Lecture Notes in Mathematics: For Information About Vols. 1-980 Please Contact Your Bookseller or Springer-Verlag

Lecture Notes in Mathematics
Vol 1008 A l g e b r a i c G e o m e t r y . P r o c e e d i n g s , 1981. E d i t e d by J. Dol-

For information about Vols. 1-980 please contact your bookseller
or Springer-Verlag. gachev V 138 p a g e s . 1 9 8 3
Vol. 9 8 1 : V a l u e D i s t r i b u t i o n Theory. P r o c e e d i n g s , 1981. E d i t e d by Vo\. :

T - C h a p m a n , Controlled Simple Homotopy Theory and
A
I. L a m e a n d S. R i c k m a n . VIII, 2 4 5 p a g e s . 1 9 8 3 . A p p l i c a t i o n s . Ill, 9 4 p a g e s . 1 9 8 3 .
Vol. 9 8 2 : Stability P r o b l e m s for S t o c h a s t i c M o d e l s . P r o c e e d i n g s , Vrjl 1010 J E. D i e s , C h a i n e s d e M a r k o v sur les p e r m u t a t i o n s . IX,

1 9 8 2 . E d i t e d by V. V. K a l a s h n i k o v a n d V. M. Zolotarev. XVII, 2 9 5 p a g e s . 226 p»9 ' - f s , 9 8 3
-
1983.
Vol. 1011: J M . S i g a l . S c a t t e r i n g T h e o r y for M a n y - B o d y Quantum
Vol. 9 8 3 : N o n s t a n d a r d A n a l y s i s - R e c e n t D e v e l o p m e n t s . E d i t e d by M e c h a n i c a l S y s t e m s . IV, 132 p a g e s . 1 9 8 3 .
A . E . H u r d . V, 213 p a g e s . 1 9 8 3 .
Vol. 1012: S. K a n t o r o v i t z , S p e c t r a l T h e o r y of B a n a c h S p a c e O p e r -
Vol. 9 8 4 : A. B o v e , J. E. L e w i s , C. P a r e n t i , P r o p a g a t i o n of S i n g u l a r i t i e s a t o r s . V, 1 7 9 p a g e s . 1 9 8 3 .
for F u c h s i a n O p e r a t o r s . IV, 161 p a g e s . 1 9 8 3 .
Vol 1013: C o m p l e x A n a l y s i s - Fifth R o m a n i a n - F i n n i s h Seminar,
Vol. 9 8 5 : A s y m p t o t i c A n a l y s i s II. E d i t e d by F. Verhulst. VI, 4 9 7 p a g e s . p a r t 1. P r o c e e d i n g s , 1981. E d i t e d by C. A n d r e i a n C a z a c u , N. B o b o c ,
1983. y\ j u r f ; h ^ s c u a n d I. S u c i u . XX, 3 9 3 p a g e s . 1983.
Vol. 9 8 6 : S e m i n a i r e d e P r o b a b i l i t y XVII 1 9 8 1 / 8 2 . P r o c e e d i n g s .
Vol 1014: C o m p l e x A n a l y s i s - Fifth R o m a n i a n - F i n n i s h Seminar,
E d i t e d by J. A z e m a a n d M. Yor. V, 512 p a g e s . 1 9 8 3 .
p a r t 2. P r o c e e d i n g s , 1981. E d i t e d by C. A n d r e i a n C a z a c u , N. B o b o c ,
Vol. 9 8 7 : C. J. B u s h n e l l , A. F r o h l i c h , G a u s s S u m s a n d p-adic D i v i s i o n M j u r o h f j s c u a n d I. S u c i u . XX, 3 3 4 p a g e s . 1 9 8 3 .
A l g e b r a s . XI, 187 p a g e s . 1 9 8 3 .
Vol. 1015: E q u a t i o n s d i f f e r e n t i e l l e s et s y s t e m e s d e R a f f d a n s le
Vol. 9 8 8 : J. S c h w e r m e r , K o h o m o l o g i e a r i t h m e t i s c h d e f i n i e r t e r G r u p - c h a T i p c o m p l e x e - II. S e m i n a r . Edited by R. G e r a r d et J. R R a m i s .
p e n u n d E i s e n s t e i n r e i h e n . Ill, 170 p a g e s . 1 9 8 3 . V ) 1 pages. 1983.
4
Vol. 9 8 9 : A. B. M i n g a r e l l i , Volterra-Stieltjes Integral E q u a t i o n s a n d Vol. 1016: A l g e b r a i c G e o m e t r y . P r o c e e d i n g s , 1 9 8 2 . E d i t e d by M.

G e n e r a l i z e d O r d i n a r y D i f f e r e n t i a l E x p r e s s i o n s . XIV, 318 p a g e s . 1 9 8 3 . R a y n a u d a n d T. S h i o d a . VIII, 5 2 8 p a g e s . 1 9 8 3 .
Vol. 9 9 0 : P r o b a b i l i t y in B a n a c h S p a c e s IV. P r o c e e d i n g s , 1 9 8 2 . Vol 1017: Equadiff 8 2 . P r o c e e d i n g s , 1 9 8 2 . Edited by H . W . K n o b l o c h

E d i t e d by A. B e c k a n d K. J a c o b s . V, 2 3 4 p a g e s . 1 9 8 3 . a n d K. S c h m i t t . XXIII, 6 6 6 p a g e s . 1 9 8 3 .
Vol. 9 9 1 : B a n a c h S p a c e T h e o r y a n d its A p p l i c a t i o n s . P r o c e e d i n g s , Vol 1018: G r a p h Theory, t a g o w 1981. P r o c e e d i n g s , 1981. E d i t e d by

1981. E d i t e d by A. P i e t s c h , N. P o p a a n d I. Singer. X, 3 0 2 p a g e s . M E t o r o w e c k i , J. W. K e n n e d y a n d M . M . S y s l o . X, 2 8 9 p a g e s . 1 9 8 3 .
1983.
Vol 10l9 :
C a b a l S e m i n a r 7 9 - 8 1 . P r o c e e d i n g s , 1 9 7 9 - 8 1 . E d i t e d by
Vol. 9 9 2 : H a r m o n i c A n a l y s i s , P r o c e e d i n g s , 1 9 8 2 . E d i t e d by G . M a u -
A S K e c h r i s , D. A. M a r t i n a n d Y. N. M o s c h o v a k i s . V, 2 8 4 p a g e s . 1 9 8 3 .
c e r i , F. R i c c i a n d G . W e i s s . X, 4 4 9 p a g e s . 1 9 8 3 .
Vol. 1 0 2 0 : N o n C o m m u t a t i v e H a r m o n i c A n a l y s i s a n d Lie G r o u p s ,
Vol. 9 9 3 : R. D. B o u r g i n , G e o m e t r i c A s p e c t s of C o n v e x S e t s w i t h the
p r o c e e d i n g s , 1 9 8 2 . E d i t e d by J. C a r m o n a a n d M. V e r g n e . V, 187
R a d o n - N i k o d y m Property. XII, 4 7 4 p a g e s . 1 9 8 3 .
pages 1983.
Vol. 9 9 4 : J.-L. Journe, C a l d e r d n - Z y g m u n d O p e r a t o r s , Pseudo-Dif-
Vol 1021: P r o b a b i l i t y Theory and Mathematical Statistics. Pro-
ferential O p e r a t o r s a n d t h e C a u c h y Integral of C a l d e r d n . VI, 129
c e e d i n g s . 1 9 8 2 . E d i t e d by K. ltd a n d J.V. P r o k h o r o v . VIII, 7 4 7 p a g e s .
pages. 1983.
1983.
Vol. 9 9 5 : B a n a c h S p a c e s , H a r m o n i c A n a l y s i s , a n d P r o b a b i l i t y Theory.
y | 0 1022: G. G e n t i l i , S. S a l a m o n a n d J.-R V i g u e . G e o m e t r y S e m i n a r
Proceedings, 1 9 8 0 - 1 9 8 1 . Edited by R . C . Blei a n d S . J . Sidney.
-Luigi B i a n c h i " , 1 9 8 2 . E d i t e d by E. V e s e n t i n i . VI, 177 p a g e s . 1 9 8 3 .
V, 173 p a g e s . 1 9 8 3 .
Vol. 9 9 6 : Invariant Theory. P r o c e e d i n g s , 1 9 8 2 . Edited by F. G h e r a r - Vol 1023: S. M c A d a m , A s y m p t o t i c P r i m e D i v i s o r s . IX, 118 p a g e s .

d e l l i . V, 159 p a g e s . 1 9 8 3 . 1983.
Vol 1024: Lie G r o u p R e p r e s e n t a t i o n s I. P r o c e e d i n g s , 1 9 8 2 - 1 9 8 3 .
Vol. 9 9 7 : A l g e b r a i c G e o m e t r y - O p e n P r o b l e m s . E d i t e d by C. C i l i -
berto, F. G h i o n e a n d F. O r e c c h i a . VIII, 411 p a g e s . 1 9 8 3 . I'dited by R. H e r b , R. L i p s m a n a n d J. R o s e n b e r g . IX, 3 6 9 p a g e s . 1 9 8 3 .
Vol. 9 9 8 : R e c e n t D e v e l o p m e n t s in t h e A l g e b r a i c , A n a l y t i c a l , a n d Vol 10215: D. Tanre, H o m o t o p i e R a t i o n n e l l e : M o d e l e s d e C h e n ,

T o p o l o g i c a l T h e o r y of S e m i g r o u p s . P r o c e e d i n g s , 1981. E d i t e d by OuiHen. S u l l i v a n . X, 211 p a g e s . 1 9 8 3 .
K . H . H o f m a n n , H. J u r g e n s e n a n d H . J . W e i n e r t . VI, 4 8 6 p a g e s . 1 9 8 3 .
Vol 1020: W. P l e s k e n , G r o u p R i n g s of Finite G r o u p s O v e r p-adic
Vol. 9 9 9 : C. P r e s t o n , Iterates of M a p s o n a n Interval. VII, 2 0 5 p a g e s . integers. V, 151 p a g e s . 1 9 8 3 .
1983.
Vol 1027: M. H a s u m i , H a r d y C l a s s e s o n Infinitely C o n n e c t e d Rie-
Vol. 1000: H. H o p f , D i f f e r e n t i a l G e o m e t r y in the L a r g e , VII, 184 p a g e s . . n a n n S u r f a c e s . XII, 2 8 0 p a g e s . 1983.
1983.
Vol 1028: S e m i n a i r e d ' A n a l y s e P L e l o n g - R D o l b e a u l t - H. S k o d a .
Vol. 1001: D.A. H e j h a l , T h e S e l b e r g T r a c e F o r m u l a f o r P S L ( 2 , nR).
A n n ( ics 1981/1983. E d i t e par R Lelong, R D o l b e a u l t et H. S k o d a .
V o l u m e 2. VIII, 8 0 6 p a g e s . 1 9 8 3 .
VIII, 3 2 8 p a g e s . 1 9 8 3 .
Vol. 1002: A. E d r e i , E.B. Saff, R.S. Varga, Z e r o s of S e c t i o n s of
Vol. 1029: S e m i n a i r e d ' A l g e b r e Paul D u b r e i l e t M a r i e - P a u l e M a l l i a v i n .
P o w e r S e r i e s . VIII, 115 p a g e s . 1 9 8 3 .
P r o c e e d i n g s , 1 9 8 2 . E d i t e par M.-R Malliavin. V, 3 3 9 p a g e s . 1 9 8 3 .
Vol. 1003: J. S c h m e t s , S p a c e s of V e c t o r - V a l u e d C o n t i n u o u s Func-
tions. VI, 117 p a g e s . 1 9 8 3 . Vol. 1030: U. C h r i s t i a n , S e l b e r g ' s Zeta-, L-, a n d E i s e n s t e i n s e r i e s .
XII 196 pages. 1983.
Vol. 1 0 0 4 : U n i v e r s a l A l g e b r a a n d Lattice Theory. P r o c e e d i n g s , 1982.
Edited by R . S . F r e e s e a n d O . C . G a r c i a . VI, 3 0 8 p a g e s . 1 9 8 3 . Vol- 1 ^ 3 D y n a m i c s a n d P r o c e s s e s . P r o c e e d i n g s , 1981. E d i t e d by
,:
Ph B i a n c h a r d a n d L. Streit. IX, 213 p a g e s . 1 9 8 3 .

Vol. 1005: N u m e r i c a l M e t h o d s . P r o c e e d i n g s , 1 9 8 2 . E d i t e d by V. Pe-
reyra and A. R e m o z a . V, 2 9 6 p a g e s . 1 9 8 3 . Vol. 1032: O r d i n a r y Differential Equations and Operators. Pro-
ceedings, 1982. E d i t e d by W . N. Everitt a n d R. T. L e w i s . XV, 521 p a g e s .
Vol. 1006: A b e l i a n G r o u p Theory. P r o c e e d i n g s , 1 9 8 2 / 8 3 . E d i t e d by
R. G o b e l , L. L a d y a n d A. Mader. XVI, 771 p a g e s . 1 9 8 3 .
Vol 1033: M e a s u r e T h e o r y a n d its A p p l i c a t i o n s . P r o c e e d i n g s , 1 9 8 2 .
Vol. 1007: G e o m e t r i c D y n a m i c s . P r o c e e d i n g s , 1981. E d i t e d by J. Palis
Ed'te^ by J M . Belley, J. D u b o i s a n d R M o r a l e s . XV, 317 p a g e s . 1 9 8 3 .
Jr. IX. 8 2 7 p a g e s . 1 9 8 3 .
continued on page 335

Lecture Notes in
Mathematics
Edited by A. Dold and B. Eckmann
1173
Hans Delfs
Manfred Knebusch
Locally Semialgebraic Spaces

UBR UBR UBR UBR UBR
069008388822
Springer-Verlag
Berlin Heidelberg New York Tokyo
%0 £l £sro - --^-7s5
Authors
Hans Deifs
Manfred Knebusch
Fakultat fur Mathematik, Universitat Regensburg
Universitatsstr. 3 1 , 8400 Regensburg
Federal Republic of Germany
Univ.-Biblfothek
Regensburg
S
Mathematics Subject Classification (1980): 1 4 G 3 0 , 54E99, 5 5 Q 0 5 , 5 7 R 0 5
ISBN 3-540-16060-4 Springer-Verlag Berlin Heidelberg New York Tokyo

ISBN 0-387-16060-4 Springer-Verlag New York Heidelberg Berlin Tokyo
This work is subject to copyright. All rights are reserved, whether the whole or part of the material
is concerned, specifically those of translation, reprinting, re-use of illustrations, broadcasting,
reproduction by photocopying machine or similar means, and storage in data banks. Under
§ 54 of the German Copyright Law where copies are made for other than private use, a fee is
payable to "Verwertungsgesellschaft Wort", Munich.
© by Springer-Verlag Berlin Heidelberg 1985
Printed in Germany
Printing and binding: Beltz Offsetdruck, Hemsbach/Bergstr.
2146/3140-543210
To C h r i s t l and Gisela
Preface
The primary occupation o f r e a l a l g e b r a i c geometry, o r b e t t e r "semialge-
braic geometry", i s t o s t u d y the s e t of s o l u t i o n s of a f i n i t e system of
polynomial i n e q u a l i t i e s i n a f i n i t e number o f v a r i a b l e s over the field
3R o f r e a l numbers. One wants t o do t h i s i n a conceptual way, not always
m e n t i o n i n g the p o l y n o m i a l d a t a , s i m i l a r l y as i n a l g e b r a i c geometry, say
o v e r €, where one most o f t e n a v o i d s working e x p l i c i t e l y w i t h the
systems o f p o l y n o m i a l e q u a l i t i e s (and n o n - e q u a l i t i e s f * 0) involved.
But a semialgebraic geometry which d e s e r v e s i t s name s h o u l d be able to
work - a t l e a s t - o v e r an a r b i t r a r y r e a l c l o s e d f i e l d R instead of the
f i e l d JR . Such f i e l d s are u s e f u l and even u n a v o i d a b l e i n semialgebraic
geometry f o r much the same r e a s o n as a l g e b r a i c a l l y c l o s e d f i e l d s of
• c h a r a c t e r i s t i c zero - a t l e a s t - a r e u n a v o i d a b l e i n a l g e b r a i c geometry
; o v e r <C, as soon as one t r i e s to avoid transcendental techniques or even
I then.
; In o r d e r t o i l l u s t r a t e t h i s we g i v e a somewhat t y p i c a l example. L e t
' f : V->W be an a l g e b r a i c map between i r r e d u c i b l e v a r i e t i e s over 1R . T h i s
y i e l d s , by restriction, a c o n t i n u o u s map f _ : V(]R) -+WCIR) between the

IK
s e t s o f r e a l p o i n t s . We assume t h a t W(3R) i s Z a r i s k i dense i n W which
.means t h a t W(3R) contains non s i n g u l a r points or, e q u i v a l e n t l y , that the
[ f u n c t i o n f i e l d 3R(W) i s formally r e a l . The generic f i b r e X of f, i . e .

1
X = f (n) w i t h n the g e n e r i c p o i n t of W ( r e g a r d i n g V and W as schemes),
\ i s an a l g e b r a i c scheme o v e r the function f i e l d 3R(W) of W, which con-
\ tains a l o t of information about f and f T O . But i t may be too difficult
[ t o study X, s i n c e the field IR(W) i s u s u a l l y very complicated. In alge-
| b r a i c geometry one often replaces X by the a l g e b r a i c v a r i e t y X obtained
I from X by e x t e n s i o n o f the base f i e l d ]R(W) t o the a l g e b r a i c c l o s u r e C
o f 3E(W) . I t i s much e a s i e r t o s t u d y the "geometric g e n e r i c fibre" X

i n s t e a d o f X, and still one may hope t o e x t r a c t r e l e v a n t i n f o r m a t i o n
about f from X. But i n s e m i a l g e b r a i c geometry t h i s p r o c e d u r e i s not ad-
visable, s i n c e most r e a l phenomena i n X w i l l be d e s t r o y e d i n X. Instead
o f X one s h o u l d study t h e v a r i e t i e s X , Q o b t a i n e d from X by base exten-
s i o n from IR(W) to the r e a l c l o s u r e s R a o f IR(W) with r e s p e c t to the
v a r i o u s o r d e r i n g s a o f the f u n c t i o n f i e l d IR(W) , and the s e t s o f ratio-

C
nal points X (R ). F o r e v e r y such a we have R^(V T) = C. Thus t h e R„
a ot a a
a r e "as near as p o s s i b l e " t o C and n e v e r t h e l e s s we may hope to d e t e c t
some o f t h e r e a l phenomena o f X, and u l t i m a t e l y of f , i n t h e s e t s
v v -
The v a r i e t y X i s the p r o j e c t i v e l i m i t o f the schemes f 1

(U) = V x ^ u with
U r u n n i n g through t h e Z a r i s k i - o p e n s u b s e t s o f W, s i n c e these U are the
Z a r i s k i neighbourhoods o f t h e g e n e r i c p o i n t ri i n W. S i m i l a r l y X i s the
p r o j e c t i v e l i m i t of the f i b r e products V ^ u , w i t h r e s p e c t t o the etale
morphisms cp : U -»W from a r b i t r a r y v a r i e t i e s U over IR to W (U * 0 , but
U(IR) may be empty), s i n c e t h e s e morphisms cp a r e the e t a l e neighbour-
hoods o f n. How about the X ? a An o r d e r i n g a of IR(W) corresponds uni-
quely t o an u l t r a f i l t e r F i n t h e Boolean lattice )T(W ( IR ) o f semialge-
braic s u b s e t s o f W(IR) such t h a t every A €F has a non empty i n t e r i o r A
i n the strong topology (= c l a s s i c a l t o p o l o g y on W(IR)), which means
t h a t A i s Z a r i s k i dense i n W, c f . [B, 8.11],[Br, § 4 ] . (A r a t i o n a l func-
t i o n h €IR(W) i s p o s i t i v e w i t h r e s p e c t t o a i f and o n l y i f h i s d e f i n e d
and p o s i t i v e on some s e t A C F ) . I t t u r n s o u t t h a t X a i s the projective
l i m i t o f t h e f i b r e p r o d u c t s Vxû w i t h r e s p e c t t o those e t a l e morphisms
cp : U ->W such t h a t cp(U(IR)) €F. (N.B. cp(U(IR)) i s semialgebraic.) This
is due t o the f a c t t h a t R a can be i n t e r p r e t e d as the u n i o n o f t h e rings
o f Nash f u n c t i o n s & (U) W on t h e v a r i o u s smooth open s e t s U €F, c f . [Ry].
Much more can be s a i d about a geometric i n t e r p r e t a t i o n over IR o f the

x R B u t
fields R,
a the v a r i e t i e s X Q and the p o i n t s i n a ( a )* t h i s would
t a k e u s t o o f a r a f i e l d . We o n l y mention t h a t t h e r e a l s p e c t r a o f commu-
tative rings i n v e n t e d by M. C o s t e and M.F. Coste-Roy provide exactly
the r i g h t language t o u n d e r s t a n d a l l t h i s , c f . [CR], [Ry], and t h e
l i t e r a t u r e c i t e d t h e r e and, f o r an i n t r o d u c t i o n t o r e a l s p e c t r a , a l s o
[L, § 4 , § 7 ] , [Br, §3, § 4 ] , [ K ] , [BCR, Chap. 7 ] .
We have been somewhat vague above. I n p a r t i c u l a r we d i d n o t make p r e -
c i s e t h e v a r i o u s d i r e c t systems which y i e l d the p r o j e c t i v e l i m i t s X
and X . We o n l y wanted t o i n d i c a t e t h a t i n s e m i a l g e b r a i c geometry

Q over
IR r e a l c l o s e d f i e l d s may come up i n a n a t u r a l and g e o m e t r i c way.
The present l e c t u r e notes give a c o n t r i b u t i o n to a b a s i c but rather
modest a s p e c t o f s e m i a l g e b r a i c geometry: t h e t o p o l o g i c a l phenomena o f
semialgebraic s e t s i n V(R) f o r V a v a r i e t y over a r e a l closed f i e l d R.
There i s a d i f f i c u l t y with t h e word " t o p o l o g i c a l " h e r e . Of c o u r s e , V(R)
i s equipped w i t h the strong topology coming from t h e t o p o l o g y of the
ordered field R. B u t , e x c e p t i n t h e case R = I R , t h e t o p o l o g i c a l s p a c e
V(R) i s totally disconnected.
These p a t h o l o g i e s can be remedied by c o n s i d e r i n g on V(R) a t o p o l o g y i n
the sense o f G r o t h e n d i e c k , where o n l y open s e m i a l g e b r a i c s u b s e t s U o f
V(R) are admitted as "open s e t s " , and f o r such a s e t U e s s e n t i a l l y only
c o v e r i n g s by f i n i t e l y many open s e m i a l g e b r a i c subsets of U a r e admitted
as "open c o v e r i n g s " .
It seems t h a t t h e c a t e g o r y o f s e m i a l g e b r a i c spaces and maps over a r e a l
closed f i e l d R, which has been i n t r o d u c e d i n o u r paper [DK ],

2 provides
the r i g h t framev/ork f o r t h i s " s e m i a l g e b r a i c topology". Already i n that
paper and l a t e r i n o t h e r ones ( [ D ] , [ D ] , [DK^],

1 [DK^], [DK^]) we found
analogues o f many r e s u l t s i n c l a s s i c a l topology. Sometimes t h i n g s a r e
even n i c e r h e r e . T h i s i s n o t a s t o n i s h i n g s i n c e , i n t h e case R =IR, t h e

s e m i a l g e b r a i c s e t s a r e r a t h e r tame from a t o p o l o g i c a l viewpoint.
I n t h e case R =IR the c a t e g o r y of semialgebraic spaces can be compared
with the category of t o p o l o g i c a l spaces, and this a f f o r d s us a new
perspective concerning t h e two b r a n c h e s o f mathematics i n v o l v e d , s e m i -
a l g e b r a i c geometry and a l g e b r a i c topology, c f . the i n t r o d u c t i o n of [B],
F o r example, a l o n g j o u r n e y along this road should g i v e a thorough un-
d e r s t a n d i n g o f why so many spaces o c c u r i n g i n u s u a l a l g e b r a i c topology
are semialgebraic s e t s .
Nevertheless the c a t e g o r y o f s e m i a l g e b r a i c spaces i s t o o restrictive
f o r some p u r p o s e s . A good i n s t a n c e where t h i s can be seen i s the theory
of semialgebraic c o v e r i n g s . I f M i s a connected affine semialgebraic
space over R, and x q i s some p o i n t i n M, we can d e f i n e t h e fundamental
group ( M , x ) i n t h e u s u a l way
Q as the s e t o f s e m i a l g e b r a i c homotopy
*)
c l a s s e s o f s e m i a l g e b r a i c l o o p s w i t h base p o i n t x Q ( c f . Ill,§6) . This
i s an honest t o goodness group, g e n e r a t e d by f i n i t e l y many e l e m e n t s s a -
tisfying f i n i t e l y many r e l a t i o n s . On the o t h e r hand we e v i d e n t l y have
t h e n o t i o n o f an (unramified) c o v e r i n g p :N -»M o f M, p being a locally
trivial s e m i a l g e b r a i c map with d i s c r e t e (= z e r o - d i m e n s i o n a l ) fibres.
Of course, one would l i k e t o c l a s s i f y the c o v e r i n g s o f M by subgroups
o f TT.j ( M , x ) . But
Q a zero-dimensional s e m i a l g e b r a i c space i s n e c e s s a r i l y
a finite s e t . Thus e v e r y s e m i a l g e b r a i c c o v e r i n g has f i n i t e degree. I t
can be shown t h a t i n d e e d the isomorphism c l a s s e s o f semialgebraic
coverings of M correspond u n i q u e l y to t h e c o n j u g a c y c l a s s e s o f sub-
groups o f f i n i t e index in ( M , x ) i n t h e u s u a l way.

Q But there should
a l s o e x i s t c o v e r i n g s o f a more g e n e r a l n a t u r e which c o r r e s p o n d to the
other subgroups o f tt^ (M,x ) . In p a r t i c u l a r

Q there should exist a uni-
v e r s a l c o v e r i n g o f M. These more g e n e r a l c o v e r i n g s can be d e f i n e d i n
*) T h i s means §6 i n C h a p t e r I I I o f t h i s book.
the category of " l o c a l l y s e m i a l g e b r a i c " spaces and maps.
A f t e r s e v e r a l y e a r s o f e x p e r i m e n t i n g w i t h l o c a l l y s e m i a l g e b r a i c spaces
we a r e c o n v i n c e d t h a t t h e s e spaces e x i s t " i n n a t u r e " . The coverings of
a f f i n e s e m i a l g e b r a i c spaces a r e r e g u l a r paracompact l o c a l l y semialge-
b r a i c spaces, t o be d e f i n e d i n I , §4. R e g u l a r paracompact spaces seem
t o be t h e "good" l o c a l l y s e m i a l g e b r a i c spaces, analogous t o the affine
spaces i n the s e m i a l g e b r a i c c a t e g o r y . F o r i n s t a n c e , f o r t h e s e spaces
t h e r e e x i s t s a s a t i s f a c t o r y cohomology t h e o r y o f sheaves, based on
f l a b b y and s o f t sheaves, which p a r a l l e l s the c l a s s i c a l t h e o r y f o r topo-
l o g i c a l paracompact s p a c e s . We w i l l not d e a l w i t h t h e s e m a t t e r s here,
e x c e p t f o r some b r i e f remarks i n Appendix A, b u t they a r e q u i t e impor-
t a n t f o r d e f i n i n g homology and cohomology groups of v a r i o u s k i n d s f o r
t h e s e s p a c e s , c f . [D], [ D ^ , [D ].
2
Although r e g u l a r paracompact spaces a r e a v e r y s a t i s f y i n g subclass of
locally s e m i a l g e b r a i c spaces one has to f a c e t h e f a c t t h a t t h e r e e x i s t
many l o c a l l y s e m i a l g e b r a i c spaces i n n a t u r e which a r e not paracompact.
( I t seems t h a t r e g u l a r i t y may be assumed i n most a p p l i c a t i o n s . ) For
example, s t u d y i n g open s u b s e t s o f q u i t e i n n o c e n t l y l o o k i n g r e a l s p e c t r a
may l e a d to r e g u l a r spaces which a r e not paracompact, c f . Appendix A.
Thus i t i s not j u s t f o r f u n o r f o r s y s t e m a t i c r e a s o n s t h a t we study
i n Chapter I more g e n e r a l s p a c e s . In t h e l a t e r c h a p t e r s we are forced
to r e s t r i c t t o paracompact s p a c e s , s i n c e o t h e r w i s e our deeper techniques
break down.
There i s one phenomenon i n our t h e o r y which may seem somewhat u n u s u a l
f o r a r e a d e r o f our p r e v i o u s p a p e r s . I n a s e m i a l g e b r a i c space M i t i s
strictly f o r b i d d e n t o work w i t h s u b s e t s o f M o t h e r than the s e m i a l g e b r a i c
subsets [DK f 2 § 7 ] . But in a locally s e m i a l g e b r a i c space M t h e r e e x i s t
two n a t u r a l c l a s s e s o f a d m i s s i b l e s u b s e t s , t h e c l a s s T(M) of locally

s e m i a l g e b r a i c s u b s e t s o f M and the s m a l l e r c l a s s JT(M) o f s e m i a l g e b r a i c
s u b s e t s o f M. The i n t e r p l a y between T(M) and T(M) i s a theme w h i c h r e -
c u r s througout the whole t h e o r y .
The g o a l o f the f i r s t volume o f our l e c t u r e n o t e s i s t o e s t a b l i s h t h e
c a t e g o r y o f l o c a l l y s e m i a l g e b r a i c spaces and maps over an a r b i t r a r y
r e a l closed f i e l d R on f i r m grounds, and t o prove enough r e s u l t s about
t h e s e spaces and maps, t h a t t h e r e a d e r w i l l f e e l well acquainted with
them and w i l l r e g a r d them as c o n c r e t e and a c c e s s i b l e o b j e c t s . The n e x t
topics, t o be c o v e r e d i n t h e second volume, a r e t h e t h e o r y o f l o c a l l y
s e m i a l g e b r a i c f i b r a t i o n s and f i b r e bundles (Chapter IV) and t h e t h e o r y
of coverings (Chapter V ) .
As background m a t e r i a l we assume o u r papers [DK 3/

