158 CHAPTER 6. FIXED POINT THEOREMS

constants, still denoted as c j such that

b=1

n+1

n

∑i=0xi = y0 +

n−1

∑j=1

c j(y j−y0

)Thus

1n+1

n

∑i=0

∑s̸=k

t0s (xi−xs) =

n−1

∑j=1

c j(y j−y0

)=

n−1

∑j=1

c j

(∑s̸=k

t jsxs−∑

s ̸=kt0s xs

)

=n−1

∑j=1

c j

(∑s ̸=k

t js (xs−x0)−∑

s ̸=kt0s (xs−x0)

)Modify the term on the left and simplify on the right to get

1n+1

n

∑i=0

∑s ̸=k

t0s ((xi−x0)+(x0−xs)) =

n−1

∑j=1

c j

(∑s̸=k

(t js − t0

s)(xs−x0)

)Thus,

1n+1

n

∑i=0

(∑s ̸=k

t0s

)(xi−x0) =

1n+1

n

∑i=0

∑s ̸=k

t0s (xs−x0)

+n−1

∑j=1

c j

(∑s ̸=k

(t js − t0

s)(xs−x0)

)Then, taking out the i = k term on the left yields

1n+1

(∑s ̸=k

t0s

)(xk−x0) =−

1n+1 ∑

i̸=k

(∑s ̸=k

t0s

)(xi−x0)

1n+1

n

∑i=0

∑s ̸=k

t0s (xs−x0)+

n−1

∑j=1

c j

(∑s ̸=k

(t js − t0

s)(xs−x0)

)That on the right is a linear combination of vectors (xr−x0) for r ̸= k so by independence,∑r ̸=k t0

r = 0. However, each t0r ≥ 0 and these sum to 1 so this is impossible. Hence cn = 0

after all and so each c j = 0. Thus [y0, · · · ,yn−1,b] is an n simplex.Now in general, if you have an n simplex [x0, · · · ,xn] , its diameter is the maximum of

|xk−xl | for all k ̸= l. Consider∣∣b−x j

∣∣ . It equals∣∣∣∣∣ n

∑i=0

1n+1

(xi−x j)

∣∣∣∣∣=∣∣∣∣∣∑i ̸= j

1n+1

(xi−x j)

∣∣∣∣∣≤ nn+1

diam(S) .

Consider the kth face of S which is the simplex [x0, · · · , x̂k, · · · ,xn]. By induction, it has atriangulation into simplices which each have diameter no more than n

n+1 diam(S). Let these

n−1 simplices be denoted by{

Sk1, · · · ,Sk

mk

}. Then the simplices

{[Sk

i ,b]}mk,n+1

i=1,k=1 are a tri-angulation of S such that diam

([Sk

i ,b])≤ n

n+1 diam(S). Do for[Sk

i ,b]

what was just donefor S obtaining a triangulation of S as the union of what is obtained such that each simplexhas diameter no more than

( nn+1

)2 diam(S). Continuing this way shows the existence of

the desired triangulation. You simply do the process k times where( n

n+1

)k diam(S)< ε.

158 CHAPTER 6. FIXED POINT THEOREMSconstants, still denoted as c; such thatn—1= b= vot Des (yy = Yo)Thusn—-1 n—-1 .LEM 2) Loum) = Lo (Eve Ee)n+ 1 iO STh jal jal \sak sHk-Yo (Leena) Yi (a« m0)j=l s#k skModify the term on the left and simplify on the right to getEE erm) +) = Eos (El 2) ee]ad =0 stk jal \ sekThus,Ya (Zu-@-m)j=l stkThen, taking out the i = k term on the left yields1 1n+l (x) (@e—@0) =~ E(r") (21 — ®o)s#k s#kn—|YY ee s—ao) + Ye; e (i! —72) ie-))HH i=0 sFk jal \sHkThat on the right is a linear combination of vectors (a, — a9) for r #k so by independence,Leek 1? = 0. However, each 1° > 0 and these sum to | so this is impossible. Hence c, = 0after all and so each cj = 0. Thus [yo,--- , y,_1,6] is ann simplex.Now in general, if you have an n simplex [ao,--- ,x,]|, its diameter is the maximum of|r, — a)| for all k #1. Consider |b —aj|. It equalsmd 1y? (aj —2x;)| = y (aj—2xj)| < diam (S).agntl igi 1 1Consider the k’” face of S which is the simplex [a9,--- ,@x,--+ ,@]. By induction, it has atriangulation into simplices which each have diameter no more than =7> diam (S). Let thesen— 1 simplices be denoted by {St md Then the simplices { ist, b yen , are a tri-angulation of S such that diam ([S*, 6) < < jf diam (5). Do for [sé 2] what was just donefor S obtaining a triangulation of S as the union of what is obtained such that each simplexSin k. 2... Lo, . .has diameter no more than (4) diam(S). Continuing this way shows the existence ofthe desired triangulation. You simply do the process k times where (4 aT )* diam (S) < €.