2

I am trying to find the condition that four points $p_1,p_2,p_3,p_4$ on the unit sphere $\mathbb{S}^1$ need to statisy in order to form a 3-orthoscheme (Tetrahedron with all faces as right angled triangles), preferably phrased in the language of vectors and rotations.

A 3-orthoscheme is basically determined by choosing 3 mutually orthogonal edges, $e_1 = \overline{p_1p_2}, e_2=\overline{p_2p_3}, e_3=\overline{p_3p_4}$. The center of the circumscribing sphere is thus the midpoint of the longest edge of the tetrahedron. Assuming that the center of the sphere is the origin, I can take unit vectors $\boldsymbol{v_i}$, corresponding to the points $p_i$. Actually all the four points lie on a hemisphere and points $p_1$ and $p_4$ can be taken to be antipodal.

The orthogonality condition means that $(\boldsymbol{v}_2-\boldsymbol{v}_1)\cdot(\boldsymbol{v}_3-\boldsymbol{v}_2) = 0$ and $(\boldsymbol{v}_3-\boldsymbol{v}_2)\cdot(\boldsymbol{v}_4-\boldsymbol{v}_3) = 0$. Decoding these gives a condition on the cosine of the angles between the edges.

If I know the lengths of the three edges, say $\ell_{i}$ which are mutually orthogonal, then it is easy to find the lengths of the remainining edges using the right angle condition. The longest edge is known to be of length 2. So this means that $$ \Sigma_{i=1}^3\ell_i^2 = \Sigma_{i=1}^{3} ||\boldsymbol{v}_{i+1}-\boldsymbol{v}_i||^2 = 4.$$ In fact if the $\ell_i$ are all same, then this precisely means $\ell_i = \dfrac{2}{\sqrt{3}}$, which corresponds to the orthoscheme that determines a cube.

Are the above set of conditions enough to say that four points form a 3-orthoscheme? Alternatively can something similar be deduced as in the case of this question for a regular tetrahedra? I am attempting to see this as a 3d generalisation of Thales theorem. Thales theorem tells us how three points on a circle need to be distributed for them to form a 2-orthoscheme, which is a right angled triangle.

Vishesh
  • 2,928

2 Answers2

1

This is a partial answer to my own question although it is not very satisfactory to me. The conditions boil down to the expressions: \begin{align} \boldsymbol{v}_1\cdot\boldsymbol{v}_2 + \boldsymbol{v}_2\cdot\boldsymbol{v}_3 +\boldsymbol{v}_3\cdot\boldsymbol{v}_4 = 1 \\ \boldsymbol{v}_1\cdot\boldsymbol{v}_2 + \boldsymbol{v}_2\cdot\boldsymbol{v}_3 = 1 + \boldsymbol{v}_1\cdot\boldsymbol{v}_3 \\ \boldsymbol{v}_2\cdot\boldsymbol{v}_3 + \boldsymbol{v}_3\cdot\boldsymbol{v}_4 = 1 + \boldsymbol{v}_2\cdot\boldsymbol{v}_4 \end{align}

But I am wondering if this determines an orthoscheme upto even similarity. Maybe a clever usage of spherical law of cosines will help, I am not sure.