2 [DK^], [DK^], some
s e c t i o n s o f [DK^], and Robson's paper [R]. Here y o u f i n d n e a r l y e v e r y -
t h i n g which we need about s e m i a l g e b r a i c spaces, w r i t t e n up i n a s y s t e -
m a t i c way compatible w i t h t h e s p i r i t o f these l e c t u r e n o t e s . Of c o u r s e ,
i t would have been more c o m f o r t a b l e f o r t h e r e a d e r i f we had s t a r t e d
the l e c t u r e n o t e s w i t h a r e v i e w o f t h e r e s u l t s o f t h o s e p a p e r s . B u t
this i s n o t r e a l l y n e c e s s a r y and would have made t h e l e c t u r e n o t e s t o o
l o n g . Of c o u r s e , the book [BCR] o f Bochnak and the Costes - as soon
as i t has appeared - will c o n t a i n most b a s i c f a c t s which a r e n e c e s s a r y
f o r an u n d e r s t a n d i n g o f t h e s e l e c t u r e n o t e s and much more.
A s u r v e y on some b a s i c r e s u l t s about s e m i a l g e b r a i c spaces has been
g i v e n i n [DK]. Another s u r v e y on b a s i c r e s u l t s about locally semialge-
b r a i c s p a c e s , which, o f c o u r s e , a l l w i l l be c o v e r e d by t h e two volumes
o f t h e s e l e c t u r e n o t e s , has been g i v e n i n [DKg] and [DK^].
We hope t h a t these l e c t u r e n o t e s , d e s i g n e d i n f i r s t p l a c e f o r t h e needs
o f s e m i a l g e b r a i c geometry, a r e a l s o o f i n t e r e s t f o r t o p o l o g i s t s . The
main r e s u l t s a r e u s u a l l y non t r i v i a l a l s o i n t h e case R = 3R and not
much e a s i e r t o be proved i n t h i s s p e c i a l c a s e . The c a t e g o r y o f l o c a l l y

s e m i a l g e b r a i c spaces o v e r 3R lies somewhat " i n between" the c a t e g o r y
TOP o f t o p o l o g i c a l H a u s d o r f f spaces and the c a t e g o r y PL o f p i e c e w i s e
linear spaces, being l e s s r i g i d t h a n PL and, i n some r e s p e c t s , l e s s
p a t h o l o g i c a l than TOP.
The c e n t r a l r e s u l t o f t h e whole volume seems t o be Theorem 4.4 in
Chapter I I , §4, which s t a t e s t h a t e v e r y r e g u l a r paracompact locally
s e m i a l g e b r a i c space M can be t r i a n g u l a t e d , and moreover a g i v e n locally
finite family of l o c a l l y s e m i a l g e b r a i c s u b s e t s of M can be triangulated
s i m u l t a n e o u s l y . Thus we may r e g a r d e v e r y r e g u l a r paracompact space as
a locally f i n i t e p o l y h e d r o n w i t h some open f a c e s m i s s i n g ( c f . the defi-
n i t i o n of s t r i c t l y locally finite simplicial complexes, i n I , §2,
which i s s l i g h t l y d i f f e r e n t from the c l a s s i c a l d e f i n i t i o n ) . But i n c o n -
trast t o P L - t h e o r y , we may subdivide s i m p l i c e s not only l i n e a r l y but
" s e m i a l g e b r a i c a l l y " . N e v e r t h e l e s s , i n the s p e c i a l case t h a t R= IR and
M i s p a r t i a l l y complete, S h i o t a and Y o k o i have r e c e n t l y proved t h a t any two
PL s t r u c t u r e s on M which r e f i n e the g i v e n s e m i a l g e b r a i c s t r u c t u r e a r e
isomorphic ([SY, Th. 4.1], they prove t h i s more g e n e r a l l y f o r s u i t a b l e
locally s u b a n a l y t i c s p a c e s ) . T h i s remarkable theorem can be extended to
partially complete r e g u l a r paracompact spaces o v e r any R, as we hope t o
e x p l a i n i n the second volume.
If S i s a real closed f i e l d c o n t a i n i n g R then, as a consequence o f
T a r s k i ' s p r i n c i p l e , we can a s s o c i a t e w i t h e v e r y l o c a l l y s e m i a l g e b r a i c
space M over R a l o c a l l y s e m i a l g e b r a i c space M(S) over S by " e x t e n s i o n
o f the base f i e l d R t o S", c f . 1.2.10. T h i s y i e l d s a v e r y good n a t u r e d
f u n c t o r M^M(S) from t h e c a t e g o r y o f r e g u l a r paracompact spaces over R
t o the c a t e g o r y o f r e g u l a r paracompact spaces over S, which i s o f c r u -
c i a l importance f o r our whole t h e o r y . The homotopy groups (cf. I l l , §6),
the homology groups (cf. I l l , §7) and a l s o the v a r i o u s K-groups o f M
(orthogonal, u n i t a r y , s y r a p l e c t i c , c f . Chapter IV i n the second volume)
are p r e s e r v e d under base f i e l d e x t e n s i o n from R t o S. These are examples

of the main message o f our whole t h e o r y , t h a t over a complicated real
closed field the l o c a l l y s e m i a l g e b r a i c spaces a r e i n many r e s p e c t s not
more c o m p l i c a t e d than over a simple field, as t h e field IR o r the field
R Q o f r e a l a l g e b r a i c numbers. We b e l i e v e t h a t t h i s message i s by no
means t r i v i a l . I t may be regarded as a v a s t g e n e r a l i z a t i o n o f T a r s k i ' s
principle for topological statements. As soon as one l e a v e s the cadre
of semialgebraic topology and works, say w i t h a l g e b r a i c functions then
t h e a n a l o g u e o f o u r message seems t o h o l d o n l y under s e v e r e restrictions
F o r example, i t i s w e l l known t h a t , i n g e n e r a l , s e m i a l g e b r a i c f u n c t i o n s
on the u n i t interval [0,1] i n R c a n n o t be approximated uniformly by
polynomials, i n c o n t r a s t to the S t o n e - W e i e r s t r a B theorem f o r R= 3R .
The book has two appendices. Appendix B (to Chapter I) c o n t a i n s some
easy but fundamental r e s u l t s i n the theory of base e x t e n s i o n . They
have n o t been i n c l u d e d i n t o Chapter I s i n c e some o f the techniques
needed t o d e r i v e them seem t o have t h e i r n a t u r a l p l a c e i n Chapter II.
A p p e n d i x A i s o f d i f f e r e n t k i n d . Here we draw the c o n n e c t i o n s between
our t h e o r y and " a b s t r a c t " s e m i a l g e b r a i c geometry which, s t a r t i n g from
the n o t i o n o f the r e a l spectrum, now i s i n a process of r a p i d develop-
ment. A p p e n d i x A i s n o t needed f o r o u r theory in a technical sense,
b u t t h e r e we will find the o c c a s i o n t o e x p l a i n some more p o i n t s o f our
philosophy about t h e " r a i s o n d'etre" of l o c a l l y semialgebraic spaces.
We thank the members o f the former Regensburger s e m i a l g e b r a i c group,
i n p a r t i c u l a r Roland Huber and Robby Robson, f o r s t i m u l a t i n g d i s c u s s i o n s
and c r i t i c i s m about the contents of these l e c t u r e notes. S p e c i a l thanks
a r e due t o J o s e Manuel Gamboa and R. Huber f o r a p e n e t r a t i n g (and very
successful) search f o r mistakes i n the f i n a l v e r s i o n o f the manuscript.
We thank M a r i n a R i c h t e r f o r her p a t i e n c e and e x c e l l e n c e i n t y p i n g the
book and R. Robson f o r e l i m i n a t i n g some o f the most annoying grammatical
mistakes. We a r e w e l l aware t h a t we c o u l d have w r i t t e n a b e t t e r book

i n o u r n a t i v e language, b u t s i n c e t h e book i s d e s i g n e d as a "topologie
generale" f o r semialgebraic geometry w h i c h s h o u l d be u s e f u l as a widely
accepted reference, we have w r i t t e n i n t h a t l a n g u a g e w h i c h w i l l be
u n d e r s t o o d by the most.
Regensburg, J u l y 1985
Hans D e l f s , M a n f r e d Knebusch
TABLE OF CONTENTS
page
Chaptier I - The b a s i c d e f i n i t i o n s 1
§1 - L o c a l l y s e m i a l g e b r a i c spaces and maps 1
§2 - Inductive limits, some examples o f l o c a l l y semialge-
b r a i c spaces 11
§3 - L o c a l l y semialgebraic subsets 27
4 2
§4 - R e g u l a r and paracompact spaces
§5 - S e m i a l g e b r a i c maps and p r o p e r maps 54
§6 - P a r t i a l l y p r o p e r maps 63
§7 - L o c a l l y complete spaces 75
Chaptter I I - Completions and t r i a n g u l a t i o n s 87
§1 - G l u i n g paracompact spaces 87
§2 - E x i s t e n c e o f c o m p l e t i o n s 94
§3 - A b s t r a c t s i m p l i c i a l complexes 99
6
§4 - T r i a n g u l a t i o n o f r e g u l a r paracompact spaces 1°
§5 - T r i a n g u l a t i o n o f weakly s i m p l i c i a l maps, maximal
complexes 113
§6 - T r i a n g u l a t i o n o f amenable p a r t i a l l y f i n i t e maps 124

3 8
§7 - S t a r s and s h e l l s 1
§8 - Pure h u l l s o f dense p a i r s 146
§9 - Ends o f spaces, t h e L C - s t r a t i f i c a t i o n 156
§10 —Some p r o p e r q u o t i e n t s 178
§11 — M o d i f i c a t i o n o f pure ends 189
§12 — T h e S t e i n f a c t o r i z a t i o n o f a s e m i a l g e b r a i c map 198
§13 — S e m i a l g e b r a i c spreads 211
§14 —Huber's theorem on open mappings 219

C h a p t e r I I I - Homotopies 22 6
§1 - Some s t r o n g d e f o r m a t i o n r e t r a c t s 22 6
§2 - S i m p l i c i a l approximations 23 2
§3 - The f i r s t main theorem on homotopy s e t s ; mapping spaces 24 3
§4 - R e l a t i v e homotopy s e t s 24 9
§5 - The second main theorem; c o n t i g u i t y c l a s s e s 257
§6 - Homotopy groups 26 5
§7 - Homology; t h e Hurewicz theorems 278
§8 - Homotopy groups o f ends 286
Appendix A - A b s t r a c t locally s e m i a l g e b r a i c spaces 295
Appendix B - C o n s e r v a t i o n o f some p r o p e r t i e s o f spaces and maps
under base f i e l d e x t e n s i o n 309
References 315
L i s t o f symbols 319
Glossary 322
Chapter I . - The b a s i c d e f i n i t i o n s
The g o a l s of t h i s c h a p t e r a r e modest. We p r e s e n t the b a s i c d e f i n i t i o n s
and some elementary o b s e r v a t i o n s needed by anyone who wants to work
with locally s e m i a l g e b r a i c s p a c e s . The c o v e r i n g s mentioned i n the p r e -
f a c e w i l l be a s p e c i a l c l a s s o f t h e " p a r t i a l l y proper maps" c o n s i d e r e d
i n §6.
§1 - L o c a l l y s e m i a l g e b r a i c spaces and maps
Our i d e a i s t o d e f i n e l o c a l l y s e m i a l g e b r a i c spaces as s u i t a b l e rin-
ged spaces i n the sense o f G r o t h e n d i e c k . Then the more d e l i c a t e
q u e s t i o n how t o d e f i n e l o c a l l y s e m i a l g e b r a i c maps becomes trivial.
These maps w i l l be simply a l l t h e morphisms between the l o c a l l y s e -
m i a l g e b r a i c spaces i n the c a t e g o r y o f r i n g e d spaces. We used the
same procedure a l r e a d y t o d e f i n e s e m i a l g e b r a i c spaces in [DK2,§7].
Definition 1. A g e n e r a l i z e d t o p o l o g i c a l space *) i s a s e t M t o g e t h e r
w i t h a s e t 'f(M) of s u b s e t s o f M, c a l l e d the "open s u b s e t s " of M, and
a set Cov M of f a m i l i e s (U la€I)

a i n ^(M) * * ) , called the " a d m i s s i b l e
c o v e r i n g s " such t h a t the f o l l o w i n g p r o p e r t i e s h o l d :
i) 0 e 7(M) , M €tf(M).
ii) If U 1 € *(M) , U 2 € T(M) , then fl U 2 € f ( M ) and U 1 UU 2 € f(M) .
iii) Every family (U la€I)

a i n ^f(M) w i t h I f i n i t e i s an element
of Cov . M
M
iv) I f (UglaGI) i s an element o f Cov^, then the u n i o n
U := U ( U l a € I ) of t h i s f a m i l y i s an element o f jP(M) .
a
*) T h i s term s h o u l d be r e g a r d e d as ad hoc.
**) In order t o guarantee t h a t Cov^. i s r e a l l y a s e t one

should o n l y a l l o w s u b s e t s o f some f i x e d l a r g e s e t
as index s e t s I . We w i l l i g n o r e a l l s e t - t h e o r e t i c
d i f f i c u l t i e s here.
For any U € T(M) we denote t h e s u b s e t o f a l l (UglaGI) € C o v M
with U(U la€I)

a = U by Cov^dJ) and we c a l l these coverings
the " a d m i s s i b l e c o v e r i n g s o f U".
v) I f ( U l a e i ) i s an a d m i s s i b l e c o v e r i n g o f U € T(M)
a and i f
V € f(M) i s a subset o f U, then (U DVla€I)

a i s an a d m i s s i b l e
c o v e r i n g o f V.
vi) I f an a d m i s s i b l e c o v e r i n g (U lcx€I) o f U € T(M)

a i s g i v e n and
f o r every a € I an a d m i s s i b l e c o v e r i n g (V pl3 £J )
a a of U i s
a
g i v e n , then (V^Î a € 1 , 3 G J ) Q i s an a d m i s s i b l e c o v e r i n g o f U.
vii) I f (U |a€I) i s a f a m i l y i n f(M) with U(U la€I) a = U € fCM) and
if (Vpl3€J) € C o v ( U ) i s a r e f i n e m e n t
M of (U la€I) a {i.e. there
e x i s t s a map A : J-» I w i t h V ^ c u ^ ^ ^ o r e v e r
^ 3 € J} , then
(U la€I) € Cov (U).

a M
viii) I f U € f(M), (U la€I) 6 C o v ( U ) and i f V i s a subset o f U w i t h

M
a M
V n U € T(M)
a f o r every a€I, then V € f (M) [and t h u s , by (v) ,
(V n U loc€I) i s an a d m i s s i b l e c o v e r i n g o f V] .
Comments.
a) We u s u a l l y j u s t w r i t e "M" f o r t h e t r i p l e (M, T(M) , Cov^) . A genera-
lized topological space M i s a ( r a t h e r s p e c i a l ) example o f a s i t e i n
the sense o f G r o t h e n d i e c k . Thus we have G r o t h e n d i e c k ' s theory of
sheaves over such spaces M a t o u r d i s p o s a l , c f . [ A ] , [SGA4, Exp. II].
L e t us r e c a l l t h e n o t i o n o f an ( a b e l i a n ) sheaf here i n our s p e c i a l
s i t u a t i o n . A p r e s h e a f ? on M i s an assignment Uv+Û) o f an a b e l i a n
group ^(U) t o every U G CT(M) equipped w i t h a r e s t r i c t i o n homomorphism
r ^ : T(U) -> ^(V) f o r every p a i r o f open s e t s U,V w i t h V c U such that
r ^ = i d and rX°r^ = r ^ f o r U D V D W . A p r e s h e a f T i s a sheaf

r i f in
U W V W
a d d i t i o n f o r any a d m i s s i b l e c o v e r i n g (U^laEI) o f any U€ T(M), the
usual sequence
o-T(u) - TT7(\J) Z 1 r T(u n u J

<x€I (a,3)€Ixi a 13
is exact. Thus a sheaf i s v e r y much the same n o t i o n as f o r u s u a l to-
pological spaces, except t h a t now only sets i n T(M) are allowed as
open s e t s and only admissible coverings are a l l o w e d as open c o v e r i n g s .
b) The last two axioms ( v i i ) and (viii) i n the d e f i n i t i o n of a gene-
ralized t o p o l o g i c a l space are l e s s s u b s t a n t i a l than the others.
Axiom ( v i i ) i s j u s t a t e c h n i c a l d e v i c e to make formal arguments
smoother. N o t i c e t h a t i f axioms (i) - ( v i ) are f u l f i l l e d , then by
enriching the s e t s C o v ( U ) , U € T(M), by a l l f a m i l i e s

M (U |a€I) i n ?(M)
which have the union U and which admit r e f i n e m e n t s lying i n Cov^dJ),
we o b t a i n a new site (M,?(M),Cov^) which f u l f i l l s (i) - ( v i i ) and
which has the same sheaves as the o r i g i n a l site (M, (M) , C o v ) .

M
The r o l e o f axiom (viii) i s more s u b t l e and will be d i s c u s s e d at the
end of t h i s s e c t i o n .
From now on R denotes a f i x e d r e a l closed field.
Definition 2. A r i n g e d space over R i s a pair (M,0 ) c o n s i s t i n g of
a g e n e r a l i z e d t o p o l o g i c a l space M and a sheaf 0?^ of commutative R-
a l g e b r a s . A morphism (cp,*f) : (M,C? ) -* (N,© ) between r i n g e d

M N spaces
(M,# ) and
M (N,C? ) over
N R i s d e f i n e d i n the o b v i o u s way: cp i s a con-
t i n u o u s map from M t o N, i . e . every open s e t V i n N has an open p r e -

-1 -1
image cp (V) and f o r every (V la€I) e C o v ( V ) the f a m i l y (cp ( V ) l a € I )
N a
i s an a d m i s s i b l e c o v e r i n g of c p " (V) . The

1
second component 3 i s a I s o -
morphism from the sheaf (0^ to the sheaf t P * ^ r e s p e c t i n g the R-alge-
bra s t r u c t u r e s . In o t h e r words, f o r any open s e t s U i n M and V in N
with cp(U) c z v we have an R-algebra homomorphism
*U,V V >! V
-<V°>
with the u s u a l c o m p a t i b i l i t i e s w i t h r e s p e c t to the r e s t r i c t i o n maps.

Example. L e t M be a s e m i a l g e b r a i c space over R as d e f i n e d i n [DI^,
§7]. Choose f o r 7(M) the s e t ?(M) o f a l l open s e m i a l g e b r a i c s u b s e t s
of M, and f o r UG f(M) d e f i n e t h e s e t C o v ( U ) t o be the s e t of a l l

M
families (LLliGI) i n f(M) such t h a t the union U ( u \ | i € I ) = U and such
that finitely many U ^ i €1 , already cover U.
The axioms (i) - ( v i i i ) are c l e a r l y fulfilled. Thus M i s a genera-
lized topological space. (N.B. In t h i s way e v e r y "restricted topo-
logical space", as d e f i n e d i n [DK>, § 7 ] , can be regarded as a gene-
ralized topological s p a c e ) . The sheaves on t h i s generalized topolo-
gical space M a r e t h e same as the sheaves on t h e " r e s t r i c t e d topolo-
gical space" M c o n s i d e r e d i n [DI< ,§7]. 2 In p a r t i c u l a r on M t h e r e i s a
sheaf C? o f R - a l g e b r a s d e f i n e d as f o l l o w s :
M I f U € flM) , then # (U) i s M
the R-algebra o f s e m i a l g e b r a i c f u n c t i o n s f:U R. F o r open s e m i a l -
g e b r a i c s e t s V c U the r e s t r i c t i o n map r^ from

r (5> (U) t o (V) i s the
r
V M M
obvious r e s t r i c t i o n of f u n c t i o n s f »-> f | v . T h i s r i n q e d space (M,# ) M
over R i s r e a l l y t h e same o b j e c t as the s e m i a l g e b r a i c space M and
will be i d e n t i f i e d with i t .
If (M,<^) i s a r i n g e d space over R then f o r any U £ 7(M) we o b t a i n
"by r e s t r i c t i o n " a r i n q e d space (U,^|U) over R as f o l l o w s : cr(U)
c o n s i s t s o f a l l V G tf(M) w i t h V c U . Cov^ c o n s i s t s o f a l l f a m i l i e s
(V laGI) £ Cov., w i t h V c(] f o r every aGI, and C?|TJ i s t h e r e s t r i c t i o n

J
a M a M
of t h e sheaf C? t o U, i . e . (C? |U)(V) = ® ^ (V) f o r e v e r y V G 7(U) .
M M M
These r i n g e d spaces (U,0 IU) a r e c a l l e d
M t h e open subspaces o f (M,0 ).
An open subset U of M i s c a l l e d an open s e m i a l g e b r a i c subset i f (U,0 IU! M
i s a s e m i a l g e b r a i c space over R, as d e f i n e d i n the example above.
Definition 3. A l o c a l l y s e m i a l g e b r a i c space over R i s a r i n g e d space
(M,C> ) over R which p o s s e s s e s an a d m i s s i b l e c o v e r i n g (M^laGI) ECov^'M)

M
such t h a t a l l M a r e open s e m i a l g e b r a i c s u b s e t s o f M.
Let us l o o k a t these d e f i n i t i o n s more c l o s e l y . Assume t h a t (M,C?^) i s
a lO'Cally semialgebraic space and that (M^lcxEI) i s an admissible co-
vering o f M by open s e m i a l a e b r a i c s u b s e t s M^. What are the o t h e r open
semialgebraic subsets of M? C l e a r l y i f U€
; f(M) i s s e m i a l g e b r a i c , then
( U H M ^ I a E I ) i s an a d m i s s i b l e c o v e r i n g of the semialgebraic space U,
and thus U i s contained i n the union of f i n i t e l y many s e t s M^. Con-
versely i f U € 7(H) and UcMot^ U ... U M 0 r f o r f i n i t e l y many i n d i c e s
, ..., a r € I then the set Ma^ U ... U M ar = W i s open semialgebraic
in M by the v e r y d e f i n i t i o n of semialcrebraic spaces. Moreover U £ T(W) .
Thus (U,C? |U) i s a l s o a s e m i a l g e b r a i c

M space. The open semialgebraic
subsets of M are p r e c i s e l y those s e t s U € f(M) which are c o n t a i n e d in
the union o f f i n i t e l y many s e t s M . We will henceforth denote the
subset of ^?(M) c o n s i s t i n g of a l l open s e m i a l g e b r a i c subsets of M by
T(M) . N o t i c e that ?-(M) = f(M) i f and only i f M i t s e l f i s semialgebraic.
We clearly have the f o l l o w i n g r e l a t i o n s between the sets r(M) and
CT(M) .
a) A subset W of M belongs to IT(M) i f and only i f for every
(U IA€A)} €Cov (U),U€ f(M) , a l l the

M intersections W D are elements
of ?(M) and W flU i s covered by f i n i t e l y many s e t s W D .
b) A subset U of M belongs to f(M) i f and only i f U PI W € y-(M) for
every W€ f(M) . A l s o U€ f(M) i f and only i f UOM^ fi(M) f o r every a€I.
It i s a l s o an easy consequence of our d e f i n i t i o n s , in particular of
the axioms v i ) , v i i ) , v i i i ) in Definition 1, that T(M) determines
the s e t Cov^ i n the f o l l o w i n g way:
c) A f a m i l y (UÎA€A) i n T'(M) b e l o n g s t o Cov^ i f and only i f for
every W € T(M) the i n t e r s e c t i o n W D U of Wwith the union U of the f a m i l y
i s covered by f i n i t e l y many s e t s W D U ^ , A G A . In f a c t , i t suffices
t h a t f o r every a € l the intersection n U i s covered by finitely
many s e t s M nu>,A€A.
Definition 4. A f a m i l y (X^|A€A) of subsets of M i s c a l l e d locally
finite i f any W€ s{M) meets o n l y f i n i t e l y many X^, i n o t h e r words„
i f W n X^ * 0 f o r o n l y f i n i t e l y many A € A . Again i t i s only necessary
to check t h a t , f o r every a € I,the set M a meets o n l y f i n i t e l y many .
As a special case o f our o b s e r v a t i o n ( c ) , we have
Proposition 1 . 1 . Every locally finite f a m i l y i n T(M) i s an element o f
Cov^. In p a r t i c u l a r , the union of t h i s f a m i l y i s an element o f T(M) .
F o r every x € M, t h e s t a l k ^ i s a local ring and t h e n a t u r a l map
from R t o t h e r e s i d u e c l a s s f i e l d 0 /+* o f (9„ i s an isomorphism.

M,x M,x M,x
Indeed, t h i s i s known t o be t r u e f o r a l l t h e s e m i a l g e b r a i c spaces
( M , < 0 l M ) and thus

a M a a l s o holds f o r (M,© ) . We i d e n t i f y ©
M M X /** M x
with the f i e l d R. I f U 6 f(M) and f GO (U) then f yields an R - v a l u e d
function f : U -> R, which maps every x C U t o t h e n a t u r a l image o f f i n

l T n e
© M x A ' M x « element f € © ( U ) M i s u n i q u e l y d e t e r m i n e d by t h i s func-
t i o n f , s i n c e t h e c o r r e s p o n d i n g f a c t i s known t o be t r u e f o r a l l t h e
r e s t r i c t i o n s r^J (f) € ( 9 ( u n M ) o f f . We i d e n t i f y f w i t h f . Thus
p M M a
we r e g a r d as a subsheaf o f t h e sheaf o f a l l R-valued f u n c t i o n s
on M. I n p a r t i c u l a r t h e r e s t r i c t i o n maps r ^ : ® ( U ) -><9 (V) a r e now

M M
the n a i v e r e s t r i c t i o n maps f »->f|v f o r f u n c t i o n s . From now on we w i l l
call t h e elements o f ^ ( U ) M the l o c a l l y semialgebraic f u n c t i o n s on U
(with r e s p e c t t o M). N o t i c e t h a t , i n t h e s p e c i a l case where U i s a
s e m i a l g e b r a i c open subset o f M, t h e s e f u n c t i o n s a r e j u s t t h e semi-
algebraic f u n c t i o n s on t h e s e m i a l g e b r a i c space (U,© IU).

M Thus i n t h i s
case t h e f CO^(\J) will a l s o be d e s i g n a t e d as t h e " s e m i a l g e b r a i c func-
t i o n s on U" .
Now l e t (N,0 ) be a second l o c a l l y s e m i a l g e b r a i c

N space o v e r R.
Definition 5. A l o c a l l y s e m i a l g e b r a i c map from (M,^) t o (N,£> ) i s N
a morphism (f : (M,e? )
M (N,# ) i n the c a t e g o r y
N of ringed spaces
over R ( c f . D e f . 2 above).
The f o l l o w i n g theorem i s known f o r s e m i a l g e b r a i c spaces [DK ,Th.7.2]

2
and e x t e n d s immediately to l o c a l l y semialgebraic spaces.
Theorem 1.2. L e t ( f , J) : (M, 6? ) -> (N,O^) be a l o c a l l y

M semialgebraic
map. F o r any open s e t s U and V o f M and N w i t h f (U) c V and any
h e e ? ( V ) we have
N
v ( h ) (x) = h ( f (x) )
for a l l x € U.
Thus (f,?) i s determined by i t s f i r s t component f and w i l l hence-
f o r t h be i d e n t i f i e d w i t h the map f from t h e s e t M t o the s e t N. C l e a r l y
a map f:M -* N i s l o c a l l y semialgebraic i f and o n l y i f f i s c o n t i n u o u s
( c f . Def. 2 above), and i f f o r every U € 'f(N) and e v e r y h € # ( U ) the N
_i
function h*f i s l o c a l l y s e m i a l g e b r a i c on f (U). N o t i c e t h a t , i n
case M and N a r e s e m i a l g e b r a i c spaces, t h e l o c a l l y s e m i a l g e b r a i c maps
from M t o N a r e j u s t the s e m i a l g e b r a i c maps from M t o N as d e f i n e d
in [DK2,§7], In g e n e r a l t h e f o l l o w i n g P r o p o s i t i o n 1.3 g i v e s a good
h o l d on l o c a l l y s e m i a l g e b r a i c maps i n terms o f s e m i a l g e b r a i c maps.
Notice t h a t f o r every continuous map f:M -> N and every admissible
covering (N^|p€J) o f N by open s e m i a l g e b r a i c sets there certainly
exists a covering (M^jaCI) of M w i t h t h e p r o p e r t i e s needed i n t h e
proposition.
Proposition 1.3. L e t f:M N be a (set t h e o r e t i c a l ) map. L e t ( M l a € I ) a
and (NplftCJ) be a d m i s s i b l e c o v e r i n g s o f M and N by open s e m i a l g e b r a i c

c N
subsets. Assume t h a t t h e r e i s a map y: I -» J such t h a t f {M^) p( ) a
f o r every a€I. Then f i s l o c a l l y semialgebraic i f and o n l y i f t h e

r e s t r i c t i o n f l M :M ->N , , o f f t o M i s a s e m i a l g e b r a i c mao f o r
a a y(a) a ^
every a E I.
P r o o f . The "only i f " d i r e c t i o n i s obvious. So assume t h a t f l M : M_ -»N , *
is semialgebraic f o r a l l ot€I. L e t U E t?(N) . Then f o r e v e r y aEI
1 1 1
f- (U) nM a = f" ( N y ( a ) nu) n M a = (flM,,)- ^^ nu)
i s an open s e m i a l g e b r a i c subset o f M . Hence, by axiom ( v i i i ) i n De-

1
finition 1, f ~ ( U ) E 7(M). L e t now (UÎAEA) be an a d m i s s i b l e covering
of U E 'f(N) . Then f o r every a£I
1 1
(f~ (U ) fl M IXEA) = ( (f l M ) "
x a a ( U fl N
x y ( a ) ) I AEA)
1 1
i s an a d m i s s i b l e c o v e r i n g o f f ~ (U) 0 M a = (f i M ) " (U fl Q ( q }), i . e . i t
1
possesses a finite refinement. We conclude that (f ( U ^ ) I A E A ) i s an
_i
admissible covering of f (U). Thus f i s c o n t i n u o u s . For a given
f u n c t i o n h E <^(U) a l l the r e s t r i c t i o n s h l U D N ,&€J, a r e s e m i a l g e b r a i c

M : M N a r e
f u n c t i o n s . S i n c e t h e maps f l a a -* y ( ) a s e m i a l g e b r a i c we see
1
t h a t a l l the f u n c t i o n s h * f | f ~ ( U ) flM ,a€I, a r e s e m i a l g e b r a i c . Thus
1
h-f E e? (f"" u) .
q.e.d.
Corollary 1.4. L e t (M,e? ) be a l o c a l l y s e m i a l g e b r a i c space over R
and U an open subset o f M. Then t h e l o c a l l y s e m i a l g e b r a i c maps from
U t o the s e m i a l g e b r a i c standard space (R,(?_) are j u s t t h e f u n c t i o n s
f €© (U).
M
T h i s i s e v i d e n t from P r o p o s i t i o n 1.3 and the c o r r e s p o n d i n g fact f o r
semialgebraic spaces. S i m i l a r l y , t h e l o c a l l y s e m i a l g e b r a i c maps from

n
(M,C? ) t o t h e s e m i a l g e b r a i c
M space ( R , 0 ) (n > 1) are t h e n - t u p l e s
R n
f
( <l' •••/ f ) o f l o c a l l y s e m i a l g e b r a i c
n functions f. , f on M.
Corollary 1.5, L e t (M,£> ) be a s e m i a l g e b r a i c space and (N,ct? ) a l o -
M N
cally semialgebraic space over R. L e t (NÎ3GJ) be an a d m i s s i b l e c o -
v e r i n g o f N by open s e m i a l g e b r a i c s u b s e t s . Then a map f:M -> N i s l o -
cally semialgebraic i f and o n l y i f t h e r e e x i s t s a f i n i t e subset J '

