\Vert x-c\Vert}{\Vert f(x)-c\Vert} = 1 - x_{(1)}(n+1). $$
  6. If $y=g(x)$ then $$ 1 - y_{(1)}(n+1) = \frac{r}{\Vert f(x)-c\Vert}(1 - x_{(1)}(n+1)). $$
  7. If $y=g(x)$ then $f(y)=f(x)$.
  8. The inverse of $g$ is $$ h(y) = c + \frac{\Vert f(y)-c\Vert}{r}(y-c) $$ (similarly to part 7, show that $z=h(y)\implies f(z)=f(y)$.)
  9. (Bonus) $r=\sqrt{1 - 1/(n+1)}$ (not required to complete the proof.)

More generally, if $X$ is star-shaped, the center of $X$ is the set $Z$ of all $c\in X$ such that, for all $x\in X$, the segment $\{(1-\theta)c + \theta x \mid 0\le\theta\le 1\}$ is included in $X$. Since the $n$-simplex is convex, hence star-shaped, and its center is open, the Theorem below implies that the $n$-simplex is homeomorphic to the $n$-ball.

Theorem. If $X\subseteq\mathbb R^n$ is compact, star-shaped and its center $Z$ has a non-empty interior, then $X$ is homeomorphic to the $n$-ball $B[0,1]\subseteq\mathbb R^n$.

Proof [sketch].

  1. After a possible translation, we can assume that $0\in \operatorname{int}(Z)$. In what follows, let $X^* = X\setminus\{0\}$.

  2. For every $x\in X^*$ define $\ell_x = \{tx \mid t\ge0\}$.

  3. Put $\bar t=\sup\{t\ge0 \mid tx\in X\}$. Since $X$ is compact, the sup is attained and we can define $$ f(x)=\bar{t}x. $$

  4. The following properties hold

    a. $f(x) \in \operatorname{cl}(X)$.

    b. The segment from $0$ to $f(x)$ is included in $X$.

    c. $\Vert f(x)\Vert\ge\delta$, where $\delta>0$ satisfies $B[0,\delta]\subseteq Z$ [cf. 1].

    d. If $z\in X^*$ is such that $f(x)$ and $f(z)$ define the same ray, then $f(x)=f(z)$.

  5. Assume momentarily that $f\colon X^*\to\operatorname{cl}(X)$ is continuous. Then, the function $g\colon X\to B[0,1]$ defined as $$ g(x) = \begin{cases} \displaystyle\frac{x}{\Vert f(x)\Vert} &\text{if } x\in X^*,\\[0.1 in] 0 &\rm otherwise \end{cases} $$ is continuous. (Hint: Assume $(x_i)_{i\ge1}\subseteq X$ converges to $x\in X$. Show that $g(x_i) \to g(x)$ by studying separately the cases $x=0$ and $x\ne0$.)

  6. Show that $g$ is injective. (Hint: Assume $x\ne y$ and analyze two cases $\ell_x=\ell_y$ and $\ell_x\ne\ell_y$.)

  7. Show that $g$ is surjective. (Hint: If $z\in B[0,1]$, put $y=\delta z$ and $y=\Vert f(x)\Vert z$. Then $g(y)=z$.)

  8. Conclude that $g$ is an homeomorphism.

  9. Now prove that $f$ is continuous, as follows:

    a. Let $(x_i)_{i\ge1}\subseteq X^*$ converging to $x\in X^*$.

    b. Since $X$ is compact we can assume $f(x_i)\to z\in\partial X$.

    c. If $f(z)\ne z$, let $H$ be the hyperplane orthogonal to $z$ and $B = H\cap B[0,\delta]$.

    d. Let $K$ be the cone with vertex $f(z)$ and base $B$. Then $K\subseteq X$ is a closed neighborhood of $z$. Contradiction.

    e. Use that $x_i$ and $f(x_i)$ belong in $\ell_{x_i}$ and $x$ and $f(x)$ in $\ell_x$ to show that $f(x)$ and $f(z)$ belong in the same ray $x/\Vert x\Vert$.

    f. Conclude that $f(x)=f(z)$ [cf. 4. d].


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.