Vishesh
  • 2,928
  • 1
    If the points $p_1$ and $p_4$ are antipodal then $\boldsymbol{v}_4=-\boldsymbol{v}_1$, so the system reduces to $\boldsymbol{v}_1\cdot\boldsymbol{v}_2 + \boldsymbol{v}_2\cdot\boldsymbol{v}_3 -\boldsymbol{v}_3\cdot\boldsymbol{v}_1 = 1$ or to $|\boldsymbol{v}_1-\boldsymbol{v}_2+\boldsymbol{v}_3|=1$. – Alex Ravsky Nov 03 '23 at 03:25
  • @AlexRavsky Thank you. By that observation actually the first of the three equations can be used to obtain both the second and the third. is that what you mean? – Vishesh Nov 03 '23 at 04:45
  • @AlexRavsky I would like to look at this as a 3d generalisation of Thales theorem which gives the condition for three points on a circle to form a right angled triangle, i.e. a 2-orthsocheme. – Vishesh Nov 03 '23 at 04:48
  • 1
    As I understood, looking for the required condition we can assume that $\boldsymbol{v}_4=-\boldsymbol{v}_1$ and then the condition transforms to $|\boldsymbol{v}_1-\boldsymbol{v}_2+\boldsymbol{v}_3|=1$. – Alex Ravsky Nov 03 '23 at 05:29
  • @AlexRavsky I see But does this give a unique 3-orthsocheme upto congruency? – Vishesh Nov 03 '23 at 05:42
  • 1
    Take antipodes $P$ & $Q$, and any distinct $R$ on the sphere. The plane through $Q$ orthogonal to $PR$ cuts the sphere in a circle having $QR$ as a diameter. For any distinct $S$ on this circle, $PQRS$ is a $3$-orthoscheme. ... Vectorially, take antipodes $P$ and $Q$, and appropriately-scaled vectors $u$ & $v$ such that $R:=P+u$ & $S:=Q+v=-P+v$ lie on the sphere. Your eqns all reduce to $u\cdot v=0$. That is, almost-any orthogonal $u$ and $v$ will do, provided $P$, $Q$, $R$, $S$ are distinct. (Neither $u$ nor $v$ can orthogonal to $PQ$; and $u$, $v$, $P-Q$ must be linearly independent.) – Blue Nov 03 '23 at 06:28
  • @Blue Wow Thanks. Can you write this as an answer so I can accept it and give the bounty? Preferably with a bit more elaboration about the vectorial approach, but it will do even otherwise. – Vishesh Nov 03 '23 at 06:46
  • 1
    I rather expect this is a description of all orthoschemes, that is I expect that vectors $\boldsymbol{v}1,\dots, \boldsymbol{v}_4\in\mathbb S^1$ form a 3-orthoscheme iff there exists a permutation $\sigma$ of ${1,\dots, 4}$ such that $\boldsymbol{v}{\sigma(4)}=-\boldsymbol{v}\sigma(1)$ and $|\boldsymbol{v}{\sigma(1)}-\boldsymbol{v}{\sigma(2)}+\boldsymbol{v}{\sigma(3)}|=1$. – Alex Ravsky Nov 03 '23 at 07:39
  • @AlexRavsky I am a bit confused about the last norm equality. I am missing something simple, I presume.I understand the permutation bit actually, its the norm one I am wondering about. – Vishesh Nov 03 '23 at 08:01
  • 1
    Concerning the norm equality, we have
    $$1=|\boldsymbol{v}_1-\boldsymbol{v}_2+\boldsymbol{v}_3|^2= (\boldsymbol{v}_1-\boldsymbol{v}_2+\boldsymbol{v}_3)\cdot (\boldsymbol{v}_1-\boldsymbol{v}_2+\boldsymbol{v}_3)=$$ $$\boldsymbol{v}_1 \cdot\boldsymbol{v}_1+\boldsymbol{v}_2 \cdot\boldsymbol{v}_2+\boldsymbol{v}_3\cdot \boldsymbol{v}_3+2(\boldsymbol{v}_1\cdot\boldsymbol{v}_3-\boldsymbol{v}_1\cdot\boldsymbol{v}_2- \boldsymbol{v}_2\cdot\boldsymbol{v}_3)=$$ $$3+2(\boldsymbol{v}_1\cdot\boldsymbol{v}_3-\boldsymbol{v}_1\cdot\boldsymbol{v}_2- \boldsymbol{v}_2\cdot\boldsymbol{v}_3).$$
    – Alex Ravsky Nov 03 '23 at 08:42
  • 1
    Thus $|\boldsymbol{v}_1-\boldsymbol{v}_2+\boldsymbol{v}_3|=1$ iff $\boldsymbol{v}_1\cdot\boldsymbol{v}_2+ \boldsymbol{v}_2\cdot\boldsymbol{v}_3-\boldsymbol{v}_1\cdot\boldsymbol{v}_3=1$. – Alex Ravsky Nov 03 '23 at 08:42
  • @AlexRavsky Thanks. that is clear enough. – Vishesh Nov 03 '23 at 08:44
1

Take antipodes $P$ and $Q$, and any distinct $R$ on the sphere. The plane through $Q$ orthogonal to $PR$ cuts the sphere in a circle having $QR$ as a diameter. (Why?) For any distinct $S$ on this circle, $PQRS$ is a $3$-orthoscheme.

After all, $R$ and $S$ being on the sphere with diameter $PQ$ guarantees (by Thales Theorem in great circles $\bigcirc PQR$ and $\bigcirc PQS$) that $\angle PRQ=\angle PSQ=90^\circ$. Taking $S$ on the plane orthogonal to $PR$ necessarily implies $\angle PRS=90^\circ$, and since $S$ is in particular on a circle with diameter $QR$ implies (by Thales again) that $\angle QSR=90^\circ$. Each face is a right triangle.


Vectorially, take antipodes $P$ and $Q$, and consider additional points $R$ and $S$ (such that all four points are distinct). Defining vectors $u:=R-P$ and $v:=S-Q$, the equations in OP's answer, with $v_1=P$, $v_2=R$, $v_3=S$, $v_4=Q=-P$ all reduce to $u\cdot v=0$. Thus, all we require is that vectors $u$ and $v$ are orthogonal.

Looking at this slightly differently, consider antipodes $P$ and $Q$, and orthogonal vectors $u$ and $v$. Let $R$ is the point where the line through $P$ in direction $u$ meets the sphere; likewise define $S$ from $Q$ and $v$. If $P$, $Q$, $R$, $S$ are distinct, then they determine a $3$-orthoscheme.

(Non-distinctness arises when $u$ or $v$ are orthogonal to $PQ$, which would make them tangent to the sphere, so that $P=R$ or $Q=S$; or, when $u$, $v$, and $PQ$ are linearly dependent, which would make $P$, $Q$, $R$, $S$ coplanar, which in turn requires $R=S$.)

Blue
  • 75,673
  • Thank you so much. Its quite clear. But do you think the other commenter had a point about the norm condition? I have accepted the answer. I just wanted your view. – Vishesh Nov 03 '23 at 08:48
  • 1
    @Vishesh: The norm condition is certainly valid, as the commenter's vector argument shows. Geometrically, given orthoscheme $PQRS$ and defining $T := P-R+S$ (the commenter's $v_1-v_2+v_3$) and $R':=-R$, we find $P-R=R'-Q=T-S$ so that $\triangle PR'T$ is an orthogonal translate of $\triangle RQS$; ie, $T$ completes a triangular prism w/other vertices $P$, $Q$, $R$, $S$, $R'$. The sphere's center is half-way between the triangles' planes, so that $T$ is on the sphere and $|T|=1$. Showing the reverse implication seems best done with the vector argument. – Blue Nov 03 '23 at 10:06
  • 1
    Thank you very much. This has been very helpful. – Vishesh Nov 03 '23 at 13:39