1
of J such t h a t f(M) i s c o n t a i n e d i n the union N o f the s e t s with
3EJ' and the map f from M t o t h e open s e m i a l g e b r a i c subspace N' of
N i s semialgebraic.
Again t h i s i s e v i d e n t from P r o p o s i t i o n 1.3. We w i l l o f t e n c a l l t h e
locally s e m i a l g e b r a i c maps from a s e m i a l g e b r a i c space (M,C^) t o a
locally semialgebraic space (N,C> ) the " s e m i a l g e b r a i c maps from

N
(M,C? ) t o (N,fl> )"

M N
We a r e ready f o r a d i s c u s s i o n o f axiom (viii) i n the d e f i n i t i o n o f
a g e n e r a l i z e d t o p o l o g i c a l space (Def. 1 ) . We have seen i n the proof
of P r o p o s i t i o n 1.3 t h a t t h i s axiom i s important to obtain a s l i c k
d e s c r i p t i o n of l o c a l l y s e m i a l g e b r a i c maps i n terms o f s e m i a l g e b r a i c
maps and spaces. On t h e o t h e r hand one may v e r i f y t h e f o l l o w i n q ob-
s e r v a t i o n concerning our d e f i n i t i o n . Assume we had d e f i n e d locally
semialgebraic spaces u s i n g axioms (i) - ( v i i ) o m i t t i n g ( v i i i ) , and
that (M, #(M),Cov ,® ) were a l o c a l l y

M semialgebraic space i n t h i s new
sense. Then we c o u l d o b t a i n a l o c a l l y semialgebraic space
(M, f* (M),Cov^,C^) i n the o l d sense as f o l l o w s . L e t f(M) be the s e t
of a l l U £ f(M) such t h a t every (U lcc€I) € C o v ( U ) has a f i n i t e r e f i n e -

M
1
ment. Define 'f (M) as the s e t of a l l U c M w i t h U H W € f(M) f o r every
W 6 j-(M) , and d e f i n e Cov^ as the s e t of a l l f a m i l i e s (U l<x€I) i n f (M)

a
such t h a t the union U o f t h e U a i s an element o f *f' (M), and such t h a t
for every W€ r(M) the s e t U flW can be covered by f i n i t e l y many U .
Then (M, 5* (M),Cov^) f u l f i l l s a l l the axioms (i) - ( v i i i ) . Moreover,
every sheaf ^ on the s i t e (M, f (M),Cov ) extends u n i q u e l y t o a sheaf

M
1
7' on the new s i t e (M, f ' (M) ,Cov^) . C l e a r l y (M,tf (M) ,Cov^,^) | w
= (M, Cf(M) ,Cov^, O ) |w f o r e v e r y W€ ^(M) , and these spaces are s e m i a l -
1
g e b r a i c . In p a r t i c u l a r (M,'J (M) , C o v ^ , ^ ) i s l o c a l l y s e m i a l g e b r a i c in
the o l d sense.
Thus, d e s p i t e i t s importance, the axiom (viii) should be r e g a r d e d as
an axiom which does not r e s t r i c t the g e n e r a l i t y of our concept of l o -
cally s e m i a l g e b r a i c spaces.
§2 - I n d u c t i v e l i m i t s ; some examples o f l o c a l l y s e m i a l g e b r a i c spaces.
The category of l o c a l l y s e m i a l g e b r a i c spaces over R c o n t a i n s the c a -
t e g o r y o f s e m i a l g e b r a i c spaces over R as a f u l l s u b c a t e g o r y . The new
c a t e g o r y may w e l l be r e g a r d e d as an enlargement of the category of
semialgebraic spaces e s t a b l i s h e d t o quarantee t h e e x i s t e n c e o f s u i t -
able i n d u c t i v e l i m i t s o f s e m i a l g e b r a i c spaces.
We f i r s t remark t h a t inductive limits exist i n the c a t e g o r y o f ge-
neralized topological spaces (with c o n t i n u o u s maps as morphisms) i n
complete generality. Indeed, l e t ( (M ,Cov ) I a€I) be a diagram o f

a
generalized topological spaces, i . e . aH* (M ,3^,Cov^) i s a f u n c t o r
from a s m a l l c a t e g o r y I i n t o the c a t e g o r y o f g e n e r a l i z e d topological
s p a c e s . L e t M denote the i n d u c t i v e l i m i t limM o f the s e t s M . F o r
e v e r y aGI l e t t p denote a t h e c a n o n i c a l map from t o M. We d e f i n e
^(M) as the s e t o f a l l s u b s e t s U o f M such t h a t t p ^ 1

(U) € f f o r every
aGI, and we d e f i n e Cov.. as the s e t o f a l l f a m i l i e s ( U , I X € A ) i n f(M)

M A
- -j
such t h a t ( c p (U-^) I A € A) G C o v
a a f o r e v e r y a€I. I t i s easy t o check
that the t r i p l e (M, 'f(M),Cov^) fulfills the axioms ( i ) - ( v i i i ) of
Definition 1 i n §1 and thus i s a g e n e r a l i z e d topological space. By
c o n s t r u c t i o n every map cp :M -* M i s a c o n t i n u o u s map from (M .'S ,Cov )

1
^ a a ^ a a a
to (M, f , C o v ^ ) , and i t i s easy t o see t h a t our new space
M (M^^cov^)
is the d i r e c t limit o f the g i v e n diagram o f g e n e r a l i z e d topological
spaces ( M g ^ ^ C o v ^ ) v i a the maps cp^. Next assume, t h a t every genera-
lized topological space M Q = (M , ^ , C O V ) q i s equipped w i t h a sheaf
<0^ o f R-valued f u n c t i o n s such t h a t ( (M , 0^) I aGI)

Q i s a diagram i n the
category of l o c a l l y s e m i a l g e b r a i c s p a c e s . We i n t r o d u c e on M a sheaf
0 XS o f R-valued f u n c t i o n s as f o l l o w s . L e t UG f(M) be g i v e n . Then
e? (U) i s t h e R - a l g e b r a o f a l l f u n c t i o n s f : U -* R such t h a t f o r every

M
aGI the f u n c t i o n
0 1 1
f (cp lcp " U)
a a : cp " (U)
a -R
1
i s an element o f © a (cp~ (U) ) . I t i s easy t o check t h a t (M,(Q ) i s the i n
M
ductive l i m i t of the diagram ((M , (0 ) I a € I) i n t h e c a t e g o r y o f r i n g e d

Q
spaces over R. The q u e s t i o n a r i s e s whether o r not (M,©^) i s a locally
s e m i a l g e b r a i c space. If this i s the case, then, of course, (M O^) i s

f
the i n d u c t i v e l i m i t o f the diagram i n the c a t e g o r y o f l o c a l l y semial-
gebraic spaces. We present a s p e c i a l case i n which (M,(£> ) i s l o c a l l y

M
s e m i a l g e b r a i c . T h i s case w i l l suffice f o r our purposes in this chapter
Lemma 2.1. Assume t h a t f o r the diagram ((Mg'^g) lot 6 I) o f l o c a l l y semi-
a l g e b r a i c spaces over R the f o l l o w i n g f o u r c o n d i t i o n s h o l d :
a) The map cp^ : M Q -• M i s i n j e c t i v e f o r every a G I
b) F o r every UGT^, with a € I , and f o r every 3 6 I the s e t <Pp~ <P (U) 1

a
is an element o f CT^.
c) F o r e v e r y (UÎACA) €Cov ,ct€I, a and f o r every 3GI the f a m i l y
(tpp ^cpÛÎAGA) is an element of Cov^.
d) F o r e v e r y U € T , a € I , a and every f€© (U) a the function
f*(P a
1
• (tpp l < P p ~ t p U ) 1
a is an element of ©^ ( c p ~ cp U)
&
1
a for 3€I.
Then <P (M ) € ^(M) a a f o r every a G I and (p i s an isomorphism from the
r i n g e d space ( M , © ) t o the open subspace

Q a (cp^ ( M ) a ,<& |cp
M a (M^)) of M.
Moreover ( c p (M ) I a G I) G Cov (M) . Thus

Q Q M (M,© ) i s l o c a l l y
M semialgebraic.
This i s completely trivial. Now assume t h a t our small category I i s
a p a r t i a l l y ordered set ( i . e . between any two o b j e c t s a , 3 of I t h e r e
exists a t most one morphism a-* 3 , and we w r i t e a <. 3 i f a morphism
exists). F u r t h e r assume t h a t f o r any a,3£I there exists some yGI
with a<Y and 3<T • In o t h e r words, ((M O^) | a G I) i s a

t "directed
system" o f l o c a l l y s e m i a l g e b r a i c spaces. F i n a l l y assume t h a t i n the
diagram M
( ( * ^ ) l a € I) the t r a n s i t i o n map
a a c p ^ : M^-^M^ i s an isomor-
phism o f ( M , © ) w i t h an open subspace

a o (cp (M ) , 0^ I t p ^ (M ) ) o f
Q a
(M ,0
R ) f o r every p a i r (a,3) with a<3. Then the c o n d i t i o n s a ) , b ) ,
c) , d) o f t h e p r e c e d i n g lemma a r e c l e a r l y fulfilled. Replacing the M a
by t h e i r images cp (M^) Q i n M we a r r i v e a t t h e f o l l o w i n g s p e c i a l case o f
Lemma 2.1.
Lemma 2.2. L e t M be a s e t and l e t (MglaEI) be a d i r e c t e d system o f
subsets o f M with U(M la€I) a = M. (M^cM^ i f a < 3 ) . Assume t h a t e v e r y
set M Q i s equipped w i t h t h e s t r u c t u r e o f a l o c a l l y semialgebraic space
(M ,(9 ) o v e r R, and t h a t
a a f o r 3<oc t h e space (M^,0^) i s an open sub-
space o f (M ,(D ) . L e t (M,© ) Q a M be t h e i n d u c t i v e l i m i t of the ( M , © ) a a
i n t h e c a t e g o r y o f r i n g e d spaces over R. Then (M^laEI) 6 Cov (M) and M
M i s a n
every space ( a /^ )
a open subspace o f (M,©^). Thus M i s a l o c a l l y
s e m i a l g e b r a i c space o v e r R.
Example 2.3. L e t (M,©^) be a l o c a l l y s e m i a l g e b r a i c space and l e t
( M l a € I ) be an a d m i s s i b l e c o v e r i n g o f M. Adding t o t h e f a m i l y
a (M la€I)
Q
all unions of f i n i t e l y many M q we o b t a i n a d i r e c t e d system (N^!3^J) of
open s u b s e t s o f M. The space (M,© ) M i s t h e i n d u c t i v e l i m i t o f t h e sub-

N N I n
spaces ( p'^' ^) • particular, every l o c a l l y s e m i a l g e b r a i c space i s
the inductive l i m i t of a directed system o f s e m i a l g e b r a i c spaces.
Example 2.4 ( E x i s t e n c e o f d i r e c t sums). L e t ( (M ,<9) I a € I) be a f a m i l y

Q
of locally semialgebraic spaces over R. Then t h e d i s j o i n t union
U(M la€I) = M o f t h e s e t s may be equipped i n a unique f a s h i o n w i t h
the structure of a locally semialgebraic space over R such t h a t (M^laED
i s an element o f Cov (M) and e v e r y M (M ,© )

a a i s an open subspace o f M.
This i s t h e s p e c i a l case o f Lemma 2.1 where t h e c a t e g o r y I i s d i s c r e t e ,
i . e . has no morphisms e x c e p t t h e i d e n t i t i e s . A s u b s e t U o f M i s open
i n M i f and o n l y i f U fl i s open i n f o r e v e r y a € I, and a f a m i l y
(UÎAEA) o f open s u b s e t s o f M b e l o n g s t o Cov^ i f and o n l y i f
(U, flM I A E A ) E Cov„ f o r every a E I . Clearly (M,O ) m i s t h e d i r e c t sum

of the family ( ( M ^ , ^ ) la e I) i n the category of l o c a l l y semialgebraic
spaces. Notice t h a t , i n the s p e c i a l case i n which a l l (M^,(0^) a r e semi-
algebraic, t h e space (M,© ) i s s e m i a l g e b r a i c

M i f and o n l y i fI i s finite
Example 2.5 ( E x i s t e n c e o f d i r e c t p r o d u c t s ) . L e t (M,© ) and (N,© ) be M N
locally semialgebraic spaces over R. We e q u i p the c a r t e s i a n product
MxN o f t h e s e t s M,N w i t h the s t r u c t u r e of a l o c a l l y semialgebraic space
as f o l l o w s . We f i r s t choose a d m i s s i b l e coverings (M !a€I) and

a (NÎ3€J)
o f M and N by open s e m i a l g e b r a i c subsets. By adding a l l unions
of f i n i t e l y many s e t s M Q to the family ( M l a € I ) we o b t a i n an a d m i s s i b l e

a
covering ( M l a € I ) o f M which i s a d i r e c t e d system o f open

a semialgebraic
subsets o f M.
Similarly, we pass from (NÎ3£J) t o an a d m i s s i b l e covering (NÎ3EJ)
o f N which i s a d i r e c t e d system o f s e m i a l g e b r a i c subsets o f N. On every
set M xNpQ with ( a , 3 ) € I x j we c o n s i d e r t h e d i r e c t product ( M

a
x N
3'^ p) a
of the semialgebraic spaces ( M

a ^^l M
a ) a n d
^ 3'®N' 3^
N N o v e r R
' a s d e
~
fined i n [DK , Th. 7.3].

2 We then equip t h e s e t MxN w i t h the structure
o f t h e r i n g e d space over R which i s t h e i n d u c t i v e l i m i t o f t h e d i r e c -
ted system o f s e m i a l g e b r a i c spaces ( (M^N^/ to

a p) I ( r 3) € I x j ) . A c c o r d i n g
a
to Lemma 2.2, t h i s space (MxN,© M x N ) i s locally semialgebraic. Moreover,
( M x N p | ( a , 3 ) G I x j ) i s an a d m i s s i b l e c o v e r i n g o f MxN. C l e a r l y
a this
ringed space (MxN,© M x N ) over R i s the unique one f o r which
( M x N p [ ( a , 3 ) € I x j ) i s an a d m i s s i b l e c o v e r i n g o f MxN and
a ( M
a
x N
3'° 3) a
is an open subspace c f (MxN,<Q M ) f o r every (a, 3 ) € I x j . The n a t u r a l
p r o j e c t i o n s p : MxN-*M and q : MxN -* N a r e l o c a l l y s e m i a l g e b r a i c maps.
We c l a i m t h a t f o r any two l o c a l l y s e m i a l g e b r a i c maps f : L->M and
g : L -»N from a l o c a l l y semialgebraic space (L,© ) t o (M,© ) and (N,£> )

L M N
t h e c o m p o s i t e map (f,g) : L-»MxN i s l o c a l l y semialgebraic from (L,© ) L
to ( M x
N,£> M x N ) . Indeed, t h e c o v e r i n g s (f " 1
(M ) I a £ I) , ( g ~ ( N ) I 3 € J)
a
1
p
o f L a r e a d m i s s i b l e . Thus by axioms v) and v i ) i n § 1 , D e f i n i t i o n 1 ,

(f~ 1
(M ) fl g
a
1
(Np) I (a, 3 ) € I x j ) i s a l s o an a d m i s s i b l e c o v e r i n g o f L.
Using axiom v i ) we may f i n d an a d m i s s i b l e c o v e r i n g (L^lyEK) o f L by
open s e m i a l g e b r a i c sets refining this covering. I f
L^cf" 1
(M ) fl g
a
1
(Np) then (f,g) I i s a semialgebraic map from L^ t o
the semialgebraic space M xN^ s i n c e the components f | L ^ : L ^ - + M and

Q a
g|L^ : L^-*N^ o f t h i s map a r e s e m i a l g e b r a i c . Thus (f,g) i s l o c a l l y semi-
a l g e b r a i c . A l l t h i s means t h a t (MxN,© M x N ) i s the d i r e c t product of the
spaces (M,(0 ) and (N,© ) i n t h e c a t e g o r y

M N of l o c a l l y semialgebraic spaces
over R v i a t h e p r o j e c t i o n maps p and q. In p a r t i c u l a r t h e space s t r u c -
ture (MxN,(D ) on t h e s e t MxN does n o t depend on t h e c h o i c e o f t h e

MxN
coverings ( M l a € I ) and ( N ^ I 3 € J ) .
a
Example 2.6. L e t (M O^) be an a f f i n e f locally complete semialgebraic
space over R. The l a t t e r c o n d i t i o n means t h a t every p o i n t o f M has a

*)
semialgebraic neighbourhood w h i c h i s a complete s e m i a l g e b r a i c space

n
If (M,€> ) M i s embedded i n some R then M i s a l o c a l l y closed semialge-
n n
braic subset of R [DK^, § 3 ] . We choose an embedding M c R with M
n
bounded i n R . L e t d : M ->R denote t h e f u n c t i o n on t h e c l o s u r e M o f M
g i v i n g t h e d i s t a n c e from 8M := M ^ M , i . e . d(x) = min(ll x - y l l ly £ 8M)
with II II t h e e u c l i d e a n norm on R . n
(N.B. The minimum e x i s t s , s i n c e 3M
i s complete, d i s a s e m i a l g e b r a i c f u n c t i o n on M. ) L e t k denote t h e
directed system o f complete s e m i a l g e b r a i c subsets K. o f M. F o r e v e r y
K € k l e t (K,0 ) denote t h e s e m i a l g e b r a i c
K subspace o f (M,(D ) M correspond-
i n g t o K, c f . [DK , p. 186]. 7 Then ( (K ,(D ) IK € k) V i s a d i r e c t e d system

0 > i
of semialgebraic spaces. Let ^ \ 0 0 ^ ^ ° denote t h e i n d u c t i v e limit
of t h i s system i n t h e c a t e g o r y o f r i n g e d spaces over R. N o t i c e that
t h i s space has the same u n d e r l y i n g s e t M as the o r i g i n a l space (M,(9 )

M
loc
m
but that the topology of M 1 q c i s d i f f e r e n t . We c l a i m t h a t ( I O C '^M )
) A c t u a l l y every locally complete s e m i a l g e b r a i c space M i s a f f i n e ,

c f . §7.
is locally s e m i a l g e b r a i c . In o r d e r to v e r i f y t h i s we i n t r o d u c e t h e open
semialgebraic subset
M(e) = {x € M|d(x) > e}
of M f o r every e > 0 i n R. N o t i c e t h a t every set K G k i s contained i n
some s e t M(e) and t h a t every M(e) i s c o n t a i n e d i n some K € ^ , namely
i n the c l o s u r e M(e) o f M(e) i n M. (N.B. d(x) >0 f o r e v e r y x€M.) The
directed system((M(e),©„, ) l e > 0 ) o f ooen subspaces o f M i s " e q u i -

x
M (e)
l o c
v a l e n t " t o t h e system ( (K,©__) IK G k) • Thus (M 0 ) i s a l s o the f M
K loC M
i n d u c t i v e l i m i t o f the spaces (M(e),© ^^). M We i n f e r from Lemma 2.2

M l 0 C i s
that ( 2_ 0 c /® M ) indeed locally s e m i a l g e b r a i c , and moreover, t h a t
( M ( e ) l e > 0 ) i s an a d m i s s i b l e c o v e r i n g o f t h i s space by open semialge-
b r a i c s u b s e t s . Up t o now we t a c i t l y assumed t h a t M i s n o t complete
since otherwise d i s not d e f i n e d . I f M i s complete we p u t

( M 1 0 C ( M
loc'°M > = 'V-
We o b s e r v e d i n §1 t h a t f o r any l o c a l l y semialgebraic space (N,© ) t h e

N
locally s e m i a l g e b r a i c maps from (N,© ) t o t h e s e m i a l g e b r a i c

N space
(R,(0 ) a r e j u s t
R the f u n c t i o n s f € © ( N ) . N What a r e t h e l o c a l l y semialge-
l o c
b r a i c maps from (N,© ) t o (R-,
XT ,©^ ) ? R_ has t h e a d m i s s i b l e cover-
' N loc' R loc
ing (]-c,c[|c>0) with c running through t h e p o s i t i v e elements o f R.
Here as u s u a l ] - c , c [ denotes t h e i n t e r v a l { x £ R|-c<x < c}. Thus i t
f o l l o w s from P r o p o s i t i o n 1.3 t h a t a f u n c t i o n f : N-»R i s a l o c a l l y

R l 0 C i f a n d o n l i f
s e m i a l g e b r a i c map from (N,(J>) t o (
N l o c ^ R ) Y f €O (N)
N
and, i n a d d i t i o n , f i s bounded on every open s e m i a l g e b r a i c subset o f N.
Example 2.7. L e t (M,© ) be a s e m i a l g e b r a i c M space o r more g e n e r a l l y a
locally s e m i a l g e b r a i c space, and l e t X be a subset o f M which i s a
u n i o n o f open s e m i a l g e b r a i c subsets o f M. {In the t e r m i n o l o g y t o be
developed i n §3 t h i s means t h a t X i s an open subset o f M i n the s t r o n g
topology}. We choose some f a m i l y (X^laGI) in such t h a t the
following holds:
(*) E v e r y X Q i s contained i n X and every W £ f'(M) with
Wcx i s contained i n t h e u n i o n o f f i n i t e l y many X - a
For example we may t a k e as f a m i l y ( X l a € l ) t h e f a m i l y o f a l l W€ r(M)

a
with WcX. As a consequence o f (*) t h e s e t X i s t h e u n i o n o f a l l X . Q
Let (Ypl&CJ) denote t h e d i r e c t e d system o f a l l f i n i t e unions of sets
X (p* < Y i f Y ft c Y ) . L e t (X,© ) denote t h e i n d u c t i v e l i m i t o f t h e d i r e c -

v
0» p Y A
t e d system o f s e m i a l g e b r a i c spaces ( (Y^,© IY^) I 3 G J) . By Lemma 2.2

M
the space (X,(9 ) i s locally

X semialgebraic. Clearly (YÎ3€J) i s an
a d m i s s i b l e c o v e r i n g o f X and thus also (XglaGI) i s an a d m i s s i b l e
covering of X with respect to t h i s space s t r u c t u r e .
I f we r e p l a c e ( X l a € I ) by t h e l a r g e r
a f a m i l y o f a l l W € f(M) w i t h WcX,
then i t follows e a s i l y from t h e c o n d i t i o n (*) t h a t t h e space struc-
t u r e on X o b t a i n e d from the new f a m i l y i s t h e same as the o l d one.
Thus t h e space (X,©^) does n o t depend on the c h o i c e o f t h e f a m i l y
(X l a € I ) . We c a l l (X,<0 ) t h e l o c a l l y
V semialgebraic space i n d u c e d on
X by t h e space (M,© ) . Of c o u r s e ,
M i f X e f (M) then t h i s space i s j u s t
the open subspace (X,© IX) o f

M (M,© ).M
Subexample 2.8. L e t (M,© ) be a s e m i a l g e b r a i c space and l e t f : M-> R

M
be some s e m i a l g e b r a i c f u n c t i o n on M. F o r every r > 0 i n R we introduce
the open s e m i a l g e b r a i c subset
M r := {x eM|-r < f ( x ) < r}
o f M. L e t v : R* T be some n o n - t r i v i a l v a l u a t i o n o f R c o m p a t i b l e with
the unique o r d e r i n g o f R ( a d d i t i v e n o t a t i o n ; c f . [P, § 7 ] , [L, § 5 ] , o r
[KW]). We f i x some X € T and c o n s i d e r t h e subets
X := {x € M | v ( f ( x ) ) > A} and
Y := {x € M | v ( f ( x ) ) > A}
of M. Then X i s t h e union o f t h e f a m i l y (M I r > 0, v (r) > A ) and Y i s

the union o f t h e f a m i l y (M I r > O, v (r) > A ) . I t i s n o t d i f f i c u l t t o
r
verify t h a t both t h e s e f a m i l i e s fulfill the c o n d i t i o n (*) i n 2.7.
(Notice t h a t f o r every W€ ^(M) t h e s e t f (W) i s s e m i a l g e b r a i c i n R. I f
WcX (resp. W c Y ) , then f (W) i s n e c e s s a r i l y bounded i n R, s i n c e
v(x)>A (resp. v ( x ) > A ) f o r e v e r y x € f(W) and v i s n o n - t r i v i a l . )
Example 2.9. The examples 2.6 and 2.7 can be r e g a r d e d as s p e c i a l cases
of a more g e n e r a l c o n s t r u c t i o n . L e t (M,© ) be a l o c a l l y s e m i a l g e b r a i c
M
space and l e t (X |a€I) be an a r b i t r a r y f a m i l y i n r(M) . Then t h e unions
of f i n i t e l y many X^ c o n s t i t u t e a d i r e c t e d system o f open subspaces o f
M, which f u l f i l l the c o n d i t i o n s o f Lemma 2.2. Thus t h e i n d u c t i v e limit
of t h i s system i s a locally s e m i a l g e b r a i c space ( X , © ^ ) . The space
(X,© ) has t h e f o l l o w i n g d e s c r i p t i o n : The s e t X i s t h e union o f t h e

x
s e t s X . fr(x) i s t h e s e t o f a l l U € r(M)
a which can be c o v e r e d by f i n i t e -
ly many X , and r(X) i s t h e s e t o f a l l s u b s e t s U o f X such

q that
U n X E j-(M) Q f o r every a € I . A f a m i l y (U IA€A) i n T(X)

A w i t h u n i o n U be-
l o n g s t o Cov^ i f and o n l y i f U n X^ i s covered by f i n i t e l y many s e t s

U n X f o r e v e r I n
A cc Y aGI. particular (X^laEI) i s an a d m i s s i b l e c o v e r -
ing o f X. F i n a l l y © v |X = O |x f o r every a € I .
Example 2.7 i s t h e s p e c i a l c a s e that
fc(X) = {U € r(M) | U c X ) .
In g e n e r a l t h e s e t H X ) may be s m a l l e r than t h e s e t on t h e r i g h t hand
s i d e . F o r i n s t a n c e , l e t M be a l o c a l l y complete - b u t n o t complete -
s e m i a l g e b r a i c space and (X |a€I) be t h e f a m i l y o f a l l open s e m i a l g e b r a i c
s u b s e t s U o f M w i t h U complete, i n d e x e d i n some way. Then X = M, b u t
J'(X) i s o n l y t h e s e t o f a l l t h e s e s u b s e t s U. We o b t a i n the space

M ( 0 1 O C
(X,© ) = ( x l o c / M ) c o n s i d e r e d i n 2.6. The spaces i n 2.6 and 2.7 are
p a r t i c u l a r l y n i c e s i n c e they a r e determined by the s e t X and t h e space
s t r u c t u r e on M.
Example 2.10 (Base f i e l d e x t e n s i o n ) . L e t R be a r e a l closed field
extension o f R. Then from any l o c a l l y semialgebraic space (M,© ) M over
R we o b t a i n a l o c a l l y semialgebraic space (M(R),(9 ) over

M R i n the f o l -
lowing way. We s t a r t by choosing an a d m i s s i b l e covering (M la£I) o f M

a
M c : M
by open s e m i a l g e b r a i c s u b s e t s which i s a d i r e c t e d system ( a 3 ^
( / 0 )
a £ I
a<3, c f . Example 2.3). Then we have t h e d i r e c t e d system ( M
a a l )
of s e m i a l g e b r a i c spaces over R w i t h (D = IM . By base e x t e n s i o n from

DK
R to R ([D, § 9 ] , [ 3 > § 4 ] ) , we o b t a i n a d i r e c t e d system
((M (R),& ) Ia€I) o f s e m i a l g e b r a i c

a spaces which f u l f i l l s the conditions
of Lemma 2.1. We then d e f i n e (M(R),© ) as t h e i n d u c t i v e l i m i t o f t h i s

M
system. Having c o n s t r u c t e d t h i s space we e a s i l y see t h a t (M(R),^^) i s
the i n d u c t i v e l i m i t o f the d i r e c t e d system o f b a s e e x t e n s i o n s of a l l
open s e m i a l g e b r a i c subspaces o f M. I n p a r t i c u l a r (M(R) ,<9) does n o t

M
depend on t h e c h o i c e o f t h e c o v e r i n g (M la€I).
Q We c a l l t h i s space "the
space over K obtained from (M,© ) by b a s e e x t e n s i o n " . F o r every W € 3~(M)

M
we i d e n t i f y t h e s e t W(R) w i t h i t s image i n M(R), which i s an open sub-
s e t o f M(R). C l e a r l y every family (UÎA€A) i n Cov M y i e l d s a family
(U (R)|A€A)
X i n Cov ^ M , but i n general Cov^ ^ a l s o contains families
which a r e n o t " d e f i n e d over R". E v e r y M a c a n be r e g a r d e d as a subset
of M (R) i n t h e n a t u r a l way (loc.cit.). Thus a l s o M can - and w i l l -
be regarded as a subset o f M(R). F o r e v e r y W € T(M) we have W (R) P.M = W.
S t a r t i n g with our analysis of l o c a l l y semialgebraic maps i n terms o f
s e m i a l g e b r a i c maps ( P r o p o s i t i o n 1.3) and u s i n g t h e t h e o r y o f base ex-
t e n s i o n o f s e m i a l g e b r a i c maps [ l o c . c i t . ] , i t i s e v i d e n t how t o ob-
tain a l o c a l l y s e m i a l g e b r a i c map
F
R :
< < >'°M(S)
M H )
^ ' V R ) '
over 8 from any l o c a l l y s e m i a l g e b r a i c map f : (M,© ) -> (N,© ) o v e r N R in

a c a n o n i c a l way. Thus we have a f u n c t o r "base f i e l d extension" from
the category of l o c a l l y semialgebraic spaces o v e r R t o the category
of locally semialgebraic spaces o v e r R. T h i s f u n c t o r p r e s e r v e s direct
sums and f i n i t e d i r e c t products.
All t h i s sounds v e r y s i m p l e . Nevertheless some c a u t i o n i s i n d i c a t e d .
Example 2.11. L e t M be a l o c a l l y complete s e m i a l g e b r a i c space o v e r R.
Then M(R) i s a locally complete s e m i a l g e b r a i c space o v e r R. The base
e x t e n s i o n M ^ ( R ) o f the space M
q c 1 q c d e f i n e d i n Example 2.6 i s u s u a l l y
different from M(R), . Indeed, M., (R) i s t h e i n d u c t i v e l i m i t o f t h e

loc ' loc
spaces M(e)(R) = M(R)(e) w i t h e running through t h e p o s i t i v e elements
of R (notation from 2.6). On t h e o t h e r hand M(R), i s the i n d u c t i v e

loc
l i m i t o f t h e spaces M(R) (e) w i t h e running through t h e p o s i t i v e ele-

