Geometry and Trigonometry 641
Note thatd(X,Y) >
√
2 if and only if
d^2 (X, Y )=
∑n
k= 1
x^2 k+
∑n
k= 1
y^2 k− 2
∑n
k= 1
xkyk> 2.
Therefore,d(X,Y) >
√
2 implies
∑n
k= 1 xkyk<0.
Now letA 1 ,A 2 ,...,Amnbe points satisfying the condition from the hypothesis, with
mnmaximal. Using the symmetry of the sphere we may assume thatA 1 =(− 1 , 0 ,..., 0 ).
LetAi=(x 1 ,x 2 ,...,xn)andAj=(y 1 ,y 2 ,...,yn),i, j≥2. Becaused(A 1 ,Ai)and
d(A 1 ,Aj)are both greater than
√
2, the above observation shows thatx 1 andy 1 are
positive.
The conditiond(Ai,Aj)>
√
2 implies
∑n
k= 1 xkyk<0, and sincex^1 y^1 is positive, it
follows that
∑n
k= 2
xkyk< 0.
This shows that if we normalize the lastn−1 coordinates of the pointsAiby
x′k=
xk
√∑
n− 1
k= 1 x
2
k
,k= 1 , 2 ,...,n− 1 ,
we obtain the coordinates of pointBiinSn−^2 , and the pointsB 2 ,B 3 ,...,Bnsatisfy the
condition from the statement of the problem for the unit sphere inRn−^1.
It follows thatmn≤ 1 +mn− 1 , andm 1 =2 impliesmn≤n+1. The example of
then-dimensional regular simplex inscribed in the unit sphere shows thatmn=n+1.
To determine explicitly the coordinates of the vertices, we use the additional information
that the distance from the center of the sphere to a hyperface of then-dimensional simplex
is^1 nand then find inductively
A 1 =(− 1 , 0 , 0 , 0 ,..., 0 , 0 ),
A 2 =
(
1
n
,−c 1 , 0 , 0 ,..., 0 , 0
)
,
A 3 =
(
1
n
,
1
n− 1
·c 1 ,−c 2 , 0 ,..., 0 , 0
)
,
A 4 =
(
1
n
,
1
n− 1
·c 1 ,
1
n− 2
·c 2 ,c 3 ,..., 0 , 0
)
,
···
An− 1 =
(
1
n
,
1
n− 1
·c 1 ,...,
1
3
·cn− 3 ,−cn− 2 , 0