M R b t h e
ments o f R. The space M ^ ( R ) q c can be o b t a i n e d from ( ) l o c Y
process d e s c r i b e d i n Example 2.9.
In o r d e r t o e x p l i c a t e our l a s t example we need some d e f i n i t i o n s o f a
c o m b i n a t o r i a l nature, c f . [DK^, § 2 ] . R e c a l l t h a t an open simplex i n
some v e c t o r space V over R i s a set

n n
o = { I t.e.|t.€R,t.>0, E t . = 1}
1 1 1 1 1
1=0 i=o
with a f f i n e l y independent p o i n t s e , . . . , e Q o f V, which a r e c a l l e d
t h e v e r t i c e s o f a. The c l o s u r e a o f a i s d e f i n e d as t h e convex
h u l l o f e ,...,e
0 i n V. T h i s n o t i o n o f c l o s u r e i s p u r e l y combina-
torial ( s i n c e we do not equip V w i t h a t o p o l o g y ) . But t h e v e c t o r
space V ( a ) spanned by a i s f i n i t e dimensional and thus i s an a f f i n e
semialgebraic space i n a c a n o n i c a l way. a i s t h e c l o s u r e o f t h e semi-
algebraic s e t a i n V(a).
Definition 1. A (geometric) s i m p l i c i a l complex over R i s a p a i r
(X,x(X)) c o n s i s t i n g o f a s u b s e t X o f some v e c t o r space V over R and
a f a m i l y Z (X) o f open s i m p l i c e s a i n V such t h a t t h e f o l l o w i n q two
properties hold:
i) X i s the d i s j o i n t u n i o n o f a l l a € I (X) .
ii) The i n t e r s e c t i o n o n T o f the c l o s u r e s o f any two s i m p l i c e s
a,T € E (X) i s e i t h e r empty o r a f a c e o f b o t h a and T .
T h e
c l o s u r e o f t h e complex (X,Z(X)) i s t h e p a i r (X,E(X)), w i t h Z (X)
the s e t o f a l l open f a c e s o f a l l o € I(X) and X the u n i o n o f a l l
T € I(X). T h i s i s another s i m p l i c i a l complex. Of c o u r s e , we c a l l the
complex (X,Z(X)) c l o s e d i f X = X, o r e q u i v a l e n t l y i f I(X) = I(X) .
Remarks.
1) Our n o t i o n o f a s i m p l i c i a l complex d i f f e r s s l i g h t l y from t h e
c l a s s i c a l one. ( C l a s s i c a l simplicial complexes a r e what we call
closed s i m p l i c i a l complexes.) The r e a s o n f o r t h i s i s that combinato-
rial semialgebraic t o p o l o g y should emphasize a t l e a s t as much open
s i m p l i c e s as c l o s e d s i m p l i c e s . In c l a s s i c a l c o m b i n a t o r i a l topology
an open simplex a i s u s u a l l y r e g a r d e d as a s i m p l i c i a l complex c o n s i -
sting o f i n f i n i t e l y many c l o s e d s i m p l i c e s , but i n o u r theory this
would confuse t h e s e m i a l g e b r a i c space a w i t h the l o c a l l y semialge-
b r a i c space a l o c introduced i n Example 2.6.
2) The complete s i m p l i c i a l complexes, as d e f i n e d i n [DK , § 2 ] ,

3 are
the c l o s e d and f i n i t e ( c f . Def. 3 below) s i m p l i c i a l complexes i n o u r
present terminology.
We w i l l o f t e n denote a s i m p l i c i a l complex (X,L(X)) s i m p l y by t h e
l e t t e r X.
Definition 2. A s i m p l i c i a l complex Y i s c a l l e d a subcomplex o f X i f
YeX and Z (Y) c Z ( X ) . F o r any two subcomplexes , Y 2 of X the i n t e r -
section Y 1 nY 2 i s the complex ( Y n Y , Z ( Y ) n Z ( Y ) ) . A subcomplex Y

1 2 1 2
of X i s c a l l e d c l o s e d i n X, i f Y i s the i n t e r s e c t i o n o f t h e subcom-
plexes Y and X o f X. T h i s means t h a t f o r e v e r y a € Z (Y) e v e r y open
f a c e T o f a which i s an element of Z(X) i s a l s o an element o f Z ( Y ) .
A subcomplex Y o f X i s c a l l e d open i n X i f the complex X^Y i s c l o s e d
i n X. Here X\Y i s d e f i n e d as the subcomplex Z w i t h Z(Z) = Z(X)\Z(Y),
hence Z = x^Y as a s e t .
For any s u b s e t A o f X the s m a l l e s t open subcomplex o f X c o n t a i n i n g A
is called t h e s t a r S t (A) of A i n X. I t c o n s i s t s o f a l l o € Z(X)

x with
o fl A * 0.
Definition 3. A s i m p l i c i a l complex X i s c a l l e d finite i f Z(X) i s a
finite set. X i s called locally finite, i f e v e r y a € Z(X) i s c o n t a i -
ned in a finite open subcomplex o f X, i . e . i f S t ( a ) x i s finite for
e v e r y a G Z(X) .
Clear-ly x i s f i n i t e i f and o n l y i f X" i s f i n i t e . But i t may happen
that X i s l o c a l l y f i n i t e and X i s n o t . When c o n s i d e r i n g t r i a n g u l a -
tions of l o c a l l y semialgebraic spaces one u s u a l l y has t o a v o i d this
slightly pathological situation. Thus we i n s e r t here y e t another d e -
finition f o r l a t e r purpose.
Definition 4. X i s c a l l e d strictly locally finite i f the complex X
is l o c a l l y f i n i t e . T h i s means t h a t every v e r t e x of X i s a vertex of
o n l y f i n i t e l y many a G Z(X) .
Here i s an example o f a s i m p l i c i a l complex X which i s l o c a l l y finite
but n o t s t r i c t l y locally finite:

P
X i s an i n f i n i t e c o l l e c t i o n o f a d j a c e n t c l o s e d t r i a n g l e s w i t h one
common v e r t e x p, and X = X \ { p } .
We now c o n s i d e r " s i m p l e x w i s e a f f i n e " maps between s i m p l i c i a l com-
p l e x e s . There seem t o be two n a t u r a l definitions.
Definition 5. L e t (X, Z(X)) and (Y,Z(Y)) be s i m p l i c i a l complexes i n
v e c t o r spaces V and W r e s p e c t i v e l y . A weakly s i m p l i c i a l map f from
(X,Z(X)) t o (Y,Z(Y)) i s a map f from t h e s e t X t o t h e s e t Y w i t h t h e
f o l l o w i n g two p r o p e r t i e s :
i) Far e v e r y a 6 E (X) the image f (a) i s an element T o f E(Y)
and f | a : a -> T i s t h e r e s t r i c t i o n o f an a f f i n e map f 0 from
o t o x, which o f c o u r s e i s u n i q u e l y determined by f .
ii) If T€Z(X) i s an open f a c e of a € Z(X) then f | T = f | x . a
The f i r s t p r o p e r t y can a l s o be e x p r e s s e d as f o l l o w s : I f a has the
v e r t i c e s e , ..., e
Q n then
n n n
f ( Z t.e.) = Z t . e! ( t > 0,Zt
± t = 1)
1 1
i=0 i=0 i=0
with e^ r u n n i n g through a l l t h e v e r t i c e s o f some T € X ( X ) , possibly
with repetitions.
D e f i n i t i o n 6. A s i m p l i c i a l map f from t h e complex (X,I(X)) t o (Y,£(Y))
i s a map f from t h e s e t X t o t h e s e t Y which extends t o a weakly simpli-
c i a l map f :X-*Y from (X,Z(X)) to ( Y , Z ( Y ) ) . Of c o u r s e , f i s u n i q u e l y
determined by f . We c a l l f the c l o s u r e o f t h e s i m p l i c i a l map f .
I f X i s c l o s e d then t h e weakly s i m p l i c i a l maps and t h e s i m p l i c i a l maps
from X to Y c o i n c i d e . The image o f such a map f i s a subcomplex Z of Y
which i s c l o s e d , and f : X -*Z i s a s i m p l i c i a l map i n the t r a d i t i o n a l
sense. But i n g e n e r a l not every weakly s i m p l i c i a l map i s s i m p l i c i a l .
Here i s a p i c t u r e o f a weakly s i m p l i c i a l map which i s n o t s i m p l i c i a l .
X c o n s i s t s o f two open t r i a n g l e s and a 2 which have a common face
(not b e l o n g i n g t o X ) . Y c o n s i s t s o f two open t r i a n g l e s and T 2
w i t h o u t a common f a c e , f maps affinely onto x ^ and a 2 affinely
onto x « 2
Remark. F o r most c o n s i d e r a t i o n s c o n c e r n i n g t r i a n g u l a t i o n s o f l o c a l l y
s e m i a l g e b r a i c s p a c e s t h e u s e f u l maps a r e the s i m p l i c i a l maps and not
the weakly s i m p l i c i a l maps.
We now a r e r e a d y t o p r e s e n t our l a s t example o f a l o c a l l y semialge-
b r a i c space.
Example 2 . 1 2 . L e t X be a l o c a l l y finite simplicial complex i n some
v e c t o r s p a c e V o v e r R. The s e t o f a l l f i n i t e open subcomplexes of X
can be r e g a r d e d as a d i r e c t e d system (X^|X€A) o f s u b s e t s o f X w i t h X
the u n i o n o f a l l X^ (u < X i f X ^ c X ^ ) . Every s e t X^ spans a f i n i t e
d i m e n s i o n a l v e c t o r space V, i n V. Every V, i s an a f f i n e semialgebraic

A A
space i n a c a n o n i c a l way, and, as a s e m i a l g e b r a i c s u b s e t o f , X^
a l s o becomes an a f f i n e s e m i a l g e b r a i c space i n a c a n o n i c a l way. I f
p < A then i s a ( c l o s e d ) s e m i a l g e b r a i c subspace o f ( i n t h e sense
of [DK / § 7 ] ) . Thus X^ i s a s e m i a l g e b r a i c subspace o f

2 and hence
a l s o a s e m i a l g e b r a i c subspace o f X^, s i n c e X^ c X^. Now X^ i s an open
subcomplex o f X^. Thus t h e s e t X^ i s open i n X^ i n t h e s t r o n g t o p o l o g y ,
which i m p l i e s t h a t X^ i s an open s e m i a l g e b r a i c subspace o f X^ [DK , 2
Prop. 7 . 5 ] . We e q u i p X w i t h t h e s t r u c t u r e o f t h e i n d u c t i v e l i m i t o f
the d i r e c t e d system o f s e m i a l g e b r a i c spaces (X^|X€A). By Lemma 2 . 2 X
is locally semialgebraic.
We a l s o c o n s i d e r t h e d i r e c t e d system (X^|K€K) o f a l l f i n i t e subcom-
plexes (KID A) i n X . A g a i n , e v e r y X ^ i s an a f f i n e s e m i a l g e b r a i c space
and X ^ i s a subspace o f X ^ i f « < p . S i n c e t h e d i r e c t e d system (X^|X€A)
is cofinal i n t h e new d i r e c t e d system ( X I K € K ) , we see t h a t o u r
space X i s a l s o t h e i n d u c t i v e l i m i t o f t h e system o f spaces (XÎKEK).
Clearly ( S t ( a ) | a € I(X))
x i s an a d m i s s i b l e c o v e r i n g o f X by open semi-
algebraic subsets. Using t h i s f a c t we see t h a t X i s a s e m i a l g e b r a i c
space i f and o n l y i f t h e complex X i s f i n i t e .
I f f : X -> Y i s a weakly s i m p l i c i a l map from X t o a l o c a l l y finite sim-
p l i c i a l complex Y t h e n i t i s easy t o check t h a t f i s a l o c a l l y semi-
a l g e b r a i c map between t h e spaces X and Y. More p r e c i s e l y t h e weakly
s i m p l i c i a l maps from X t o Y a r e t h o s e l o c a l l y s e m i a l g e b r a i c maps
from X to Y which map open s i m p l i c e s a f f i n e l y onto open simplices.

F i n i t e s i m p l i c i a l complexes a r e e x t r e m e l y i m p o r t a n t i n t h e t h e o r y o f
s e m i a l g e b r a i c spaces, s i n c e t h e f o l l o w i n g r e s u l t h o l d s w h i c h has been
proved : n [DK^, §2] and w i l l a l s o appear i n [BCR, Chap. 9] (cf. [Hi],
[Lo], [Ci] f o r the case R = IR ) .
Theorem 2.13. For any a f f i n e s e m i a l g e b r a i c space (M,(0^) and finitely
many s e n i a l g e b r a i c s u b s e t s A ,...,A
1 r of M t h e r e e x i s t s a f i n i t e sim-
p l i c i a l complex X t o g e t h e r w i t h subcomplexes Y.j,...,Y r and a s e m i a l -
g e b r a i c isomorphism cp : (X,© ) -> (M,0 ) w i t h cp(Y.) = A. f o r i =

v M 1,...,r
We c a l l such an isomorphism cp a s i m u l t a n e o u s t r i a n g u l a t i o n o f M,
A
1''--/^ .r In the next c h a p t e r we w i l l g e n e r a l i z e t h i s theorem to a
t r i a n g u l a t i o n theorem f o r a prominent class of l o c a l l y semialgebraic
spaces v s i n g s t r i c t l y locally finite simplicial complexes.

§3 - L o c a l l y semialgebraic subsets
From now on a l o c a l l y semialgebraic space (M,©^) w i l l o f t e n be denoted
by just the l e t t e r M.
e a
L e t M be a l o c a l l y semialgebraic space o v e r R and l e t (M^laGI) *>
fixed admissible c o v e r i n g o f M by open s e m i a l g e b r a i c subsets.
Definition 1. A s u b s e t X of M i s called locally semialgebraic i f , for
every W€ ?*(M) , t h e s e t X n W i s s e m i a l g e b r a i c i n the semialgebraic
space W. T h i s a l r e a d y h o l d s i fxnM Q i s semialgebraic in M a f o r every
a € I . The s e t o f a l l l o c a l l y semialgebraic subsets o f M i s denoted
by T(M) .
It i s evident t h a t T(M) has t h e f o l l o w i n g t h r e e properties:
1) f (M) e T(M) .
2) X G r(M) M s X G T(M)
3) The u n i o n and the i n t e r s e c t i o n o f any l o c a l l y finite family
(XÎXGA) i n T(M) ( c f . D e f i n i t i o n 4 i n §1) i s an element of T(M).
Proposition 3.1. L e t f : M -»N be a l o c a l l y s e m i a l g e b r a i c map between
locally semialgebraic s p a c e s . Then t h e preimage f ^ (Y) o f any l o c a l l y
semialgebraic subset Y of N i s a locally semialgebraic subset o f M.
P r o o f . L e t W6 r(M) be g i v e n . There e x i s t s some v e f ( N ) with f (W) <=V.
The map g : W-»V o b t a i n e d from f by r e s t r i c t i o n i s semialgebraic. Since

-1 -1
Wnf (Y) = g ( V f l Y ) and V n Y i s s e m i a l g e b r a i c we c o n c l u d e thcit

1
Wflf (Y) i s s e m i a l g e b r a i c . q.e.d.
Caution. In c o n t r a s t t o t h e s e m i a l g e b r a i c theory t h e image o f a l o -
cally semialgebraic s e t under a l o c a l l y s e m i a l g e b r a i c map i s n o t
necessarily locally semialgebraic, c f . 6.3.a below.

We now endow a g i v e n s e t X € T(M) w i t h the s t r u c t u r e o f a l o c a l l y s e -
mialgebraic space. We e n l a r g e our a d m i s s i b l e covering (M^lcxEI) o f M
to the d i r e c t e d system o f open s e m i a l g e b r a i c subsets (NÎ3€J) c o n s i -
s t i n g o f a l l f i n i t e unions o f s e t s M . N o t i c e q that (NÎ3€J) i s a g a i n
an a d m i s s i b l e c o v e r i n g o f M. F o r every 3 € J the i n t e r s e c t i o n N^ fl X
will be regarded as a s e m i a l g e b r a i c subspace o f t h e s e m i a l g e b r a i c
space N^ [ D K , § 7 ] .2 L e t (X,© ) be the i n d u c t i v e l i m i t o f the d i r e c t e d

x
system o f s e m i a l g e b r a i c spaces (N nxl3€J).

o By Lemma 2.2 (X,£? ) i s
locally s e m i a l g e b r a i c . We c a l l these spaces (X,C? ) t h e ( l o c a l l y semi-

A
algebraic) subspaces o f ( M , C ? ) . M In t h e s p e c i a l case t h a t X €tf(M)the
space (X,© ) i s j u s t the open subspace

x (X,6> IX) o f ( M , # ) as c o n s i -
M M
dered i n §1 .
C l e a r l y t h e i n c l u s i o n map j:X -» M i s a l o c a l l y semialgebraic map from
(X,© ) X t o {M O f ). U s i n g our a n a l y s i s o f l o c a l l y semialgebraic maps
i n terms o f s e m i a l g e b r a i c maps ( P r o p o s i t i o n 1.3) i t i s a t r i v i a l
matter t o prove
Proposition 3.2. L e t f : L -> M be a l o c a l l y s e m i a l g e b r a i c map from a
locally semialgebraic space L t o M. Assume t h a t f ( L ) c X . Then t h e
map g:L -> X o b t a i n e d from f by r e s t r i c t i o n o f the image space i s a
locally s e m i a l g e b r a i c map from L t o t h e subspace X o f M.
In p a r t i c u l a r , our space (X,0 ) does not depend on t h e c h o i c e o f the

V
A
admissible covering (M | a € I ) o f M.
Example 3.3. I f Y i s a subcomplex o f a l o c a l l y finite simplicial com-
p l e x X, then Y i s a subspace o f X, where both X and Y have t h e i r c a -
nonical s t r u c t u r e as l o c a l l y semialgebraic spaces (Example 2.12).
In e v e r y one o f t h e f o l l o w i n g f o u r p r o p o s i t i o n s t h e f i r s t statement
i s an immediate consequence o f the analogous statement f o r semialge-

braic s p a c e s , and t h e second statement then f o l l o w s from Proposition
3.2 i n a p u r e l y f o r m a l way.
Proposition 3.4. L e t X be a l o c a l l y semialgebraic subset o f M and l e t
Y be a s u b s e t o f X . Then Y € T ( X ) i f and o n l y i f Y € r(M) . In t h i s case
the s u b s p a c e s t r u c t u r e on Y w i t h r e s p e c t t o X i s t h e same as t h e sub-
space s t r u c t u r e w i t h r e s p e c t to M.
Proposition 3.5 (Existence of f i b r e products). L e t g> : M->S, i|> : N S be
locally semialgebraic maps o v e r R. The subset M x N o f MxN c o n s i s t i n g

g
of a l l p a i r s (x,y) i n MxN w i t h cp(x) = (y) i s a l o c a l l y semialgebraic
s u b s e t o f MxN and hence a l o c a l l y semialgebraic space over R. The
diagram
where p and q a r e r e s p e c t i v e l y the r e s t r i c t i o n s of the p r o j e c t i o n s
from MxN t o M and N , i s a p u l l back i n t h e c a t e g o r y o f l o c a l l y semi-
algebraic spaces.
Proposition 3.6. With n o t a t i o n as i n P r o p o s i t i o n 3.5, l e t X and Y be
respectively locally semialgebraic s u b s e t s of M and N . Then t h e sub-
set
Xx Yg = {(x,y) e X x Y l c p ( x ) = \\)(y)}
of MxgN i s l o c a l l y semialgebraic i n t h e space M x N . The subspace

g
s t r u c t u r e on t h i s s e t with r e s p e c t to M x N coincides with the s t r u c -

g
ture as the f i b r e p r o d u c t o f X and Y w i t h r e s p e c t t o the l o c a l l y semi-
a l g e b r a i c maps cp|X and \\>\Y.

P r o p o s i j o n 3.7. L e t f : M->N be a l o c a l l y s e m i a l g e b r a i c map. Then t h e
graph Tf) i s a l o c a l l y s e m i a l g e b r a i c s u b s e t o f M x N . The map x ~ ( x , f (x))
from M 0 r ( f ) i s an isomorphism from t h e l o c a l l y semialgebraic space
M t o th subspace T (f) o f M x N .
We now ook f o r l o c a l l y semialgebraic subsets which a r e " s m a l l " .
Definitpn 2. L e t M be a l o c a l l y semialgebraic space o v e r R. A subset
X of M s called semialgebraic, i fX is locally semialgebraic and i f
the subpace (X,iD )

v of M i s a semialgebraic space. The s e t o f a l l
semialgbraic subsets o f M i s denoted by V(M) .
Clearlyour previous s e t f (M) i s contained i n V(M) .
P r o p o s i j o n 3.8. A l o c a l l y semialgebraic subset X of M i s semialgebraic
if and n l y i f X i s c o n t a i n e d i n some W G Tin) .
Proof, f x i s contained i n a s e t WG r(M) , then by P r o p o s i t i o n 3.4 the
space (,© ) x i s a subspace o f t h e s e m i a l g e b r a i c space (W,© ) and thus

w
(XjO^) t s e l f i s s e m i a l g e b r a i c . Assume now t h a t X G f ( M ) . L e t ( M l a € I )

a
be an a m i s s i b l e c o v e r i n g o f M by open s e m i a l g e b r a i c subsets. Then
(M nxiei) i s an a d m i s s i b l e c o v e r i n g o f the subspace X o f M . S i n c e
this spce i s s e m i a l g e b r a i c i t i s already covered by f i n i t e l y many
sets M i x . Q The u n i o n W o f t h e c o r r e s p o n d i n g sets M a i s an element o f
r(M) an c o n t a i n s X. q.e.d.
Using t i s p r o p o s i t i o n one e a s i l y verifies
Corollay 3.9. y ( M ) i s t h e s m a l l e s t s e t o f s u b s e t s o f M with the
followig three p r o p e r t i e s :
i) r(M) czt
ii) X,Y e t » > x u Y e r
iii) X,Y 6)T =* X s Y G t
From t h e t h e o r y o f semialgebraic spaces and from C o r o l l a r y 1 .5 we infer
Proposition 3.10. L e t f : M -+ N be a l o c a l l y s e m i a l g e b r a i c map ai*d X
be a semialgebraic subset o f M . Then f (X) i s a s e m i a l g e b r a i c subset
of N.
From the theory of semialgebraic spaces ( c f . [DK ,7.8]) a l s o zh&

2 fol-
lowing i s evident.
Proposition 3.11. L e t X be a subspace o f a l o c a l l y semialgebraic
space M . Then f o r e v e r y U € T(X) t h e r e e x i s t s some V € JT(M) wich
V n X = U.
e b r a i c
We now i n t r o d u c e the s t r o n g t o p o l o g y on a g i v e n l o c a l l y semialg
space M over R. T h i s i s the t o p o l o g y , i n the c l a s s i c a l sense,on the
s e t M which has T(M) as a b a s i s o f open s e t s . Thus the open sets i n
the s t r o n g t o p o l o g y are the unions o f a r b i t r a r y f a m i l i e s i n f(M) (° r
in !T(M) ) .
N o t i c e t h a t the s t r o n g t o p o l o g y on the d i r e c t p r o d u c t MxN of two lo-
cally semialgebraic spaces M , N i s j u s t the u s u a l d i r e c t product of
the s t r o n g t o p o l o g i e s on M and N. A l s o the s t r o n g t o p o l o g y on a

sub-
space X o f M i s the u s u a l subspace t o p o l o g y o f the s t r o n g topology
on M, as f o l l o w s from P r o p o s i t i o n 3 . 1 1 .
P r o p o s i t i o n 3 . 1 2 . A s e t X G T(M) i s an element of T?(M) i f and o n l y i f
X i s open i n the s t r o n g topology.

P r o o f , f course, i f X € tf(M) then X i s open i n the s t r o n g topology.
Assume DW t h a t X € T(M) i s open i n t h e s t r o n g t o p o l o g y . Let (M^laGI)
be an a ^ i s s i b l e c o v e r i n g o f M by s e t s M G f(M) . Then X O M ^ e H M ^ )
a
and X fl ^ i s open i n M^ w i t h r e s p e c t t o the s t r o n a t o p o l o g y of M a
f o r eve^ a € 1 . Thus by the s e m i a l g e b r a i c theory [DK ,Prop. 7.5]

2
XflM a € -(M ) f o r every a G I . T h i s i m p l i e s t h a t X G f(M) by axiom viii)
i n § 1 , e f i n i t i o n 1. q.e.d.
Hencefoth we use the f o l l o w i n g t e r m i n o l o g y . The words open, c l o s e d ,
dense, t c . a l l r e f e r t o the s t r o n g t o p o l o g y on M . The s e t s U G CT(M)
are r e f r e d to as the "open l o c a l l y s e m i a l g e b r a i c " subsets o f M i n
accordace w i t h P r o p o s i t i o n 3.12. The s e t s UG fr(M) a r e c a l l e d the
"open J m i l a l g e b r a i c " subsets of M. A map f :M -»• N between l o -
cally s t i i a l g e b r a i c spaces w i l l be c a l l e d "continuous", i f f i s con-
tinuoustfith r e s p e c t t o the s t r o n g t o p o l o g i e s , and " s t r i c t l y conti-
nuous",if f i s a continuous map from the g e n e r a l i z e d t o p o l o g i c a l
space 0, ? ( M ) , C o v ) t o ( N , J ( N ) , C o v ) . By d e f i n i t i o n
M N every locally
s e m i a l g b r a i c map f : M -* N i s s t r i c t l y continuous. Every s t r i c t l y con-
tinuousnap i s of course continuous.
From C o o l l a r y 1.5, P r o p o s i t i o n 3.10, and t h e s e m i a l g e b r a i c theory
[DK , 7 . ]
2 w e infer
P r o p o s i j o n 3.13. A map f:M -> N from a s e m i a l g e b r a i c space M t o a
locallysemialgebraic space N over R i s semialgebraic (= l o c a l l y se-
mialgebaic, c f . § 1 ) i f and o n l y i f f i s continuous and r (f) i s a
s e m i a l g b r a i c subset o f M x N .
Caution A continuous map f :M - N between l o c a l l y semialgebraic
spaces ' i t h a l o c a l l y s e m i a l g e b r a i c graph r (f) c M x N i s not necess<
rily l o a l l y semialgebraic. Consider f o r example the i d e n t i t y map

c
g:R^ oc -* R f x x, (Example 2.6), which i s l o c a l l y semialgebra •
The maip f = g ^ i s c o n t i n u o u s , since the strong topology on R^ , ( c ^ s
the sairae a s t h e s t r o n g t o p o l o g y on R. The graph r (f) i s l o c a l l '

R X R
semial<gebraic in -L o c s i n c e r (g) i s l o c a l l y semialgebraic i n
R x R
l o c - But f i s n o t l o c a l l y s e m i a l g e b r a i c . T h i s example a l s o >hows
that P r o p o s i t i o n 3.13 i s f a l s e i f we o n l y demand t h a t r ( f ) be p c a l l y
semialgebraic.
Proposition 3,14.
a) F o r e v e r y locally semialgebraic subset X of a l o c a l l y semialge-
braic space M the i n t e r i o r X and t h e c l o s u r e X o f X ( i n M, wit* r e s -
pect t o the strong topology) are l o c a l l y semialgebraic subsets o f M.
b) I f U i s an open l o c a l l y semialgebraic subset o f t h e subspace X o f
M then t h e r e e x i s t s some V € f(M) w i t h V fl X = U.
D K
Proof. Statement (a) i s e v i d e n t from t h e s e m i a l g e b r a i c theory - 2'
7.7]. In o r d e r t o prove (b) c o n s i d e r t h e c l o s u r e A o f XMJ i n M Then
A € y(M) . Thus V := MÂ€ T(M) . S i n c e V i s open i n M, we have V 6 ^(M) .
Clearly ADX = XÛ. Thus V O X = U. q.e.d.
I f X i s s e m i a l g e b r a i c , then, of course, a l s o X i s s e m i a l g e b r a i c - But
i t may happen t h a t X i s not s e m i a l g e b r a i c , as t h e f o l l o w i n g example
shows..
Example 3 . 1 5 . We s t a r t w i t h t h e s e m i a l g e b r a i c space
M := ] 0 , 1 [ x ] 0 , 1 [ over R and c o n s i d e r the f a m i l y (X l e € [ 0 , 1 [ ) ^ f

°P e n
s e m i a l g e b r a i c s u b s e t s o f M d e f i n e d as f o l l o w s :
X £ := ]E,1[ x]0,1[ i f e >0,
X Q := U x , y ) € ]0,1[ x]0,1[|x 2
+ (y - ± ) 2
<j] .
The u n i o n o f these s e t s i s M. We a p p l y t h e p r o c e d u r e o f E x a m p l e 2
- 9 fc
t h i s family and o b t a i n on t h e s e t M a new s t r u c t u r e o f a l o c a l l y semi-
f
a l g e b r a i c space. We c a l l t h i s nev; space M . N o t i c e t h a t the s t r o n g t o -
1
p o l o g i e s of M and M* c o i n c i d e . The s e t X Q i s open s e m i a l g e b r a i c in M
by c o n s t r u c t i o n . But i t s c l o s u r e X Q i n M' i s not c o n t a i n e d i n the
u n i o n of f i n i t e l y many s e t s X . Thus X Q i s not a s e m i a l g e b r a i c subset

1
of M .
We now s t a t e a f a c t which i s v e r y u s e f u l f o r c o n s t r u c t i n g l o c a l l y
s e m i a l g e b r a i c maps. I t s t r e s s e s t h e importance o f c l o s e d locally
semialgebraic sets.
Proposition 3.16 (Gluing p r i n c i p l e f o r l o c a l l y semialgebraic maps).
Let ( X |X6A) be a f a m i l y o f c l o s e d l o c a l l y
A semialgebraic subsets of
M such t h a t e v e r y WE ?"(M) i s c o n t a i n e d i n the u n i o n of f i n i t e l y many
X . Let f
A : M _>
N be a map from M t o a second l o c a l l y semialgebraic
space N. Assume t h a t f |X i s a l o c a l l y A s e m i a l g e b r a i c map from X A to
N f o r every X E A. Then f i s l o c a l l y semialgebraic.
Proof. L e t (M laEI) be an a d m i s s i b l e c o v e r i n g o f M by open semialge-

a
braic subsets. I t s u f f i c e s t o prove t h a t f | M a i s locally semialge-
braic f o r every a E I . There e x i s t s a f i n i t e subset J o f I such t h a t
M a i s contained i n U ( X A |XEJ).f(M nx )
a A i s a s e m i a l g e b r a i c subset o f N
for every A E J . Thus N := f ( M ) i s a s e m i a l g e b r a i c subset o f N. I t

r e
suffices t o know t h a t t h e map g :M N obtained from f |M by ~
striction o f t h e range space i s s e m i a l g e b r a i c . But t h i s i s e v i d e n t

M n x i s
from t h e s e m i a l g e b r a i c t h e o r y , since 9 ' a a x semialgebraic f o r
every X €J. q.e.d.
t o
Example 3.17. A map f from a l o c a l l y f i n i t e simplicial comple* *
a locally s e m i a l g e b r a i c space N i s l o c a l l y s e m i a l g e b r a i c i f and o n l y
if f o r every aEI(X) the r e s t r i c t i o n f | a n X i s a s e m i a l g e b r a i c map
from a n X t o N.
Definition 3. A p a t h i n a locally s e m i a l g e b r a i c space M i s a s e m i a l -
g e b r a i c map from t h e u n i t intervall [0,1] i n R t o M. The path compo-
nent o f a p o i n t x € M i n M i s the s e t o f a l l y € M such that there exists
a path Y i n M w i t h y(0) = x and y d ) = y.
Proposition 3.18. Every path component M' o f M i s a c l o s e d and open
locally s e m i a l g e b r a i c subset o f M. The f a m i l y (MÎXGA) o f a l l path
components o f M i s l o c a l l y f i n i t e . In p a r t i c u l a r i t i s an a d m i s s i b l e
c o v e r i n g o f M. Thus M i s t h e d i r e c t sum ( c f . Example 2.4) o f the sub-
spaces .
Proof. F o r any W€ f\M) t h e i n t e r s e c t i o n M' flW i s a union o f path
components o f t h e s e m i a l g e b r a i c space W. Thus M' fl W i s an open and

m
closed s e m i a l g e b r a i c subset o f W, c f . [DK 2/ §11 and § 1 2 ] . T h i s i ~
1
plies that M i s an open and c l o s e d l o c a l l y semialgebraic subset of
M. Since f o r any WE f(M) and every X € A t h e i n t e r s e c t i o n WflM^ con-
s i s t s o f f u l l path components o f W, t h e r e can be o n l y f i n i t e l y many
X with q.e.d.
As i n the s e m i a l g e b r a i c t h e o r y , i t i s now c l e a r t h a t every p a t h com-

1
ponent M o f M i s connected, i . e . t h e r e does not e x i s t a p a r t i t i o n
1
of M ij.to two non empty d i s j o i n t open l o c a l l y semialgebrai-c subsets.
Thus we c e l l the path components o f M a l s o the c o n n e c t e d components
o f M. We denote the s e t o f connected components o f M by TT (JM) . q
Example J3_. 19. We c a l l a locally s e m i a l g e b r a i c space M d i s c r e t e , i f
{x} CÛ) f o r every x €M (hence {x} € f(M) ) , and i f ({x}|x €M) i s an ad-
m i s s i b l y c o v e r i n g o f M. T h i s means t h a t M i s the d i r e c t sum o f t h e one
p o i n t spaces {x}, x €M. Clearly a locally semialgebraic space M i s d i s -
c r e t e i - and o n l y i f a l l i t s connected components a r e one p o i n t sets.
Every s*t can be equipped with the s t r u c t u r e of a d i s c r e t e locally semi-
algebra.c space over R.
Paths c<n be u s e f u l f o r t e s t i n g p r o p e r t i e s o f spaces and maps. We give
an exam})i^. Other examples w i l l come up l a t e r (Th. 6.7, Th. 6.9, II,§9,
. . .) .
P r o p o s i t i o n 3.20 (Path c r i t e r i o n f o r c o n t i n u i t y ) . L e t f : M -»N be a map
between l o c a l l y s e m i a l g e b r a i c spaces and l e t x be a p o i n t o f M. L e t U
be a s e m i a l g e b r a i c open neighbourhood o f x i n M. Assume t h a t f|U :U ->N
has a s e m i a l g e b r a i c graph. Assume f u r t h e r t h a t , f o r every path
T : [0,1. - * u i t h W y(0) = x, r ( t ) *x f o r t >0, t h e map f«Y : [0,1] ->N i s
continuous a t t = 0. Then f i s c o n t i n u o u s a t x.
P r o o f . Ve may assume U = M s i n c e c o n t i n u i t y i s a l o c a l p r o p e r t y . Suppose
t h a t f ;. not c o n t i n u o u s
s a t x. Then t h e r e e x i s t s a s e m i a l g e b r a i c n e i g h -
bourhooa W o f f ( x ) such t h a t every neighbourhood V o f x c o n t a i n s a
point z w i t h f ( z ) £W. The s e t A f ~ (N sW)

1
i s semialgebraic since f
has a s e m i a l g e b r a i c graph. The p o i n t x l i e s i n A Â. By the semialge-
b r a i c curve s e l e c t i o n lemma <[DK , § 1 2 ] , [DK^, §2]) t h e r e e x i s t s a path

2
Y : [0,1j -*M w i t h y (0) = x and y(t) € A f o r t >0. We have f'Y(O) = f ( x )

e
but f Y ' t ) €N ^W f o r t >0. Thus f-Y i s not continuous
a t t = 0, c o n t r a r y t o o u r h y p o t h e s i s . T h i s p r o v e s t h a t f i s continuous
a t x. q.e.d-
This proposition s h o u l d be r e g a r d e d as an analogue o f t h e c r i t e r i o n
that a map f :M N between m e t r i c spaces i s c o n t i n u o u s a t a p o i n t x€M
if and o n l y i f f o r e v e r y sequence of points (x ln€3N)

n i n M^{x} which
converges t o x t h e image sequence (f (x^) I n£JSI ) converges t o f (#) • " " N o
tice that i t may happen t h a t o u r base field R does n o t c o n t a i n a
sequence of positive elements c o n v e r g i n g t o z e r o . Over such a f i e l d i t
is, o f c o u r s e , i m p o s s i b l e t o t e s t c o n t i n u i t y by sequences.
Definition 4. The d i m e n s i o n dimM o f a non empty l o c a l l y semialgebraic
space M i s d e f i n e d as t h e supremum (an i n t e g e r > 0 o r ») o f dimensions
o f a l l open s e m i a l g e b r a i c s u b s e t s o f M. F o r t h e empty space 0 we define
dim 0 = -1 . F o r X € T(M) we u n d e r s t a n d by dim X t h e dimension o£ t h e
subspace X o f M. F o r e v e r y x € M t h e l o c a l d i m e n s i o n dim M x i s de^f i n e d
as t h e infimum o f t h e dimensions o f a l l W€ f(M) w i t h x € W. The space M
is called p u r e o f d i m e n s i o n n, i f dim M = n f o r e v e r y x € M.
x (Then n e c e s -
s a r i l y dimM = n < «>, see statement 3.21.f b e l o w ) .
From the s e m i a l g e b r a i c t h e o r y one e a s i l y d e r i v e s the f o l l o w i n g facts
about d i m e n s i o n s , c f . [DK , §8] f o r s t a t e m e n t s b) - f ) and [CKI-R, § 1 3

2
f o r statement g ) . Statement a) i s e a s i l y seen by c o n s i d e r i n g pciths i n M.
Proposition 3.21. L e t M be a non empty l o c a l l y semialgebraic space.
a) dimM = 0 i f and o n l y i fM i s discrete.
b) I f ( X | A € A ) i s a f a m i l y
X i n T(M) w i t h u n i o n M and t h e p r o p e r t y t h a t
every W€ f(M) i s a l r e a d y c o n t a i n e d i n t h e u n i o n o f f i n i t e l y many
X^ then
A
dimM = sup (dim X^ | A € A ) .
c) Assume t h a t X € T(M) has dimension dim X = dimM < » . Then tfte i n -

t e r i c r x o f X i s not empty and dim X = dimM .
d) dim ( X - x ) < dim X f o r every X € T(M) w i t h dim X < «> and X n o t empty.
e) I f there e x i s t s an i n j e c t i v e locally s e m i a l g e b r a i c map from M t o
a secor.d l o c a l l y s e m i a l g e b r a i c space N, then dimM < dim N .
f) dimM is t h e supremum o f a l l t h e l o c a l d i m e n s i o n s dim M,x € M. x
g) F o r every i n t e g e r n > 0 the s e t
I (M) := { x € M l d i m M>n}
n x —
i s a closed l o c a l l y s e m i a l g e b r a i c s u b s e t o f M. I f t h e s e t
I°(M) := I (M) ^ I . (M) € T(M)

n n nf i 1
i s not empty then this subspace o f M i s pure o f d i m e n s i o n n. The
family (i°(M)ln>0) i s l o c a l l y finite,

n —
Caution, i f there e x i s t s a s u r j e c t i v e locally s e m i a l g e b r a i c map from
M t o N then i t i s not always t r u e t h a t dimM > dimN , i n contrast to
the semialgebraic case.
We f i n a l l y d i s c u s s the behavior o f subspaces o f a l o c a l l y semialge-
braic space under e x t e n s i o n o f t h e base f i e l d R ( c f . Example 2 . 1 0 f o r
definitions).
Proposition 3 . 2 2 . L e t M and N be l o c a l l y s e m i a l g e b r a i c spaces over R
and l e t f M->N be a l o c a l l y
: s e m i a l g e b r a i c map. L e t R be a r e a l closed
overfield o f R.
a) If x a ( M ) , then t h e map i g : X (R) - M ( R ) , o b t a i n e d from t h e i n c l u -
s i o n map i : x ^ M by base f i e l d e x t e n s i o n i s an isomorphism o f t h e
space X(R) onto some subspace o f M(R). We i d e n t i f y henceforth X(R)
with this subspace, thus X(R) € T ( M ( R ) ) . We have X (R) fl M = X.
b) I f X € f ( M ) , then X (R) € r(M (R) ) .
c) If X ^ ( M ) , then X (R) i s t h e same as t h e i n t e r i o r 6

X(R) o f X(R) i n
M(R), and X(R) i s t h e same as t h e c l o s u r e X(S) o f X(R) i n M(R).

d) L e t
MxgN
M-
be a p u l l back i n the category of l o c a l l y s e m i a l g e b r a i c speces over
R. Then
(Mx N)» (K)

s • N(S)
^R
PR
M(S).
is a p u l l back i n the category of l o c a l l y s e m i a l g e b r a i c speces over
R. H e n c e f o r t h we i d e n t i f y (Mx N) (R) w i t h M (R) x

g g N (R) .
e) The s u b s e t s r (f) (S) and r(fg) o f M(R)xN(S) = (MxN) (R) a r e equal.
f) L e t X be a s e m i a l g e b r a i c subset o f M. Then f (X) (R) = f g ( X ( ? ) ) -
g) L e t Y be a l o c a l l y s e m i a l g e b r a i c s u b s e t o f N . Then f " ( Y ) ( f )
1 =
1
f^"" (Y (R)) .
h) The map f ^ i s i n j e c t i v e i f f the map f i s i n j e c t i v e .
i) I f (M laGI) i s t h e f a m i l y o f connected
a components o f M, then
(M (K)|a€I) i s t h e f a m i l y o f connected components o f M(R)
j) dimM = dimM(K) .
k) £ (M(H)) = E ( M ) ( K ) n f o r every n>0.
C a u t i o n . I f t h e map f i s s u r j e c t i v e , then the map f g i s not n e c e s s a r i -
l y s u r j e c t i v e . Consider e.g. the map f : ^ l o c -> ^ t XH+X. Then for every
real closed f i e l d extension R of E t h e image o f f g i s t h e v a l u a t i o n
r i n g O o f a l l elements o f R whose a b s o l u t e v a l u e s a r e n o t i n f i n i t e l y
l a r g e with r e s p e c t t o IR . T h i s example a l s o shows t h a t statement f ) i s
false i n general f o r l o c a l l y semialgebraic subsets.

Proof, y M and N a r e a f f i n e s e m i a l g e b r a i c t h e s t a t e m e n t s a) - h) a r e
e a s i l y poved by use o f T a r s k i ' s p r i n c i p l e ( c f . [DK , § 9 ] , [ DK, § 4 ] ,

3
[BCR, C h . 5 ] ) . The o t h e r s t a t e m e n t s i ) - k) can be q u i c k l y

p deduced
by use c T a r s k i ' s p r i n c i p l e and some t r i a n g u l a t i o n o f M (Thieorem 2.13).
F o r statement i ) i t i s c o n v e n i e n t t o use t h e c o m b i n a t o r i a l dlescription
o f connectedness i n s i m p l i c i a l complexes, c f . Chap. I I , §7 b>elow. { i n
[D] and t>K ] we o b t a i n e d

3 i ) by use o f homology, a t o o l n o t r e a l l y
necessai/ f o r t h i s simple statement.}
Startincf r o m the semialgebraic c a s e we o b t a i n t h e g e n e r a l statement
i n a stiiightforward w a y . We i n d i c a t e the p r o o f o f t h e f i r s t two
statemer :s leaving the r e s t to the reader. F i r s t assume t h a t X i s semi-
algebrai:. w e choose some U € r(M) w i t h X c U . A s s e r t i o n a) is; t r u e f o r
the i n c i s i o n map X ^ U , s i n c e X and U a r e s e m i a l g e b r a i c . By definition
o f t h e sâce M(R), t h i s a s s e r t i o n i s a l s o t r u e f o r t h e i n c l u s i o n U«-»M.
Thus the a s s e r t i o n h o l d s f o r X-*M. S i n c e U(R) € f(M(R)) t h e a s s e r t i o n b)
is now €/ident. To prove a s s e r t i o n a) f o r any X £ T(M) we c h o o s e an ad-
missible covering (U la€I)

a o f M by open s e m i a l g e b r a i c s e t s . Adding
all unicis o f f i n i t e l y many U a to this f a m i l y we assume t h a t (U lct€I)

a
is i n a c u t i o n a d i r e c t e d system o f open s e m i a l g e b r a i c subsets o f M.
Again b} d e f i n i t i o n o f t h e space M(R), we know t h a t (U (R)la6I)

a i s an
a d m i s s i v e c o v e r i n g o f M(R) by open s e m i a l g e b r a i c s . Now a s s e r t i o n a)
holds fee a l l t h e i n c l u s i o n maps U ^ X ^ M . A l s o (U DX|a€I) € Cov .

a x
Thus X(?) i s the d i r e c t l i m i t o f t h e spaces ( U n x ) (R) . As a s e t , X (R)

f l
i s the ^ i o n Q f t h e d i r e c t e d system o f ( U fl X) (R) . The map i g : X(R) -M(R)

Q
is i n j e c t i v e on every subset ( U fl X) (R) . Thus i g i s i n j e c t i v e on t h e

a
whole set x (R) . We r e g a r d X (R) as a subset o f M(R) . F o r any a G I we
have u g) nx(R) = ( U D X) (R) . Indeed, g i v e n any x C U (R) n X (R) , we choose
some 3 > o w i t h x G ( x n u ) ( R ) and c o n c l u d e

o from t h e s e m i a l g e b r a i c case t h a t
x i s c o r t a i n e d i n (XflU nu ) (R) = (XflU ) (R) . Thus X (R) € T(M (R)) , and i t i s

Q
also cilear that the g i v e n space s t r u c t u r e on X (R) c o i n c i d e s with the
subspacceB s t r u c t u r e i n M(R). In the same way we see t h a t X (R) n M '- X-

§4 - Regular and paracompact spaces
A loca.'iy s e m i a l g e b r a i c space M i s c a l l e d separated, i f M i s Hausdorff
in the s t r o n g t o p o l o g y , i n other words, i f t h e d i a g o n a l A of MxM i s
c l o s e d i n MxM. (N.B. A i s l o c a l l y s e m i a l g e b r a i c by P r o p . 3.7). Hence-
forth v e t a c i t l y assume t h a t a l l our l o c a l l y s e m i a l g e b r a i c spaces a r e
separated, and we w i l l call them "spaces".
Nevertheless, the f o l l o w i n g s t r o n g e r separation property appears t o be
the t r \ l y u s e f u l one.
Definition 1 . A space M i s c a l l e d r e g u l a r , i f f o r every closed l o c a l l y
semialgebraic subset A o f M and e v e r y p o i n t x i n MÂ there exist sets
U,V€ C?M) w i t h x € U , A c V and U fl V = 0. In o t h e r terms, every x €M
has a fundamental system o f c l o s e d s e m i a l g e b r a i c neighbourhoods.
Robsor's embedding theorem [R] s t a t e s t h a t a s e m i a l g e b r a i c space X i s
affine i f and o n l y i f X i s r e g u l a r . S i n c e every subspace o f a r e g u l a r
locally semialgebraic space i s a l s o r e g u l a r t h i s implies
Proposition 4.1. I f M i s r e g u l a r then every semialgebraic subspace o f
M i s a'fine.
We do not know wether the c o n v e r s e o f t h i s statement i s t r u e i n gene-
ral. We w i l l prove t h e c o n v e r s e i n a s p e c i a l case below (Prop. 4.7).
Examples 4.2. a) The spaces M,1

introduced i n Example 2.6 are r e g u l a r .
loc
More g e n e r a l l y , the space X c o n s i d e r e d i n Example 2.9 i s r e g u l a r i f
the ambient space M i s r e g u l a r . Indeed, every x € X i s contained i n
some s e t X A o f the f a m i l y ( X l a € I ) considered

0 t h e r e . Every closed
semialgebraic neighbourhood o f x i n M, which i s c o n t a i n e d in X ,is

A
a closed semialgebraic neighbourhood o f x i n X since the strong
topology o f X i s the subspace topology o f the s t r o n g t o p o l o g y of M.
Thus x h a s a fundiamental system o f c l o s e d s e m i a l g e b r a i c neighbor-
hoods i n X.
b) As a l r e a d y o b s e r v e d above, any l o c a l l y semialgebraic subspace o f a
r e g u l a r space i s r e g u l a r .
c) The d i r e c t p r o d u c t o f f i n i t e l y many r e g u l a r spaces i s

x m o f t w o r e c
r e g u l a r . The f i b r e p r o d u c t M-j N 2 *ular spaces M-j and
with respect t o arbitrary locally s e m i a l g e b r a i c maps f ^ : -»N
and f2 :M 2 -»N i s a l s o r e g u l a r , s i n c e i t i s a (closed) subspace o"
M xM .
1 2
Question 4.3. Is,, f o r M a r e g u l a r space o v e r R, the e x t e n s i o n M(R^
with r e s p e c t t o any r e a l closed overfield R ( c f . Example 2.10)
again regular?
At f i r s t glance r e g u l a r spaces might l o o k o n l y s l i g h t l y more diffi-
c u l t t o work w i t h than a f f i n e semialgebraic spaces. But t h e ease w i t h
which we t r a n s f e r e d r e s u l t s from s e m i a l g e b r a i c to l o c a l l y semialge-
b r a i c spaces i s d e c e p t i v e and d i s a p p e a r s as soon as we s t u d y deeper
t o p o l o g i c a l - i n p a r t i c u l a r cohomological - p r o p e r t i e s o f such
spaces. I t seems t o us t h a t , a t l e a s t a t p r e s e n t , regular locally
semialgebraic spaces are i n general t o t a l l y inaccessible objects.
The b e s t we can do i s t o study r e g u l a r spaces which a r e a l s o "para-
compact" 4 F o r t u n a t e l y , these spaces seem t o s u f f i c e f o r many problems
a r i s i n g from t h e t h e o r y o f a f f i n e s e m i a l g e b r a i c spaces.
D e f i n i t i o n 2. A l o c a l l y semialgebraic space M i s c a l l e d paracompact/
if there e x i s t s a l o c a l l y finite ( c f . D e f . 4 i n §1) c o v e r i n g (MâGI)
o f M by open s e m i a l g e b r a i c subsets. (Any such c o v e r i n g i s a d m i s s i b l e ,
c f . Pro?. 1.1).
Example_j.4. a) Assume t h a t i n our base f i e l d R there e x i s t s a count-
a b l e fundamental system o f neighbourhoods o f z e r o . T h i s means there
e x i s t s • sequence e >e, > c . > ... o f p o s i t i v e elements i n R w i t h

o 1 2
lime = o . Then f o r e v e r y l o c a l l y complete s e m i a l g e b r a i c space M o v e r
n-*»
R the s,ace M i n t r o d u c e d i n Example 2.6 i s paracompact. Indeed,
loc
the set;
M := {x €Mld(x) > t
o i
M := { x € M l e , >d(x) >e -} (n>1)
n n- I n-t-«
are ope, s e m i a l g e b r a i c i n M 1 q c . They form a l o c a l l y finite covering
of M. N>tice t h a t n e a r l y a l l r e a l c l o s e d f i e l d s which o c c u r i n p r a c -
t i c e ha-e a c o u n t a b l e fundamental system o f neighbourhoods o f z e r o .
For exaiple e v e r y r e a l c l o s u r e R o f a f i n i t e l y generated f i e l d i s even

n
" m i c r o b a l " , i . e . R c o n t a i n s an element J >0 w i t h l i r a * f = 0 [Du] .
n-*»
Every r>al c l o s e d f i e l d R i s the u n i o n o f a d i r e c t e d system o f m i c r o -
b i a l re«l c l o s e d subfields.
b) L e t X be a s e m i a l g e b r a i c space and ( X j a G I ) a f a m i l y o f open semi-
algebra.c s u b s e t s w i t h u n i o n X such t h a t e v e r y WG KM) with WcX meets
o n l y f i i i t e l y many s e t s X . Then t h e l o c a l l y s e m i a l g e b r a i c space i n -
troduce! on t h i s s e t X i n Example 2.7 i s paracompact. As an example,
the sma.lest v a l u a t i o n r i n g cr o f R which i s c o m p a t i b l e w i t h the o r d e -
ring of R namely,
<r={x€R| |x| < n f o r some n € JN }
i s a paacompact - and a l s o r e g u l a r - l o c a l l y s e m i a l g e b r a i c space.
c) More g e n e r a l l y the space X c o n s i d e r e d i n Example 2.9 i s paracompact
as soon as the g i v e n f a m i l y ( X |a€I) i n f(M) has the f o l l o w i n g p r o p e r t y
For any a e i the s e t o f a l l 3 €1 w i t h X a nx p *0 is finite.
d) Ever} subspace o f a paracompact space i s paracompact.
e) D i r e c t p r o d u c t s o f f i n i t e l y many paracompact spaces are paracompact.
More g e i e r a l l y , a f i b r e product M 1 x M
N 2 i s paracompact i f both M 1 and
M 2 are laracompact.
f) Any e x t e n s i o n M(R) o f a paracompact space M over R w i t h respe-t t o
a real c l o s e d c o v e r f i e l d R r>R i s paracompact.
We b e g i n a bas:ic s t u d y o f paracompact spaces. So l e t M be a pcra-ompact
space and l e t ( M ^ l a e i ) be a l o c a l l y f i n i t e c o v e r i n g o f M by open semi-
a l g e b r a i c subs<ets.
Proposition 4 . 5 . If (U^|X€A) i s an a d m i s s i b l e c o v e r i n g of some U £ 3"(M) ,
then t h e r e exi:sts a l o c a l l y finite family ( V I K£K) o f open senialge-

V
braic s u b s e t s V' o f U which c o v e r s U and r e f i n e s t h e c o v e r i n g (UÎ^GA) .
K
Proof. For every a € I we choose a f i n i t e subset A a o f A such that
(UJAEAJ covers Mrt n u . Then (M nuJa€I,A€A) i s a locally finite

A 01 CX Ot A Ot
c o v e r i n g o f U by open s e m i a l g e b r a i c s e t s which r e f i n e s the cohering
(U IX€A) .
X
P r o p o s i t i o n 4 . 6 . F o r every semialgebraic subset X o f M t h e closure X
in M i s again semialgebraic.
According t o t h i s p r o p o s i t i o n t h e r e g u l a r space c o n s i d e r e d i n Example
3.15 i s c e r t a i n l y n o t paracompact.
Proof of P r o p o s i t i o n 4 . 6 . I t s u f f i c e s to v e r i f y that the sets M are q
semialgebraic since this i m p l i e s t h a t t h e c l o s u r e o f t h e unicns of
f i n i t e l y many M q are semialgebraic, and every X i s c o n t a i n e d jn such
a s e t . L e t some a € I be g i v e n . We denote by 1 ( a ) t h e s e t o f a J l 3 € 1
with M HM- * 0 . T h i s s e t i s f i n i t e . Clearly M does n o t meet en* M
w i t h y £ I (a) . Thus i s contained i n t h e open s e m i a l g e b r a i c set
U (Mp I 3 € I ( a ) ) , which i m p l i e s t h a t M q itself i s semialgebraic. q.e.d.
We now d i s c u s s r e g u l a r paracompact s p a c e s . These a r e t h e "gooc" s p a c e s
i n our theory.
Proposcion 4 . 7 . A g a i n , l e t M be a paracompact space anid (M^laGI) a
locall f i n i t e covering o f M by open s e m i a l g e b r a i c s e t s . The f o l l o w i n g
are eqivalent:
a) M i r e g u l a r
b) Eve^ open s e m i a l g e b r a i c subset o f M i s a f f i n e .
c) Eve^ f i n i t e u n i o n of s e t s i s affine.
Proof, a) b) i s c o n t a i n e d in Proposition 4 . 1 . b) => c) i s t r i v i a l .
Assume-iow t h a t c) holds t r u e . We want t o prove t h a t M i s r e g u l a r . L e t
x be a p o i n t of M and V some open s e m i a l g e b r a i c n e i g h b o u r h o o d o f x. As
explaied i n the p r o o f of P r o p o s i t i o n 4 . 6 , there exist finite subsets
J c K o I such that
V z U M c U M c U M .
The spce U(M^ I3€K) = N i s a f f i n e , hence r e g u l a r . Thus t h e r e exists
some s m i a l g e b r a i c neighbourhood W o f x i n N w i t h W D N c V . But W i s con
tainedin U ( M l a € J ) , hence i n N, and we c o n c l u d e t h a t W c V .

a q.e.d.
Exampl_4.8. Every l o c a l l y finite simplicial complex X ( c f . Example
2.12) s a r e g u l a r paracompact s p a c e . Indeed, t h e c o v e r i n g (St (o) I

x
by a l l open s t a r s i s l o c a l l y finite and f u l f i l l s property c)
of theproposition.
If (M 2tei) i s a l o c a l l y
a finite covering of M f u l f i l l i n g property c)
of P r o o s i t i o n 4 . 7 , then f o r any r e a l closed field R^R the f a m i l y
(M (R) ei)
a a i s a locally finite covering o f M(R) w i t h the same p r o p e r -
ty. T h s we o b t a i n a partial answer t o Q u e s t i o n 4 . 3 .
Corollary 4 . 9 . F o r any r e a l c l o s e d field R => R the base e x t e n s i o n M(R)
of a r g u l a r paracompact space M o v e r R i s a l s o r e g u l a r paracompact.

Proposition 4. 110 ( P a r t i t i o n o f 1, c r u d e v e r s i o n ) . L e t M be r e g v l a r
paracompact an<3 l e t (u^ IAEA) be a l o c a l l y f i n i t e c o v e r i n g o f M J>y open
semialgebraic subsets. Then t h e r e exists a family

(cp j A G A) o f l o c a l l y
A
1
semialgebraic f u n c t i o n s cp :M -> [0,1] w i t h c ? ~ ( ] 0 , 1 ] ) = U and
x x x
E cp (x) = 1
A
AEA
for every x EM-
Proof. According t o P r o p o s i t i o n 4.6 e v e r y c l o s u r e TJ i s s e m i a l -

A
gebraic, hence affine s e m i a l g e b r a i c . We choose f o r e v e r y A a s e n i a l g e -
1
braic f u n c t i o n X : u \ - [0,1] w i t h x X A ~ ( 0 ) = u\Û, . (For example embed
A A A
n e t h e
U x into some R a s a bounded s e m i a l g e b r a i c s u b s e t and l e t ^
A
e u c l i d e a n d i s t a n c e from ^ x ^ x multiplied by some c o n s t a n t ) . By a m i l d
a p p l i c a t i o n o f P r o p o s i t i o n 3.16 ( g l u i n g o f l o c a l l y s e m i a l g e b r a i c maps)
a n i
we know t h a t tJhe f u n c t i o n :M [0,1] d e f i n e d by <K IU = x^
^ A A A
, l y N U = 0 i s l o c a l l
Â A Y s e m i a l g e b r a i c . S i n c e t h e c o v e r i n g (U. |XHA) i s
A
locally finite, the f u n c t i o n = E iK on M i s w e l l d e f i n e d and l o c a l l y
A
AEA ,
semialgebraic. This f u n c t i o n has no z e r o s . The f u n c t i o n s cp i- ^ ^x x
fulfill the r e q u i r e m e n t s o f t h e p r o p o s i t i o n . q.e.d.
Theorem 4.11 ( S h r i n k i n g o f c o v e r i n g s ) . Assume t h a t M i s r e g u l a r and

2 b r a i c
that (U IA€A)
X i s a locally f i n i t e c o v e r i n g o f M by open semialg
sets. (In p a r t i c u l a r M i s paracompact). Then t h e r e e x i s t s a covering
(V x IAEA) o f M by open s e m i a l g e b r a i c s e t s with V^cU^.
Proof. For every AEA we i n t r o d u c e the f i n i t e s e t
A (A) := {yEA |U OU, * 0}

M A
and the open s e m i a l g e b r a i c s e t
W, := U(U lyEA(A)).
A M
Then U^cW^. We further introduce the f i n i t e s e t

IX) := {p€A |U DW x * 0)
= U(A(y) ly€A(X) ) .
and th> * . (Remark: (M, | X € A ) i s s t i l l a l o c a l l y f i n i t e

c M covering
X X x
o f M). (• | i ( X ) ) w i t h := U i s a f i n i t e cohering of t h e a f f i n e
u e p
semialje.raic space M . By t h e s h r i n k i n g lemma f o r a f f i n e s>emialge-

X
b r a i c p.ces [DK , Lemma 1 . 4 ] t h e r e

3 5 e x i s t s a covering ( V ^ !yi€I (X) ) o f
M A by , p < n s e m i a l g e b r a i c subsets V X y c f W^ w i t h V ^ H M ^ W ^ . We choose
such a covering f o r every X € A . F o r e v e r y y £ A we d e f i n e the open semi-
algebrî, s e t
V : = U (V, |X€A(u) )
Xy
. r: and t h a t i n t h i s c a s e y€I(X))
(Notice h a t xeA(y) i f and o n l y i f u£A(X), ana tnaL
For ev . x e A ( y )
Br we have
U
i c w
x c W
x c M
x
hence
\y Xy y X X
and thar.foj-e
V V
X y n M
X c W
X y =
V
Thus v c u . we c l a i m t h a t M i s c o v e r e d by the s e t s V . Indeed, l e t
J y
x€M te i v e n . Then x E U ^ f o r some X € A . Now

f
\ CM x = u(v X y lyei(X)).
Thus x r € f o r some y £ I ( X ) . Since V, cU , we know t h a t U x nu i s not
Xy A y M
empty, i . e - r X€A(u)- Thusx€V . y q.e.d.
The re a d , r s h o u l d d r a w a schematic p i c t u r e i n order t o understand the
proof.
0
For e v e r y liocally semialgebraic f u n c t i o n f :M->R the s e t ( x £ M * '
ure
i s open amdl l o c a l l y semialgebraic i n M. As u s u a l we c a l l the .o;
of t h i s s<et: t h e s u p p o r t supp(f) of f . I t i s a closed l o c a l l y ^ a l -
gebraic s<ett. As a consequence o f t h e l a s t two r e s u l t s Prop. 4.1 a i ^
Theorem 4 .111 we have
Theorem 4.12 (Partition o f 1, f i n e v e r s i o n ) . L e t M be r e g u l a paracom-
pact and l e t ( U ^ I X e A ) be a l o c a l l y finite c o v e r i n g o f M by op* : e m i

~
algebraic s e t s . Then t h e r e e x i s t s a f a m i l y (cp^lXEA) o f l o c a l l sîal-
gebraic f u n c t i o n s cp^:M [ 0 , 1 ] with supp(tp^) c U ^ f o r e v e r y \ (\ ' n d
I cp (x) = 1
X€A A
for every :x € M.
As u s u a l we c a l l such a f a m i l y (cp^ | X € A ) a ( l o c a l l y semialgebri-C
t i t i o n o f u n i t y s u b o r d i n a t e t o the c o v e r i n g (U,IX€A).
• — ~ A
a n d
Theorem 4.13 ( " T i e t z e ' s e x t e n s i o n theorem") . L e t M be paracopa-'t
regular. L e t f : A -» K be a l o c a l l y semialgebraic f u n c t i o n on a ^ l ^ ^
e d
locally semialgebraic subset A of M with v a l u e s i n some g e n e r l i '
n
interval o f R ( i . e . some convex s e m i a l g e b r a i c subset K o f R) . r h * there
exists a locally semialgebraic f u n c t i o n g:M -+ K w i t h g IA = f .
P r o o f . We choose a l o c a l l y finite covering (M |a€I) o f M by o^n semi-

i a t e
algebraic subsets and then a partition of u n i t y (cp |a€I)
a suboli
i n e
to t h i s c o v e r i n g . S i n c e "Tietze's extension theorem" h o l d s f o a - f
semialgebraic spaces [DK , Th.4.5], t h e r e

5 e x i s t s f o r every a€ a semi-
algebraic f u n c t i o n g^'-M^ K which e x t e n d s f |A fl . Consequent
g := S <4> 9 i s a w e l l d e f i n e d l o c a l l y
r semialgebraic f u n c t i o n ^ M
a€I
which extends f and has v a l u e s i n K.
Uriiv.-Bibllothek
£c?er sburg
;
Example 4.14. I f A and B a r e two d i s j o i n t closed l o c a l l y semialgebraic
subspacjs o f a r e g u l a r paracompact space M then there e:xists a l o c a l l y
semialgebraic f u n c t i o n f:M -* [0,1] which has c o n s t a n t values 0 and 1
on A,B r e s p e c t i v e l y . In p a r t i c u l a r A and B can be s e p a r a t e d by two d i s -
j o i n t coen l o c a l l y semialgebraic sets.
In f a c t a stronger statement i s t r u e .
Theoreir_4. 1 5 . i f A and B a r e d i s j o i n t closed l o c a l l y semialgebraic sub-
s e t s of a r e g u l a r paracompact space M then there e x i s t s a locally semi-

-1 -1
algebraic f u n c t i o n f:M -+ [0,1] w i t h f*~ (0) = A and f (1) = B.
T h i s i s known t o be t r u e f o r M a f f i n e semialgebraic [DK. ^, Th. 1.6],
and thus f o l l o w s f o r M r e g u l a r paracompact by use o f a p a r t i t i o n of
u n i t y as i n the proof o f Theorem 4.13.
In c l a s s i c a l topology i t i s o f t e n necessary t o assume some axiom o f
" c o u n t a D i i i t y " . In our c o n t e x t i t t u r n s o u t t h a t such a c o u n t a b i l i t y
conditiDn i s a u t o m a t i c a l l y s a t i s f i e d by c o n n e c t e d paracompact spaces.
D e f i n i t i o n 3. A l o c a l l y semialgebraic space M i s c a l l e d Lindelof i f M
posesses an a d m i s s i b l e open c o v e r i n g (M I G IN)

n by c o u n t a b l y many s e t s
M n € .
Here tha word " L i n d e l o f " a l l u d e s t o t h e L i n d e l o f - p r o p e r t y f o r c l a s s i c a l
topological spaces [Ke, p. 50] . T h i s i s j u s t i f i e d by t h e f o l l o w i n g p r o -
position.
P r o p o s i c i o n 4.16. A space M i s L i n d e l o f i f and o n l y i f f o r every admis-
s i b l e ODen c o v e r i n g (U^ IX6A) o f M t h e r e e x i s t s a c o u n t a b l e refinement

(V lnexO € Cov (fM).
n M i n t h i s c a s e the V n may always be chosen to be i n
r(M) .
We l e a v e t h e eas;y p r o o f t o the r e a d e r .
Theorem 4.17. E v e r y connected paracompact space M i s L i n d e l o f .
P r o o f . We choose a locally finite covering (M !a€I) o f M by open semi-
algebraic s e t s M; A . We want t o prove t h a t I i s c o u n t a b l e . For every
index y € I we d e f i n e f i n i t e subsets
J (Y)
Q cJ 1 ( r ) <= J 2 () c . . .
Y
of I i n d u c t i v e l y as f o l l o w s : J ( Y ) := ( Y > ,
Q
J
n+1 ( t ) : =
{aeilM^nM * 0 f o r some (jeJ (y)}.
n
L e t J ( ) denote T the union o f t h e J (Y) f o r a l l n > 0, and l e t N(Y) de-
note the u n i o n o f the s e t s w i t h a e J (y) . S i n c e t h e f a m i l y (M |CX€J(Y))
is locally finite, N ( Y ) i s an open l o c a l l y s e m i a l g e b r a i c subset of M.
Moreover, f o r any a€I e i t h e r M c N ( ) or M

T n N(y) i s empty. In the
ot Ci
first case c l e a r l y N(a) = N ( ) , w h i l e Y i n t h e second case N(a) n N ( y ) = 0.
T h i s means t h a t e i t h e r J ( a ) = J ( ) o r J ( a ) D J ( ) = 0 . We choose a sub-

T T
set K o f I such t h a t I i s the d i s j o i n t u n i o n o f t h e s e t s J ( Y ) with € K .

T
Then (N(Y) I Y € K ) i s an a d m i s s i b l e c o v e r i n g o f M by p a i r w i s e disjoint
open l o c a l l y s e m i a l g e b r a i c s u b s e t s . S i n c e M i s assumed t o be connected,
we conclude t h a t K i s a one p o i n t s e t . Thus I = J ( ) Y f o r any y t l , and
I i s countable. q.e.d.
From t h i s theorem i t i s c l e a r t h a t a paracompact space M i s L i n d e l o f
if and o n l y i f M has a t most c o u n t a b l y many connected components. One
might ask whether e v e r y L i n d e l o f space i s a l r e a d y paracompact. The space
M' c o n s i d e r e d i n Example 3.15 i s a c o u n t e r e x a m p l e . We have a l r e a d y ob-
served t h a t t h i s space i s not paracompact s i n c e the c l o s u r e X Q of the

1
semialg b r a i c s u b s e t X Q of M ( c f . 3.15) i s n o t s e m i a l g e b r a i i c c . On the
other h n d / i f R contains a sequence (e (n) In € 3N) of p o s i t i v e elements
converg n g t o z e r o , then M' has an a d m i s s i b l e c o v e r i n g by t:hie s e t s X Q
a n d x
e ( ) ,ne3N. Thus i n t h i s c a s e M' i s Lindelof.
This e x m p l e i s typical, s i n c e the f o l l o w i n g h o l d s .
Proposi i o n 4. 1 8 < L e t M be a L i n d e l o f s p a c e . Assume t h a t , f o r every
U € r ( M ) h e c l o s u r e U o f U i s s e m i a l g e b r a i c . Then M i s paraiccompact.
t
Proof. e t (uîneiN) be an a d m i s s i b l e c o v e r i n g o f M by o p e n semialge-
braic s t s . Replacing every U R by the s e t U 1 U • • • UU , n we rma;y assume
that U z
11
U f o r every n € IN .
n+1
By a s s u i p t i o n / e very U n i s contained i n some U m w i t h m > n. (Choosing
a suitaj e sequence
k
< ) < k(2) < k(3) <
in 3N , , e o b tain an a d m i s s i b l e covering ( V In e IN) o f M, V

n R := U k ( n ) ,
such t h t v e ft M) a n d v cv n n + 1 f o r every n € 3N . D e f i n e open semial-
gebraic s e t s W , n € 3 N , as f o l l o w s :
n
= forn 3
W
1 =V W
2 V W
n =
W 2 i'
Then (W | n € J N ) i sa locally finite c o v e r i n g o f M. Indeed, 0V m = 0
i f n > 1+2 . q.e.d.
in the ) r o o f w e constructed a particularly n i c e open c o v e r i n g of M.
For lat, r u s e we s t a t e a consequence o f t h i s p r o o f , Theorem 4.17 and
Proposi : i o n 4 # 6 .
Corolla. y 4.19. L e t M be a c o n n e c t e d paracompact space. Then M has a
locally f i n i t e covering (M R |n € 3N) by open s e m i a l g e b r a i c subsets M n
such th xt M n nM m = 0 f o r |n-m| > 2.

N
Theorem 4..V7 a l s o t e l l s us t h a t w i t h o u t the assumption about R
c o m
Example 4.. 4 we c a n n o t f o r c e t h e space M 1 q c c o n s i d e r e d t o be p a r *~
pact.
a c e
Remark 4.20>. L e t M be an a f f i n e l o c a l l y complete s e m i a l g e b r a i c P
h e n R
over R whilc:h i s not c o m p l e t e . Assume t h a t M L Q C i s paracompact.
contains a scequence £ <| >e 2 > ... of p o s i t i v e elements w i t h l i m n = °-

n-*»
n e c t e d
Indeed, we may assume t h a t M i s c o n n e c t e d . Then M i s also c o
loc
i n
By Theoreim 4.17, M l o c i s L i n d e l o f . We choose an a d m i s s i b l e cove 9
(U In 6IN) off M by open s e m i a l g e b r a i c s u b s e t s w i t h U cU „ o r
m a
every n € IN . We f u r t h e r choose an embedding o f M i n t o some R a
i o n d
bounded s e i m i a l g e b r a i c s u b s e t , and we c o n s i d e r the d i s t a n c e f u n c
e m i n i
on M from t h e boundary M ^M, c f . Example 2.6. L e t e (n) denote t
n c e
mum of d o n t h e complete s e m i a l g e b r a i c subset U o f M. The seau
n
(e(n)ln€INl) must converge t o z e r o , s i n c e d a t t a i n s a r b i t r a r i l y s m a l l
p o s i t i v e v a l i u e s on M.
§5 - Semialgebraic maps and p r o p e r maps
The i n t e r p l a y between s e m i a l g e b r a i c and l o c a l l y semialgebraiic sets i n
our thecry becomes more t r a n s p a r e n t i f we a l s o introduce "semialgebraic
maps" as a c o u n t e r p a r t to l o c a l l y s e m i a l g e b r a i c maps. In p a r t i c u l a r this
will help us to u n d e r s t a n d "proper maps" between paracompac:t spaces.
(These naps w i l l be d e f i n e d i n t h e o b v i o u s way - s e e Def. 21 b e l o w ) .
Definition 1 . A map f : M -* N between l o c a l l y semialgebraic spaces i s
called semialgebraic (resp. a f f i n e semialgebraic) i f f i s Locally

1
s e m i a l g e b r a i c and i f t h e preimage f " (X) o f any s e m i a l g e b r a i i c (resp.
a f f i n e semialgebraic) subset X of N i s semialgebraic (resp. affine
semialgebraic) i n M.
We s h a l J see t h a t s e m i a l g e b r a i c maps a r e p l e a s a n t and easy t o be
h a n d l e d from a f o r m a l v i e w p o i n t . A f f i n e s e m i a l g e b r a i c maps a r e much
more d i f f i c u l t , i f no e x t r a h y p o t h e s e s a r e imposed on t h e s p a c e s .
Examples 5.1.
a) L e t x and Y be l o c a l l y finite s i m p l i c i a l complexes and L e t f : X - Y
be a weekly s i m p l i c i a l map. Then f i s semialgebraic i f and o n l y i f ,
f o r every open simplex T o f Y, t h e preimage f ~ ( T ) 1

is a finite subcom-
p l e x o f x . In t h i s case f i s , i n f a c t , affine semialgebraic.
b) L e tftbe a l o c a l l y complete s e m i a l g e b r a i c space, which i s not com-
p l e t e . Then t h e n a t u r a l map M 1 Q C - * M , X ~ X , i s not semialgebraic.
c) I f M i s a s e m i a l g e b r a i c space (resp. a f f i n e semialgebraic space),
then every locally semialgebraic map w i t h domain M i s s e m i a l g e b r a i c
( r e s p . e f f i n e s e m i a l g e b r a i c ) . Thus our usage o f t h e word "semialge-
braic" f o r t h e s e maps i n t h e p r e c e d i n g s e c t i o n s i s i n harmony with
D e f i n i t i o n 1.
We b e g i n o u r s h o r t s t u d y o f s e m i a l g e b r a i c maps w i t h some obvious
remarks.
Remarks 5.2.
a) F o r e v e r y X € T(M) t h e i n c l u s i o n map i : X -> M i s a f f i n e semialgebraic.
b) L e t ( N l a £ I ) be an a d m i s s i b l e c o v e r i n g o f N by open
a semialge-
n e
braic sets N . Then a map f : M-» N i s s e m i a l g e b r a i c i f and o n l y i f t
following two c o n d i t i o n s are f u l f i l l e d :
1 N
i) (f~~ ( >) l a € I) i s an a d m i s s i b l e c o v e r i n g o f M by open
a semialge-
braic sets.
ii) The r e s t r i c t i o n map
1 1
f | : f " (N ) Q : f ~ (N ) -> N
a Q
1 N
i s a s e r m i a l g e b r a i c map between t h e s e m i a l g e b r a i c spaces f ( ) a
and N Q f o r every a £I.
c) I f f : M -> IN i s s e m i a l g e b r a i c then, f o r every locally semialgebraic
map g : N f
-> N , t h e map f ' : Mx^N' ->N* o b t a i n e d from f by base extension
with respect t o g i s again semialgebraic.
d) I f f : M-*N i s s e m i a l g e b r a i c and N i s paracompact, then M i s a l s o
paracompact. Indeed, every locally finite covering (N la€I)

a o f N by
1
s e t s N € r(N)
a yields a locally finite covering (f~ (N )IcxGI) o f M by
s e t s i n f(M) .
e) I f f : M -> N i s a f f i n e s e m i a l g e b r a i c and N i s r e g u l a r paracompact then
M i s r e g u l a r paracompact (Use P r o p o s i t i o n 4.7).
Proposition 5.3. L e t f : M ->N be a s e m i a l g e b r a i c map. Then t h e image
f(X) of any l o c a l l y semialgebraic subset X o f M i s again locally semi-
algebraic.
1
Proof. L e t U € f(N) be g i v e n and l e t g : f " (U) U be t h e r e s t r i c t i o n o f
1 1
f t o f " ( U ) . Then f " (U) i s a semialgebraic space and g i s a s^mialge-
b r a i c mjp x fl f ~ (U) i s a s e m i a l g e b r a i c s u b s e t o f f
1 1
(U) , amd thus
g ( x n f " 1 (u)) = f (x) n u
i s a s e n i . i g e b r a i c subset of U. T h i s means t h a t f (X) € T(N) . q.e.d.
In the 3a ;e t h a t M i s L i n d e l o f (e.g. M connected and paraco>mpact, c f .
Th. 4.17) we can prove a s t r o n g converse of t h i s p r o p o s i t i o n .
Proposi:j m 5.4. L e t f : M -> N be locally s e m i a l g e b r a i c . Assuime, t h a t
every c:>m>onent of M i s L i n d e l o f . Assume, f u r t h e r m o r e , that
a) a l l f i j r e s o f f are s e m i a l g e b r a i c ,
b) ever/ c l o s e d ) zero d i m e n s i o n a l locally semialgebraic subset A of M
nas a . o c a l l y s e m i a l g e b r a i c image f ( A ) .
Then f Ls s e m i a l g e b r a i c . I f i n a d d i t i o n M i s r e g u l a r , then f is affine
semialgBb;aic.
P r o o f . * e a s i l y reduce t o the c a s e where N i s a f f i n e

e semialgebraic
and mus- : h e n show t h a t M i s s e m i a l g e b r a i c . I f , i n a d d i t i o n M i s r e g u -
lar, t h n M i s a f f i n e s e m i a l g e b r a i c by Robson's embedding theorem

B [R].
Let ( M i i : i ) be the f a m i l y of connected

i components o f M. Then we choose
admission coverings (M. |n€3N) o f a l l components M i by open semialge-

r, n
braic sabots. Replacing some o f t h e s e t s M ^ n by the empty s e t we may
assume tint e i t h e r M ± n i s not c o n t a i n e d i n IVL^ U ... U ^ ^ ^ or M^

± n
is empt/. e W c l a i m t h a t a l l but f i n i t e l y many s e t s M ^ ( i C I , n € 1 0

± n are
empty. r h m c l e a r l y M i s s e m i a l g e b r a i c and P r o p o s i t i o n 5.4 i s proved.
Assume -hit A := { ( i , n ) € i x K |M. * 0 } i s i n f i n i t e . For every ( i , n ) € A

i, n
we choose an element x. £ M. ^ (M. .U...UM . , ) . C o n s i d e r the s e t
i,n j.,xi f
A := {x, I (i,n) € A } ,
l, n
C l e a r l y A has a finite i n t e r s e c t i o n with every M ^ - ± n Thus A i s a
(closed) ..ocally s e m i a l g e b r a i c s u b s e t of dimension zero i n M. Of course,
t h e samo s t r u e f o r every non-empty subset A' o f A. By hypothesis b)

the image B = f (A) i s a locally s e m i a l g e b r a i c , hence s e m i a l g e l r a i c ,
1 B I f
subset o f N, and t h e same goes f o r e v e r y non-empty subset B cf *
dimB > 0, thesn B would c o n t a i n a c l o s e d s e m i a l g e b r a i c subspac< isomor-

n u s
phic t o [O, 1 ]] . B u t n o t e v e r y subset o f [0,1] i s s e m i a l g e b r a i c 1
dimB = 0„ i..e. B i s f i n i t e . T h i s i m p l i e s t h a t A c o n t a i n s an nfimte
subset C whicch i s mapped by f t o one p o i n t y. Now C i s a closed z e r o -

1
dimensional ^locally semialgebraic subset of the f i b r e f " ( y ) . Since
1
f~ (y) i s asssumed t o be s e m i a l g e b r a i c C must be f i n i t e . This cortradic-
t i o n p r o v e s ithat i n d e e d M i s s e m i a l g e b r a i c . q.ed
Notation. Lett M be a l o c a l l y semialgebraic space. F o r convenience we

a n d t n e
denote t h e s<et o f a l l c l o s e d s e m i a l g e b r a i c subsets o f M by r ( 0
set of a l l c l o s e d l o c a l l y semialgebraic subsets o f M by T(M) .
Definition 2 . A locally s e m i a l g e b r a i c map f : M -> N i s c a l l e d proper, i f

1
for every lo«cally s e m i a l g e b r a i c map g : N -> N t h e f o l l o w i n g ho-d;J: The
1 1
locally s e m i a l g e b r a i c map f : Mx^N' -> N obtained from f by base e x t e n -
1
sion with r e s p e c t t o g maps every X 6 T(Mx N') o n t o a s e t f ' ( X e ^ C N ) .
N
: o m M
A locally?' s e j m i a l g e b r a i c space M i s c a l l e d complete i f t h e map f :
to t h e o n e p o i n t space i s p r o p e r .
Here, as e l s e w h e r e ([EGA I I , § 5 . 4 ] , [DK , § 9 ] ) ,

2 i t i s easy t o v e r i f y
that the c l a s s of proper maps has t h e f o l l o w i n g f o r m a l proper:i£S.
(Recall that a l l our spaces a r e s e p a r a t e d ) .
Remarks 5.5.
i) Every c l o s e d embedding M N (i.e. locally semialgebraic Isomor-
phism from M onto a c l o s e d subspace o f N) i s p r o p e r .
ii) The c o m p o s i t i o n o f p r o p e r maps i s p r o p e r .

1
iii) Any map f ' : Mx^' -* N obtained from a p r o p e r map f : M -+ N by base
extension i s proper.
iv) I f f : M-*N,g : M - » N1 1
a r e proper maps "over" a space- S ( i n t h e u s u a l
sense) then f x g M x M ' -+ Nx N i s p r o p e r .

:
1
b b t>
Let f : M-*N and g : N -* L be l o c a l l y s e m i a l g e b r a i c maips.
v) I f g.f i s p r o p e r then f i s p r o p e r . In p a r t i c u l a r (L, = p o i n t ) every j
l o c a l l y s e m i a l g e b r a i c map whose domain i s a c o m p l e t e space i s
proper.
vi) I f gof i s p r o p e r and f i s s u r j e c t i v e then g i s p r o p e r . I n p a r t i c u -
l a r t h e image o f a complete space M under any l o c a l - l y s e m i a l g e b r a i c
map f : M - N i s a complete subspace f (M) o f N . ( N . B . . f (M) G T(N) ) .
P r o p o s i t i o n 5.6. L e t f be a l o c a l l y s e m i a l g e b r a i c map ffrom a s e m i a l g e -
b r a i c space M t o a space N. Then f i s p r o p e r i f and onlly i f t h e s e m i a l -
g e b r a i c s e t f (M) i s c l o s e d i n N and i f t h e map " f " : M-» f (M) obtained
from f by r e s t r i c t i o n o f t h e image space i s p r o p e r i n t h e c a t e g o r y o f
s e m i a l g e b r a i c spaces, as d e f i n e d i n [DK , § 9 ] . 2
P r o p o s i t i o n 5.7. L e t f : M -> N be a l o c a l l y s e m i a l g e b r a i c map and l e t
( U l a € I ) be an a d m i s s i b l e open c o v e r i n g o f N . Then f i s proper i f and

a
1
o n l y i f a l l the maps f ~ (U ) U , a € I , o b t a i n e d from £ by r e s t r i c t i o n
are proper.
We omit t h e easy p r o o f s o f these two p r o p o s i t i o n s .
Lemma 5.8. A complete locally s e m i a l g e b r a i c space M h a s o n l y finitely
many c o n n e c t e d components.
P r o o f . The f a m i l y ( M l a € I ) o f connected
a components o f M i s an a d m i s s i b l e
c o v e r i n g o f M. I n e v e r y s e t M a we choose some p o i n t x „ The s e t

a
A := { x l a € l }
a
meets e v e r y i n one p o i n t . Thus A i s c l o s e d and l o c a l l y s e m i a l g e b r a i c .
The subspace A o f M i s complete and d i s c r e t e . The same goes f o r every
s u b s e t o f A. Assume t h a t I i s i n f i n i t e . We choose a sequence (a (n) I n€lNJ)

of pairwise diffcerent i n d i c e s i n I and c o n s i d e r t h e complete and d i s -
crete subspace
B : = { x l n e i ]
a(n) *
of M. The f u n c t i . o n f : B -* R d e f i n e d by f ( x a( n ) ) = n i s locally semialge-
braic and proper:. B u t i t s image jfl i s n o t s e m i a l g e b r a i c i n R. T h i s con-
tradiction proveis t h a t I i s f i n i t e . q.e.d.
We now come t o t:he main r e s u l t o f t h i s s e c t i o n .
Theorem 5.9. Let: M be a space a l l whose connected components a r e L i n d e -
lof (e.g. M paraicompact, c f . Th. 4.17). Then every proper map f :M-*N
from M to an a r b i t r a r y space N i s a f f i n e semialgebraic.
We f i r s t prove t h e theorem i n t h e s p e c i a l case t h a t N i s a p o i n t . Then
the theorem meams t h e f o l l o w i n g .
Corollary 5.10. L e t M be a complete space a l l whose connected components
a r e L i n d e l o f (e..g. a complete paracompact s p a c e ) . Then M i s a f f i n e semi-
algebraic.
In o r d e r t o p r o v e t h a t M i s s e m i a l g e b r a i c we choose an a d m i s s i b l e cover-
ing (M ln€W)
n off M by open s e m i a l g e b r a i c s u b s e t s M n which i s p o s s i b l e
by Lemma 5.8. Suppose M i s n o t s e m i a l g e b r a i c . Then M cannot be c o v e r e d
by f i n i t e l y many M . As i n t h e p r o o f o f P r o p o s i t i o n 5.4 we assume t h a t

n
u M f o r
M n i s not c o n t a i n e d i nM^... n _ i every n> 1, and c o n s i d e r a s e t
A := {x ln€]N} n
w i t h x € M ^ (M U . . . U M ^ ) . A g a i n A i s c l o s e d and l o c a l l y
n n 1 semialge-
b r a i c i n M, and d i m A = 0. Thus A i s a d i s c r e t e complete space. This
c o n t r a d i c t s Lemma 5.8, and we see t h a t M i s i n d e e d semialgebraic.

}
I t remain: t o prove t h a t M i s a f f i n e * . By Robson's embedading theorem
[R] i t s u f i c e s t o v e r i f y t h a t M i s r e g u l a r . L e t a p o i n t > x € M and an
a f f i n e opai s e m i a l g e b r a i c neighbourhood U o f x be g i v e n . V We c l a i m that
t h e r e e x i i t s a complete neighbourhood K o f x w i t h K c U . I£f we have
proved th.s we are done, s i n c e K w i l l be c l o s e d i n M. We c choose an i s o -
morphism t : U ^ Z from U t o a bounded s e m i a l g e b r a i c subsetet Z o f some

n
R . We ap)ly the f o l l o w i n g g e n e r a l lemma, which w i l l be pr>roved below.
Lemma 5.1 . L e t X and Y be l o c a l l y semialgebraic s u b s e t s c o f some spaces
M and N, aid l e t ip : X Y be an isomorphism. Assume t h a t M is.s complete and
t h a t X i s open i n i t s c l o s u r e X . Then Y i s open i n i t s c l c o s u r e Y.
By t h i s latima Z i s open i n i t s c l o s u r e Z i n R . L e t p := cicp(x) and

n
r := d(p, z ) , the e u c l i d e a n distance o f p from 3Z = Z - . Z . . We have r > 0 .
The s e t
L := {z € Z|d(z,8Z) > |}
i s a comp.ete neighbourhood o f p i n Z . Thus K := cp (L) i i s a complete 1
neighbounood of x i n U.
I t remain; t o prove Lemma 5.11. R e p l a c i n g M and N by t h e ( c l o s u r e s X and
Y o f X ani y we assume t h a t X i s dense i n M and Y i s densse i n N. L e t
ZcXxY dmote t h e graph o f cp and Z i t s c l o s u r e i n MxN. LGet f u r t h e r p
and q demte t h e p r o j e c t i o n s o f Z onto M and N. q i s a prcoper s u r j e c -
t i v e map. We have a commutative triangle
w i t h q 12 in isomorphism and j t h e i n c l u s i o n map from Z to:> q (Y) . Now
j i s proper, s i n c e q l Z i s p r o p e r ( c f . 5.5.v. We d i d n o t uase y e t t h a t q
*) We thaxk R. Robson f o r communicating t h i s proof t o u s . The f i r s t

p r o o f , t o o u r knowledge, has been g i v e n by N. Schwartz ! [Sch, Th. 157].
1
i s p r o p e r . ) Siiince j i s a dense embedding, we conclude t h a t Z = q" (Y) .
1 1
Replacing cp b y cp' we see t h a t Z = p " (x) . Assume now t h a t X i s open i n
1
M. Then Z = p (X) i s open i n Z. S i n c e Z = q" (y) we have
N \ Y = q(Z \ Z) ,
and this set i:s c l o s e d , s i n c e q i s p r o p e r . T h i s f i n i s h e s the proof of
Lemma 5.11 and. o f C o r o l l a r y 5.10.
We now come t o . t h e p r o o f of Theorem 5.9 i n g e n e r a l . From C o r o l l a r y 5.10
we know t h a t ewery f i b r e o f f i s ( a f f i n e ) s e m i a l g e b r a i c . By P r o p o s i t i o n
5.4 this i m p l i . e s t h a t f i s s e m i a l g e b r a i c . Thus, f o r every a f f i n e semi-

1
algebraic subset X o f N the preimage f " (X) i s s e m i a l g e b r a i c . The re-
—1
1
striction f CX) -> X o f f i s p r o p e r . We want t o see t h a t f " (X) i s
a f f i n e . Thus wre have t o prove the f o l l o w i n g g e n e r a l fact.
Theorem 5.12. L e t M be a s e m i a l g e b r a i c space. Assume t h a t t h e r e exists
a p r o p e r map f:rom M t o an a f f i n e s e m i a l g e b r a i c space N. Then M i s
affine.
A proof of thi_s f a c t has been g i v e n by N. Schwartz u s i n g h i s t h e o r y of
"real closed spaces" [Sen, Cor. 162]. (These are the "abstract l o c a l l y
semialgebraic spaces" i n our t e r m i n o l o g y i n Appendix A below.) A proof
by our "geometric methods" ( c f . App. A) would be d e s i r a b l e but seems
to be d i f f i c u l t . A major p o i n t i n Schwartz's p r o o f i s t h a t he first
v e r i f i e s C o r o l l a r y 5.10 f o r M s e m i a l g e b r a i c and then a p p l i e s t h i s to
the f i b r e s o f the a b s t r a c t i o n f : M-*N of f ( c f . App. A ) , which can be
interpreted as s e m i a l g e b r a i c spaces o v e r suitable real closed fields
containing R.
Theorem 5.9 says i n p a r t i c u l a r t h a t , i f M i s not too w i l d , say paracom-
p a c t , then a l o c a l l y s e m i a l g e b r a i c map f : M-»N i s proper i f and only
i f , for a given admissible covering (N la€I)

a o f N by open semialgebraic
1 1
s e t s , the preimages f ~ (N )
a are s e m i a l g e b r a i c and the maps f " (N ) - N a a
o b t a i n e d from f by r e s t r i c t i o n a r e proper maps i n t h e c a t e g o r y o f semi-
a l g e b r a i c spaces. T h i s may be r e g a r d e d as a n e g a t i v e r e s s u l t : Essential-
l y n o t h i n g new happens i f we study proper l o c a l l y s e m i a i l g e b r a i c maps
between paracompact spaces i n s t e a d o f p r o p e r semialgebr<-aic maps between
s e m i a l g e b r a i c spaces. On t h e o t h e r hand p r o p e r n e s s i s ccommonly regarded
as a c o n d i t i o n which i s o f t e n n e c e s s a r y t o o b t a i n good r e s u l t s on "fami-
l i e s o f spaces". Should we admit t h a t l o c a l l y s e m i a l g e b : r a i c spaces a r e
o n l y a moderately useful g e n e r a l i z a t i o n o f semialgebraicc spaces?
F o r t u n a t e l y t h e r e i s a way o u t o f t h i s dilemma. I n c o n t r a s t t o the
classical space c a t e g o r i e s i t i s here p o s s i b l e t o defin<e "partially
p r o p e r maps" which a r e more g e n e r a l than proper maps b u t behave n e a r l y
as w e l l as p r o p e r maps. These maps w i l l be d i s c u s s e d i n t h e n e x t
section.
§6 - P a r t i a l l y p>roper maps
Recall that, f o r - any space M, we denote the set of a l l closed semialge-
braic s u b s e t s of: M by r (M) and t h e s e t o f a l l c l o s e d l o c a l l y semialge-
braic s u b s e t s of: M by T (M) .
Definition 1 .
i) A locally s s e m i a l g e b r a i c map f : M -*N i s c a l l e d p a r t i a l l y proper i f
for e v e r y P\ € V (M) , t h e r e s t r i c t i o n f I A : A ->N i s proper. Accord
ing to P r o p o s i t i o n 5.6 t h i s means t h a t we have f (A) £?(N) for
e v e r y A £ H ( M ) , and t h e map A -»f (A) o b t a i n e d from f by r e s t r i c t i o n
i s proper semialgebraic.
ii) A locally s e m i a l g e b r a i c space M i s c a l l e d p a r t i a l l y complete i f
the map f r o m M t o t h e one p o i n t space i s partially proper. This
means t h a t e v e r y c l o s e d s e m i a l g e b r a i c subset o f M i s a complete
semialgebraic space.
Of c o u r s e e v e r y p r o p e r map i s p a r t i a l l y p r o p e r and every complete
space i s p a r t i a l l y complete.
Examples 6.1. Here a r e examples o f p a r t i a l l y complete spaces:
i) Every d i s c r e t e space (Ex. 3.19) i s p a r t i a l l y complete.
ii) The space M ±qc c o n s i d e r e d i n Example 2.6 i s p a r t i a l l y complete
for every l o c a l l y complete s e m i a l g e b r a i c space M.
iii) Every c l o s e d l o c a l l y f i n i t e s i m p l i c i a l complex (Ex. 2.12) i s a
p a r t i a l l y complete space.
Relevant examples o f p a r t i a l l y proper maps w i l l emerge n a t u r a l l y i n
the course o f t h i s section.
*) The term "proper i n p a r t s " would r e f l e c t this situation better,

but i s c l u m s i e r .
Some formal p r o p e r t i e s o f p a r t i a l l y p r o p e r maps f o l l o w e a s i . l y from t h e
corresponding p r o p e r t i e s o f p r o p e r maps and/or d i r e c t l y fro>m t h e d e f i n i -
tions .
Remarks 6.2.
i) The composition o f two p a r t i a l l y p r o p e r maps i s p a r t i a . l l y proper.
Let f :M -*N and g : N ->L be l o c a l l y s e m i a l g e b r a i c maps.
ii) I f g.f i s p a r t i a l l y proper, then f i s p a r t i a l l y p r o p e r . In p a r t i -
c u l a r every l o c a l l y s e m i a l g e b r a i c map w i t h domain a p a . r t i a l l y com-
p l e t e space i s p a r t i a l l y proper.
iii) I f q.f i s p a r t i a l l y p r o p e r and f i s s u r j e c t i v e and s e m i i a l g e b r a i c ,
then g i s p a r t i a l l y proper.
iv) I f f :M-N i s partially p r o p e r t h e n f o r every X € T ( N ) the r e s t r i c -
t i o n f - (X) ->x o f f i s p a r t i a l l y p r o p e r . I n p a r t i c u l a r ,
1
a l l fibres
of a p a r t i a l l y p r o p e r map a r e p a r t i a l l y complete.
P a r t i a l l y proper maps w a r r a n t more c a u t i o u s t r e a t m e n t than p r o p e r maps.
Counterexamples 6.3.
a) One can c o n s t r u c t a p a r t i a l l y p r o p e r " s p i r a l map" h from IR l o c
2
into IR . ( I n f i n i t e s p i r a l with center 0 ) . h ( I R ) i s n o t even a
l o c
2
s e m i a l g e b r a i c s u b s e t o f t h e s e m i a l g e b r a i c space IR .
b) L e t M be a l o c a l l y complete s e m i a l g e b r a i c space which i s not complete
As s t a t e d above, t h e space M 1 q c c o n s t r u c t e d i n Ex. 2.6 i s p a r t i a l l y
complete. L e t f be t h e l o c a l l y s e m i a l g e b r a i c map x »-»x from M 1 q c to
M, and l e t g be the map from M t o t h e one p o i n t space. Then g»f i s
partially p r o p e r and f i s s u r j e c t i v e . But g i s n o t p a r t i a l l y
p r o p e r . Thus we see t h a t i n s t a t e m e n t 6.2. i i i we need some e x t r a
c o n d i t i o n on f i n a d d i t i o n t o s u r j e c t i v i t y . T h i s a l s o shows t h a t t h e
image o f a p a r t i a l l y complete space under a l o c a l l y s e m i a l g e b r a i c
map i s n o t n e c e s s a r i l y p a r t i a l l y complete - even i f i t i s a l o c a l l y

semialgebiraic s e t .
The question as t o whether p u l l backs o f p a r t i a l l y proper maps remain
partially prcoper i s delicate. We want t o prove t h a t t h i s i s indeed t h e
c a s e , b u t t h : i s w i l l o n l y be f e a s i b l e after some p r e p a r a t i o n .
Theorem 6.4. A locally s e m i a l g e b r a i c map f :M ->N i s p a r t i a l l y proper i f
and only i f the f o l l o w i n g c o n d i t i o n s h o l d :

:
i) f (A) Gf (N) f o r every A €?(M).
1
ii) A l l f ib)res f (y) , y €N, a r e p a r t i a l l y complete spaces.
P r o o f . C l e a r l y c o n d i t i o n s i ) and i i ) h o l d i f f i s p a r t i a l l y proper. We
now assume i ) and i i ) and want t o p r o v e t h a t f o r a g i v e n A € V (M) , t h e
restriction f l A : A ->N i s p r o p e r . By assumption f (A) G~F(N) . I t remains
to be shown t h a t t h e s e m i a l g e b r a i c map g : A ->f (A) o b t a i n e d from f by r e -
striction i s ; proper. The image o f e v e r y c l o s e d s e m i a l g e b r a i c subset o f

r , A 1 S
A i s closed i n f (A) . F o r e v e r y y Gf (A) t h e f i b r e g (y) = f (y)
complete. Thius we know from [DK , Th. 12.5] t h a t g i s indeed

2 proper.
q.e.d.
C o r o l l a r y 6.5. L e t f : M -»N be a l o c a l l y s e m i a l g e b r a i c map. Assume t h a t
every p o i n t y o f N has a l o c a l l y s e m i a l g e b r a i c neighbourhood L such

_ -I
t h a t the r e s t r i c t i o n f (L) -*L of f i s p a r t i a l l y proper. Then f- i s
partially proper.
Proof. A l l f i b r e s o f f a r e p a r t i a l l y c o m p l e t e . Thus i t s u f f i c e s to v e r i f y
t h a t , f o r a g i v e n A Gf(M), t h e image f ( A ) i s c l o s e d i n H. L e t y be a
g i v e n p o i n t i n t h e c l o s u r e f(A) . We choose a neighbourhood L G T(M) of

-1 -1
y such t h a t f (L) -»L i s p a r t i a l l y p r o p e r . Then f (A) flL = f (A Of (L) )
i s c l o s e d i n L. Thus y Gf (A) 0L and a f o r t i o r i y Gf (A) . q.e.d.
In order to prove t h a t p u l l backs o f p a r t i a l l y proper maps remain p a r -

tially proper, we have t o cope w i t h t h e d i f f i c u l t y that i n general the
c l o s u r e o f a s e m i a l g e b r a i c s e t may not be s e m i a l g e b r a i c (Example 3.15).
But t h e r e i s a s p e c i a l case where t h i s i s t r u e and t h a t will suffice
for our proof.
D e f i n i t i o n 2. An incomplete path i n a space M i s a s e m i a l g e b r a i c map
a : [0,1[->M from t h e h a l f - o p e n interval [0,1[ t o M.
Lemma 6.6. i f a : [0,1[ -> M i s an incomplete path then thee c l o s u r e A o f
A:=oc([0,1[) i n M i s e i t h e r A o r A U ( p ) w i t h some p o i n t p G M. Thus A
is semialgebraic.
P r o o f . Assume t h a t t h e r e e x i s t s some p o i n t p G A ^ A . L e t U be a s e m i a l -
1
g e b r a i c neighbourhood o f p. Then t h e s e m i a l g e b r a i c s u b s e t a (U) o f [0,1 [
c o n t a i n s some i n t e r v a l ]a,1[ w i t h 0 < a < 1 . Indeed, otherwise
U n a([0,1[) c a ( [ 0 , a ] ) f o r some a G ] 0 , 1 [ . But a ( [ 0 , a ] ) i s a c l o s e d sub-
s e t o f M which does n o t c o n t a i n p, w h i l e p i s i n t h e c l o s u r e o f
Ufl ot([0,1[). Now i f q were a p o i n t i n A ^ A d i f f e r e n t f r o m p, then we
c o u l d choose open s e m i a l g e b r a i c neighbourhoods U and V o f p,q r e s p e c -
t i v e l y w i t h U n V = 0 . But a ' (U) z> ]a,1[ and a " (V) => ]b,1 [ w i t h a,b G ] 0 , 1 [ .
1 1
Since ]a,1[ n ]b,1[ i s n o t empty, t h i s i s absurd. Thus A \ A c o n t a i n s a t
most one p o i n t p. q.e.d.
We s h a l l a l s o need t h e f o l l o w i n g f a c t about " l i f t i n g " incomplete paths.
Lemma 6.7. L e t f : M -> N be a s e m i a l g e b r a i c map between a r b i t r a r y spaces
and l e t a : [0, 1 [ -»N be an i n c o m p l e t e path w i t h a ( [ 0 , 1 [ ) c f (M) . Then
t h e r e e x i s t s some c G [0,1[ and a s e m i a l g e b r a i c map 3 : [c,1[->M w i t h
f = al[ ,1[.
c
P r o o f . We l o o k a t t h e c a r t e s i a n square
L := [0,1 [ X M N 2 -» M
[0,1[ ^ N
a
w i t h p , q denotiimg t h e n a t u r a l p r o j e c t i o n s . The lemma means t h a t t h e r e
exists some c £. L 0 , 1 [ s u c h t h a t t h e r e s t r i c t i o n p " ( [ c , 1 [) -* [c, 1 [ o f p

1
admits a s e m i a l l g j e b r a i c s e c t i o n . The map p i s s u r j e c t i v e . I t i s also
semialgebraic, hience t h e space L i s s e m i a l g e b r a i c . L e t L ^ , . . . , L r be
f i n i t e l y many (open) a f f i n e s e m i a l g e b r a i c s u b s e t s o f L which cover L.
Then [ 0 , 1 [ i s thie u n i o n o f t h e s e m i a l g e b r a i c s e t s p(L ),...,P(L )•

1 r
There exists scorme a € [ 0 , 1 [ such t h a t one o f these s e t s , say p(L,j), con-
tains t h e inte:rv7al [a,1[. We now a p p l y H a r d t ' s theorem ([DK , Th. 6.4],

3
[H] f o r R = 1 R , c f . I I , Th. 6.3 below) t o t h e r e s t r i c t i o n TTIL^P ([a,1[)
-+ [ a , 1 [ , which, i l s a s u r j e c t i v e s e m i a l g e b r a i c map between a f f i n e spaces.
By t h i s theoreim,,. [ a , 1 [ c o n t a i n s a dense open s e m i a l g e b r a i c s u b s e t U

1
such t h a t TT i s t r i v i a l o v e r U, i . e . TT"" (U) I s i s o m o r p h i c t o a p r o d u c t
UxF o v e r U . C l e a r l y TT has a s e m i a l g e b r a i c s e c t i o n o v e r U, and U c o n t a i n s
an i n t e r v a l [c,1![. q.e.d.
We now a r e r e a d y t o prove o u r main r e s u l t about partially p r o p e r maps.
Theorem 6.8 ( R e l a t i v e path completion c r i t e r i o n ) . For a l o c a l l y semi-
a l g e b r a i c map f : M-*N the following are equivalent.
i) f is partially proper.
ii) For any p a t h a : [ 0 , 1 ] -* N and any i n c o m p l e t e path 3 : [0,1[->M with
f « 3 = a I [ 0 , 1 [ t h e r e e x i s t s a (unique) p a t h 3 : [0,1]->M such t h a t 3
extends 3 (hence f-"3 = a).
P r o o f , i ) •> i i ) : L e t a : [ 0 , 1 ] N be a p a t h and 3 : [ 0 , 1 [ ->M an i n c o m p l e t e
path w i t h f « 3 = oil [ 0 , 1 [. We want t o extend 3 t o a path i n M. L e t B de-
note the c l o s u r e o f 3 ( [ 0 , 1 [ ) i n M. By Lemma 6.6 t h i s s e t i s semialge-
b r a i c . Thus f (B) €r(N) and t h e map f |B : B -» f (B) i s p r o p e r s e m i a l g e b r a i c .

We can read a as a p a t h i n f (B) and 3 as an i n c o m p l e t e paath i n B. We
now know from the semialgebraic theory [DK , C o r . 2.3] t h i a t

4 3 extends t o
a path 3 : [0,1 ] ->B.
ii) => i ) : Given some L € 7 (M) we have t o prove that the ^ r e s t r i c t i o n
fIL o f f i s proper. C o n d i t i o n i i ) h o l d s f o r f IL s i n c e i t h o l d s for f.
Thus, r e p l a c i n g M by L, we may assume from the beginning t h a t t h e space
M i s s e m i a l g e b r a i c . Now t h e map f : M-> N i s s e m i a l g e b r a i c . In o r d e r t o
prove t h a t f i s p r o p e r we have t o c o n s i d e r a c a r t e s i a n square
MxN ' 3
^ N'
N
I
g
and some A e T ( M x N*), and we have t o v e r i f y t h a t q (A) € r ( N T ) . Since q i s
semialgebraic (5.2.c) we know a l r e a d y t h a t q (A) € T ( N ' ) . ILet y be a p o i n t
i n t h e c l o s u r e o f q ( A ) . By t h e c u r v e s e l e c t i o n lemma ([DK< , Th. 12.1],

2
[DK , 4 §2]) t h e r e e x i s t s a p a t h a i n N' w i t h a ( [ 0 , 1 [ ) c q ( A ) i and a(1) = y .
By Lemma 6.7 t h e r e e x i s t s some c £ [0,1 [ and a s e m i a l g e b r a i c map
3 : [ c , 1 [ -+ A such t h a t q«3 = a l [ c , 1 [ . A p p l y i n g c o n d i t i o n i i ) t o t h e
map p o ^ : [ ,1 [ -*M and t h e map

c (g-a) I [c, 1 ] : [c,1 ] -* N we s e e t h a t p.3
extends s e m i a l g e b r a i c a l l y ( i . e . c o n t i n u o u s l y ) t o [ c , 1 ] . S i n c e a l s o q«»3
extends s e m i a l g e b r a i c a l l y t o [ c , 1 ] , t h e map 3 extends seimialgebraically
t o a map 3 : [ c , 1 ] -»Mx N" . Then q°3 = a I [c, 1 ] . In p a r t i c u l a r , q(3 (1)) =
<*(D = y . S i n c e A i s c l o s e d i n Mx N' we have 3 d ) £ A, hence y € q (A) ,
and we s e e t h a t q ( A ) i s c l o s e d i n N'. q.e.d.
In t h e s p e c i a l case that N i s the one-point space t h e theorem means
the f o l l o w i n g .
Corollary 6.9 (Absolute path completion c r i t e r i o n ) . A space M i s p a r -
tially complete i f and o n l y i f every incomplete path a : [0,1[ ->M ex-
tends t o a path ot : [0,1] - > M .

For later use we s t a t e a consequence of (the e a s i e r p a r t of) t h i s
corollary.
Corollary 6.10.. I f M i s a p a r t i a l l y complete subspace of some space N,
then t h e s e t M i s closed i n N.
P r o o f . L e t x bee a p o i n t i n t h e c l o s u r e M o f M. We choose a path
a : [0,1] -*N w i i t h a ([0,1 [) C M and a(1) = x. By C o r o l l a r y 6.9 the map
a l [ 0 , 1 [ from ![0,1[ t o M can be completed t o a path i n M. T h i s means,
t h a t x = a(1) e .M. :
q.e.d.
As another c o n s e q u e n c e o f Theorem 6.8 we o b t a i n a c h a r a c t e r i z a t i o n of
partially propter maps by a preimage c o n d i t i o n . T h i s c h a r a c t e r i z a t i o n i s
remarkable slmce no c o n d i t i o n s on the image space (e.g. " l o c a l l y com-

*)
p l e t e " ) i s neetded , i n c o n t r a s t t o the t o p o l o g i c a l case, where one
needs l o c a l c o m p a c t n e s s [Bo, Chap. I, §10, No. 3].
Corollary 6.11 . For a l o c a l l y s e m i a l g e b r a i c map f : M -> N the following
are equivalent .
a) f i s p a r t i a l l y proper.
b) The p r e i m a g e f (L) o f every p a r t i a l l y complete locally semialge-
braic subset L of N i s p a r t i a l l y complete.
c) The preimage of every complete s e m i a l g e b r a i c subset L o f N w i t h
dimL < 1 i s partially complete.
P r o o f , a) b) : The map f (L) -» L i s p a r t i a l l y proper by Corollary

1
6.10 and 6 . 2 . i v . The space L i s p a r t i a l l y complete. Thus f"" (L) i s
partially conuplete.
The i m p l i c a t i o n b) => c) i s t r i v i a l . The i m p l i c a t i o n c) => a) i s e v i d e n t
by t h e path c o m p l e t i o n c r i t e r i a 6.8 and 6.9.
*) T h i s f a c t has escaped out a t t e n t i o n i n [DKr, § 3 ] .

U s i n g the r e l a t i v e path c o m p l e t i o n criterion we a r e a l s o able to v e r i f y
t h a t p u l l backs o f p a r t i a l l y proper maps a r e a g a i n p a r t i a i l l y :>er.

prop*
M x N 3 N
Theorem 6.12. L e t N ' " '
I
g
be a c a r t e s i a n square o f l o c a l l y s e m i a l g e b r a i c maps. Assume t h a t f i s
partially proper. Then q i s a l s o p a r t i a l l y proper.
P r o o f . We v e r i f y c o n d i t i o n ( i i ) o f Theorem 6.8. Thus we aire g i v e n semi-
a l g e b r a i c maps a : [0,1 ] -N' and 3 : [0,1 [ - M x ^ ' , w i t h q<>3i = a I [0,1 [ and
we want t o extend 3 t o [ 0 , 1 ] . We i n t r o d u c e t h e maps Y := g»a from [0,1]
to N and 6 := p « 3 from [0,1[ t o M. C l e a r l y f°6 = Y l t 0 , 1 [ . S i n c e f i s
partially proper, 6 extends t o a path 6 : [0,1] -> M w i t h f - 6 = y. Now
3(t) = (6(t),a(t)), 0 < t < 1,
i s a path i n Mx N' w i t h p i [ 0 , 1 [ = 3 and q - 3 = a. q.e.d.
Proposition 6.13. Every partially proper s e m i a l g e b r a i c maip i s p r o p e r .
P r o o f . L e t f : M-*N be p a r t i a l l y proper and s e m i a l g e b r a i c . Choose an
admissible covering (N l a € l ) o f N by open s e m i a l g e b r a i c s u b s e t s . The

1
s e t s f ~ ( N ) a r e open s e m i a l g e b r a i c i n M. By Theorem 6.12! the maps
1
f~ (N ) -»N
a are p a r t i a l l y p r o p e r , which i n t h i s c a s e means they a r e
proper. Thus f i s p r o p e r . q.e.d.
I f M i s paracompact t h e n i t i s now c l e a r from Theorem 5.9) t h a t a p a r t i -
ally proper locally s e m i a l g e b r a i c map f : M-> N i s p r o p e r i f and o n l y i f
f i s semialgebraic.
We v e r i f y t h a t an i m p o r t a n t class of locally s e m i a l g e b r a i c maps con-
sists of p a r t i a l l y p r o p e r maps.
Definition 3. /A l o c a l l y s e m i a l g e b r a i c map f : M -+ N i s c a l l e d weakly l o -
cally trivial i i f every p o i n t y o f N has a s e m i a l g e b r a i c neighbourhood
L such t h a t t h c e r e e x i s t s a commutative triangle
w i t h cp a l o c a l l y semialgebraic isomorphism and p r 1 the natural projec-
tion .
Remark. T h i s m o t i o n i s t o o weak f o r many - b u t n o t a l l - purposes. We
call f : M-*N ltocally trivial i f N has an open a d m i s s i b l e covering

1
(N lct€I)
a such t h a t every f ~ (N ) i s i s o m o r p h i c t o a product N *F
A A over
N
a
Example 6.14. E v e r y weakly l o c a l l y t r i v i a l map f : M - N w i t h partially
complete f i b r e : s i s partially proper.
Indeed, i n the* t r i a n g l e (*) above t h e map p r 1 i s partially p r o p e r by

1
Theorem 6.12. Thus f | f ~ ( L ) i s partially proper f o r some neighbourhood
L o f any y € N . By C o r o l l a r y 6.5 the map f i s p a r t i a l l y proper.
The "locally s i e m i a l g e b r a i c c o v e r i n g " which we mentioned i n t h e p r e f a c e
w i l l be d e f i n e d as t h e l o c a l l y trivial maps w i t h d i s c r e t e (Ex. 3.19)
fibres. These maps a r e c e r t a i n l y partially proper, a fact which i s o f
good use t o t h e t h e o r y of coverings, c f . Chapter V.
We t u r n t o a n o t h e r example. L e t f : X-* Y be a s i m p l i c i a l map between
strictly locally finite simplicial complexes (cf. d e f i n i t i o n s a t the
end o f § 2 ) . Under which c o m b i n a t o r i a l c o n d i t i o n s i s f p a r t i a l l y proper
Let f : X->Y denote t h e c l o s u r e o f f . T h i s map i s always p a r t i a l l y p r o -

per s i n c e X i s p a r t i a l l y complete. Thus t h e s i m p l i c i a l maj.p g : f (Y) -* Y
r
o b t a i n e d from f by r e s t r i c t i o n i s also p a r t i a l l y proper ( Th. 6.12).
Looking a t t h e commutative triangle
1
X 2 f" (Y)
w i t h j the i n c l u s i o n map we see t h a t f i spartially proper i f and o n l y
if j i spartially proper (Remarks 6 . 2 . i , i i ) . Now i f j i s p a r t i a l l y
proper then, f o r every open simplex o o f X, t h e s e m i a l g e b r a i c s e t a f l X

1 1
must be c l o s e d i n f ~ (Y) . T h i s means a D X = a fl f " (Y) , a n d we see t h a t
f i spartially proper i f and o n l y i f X = f ~ ' (Y) . S i n c e X i s paracompact
the map f ±$ p r o p e r i f and o n l y i f i n a d d i t i o n f i s s e m i a . l g e b r a i c . T h i s
means t h a t t h e preimage o f every f i n i t e subcomplex o f Y uinder f i s a
finite subcomplex o f X. Thus we a r r i v e a t t h e f o l l o w i n g
Example 6.15. L e t f : X -» Y be a s i m p l i c i a l map between s t r i c t l y locally
finite complexes. Then f i s p a r t i a l l y proper i f and o n l y i f the f o l l o w -
ing h o l d s f o r e v e r y open simplex o€I(X): i f T i s an open face of o
w i t h f (T) € I ( Y ) , then T € I ( X ) . f i s proper i f and o n l y i f , i n a d d i t i o n ,
t h e r e e x i s t s o n l y f i n i t e l y many s i m p l i c e s a € I ( X ) with f ( a ) = p f o r
every p G I (y) .
A g r e a t amount o f o u r work i n t h e s e L e c t u r e Notes w i l l be spent on a
rather special class of partially p r o p e r maps, t h e p a r t i a l l y finite
maps.
Definition 4. A l o c a l l y s e m i a l g e b r a i c map f:M->N i s calle<d f i n i t e , i f
f i s p r o p e r and has d i s c r e t e - hence f i n i t e ( c f . Lemma 5.8) - f i b r e s ,
f i scalled partially finite, i f f i spartially proper a n d has d i s c r e t e
fibres. As i s e a s i l y seen f i s p a r t i a l l y f i n i t e i f and o n l y i f the r e -
striction f | A : A-»N o f f t o e v e r y c l o s e d s e m i a l g e b r a i c s u b s e t o f M i s
finite. (Hint: lUse t h e f a c t t h a t a space which does not admit any non
constant path iis d i s c r e t e . )
C a u t i o n . A p a r r t ; i a l l y p r o p e r map w i t h f i n i t e f i b r e s i s not n e c e s s a r i l y
finite. F o r e x a m p l e , assume t h a t R c o n t a i n s a sequence o f p o s i t i v e e l e -
ments c o n v e r g i n g t o zero, and t h a t M i s an a f f i n e l o c a l l y complete - but
not complete - s e m i a l g e b r a i c space. The space ^ l Q C is partially complete
Thus t h e n a t u r a i l map p : M 1 Q C - M i s p a r t i a l l y p r o p e r . But p i s not proper
hence n o t f i n i t i e . Indeed, M 1 q c i s paracompact (Ex. 4.4.a). I f p would
be p r o p e r t h e n , by Theorem 5.9, M 1 q c would be s e m i a l g e b r a i c , which i s
not true.
Looking f o r examples of p a r t i a l l y f i n i t e maps we s t a t e
Proposition 6.1 6. L e t f : X - Y be a s i m p l i c i a l map between s t r i c t l y l o -
cally finite s i l m p l i c i a l complexes. The f o l l o w i n g a r e e q u i v a l e n t :
i) f i s i n j e c t i v e on t h e s e t E(a) o f v e r t i c e s o f e v e r y o€I(X).
ii) f has d i s c r e t e fibres.
iii) f : X->Y iss p a r t i a l l y finite.
P r o o f
r iii-)* f maps t h e c l o s u r e a o f e v e r y a € I ( X ) isomorphically
onto t h e c l o s e d simplex f (a) , hence a l s o t h e c l o s u r e c o f e v e r y a £ I (X)
isomorphically onto the c l o s e d simplex f ( a ) . Thus e v e r y f i b r e o f f meets
every c l o s e d simplex i n a t most one p o i n t . A f o r t i o r i i t meets every
open simplex i n a t most one p o i n t . S i n c e X i s t h e u n i o n o f i t s f i n i t e
open subcomplexes, a l l f i b r e s o f f a r e d i s c r e t e . The i m p l i c a t i o n i i i ) =>
ii) is trivial- i i ) => i ) : Suppose t h e r e e x i s t s an open s i m p l e x a inX
such t h a t f i s n o t i n j e c t i v e on E ( o ) . Then a l l f i b r e s o f t h e a f f i n e map
f |a from a t o f ( a ) c o n t a i n l i n e segments and thus have a t l e a s t dimen-
s i o n one. T h i s c o n t r a d i c t s t h e assumption t h a t f has d i s c r e t e fibres.

q.e.d.
Example 6.17. L e t a g a i n f : X Y be a s i m p l i c i a l map bettween s t r i c t l y
locally finite simplicial complexes. Then we s e e from E x a m p l e 6.15 and
the p r e c e d i n g P r o p o s i t i o n 6.16 t h a t f i s p a r t i a l l y finiite i f f f is in-
jective on t h e s e t o f v e r t i c e s E(o) o f every a € I (X) amd i n addition
T€I(X) f o r every open f a c e T of every a € Z ( X ) with f(xr) € I ( Y ) . The map
f i s finite, i f f , i n a d d i t i o n , f o r every p€I(Y), theres e x i s t only fini-
t e l y many a € Z ( X ) w i t h f ( a ) = p. Then f i s a l s o a f f i n e semialgebraic.
Remark 6.18. The statement i n Examples 6.15 and 6.17 reamain t r u e i f t h e
complexes X and Y a r e o n l y l o c a l l y f i n i t e instead of s t r i c t l y l o c a l l y
finite. T h i s can be seen f o r the f i r s t statement i n 6.115 by use o f t h e
path c o m p l e t i o n criterion f o r p a r t i a l properness (Th. 66.8). Then t h e
other statements can be deduced i r much t h e same way a s b e f o r e .
We mention y e t another c l a s s o f examples o f f i n i t e a f f i l n e semialgebraic
maps. Every f i n i t e morphism cp : V-*W between a l g e b r a i c w a r i e t i e s V and W
o v e r R ( i n t h e sense o f a l g e b r a i c geometry) y i e l d s a fiinite affine semi-
a l g e b r a i c map f = cp • V ( R ) -+W(R). As we s h a l l e x p l i c a t e i n t h e next s e c -
tion, t h e s e m i a l g e b r a i c spaces V(R) and W(R) a r e l o c a l l y complete (7.1.c
and f may be a l s o r e g a r d e d as a l o c a l l y s e m i a l g e b r a i c imap f l Q c from t h e

w R T h e
partially complete space V ( R ) 1 q c to ( ) l o c s e m i a l g e b r a i c map f i s
proper. T h i s i m p l i e s , by t h e d e f i n i t i o n of V ( R ) l Q C and W ( R ) 1 q c that
the l o c a l l y s e m i a l g e b r a i c map f l Q C i s affine s e m i a l g e b r a i c . As a p a r -
tially proper s e m i a l g e b r a i c map f i s even p r o p e r . Thus f l Q c isa
finite affine s e m i a l g e b r a i c map.
There remains t h e q u e s t i o n , how our c e n t r a l n o t i o n s "proper", "partially
proper", "finite", "partially f i n i t e " behave under e x t e n s i o n o f t h e base
field R t o some r e a l closed overfield R. T h i s seems t o be more d i f f i c u l t
than t h e c o n s i d e r a t i o n s about e x t e n s i o n o f t h e base f i e l d , say, a t t h e
end o f § 3 . We s h a l l deal with t h i s q u e s t i o n i n Appendix B.

§7 - L o c a l l y compjlete s p a c e s
As b e f o r e t h e worcd "space", i f used without f u r t h e r specification,
means " l o c a l l y s e s m i a l g e b r a i c space" (of c o u r s e separated) over our
f i x e d r e a l c l o s e d l base f i e l d R.
Definition 1 . A sipace M i s c a l l e d l o c a l l y complete, i f e v e r y x E M has
a s e m i a l g e b r a i c n e i g h b o u r h o o d which i s a complete s e m i a l g e b r a i c space.
By C o r o l l a r y 5.10) e v e r y such neighbourhood K i s an a f f i n e semialgebraic
space. I t i s them c l e a r from the r e g u l a r i t y o f K t h a t x has a fundamen-
tal system o f cormplete s e m i a l g e b r a i c neighbourhoods i n K, hence i n M.
In p a r t i c u l a r , every l o c a l l y complete space M i s r e g u l a r . I f M i s a l s o
s e m i a l g e b r a i c , thien M i s a f f i n e .
L o c a l l y complete s e m i a l g e b r a i c spaces have been c o n s i d e r e d b e f o r e i n
t h i s c h a p t e r (Exaimples 2.6, 4.4; Remark 4.20). There t h e p o i n t was that
such a space M l e a d s n a t u r a l l y t o a p a r t i a l l y complete r e g u l a r space

M w n i c n u n d e r :
loc ' f a v o u r a b l e c i r c u m s t a n c e s , i s a l s o paracompact. We
shall generalize the c o n s t r u c t i o n of M 1 q c below to an a r b i t r a r y locally
• complete space M.
Examples 7.1. a) Every p a r t i a l l y complete r e g u l a r space M i s l o c a l l y
complete.
b) Every c l o s e d o r open subspace o f a l o c a l l y complete space M i s l o -
c a l l y complete. Thus e v e r y l o c a l l y c l o s e d subspace X o f M (X E T(M) , X
open i n X) i s l o c a l l y complete.
c) For any a l g e b r a i c v a r i e t y V o v e r R the s e m i a l g e b r a i c space V(R) i s
locally complete.
Indeed, i f V i s a f f i n e , V can be embedded as a c l o s e d subvariety into

1
an a f f i n e s t a n d a r d v a r i e t y A J . Then V(R) i s a closed semialgebraic
n n
subspace o f R . S i n c e R i s l o c a l l y complete, V(R) i s alsso l o c a l l y com-
p l e t e , i n g e n e r a l , we c o v e r V by f i n i t e l y many Z a r i s k i - o j p e n subvarieties
V
1'***' v
r ' s o R
V ( ) i s covered by t h e f i n i t e l y many open s s e m i a l g e b r a i c
subsets V ( R ) which a r e l o c a l l y complete. Thus V(R) i s l c o c a l l y

i complete.
V(R) i s a l s o s e m i a l g e b r a i c , hence a f f i n e s e m i a l g e b r a i c . VWe s h a l l explain
i n the second volume t h a t f o r any v a r i e t y W o v e r RCv^T) = C the s e t of
geometric p o i n t s W(C) i s a l s o a s e m i a l g e b r a i c space o v e r R i n a n a t u r a l way
(cf. [K r § 1 ] ) . i t then w i l l be c l e a r t h a t W(C) i s l o c a l J l y complete.
Additionally, t h e i n f i n i t e c o v e r i n g s o f V ( R ) ( r e s p . W(C)) - t o be s t u d i e d
i n Chapter V - a r e l o c a l l y complete, s i n c e they a r e " l o c e a l l y isomorphic"
to V(R) (resp. W(C)). These spaces a r e no l o n g e r s e m i a l g e e b r a i c .
d) D i r e c t p r o d u c t s o f f i n i t e l y many l o c a l l y complete spacces a r e l o c a l l y
complete. The f i b r e p r o d u c t * M
N 2
o f t w o
l o c a l l y complette spaces IVL
w i t h r e s p e c t t o any l o c a l l y s e m i a l g e b r a i c maps -*N ( i = 1 , 2 ) i s l o -
c a l l y complete s i n c e i t i s a c l o s e d subspace o f M x M . 1 2
e) Going back t o Example 2 . 9 , assume t h a t t h e space M theere i s l o c a l l y
complete. Then the space X c o n s t r u c t e d t h e r e i s a l s o l o c a l l y complete.
f) We s h a l l see i n Appendix B t h a t , f o r any l o c a l l y c o m p l e t e space M
o v e r R and any r e a l c l o s e d o v e r f i e l d R o f R, t h e space M'(K) i s locally
complete.
P r o p o s i t i o n 7 . 2 . L e t M be a dense locally semialgebraic subset of a
space N and x €M. Then e v e r y complete l o c a l l y s e m i a l g e b r a i c neighbour-
hood K o f x i n M i s a l s o a neighbourhood of x i n N. In p a r t i c u l a r , i f
M i s a l o c a l l y complete subspace o f a space L, then M i s open i n M,
the c l o s u r e o f M i n L, i . e . , M i s l o c a l l y c l o s e d i n L.
P r o o f . S i n c e K i s complete, K i s a l s o c l o s e d i n N. Hence N i s the u n i o n
of K and t h e c l o s u r e M \K of M \K. Now x CM \K and thus K i s a l s o a
neighbourhood o f x i n N. Q.e.d.
L e t X be a l o c a l l l y finite simplicial complex o v e r R ( c f . § 2 ) . We look
for a "combinatorial" criterion t h a t X - r e g a r d e d as a space - i s l o -
cally complete. A small d i f f i c u l t y a r i s e s from the f a c t t h a t X needs
n o t be l o c a l l y : f i n i t e and t h e r e f o r e needs not have an interpretation
as a s p a c e .
Proposition 7.3 . The f o l l o w i n g a r e e q u i v a l e n t :
a) The s p a c e X i s l o c a l l y complete.
b) F o r e v e r y op»en s i m p l e x T o f X t h e s t a r s S t ( x ) and S t ( x ) a r e e q u a l .
x x
c) The complex :X \X i s closed.
P r o o f . The e q u i v a l e n c e b) c) i s f a i r l y o b v i o u s . X i s t h e union o f
the subcomplexeis S t ( ) , x T T €l(X), and t h e s e a r e open i n X. I f X is
closed, i.e.X i s an open subcomplex o f X, then a l l s t a r s St (T) are
open subcomplexies o f X. But S t ^ x ) i s t h e s m a l l e s t open subcomplex o f X
containing T . T?hus S t ( ) x T = St ( )
x T f o r every T €I(X). Conversely, i f
: this i s t r u e , t.hen the complexes S t ( T ) X a r e open i n X. Thus a l s o X i s
\ open i n X.
a =* b) : F i x sonne T €I(X) and l e t Y d e n o t e the f i n i t e open subcomplex
St (r)x of X. S i n c e X i s l o c a l l y complete, the open subspace Y o f X i s
a l s o l o c a l l y coimplete. By p r o p o s i t i o n 7.2, Y i s open i n Y. Now let a
be an open simp>lex o f S t - ( x ) , i . e . a £ I (X) and x < a . Choose some
p G I ( X ) with a a ) : The f a m i l y ( S t ( x ) , T €I(X))

x i s an (admissible) c o v e r i n g of the
• space X by open semialgebraic subets. I t s u f f i c e s to v e r i f y that, f o r
; any f i x e d x €I(X), t h e space Y := S t ( x ) x is locally complete. By assump-
? tion, Y = S t ( x ) . x Thus Y i s an open subcomplex o f X. A f o r t i o r i , Y is
an open subcomplex o f t h e f i n i t e closed complex Y. We conclude that Y,
as a space, i s open s e m i a l g e b r a i c i n t h e complete space Y, and there-
f o r e Y i s l o c a l l y complete. q.e.d.
We s h a l l see i n Chapter I I , §2 that e v e r y r e g u l a r paraicompact space i s
isomorphic t o a subspace o f some p a r t i a l l y complete r e g u l a r paracompact
space. Once we know t h i s , i t i s c l e a r from P r o p o s i t i o n ! 7.2 t h a t t h e l o -
c a l l y complete r e g u l a r paracompact spaces a r e , up t o iisomorphism, j u s t
the open subspaces o f t h e p a r t i a l l y complete r e g u l a r pâracompact spaces
In the case o f a s e m i a l g e b r a i c l o c a l l y complete space M we can do b e t -
t e r . We d e f i n e a g e n e r a l i z e d t o p o l o g i c a l space ( c f . §1 , Def. 1)
(M , H M ) , C O V ,) as f o l l o w s . The s e t M i s t h e d i s j o i m t union o f M and

+ +
one f u r t h e r p o i n t ». A subset U o f M i s an element of: ^(M ) i f e i t h e r
U € r(M) o r U = (M \K) U {<»} w i t h K a complete s e m i a l g e b r a i c subset o f

+
M. A f a m i l y (U la€I) i n r(M ) i s an element o f Cov . i . f the u n i o n U o f
T
the U a i s an element o f r(M ) and U i s a c t u a l l y the umion o f f i n i t e l y
+ 4
many U a
i t i s e a s i l y checked t h a t t h e t r i p l e (M , r (M "),Cov ,) i s i n - +
M
+
deed a g e n e r a l i z e d t o p o l o g i c a l space. F o r every U €r(Ml ) we d e f i n e a
ring ° M + ( U ) o f R-valued f u n c t i o n s on U as f o l l o w s . I f U € r(M) then
© (U)
M + : = © (U). M I f U = (M \K) U{°o}, then a f u n c t i o n f: : U -*R i s an e l e -
ment °f (D ( U ) i f t h e r e s t r i c t i o n f l M ^ K i s an element o f © ( M ^ K ) and

+ M
i f f i s a l s o c o n t i n u o u s a t °° i n t h e s t r o n g t o p o l o g y o f U, i . e . for
every e >0 i n R there e x i s t s some complete s e m i a l g e b r a i i c s e t L z>K i n
M with lf(x) - f ( ~ ) | < e f o r a l l x GM^L. C l e a r l y the assignment

U ( u i s a + +
) sheaf (D _,_ of r i n g s o f f u n c t i o n s on ( M ^ ( M ) ,Cov ,) .
+ +
H e n c e f o r t h we s i m p l y denote the r i n g e d space (M ,r(M ) ,Cov ,,€>,) by
M .
Theorem 7.4. M i s a complete s e m i a l g e b r a i c space o v e r R.
+
We c a l l M t h e o n e - p o i n t c o m p l e t i o n o f M. I f M i s a l r e a d y complete,
+
then M i s the d i r e c t sum o f M and t h e one p o i n t space {oo}. The theo-
rem i s t r i v i a l i n t h i s case.
In o r d e r t o prove? t h e theorem we s t a r t by c o n s i d e r i n g the special
n n n + 1
case M = R . L e t :S denote t h e u n i t sphere i n R , l e t e denote t h e
n n n
north pole (0,0,. ..,1) o f S , and l e t p : S ^ { e } ^ R denote t h e s t e r e o -
graphic p r o j e c t i o n , p{x^ , . . . ,x ^) = (y^,*-.,Y ) n with
y ± = x (1-x i m + 1 )" 1
, 1 < i < n.
This i s a semialgrebraic isomorphism w i t h the i n v e r s e map (y^f---/Y ) n
( x X }
" 1 n + 1 '
2 2 1 2 1
x n + 1 = ( r - 1 ) ( r + 1 ) ~ , x. = 2 y , ( r + 1 ) "
2 2 2 —
for 1 emialgebraic space S with the ringed space (R ) «
n + n
H e n c e f o r t h we i d e m t i f y ( R ) with S u s i n g t h i s map p.
Lemma 7.5. Any l o c a l l y complete s e m i a l g e b r a i c space M can be embedded

n
i n some R as a d o s e d subspace.
m
P r o o f . Embed M somehow i n a space R . Then M i s open i n i t s c l o s u r e M.
m 1
Now embed M i n t o R *R as t h e graph o f t h e f u n c t i o n x »-»d(x,M\M) with
d(-,M\M) t h e e u c l i d e a n d i s t a n c e f u n c t i o n from M ^M. (This very classi-
c a l proof, also contained i n [DK^, § 3 ] , i s r e p e a t e d h e r e f o r t h e con-
venience of the r e a d e r ) .
U s i n g t h i s lemma,, t h e p r o o f o f Theorem 7.4 i s easy. We may assume t h a t

n +
M i s a c l o s e d subspace o f R f o r some n >1. We r e g a r d M as a s u b s e t
n + + n +
of ( R ) identifying the point « i n M w i t h t h e p o i n t «> i n ( R ) - This
n +
subset i s c l o s e d and s e m i a l g e b r a i c i n ( R ) . I t i s now e a s i l y checked
+ n +
t h a t the subspace s t r u c t u r e on M i n ( R ) coincides with
+
(f(M ) , C O V , , O M + ) .
P r o p o s i t i o n 7.6. L e t N be a p a r t i a l l y complete space and l e t V be an

open l o c a l l y semialgebraic subset o f N. L e t f : V -» M be a locally
s e m i a l g e b r a i c map from V t o a l o c a l l y complete s e m i a l g e b r a i c space M.

+
Extend f to a map g : N - » M by g ( N \ V ) = {«>} . Then t h e maip g i s l o c a l l y
semialgebraic ( i . e . a morphism o f r i n g e d spaces) i f and o n l y i ffis
partially proper.
1
P r o o f . We have g"" (M) = V. I f g i s a morphism, then g i s ; p a r t i a l l y
proper, since N i s p a r t i a l l y complete. T h i s i m p l i e s that: t h e r e s t r i c -
tion f :g (M) -»M o f g i s a l s o p a r t i a l l y proper. On t h e o t h e r hand,
assume t h a t f i s p a r t i a l l y proper. Once we know t h a t g i i s c o n t i n u o u s ,
then i t i s evident that g i s l o c a l l y s e m i a l g e b r a i c . I n d e e d , f o r any
semialgebraic s u b s e t W o f N, the graph o f giW i s t h e u n i o n o f t h e graph
of f | W n v and g|W^V, and thus i s s e m i a l g e b r a i c . To v e r i f y continuity

+ 1
we have t o check, t h a t f o r any U€ r ( M ) t h e preimage g ((.U) i s open i n
+
N. T h i s i s o b v i o u s i f U c M . Otherwise, M \ U i s a complete semialgebraic
1 1
subset K o f M. Then g " (K) = f"" (K) i s a partially complete locally
1
semialgebraic s u b s e t o f N. In p a r t i c u l a r , g " ( K ) i s c l o s e d I n N, and
1
g ( U ) i s open i n N.
Example_7.7. Any p r o p e r s e m i a l g e b r a i c map f : N M between l o c a l l y com-
plete semialgebraic spaces extends t o a (proper) s e m i a l g e b r a i c map

+ + +
f :N ->M+ w i t h f («>) = oo.
It i s also clear from P r o p o s i t i o n 7.6 t h a t , i f a l o c a l l y complete semi-
a l g e b r a i c space M i s embedded i n a complete s e m i a l g e b r a i c space M w i t h
M\M a one p o i n t s e t , then M may be r e g a r d e d as t h e one p o i n t comple-
tion M o f M. Indeed, t h i s istrivial i f M i s complete. Otherwise, just
apply t h e p r o p o s i t i o n t o N = M, V = M, f t h e i d e n t i t y map M-*M, and
recall that a b i j e c t i v e proper map i s an Isomorphism.
Given a l o c a l l y complete space M we now d e f i n e a p a r t i a l l y complete
space M, , g e n e r a l i z i n g t h e c o n s t r u c t i o n i n Example 2.6. T h i s can be

done i n a s t r a i g h t f o r w a r d way. L e t ^ . ( M ) denote t h e s e t o f a l l complete
semialgebraic ssubsets o f M, and l e t ^ (M) denote t h e s e t o f a l l open
semialgebraics U € f(M) w i t h U € t (M) . There i s a t i n y p o i n t here - i f
M i s n o t p a r a c o m p a c t , we do n o t know whether U i s s e m i a l g e b r a i c f o r
every U € ?*(M) ((cf. Example 3.15). T h i s s u b t l e t y w i l l not d i s t u r b u s .
Both s e t s ^ (M)) and

c (M) , o r d e r e d by i n c l u s i o n , a r e d i r e c t e d systems
*)
of subspaces ore M
We d e f i n e M.. as t h e i n d u c t i v e l i m i t o f t h e system u r (M) i n t h e c a t e -

loc c
gory o f r i n g e d spaces over R w i t h u n d e r l y i n g s e t M. As observed i n
g r e a t e r g e n e r a l i t y i n Examples 2.9 and 4.2. a, ^ l Q C i s a regular locally
semialgebraic s p a c e and (M) i s an a d m i s s i b l e c o v e r i n g o f M ^ q c by
open s e m i a l g e b r r a i c s . Furthermore, f o r each U e f (M) the subspace s t r u c -
t u r e s on U with, r e s p e c t t o M and IVL a r e t h e same. From t h i s we c o n -

loc
elude t h a t i n j f a c t f" (M) = r(JVL ), and t h a t IVL i s l o c a l l y complete.
c loc loc
Every U € t (M)
Q i s contained i n t h e s e t U € r (M) . A l s o , every K € lf* (M) c
i s contained im some V € t (M) . Indeed, K i s c o n t a i n e d i n some W € ?"(M) ,

c
and t h e open suibspace W o f M i s l o c a l l y complete and s e m i a l g e b r a i c .
We have seen im Example 2.6 t h a t K has a complete s e m i a l g e b r a i c neigh-
bourhood A i n M . Thus A € )T (M) and A D K . Hence t h e d i r e c t e d systems o f

Q
spaces f" (M) a n d

c (M) a r e e q u i v a l e n t , and we c o n l c u d e that
(7.8) M l Q C = limJT (M).
c
Moreover, we s e e t h a t every K € JT (M) i s - i n i t s g i v e n space s t r u c t u r e
a complete subspace of M,
c
, and t h a t T (M ) =
n JT (M) .
loc c loc c
The i n c l u s i o n maps K <-» M w i t h K r u n n i n g through t (M)

Q yield - by t h e
very nature o f t h e i n d u c t i v e l i m i t (7.8) - a c a n o n i c a l l o c a l l y semi-
*) Every s e t o f subsets o f M w i l l be r e g a r d e d as a f a m i l y o f s u b s e t s
of M, each member of the s e t b e i n g indexed by i t s e l f .
algebraic map
^M loc
We want to c h a r a c t e r i z e the map p^ by a u n i v e r s a l p r o p e r t y . For this
we need a new definition.
D e f i n i t i o n 2. A space M i s c a l l e d t a u t , i f the c l o s u r e X o f any semi-
a l g e b r a i c subset X o f M i s a g a i n semialgebraic.
Examples 7.9. a) As observed i n P r o p o s i t i o n 4.6, every paracompact
space i s t a u t . The p r o p e r t y " t a u t " s h o u l d be c o n s i d e r e d as a weakening
of the p r o p e r t y "paracompact", needed f o r t e c h n i c a l r e a s o n s .
b) I f N i s a t a u t space and f : M -*N i s a s e m i a l g e b r a i c map, then M i s
t a u t . Indeed, g i v e n some X € f(M) , the s e t Y := f(X) i s s e m i a l g e b r a i c
i n N. Thus Y £ JT(N) . S i n c e f i s s e m i a l g e b r a i c , we conclude that

1
f"" (Y) eT(M) , and then t h a t X €?(M) .
c) Every subspace o f a t a u t space i s t a u t .
d) The d i r e c t p r o d u c t M^ x M 2 of t a u t spaces and M 2 i s t a u t . More
g e n e r a l l y , t h e f i b r e product M^x^M w i t h r e s p e c t t o a r b i t r a r y
2 locally
X M i s a s u b s a c e
s e m i a l g e b r a i c maps M^ -*N and M 2 -+N i s t a u t , s i n c e M-] N 2 P
o f M^ x M . 2
e) I f M i s a t a u t space over R, then f o r any real closed f i e l d R z>R
the space M(R) i s t a u t . Indeed, f o r any semialgebraic subset X of M(R)
t h e r e e x i s t s some open s e m i a l g e b r a i c s u b s e t U o f M w i t h X c U ( R ) . The
c l o s u r e U of U i s s e m i a l g e b r a i c . I t s e x t e n s i o n U(R) i s semialgebraic
and c o n t a i n s X. Hence X er(M(R)) . C o n v e r s e l y , i f M(R) i s taut, then
c e r t a i n l y M i s taut ( c f . Prop. 3.22.b) and c) and Th. B . 1 . i i n App. B).
Remark 7.10. Every t a u t and p a r t i a l l y complete space N i s l o c a l l y com-
p l e t e . Indeed, g i v e n some p o i n t x i n N, we choose an open semialgebraic
neighbourhood U of x. Then the c l o s u r e U o f U i s a complete semialge-
b r a i c neighbourhood of x. In p a r t i c u l a r , N i s r e g u l a r .
:M > M rM a n
We come back t o t h e n a t u r a l map p ^ i o C " f° arbitrary locally
complete space o v e r R.
Theorem 7 . 1 1 , The space M ^ O C i s t a u t and p a r t i a l l y complete. p M has
the following universal property. Given any l o c a l l y semialgebraic map
f : N ->M w i t h N a t a u t and p a r t i a l l y complete space over R , t h e r e exists
a unique l o c a l l y semialgebraic map g :N - » M ^ qc such t h a t f = Pj4*^*
M,l o c
P M
N 1 *M
Proof. L e t U be an open s e m i a l g e b r a i c s u b s e t o f IVL . Then U € ^ (M) .

^ loc c
The c l o s u r e U o f U i n M i s a l s o t h e c l o s u r e o f U i n M, , s i n c e the
loc'
s t r o n g t o p o l o g i e s o f M and M ^ a r e t h e same. We have U €
q c (M) =
T (M. ) . In p a r t i c u l a r , L U i s semialgebraic i n M, , and we see t h a t

c loc ^ loc
M i s t a u t
loc -
M T n e n A s
L e t A be a c l o s e d semialgebraic s u b s e t o f -j_oc" ^ c l o s e d i n M,
again since the strong topologies of M ^ q c and M a r e the same. More-
o v e r , A e)T(U) f o r some U € ^ (M) . Thus A € r ( K ) f o r K := U € r (M) . We

c c
c o n c l u d e t h a t A £f (M) = t (M, ). T h i s proves t h a t M, i s partially r c
c c loc loc
complete.
Now c o n s i d e r a locally semialgebraic map f :N ->M w i t h N t a u t and p a r -

> M
tially complete. We a r e l o o k i n g f o r a map g :N " ]_ o c with f = P °g M
M
and are, of course, f o r c e d t o d e f i n e g :N - * l o c t o be t h e same as f
as a map between s e t s . We have t o v e r i f y t h a t t h i s map g i s l o c a l l y
semialgebraic. I t suffices t o check t h a t , f o r any g i v e n U € <f(N) , t h e

M i s
map g l U :U -*-|_oc semialgebraic. S i n c e N i s t a u t and p a r t i a l l y com-
plete, the c l o s u r e L := U o f U i n N i s s e m i a l g e b r a i c and complete. I f

we know t h a t g l L i s s e m i a l g e b r a i c , then, o f c o u r s e , g l U must a l s o be
semialgebraic. But t h i s is trivial. K := f ( L ) i s an element o f r (M),
and g|L i s t h e composite o f t h e s e m i a l g e b r a i c map h : L -»K, obtained

< L M
from f by r e s t r i c t i o n , and t h e i n c l u s i o n map K * ^ o c which i s a l s o
semialgebraic. q.e.d.
It f o l l o w s from Theorem 7.11 t h a t any l o c a l l y semialgebraic map f :M -*N
between l o c a l l y complete spaces over R y i e l d s a unique l o c a l l y semial-
g e b r a i c map f , :M ->N. with n p ° f,

XT = f ° P „ . Moreover, t h e u n i -
^ * lOC loc loc MSI lOC ~M
v e r s a l property described i n t h e theorem i s q u i t e u s e f u l f o r a n a l y z i n g
the b e h a v i o r o f t h e f u n c t o r M from t h e c a t e g o r y o f l o c a l l y com-
p l e t e spaces over R t o t h e f u l l subcategory of t a u t p a r t i a l l y complete
spaces o v e r R. We s t a t e two f a c t s about t h i s functor.
L e t f ^ : M^ ->N ( i = 1,2) be l o c a l l y semialgebraic maps between locally
x r = M b e t h e fi t ) r e
complete spaces over R and l e t M^ ^ 2 p r o d u c t o f M^
and M 0 with respect to the f . . Also l e t P := M 1 , x N

M o n be t h e
2 l 1 / loc j _ o c 2,xoc
f i b r e product of M 1 1 q c and M 2 l Q c with respect to t h e maps (fi)i o c .
By Example 7.1.d t h e space M i s l o c a l l y complete and by Example 7.9.d
and §6 t h e space P i s t a u t and p a r t i a l l y complete.
Proposition 7.12. The n a t u r a l map p : P -*M i n d u c e d by t h e maps

:N Nn a s t n e
p^_ : (M^) j L o c and p N ]_ o c ~* u n i v e r s a l property for M
described i n Theorem 7.11. In s h o r t ,
M x M
W-lWloc = 1,loc N l o c 2,loc
This i s easily verified.
P r o p o s i t i o n 7.13. L e t f : M -»N be a p a r t i a l l y p r o p e r map between spaces
over R. Assume t h a t N i s l o c a l l y complete. A l s o assume t h a t e i t h e r M
is taut or that f i s semialgebraic. Then M i s l o c a l l y complete, and

i s a pullback i n the category o f spaces over R. In s h o r t
IOC N loc
P r o o f . L e t x €M be g i v e n . We choose a complete s e m i a l g e b r a i c neigh-
bourhood A o f f ( x ) i n N. Then f (A) i s a c l o s e d neighbourhood o f x
i n M, and t h e r e s t r i c t i o n f ( A ) -•A o f f
1
i s p a r t i a l l y proper. If f i s
semialgebraic, then f ^ (A) i s s e m i a l g e b r a i c and complete. I f M i s t a u t ,
then x c e r t a i n l y has a c l o s e d s e m i a l g e b r a i c neighbourhood D. A l s o ,
B := D n f ( A ) i s a c l o s e d s e m i a l g e b r a i c neighbourhood o f x and t h e
1
r e s t r i c t i o n B ->A o f f i s p r o p e r . We conclude t h a t B i s complete. Thus,
i n both c a s e s , M i s l o c a l l y complete. We c o n s i d e r t h e p u l l b a c k diagram
f o r the maps f and p^. The map g i s p a r t i a l l y proper. Thus P i s p a r t i -
a l l y complete. I f f i s s e m i a l g e b r a i c , then g i s s e m i a l g e b r a i c , and we
conclude from t h e t a u t n e s s o f N ^ q c that P i s taut (Example 7.9.b). I f
M i s t a u t , then P i s a l s o t a u t , s i n c e i t i s t h e f i b r e product o f two
t a u t spaces (Example 7.9.d). Now i t i s e a s i l y checked t h a t q has t h e
u n i v e r s a l p r o p e r t y d e s c r i b e d i n Theorem 7.11 s i n c e p N has t h i s property.
Thus q can be i d e n t i f i e d with p . M q.e.d.
The constructions i n this s e c t i o n enable us t o s t r e n g t h e n one o f the
counterexamples i n 6.3 f o r p a r t i a l l y proper maps.

Counterexample 7,14. I f f : N-* M and g : M -* S a r e l o c a l l y semialgebraic
maps w i t h f s u r j e c t i v e and g«f p a r t i a l l y f i n i t e then i t may happen
that g i s not p a r t i a l l y f i n i t e . Choose f o r example some semialgebraic
locally complete space M which i s not complete. L e t f be t h e n a t u r a l

M M a n d b e t n e +
map p M : l o c ~> 9 i n c l u s i o n map M^M . Then f i s s u r j e c -
1 S
tive and g-f i s p a r t i a l l y p r o p e r , s i n c e ^-^ QC partially complete.
The f i b r e s o f g«f a r e f i n i t e (empty o r o n e - p o i n t s e t s ) . Thus g«f i s
partially f i n i t e . But g i s n o t p a r t i a l l y finite.
We have seen t h a t the n o t i o n of tautness i s - besides partial complete-
ness - t h e key t o understand t h e spaces Mô C from a c a t e g o r i a l view
point (Th. 7.11). We s h a l l later r e s t r i c t o u r a t t e n t i o n t o paracompact
spaces i n s t e a d o f t a u t spaces. The f o l l o w i n g p r o p o s i t i o n makes c l e a r
t h a t t h i s means t o add a h y p o t h e s i s o f c o u t a b i l i t y .
P r o p o s i t i o n 7.15. F o r a space M t h e f o l l o w i n g a r e e q u i v a l e n t :
i) M i s paracompact and has c o u n t a b l y many c o n n e c t e d components.
ii) M i s t a u t and L i n d e l o f ( c f . §4, Def. 3).
Proof. The i m p l i c a t i o n i.) => i i ) i s c o n t a i n e d i n Proposition 4.6 and
Theorem 4.17. In o r d e r t o prove i i =» i ) we choose an a d m i s s i b l e covering
(X |n€3N)
n o f M by open s e m i a l g e b r a i c s u b s e t s X^. We then choose a
family ( U |n£W) i n fr(M) such t h a t U D X and U ^ x U U - f o r n > 2 . 1 1
n' I T n n n- i —
V : = U V : = =
We f i n a l l y d e f i n e open s e m i a l g e b r a i c s e t s V-j :=U^, 2 2' n
U Û ( n > 3 ) . The V cover t h e s e t M and every V meets a t most

n n-z 0
— n -* n
t h r e e o f t h e s e t s X . Thus t h e c o v e r i n g (V ln€U) of M i s l o c a l l y
x v
n
f i n i t e , which proves t h a t M i s paracompact. Of c o u r s e , M has o n l y
c o u n t a b l y many connected components, s i n c e every X n meets o n l y finitely
many connected components. q.e.d.

C h a p t e r I I . - Completions and t r i a n g u l a t i o n s
§1 - Gluing paracompact spaces
Whereas i n t h e f i r s t c h a p t e r we were m o s t l y concerned w i t h t h e a n a l y s i s
and e x p l i c a t i o n o f t h e fundamental n o t i o n s , we now emphasize construc-
tions. In t h i s s e c t i o n we d e a l w i t h t h e problem o f g l u i n g l o c a l l y semi-
a l g e b r a i c spaces along closed subspaces. Our c e n t r a l r e s u l t , Theorem
1.3 below, w i l l be u s e f u l f o r c o n s t r u c t i n g " c o m p l e t i o n s " o f paracom-
pact s p a c e s , c f . §2 below. I t i s a l s o i n t e r e s t i n g i n i t s own r i g h t .
For t e c h n i c a l r e a s o n s , we now do n o t assume t h a t l o c a l l y semialgebraic
spaces a r e always u n d e r s t o o d t o be s e p a r a t e d . But i f a space i s c a l l e d
"paracompact" then we mean " s e p a r a t e d and paracompact", as i n C h a p t e r I.
U l t i m a t e l y we a r e o n l y i n t e r e s t e d i n separated spaces, o r b e t t e r , i n
regular s p a c e s . We have t o make s u r e t h a t g l u i n g s e p a r a t e d (resp.
regular) spaces always r e s u l t s i n s e p a r a t e d (resp. r e g u l a r ) spaces.
This i s g u a r a n t e e d by
Lemma 1.1. L e t M be a l o c a l l y semialgebraic space o v e r R, and l e t
( M ( a € l ) be a l o c a l l y
a finite family i n T(M) w i t h M = u t M ^ l a C l ) . Assume
t h a t t h e subspaces of M are a l l separated (resp. r e g u l a r ) . Then M
is also separated (resp. r e g u l a r ) .
Proof. Assume f i r s t that a l l a r e r e g u l a r . L e t x be a p o i n t inM
and B be a c l o s e d l o c a l l y semialgebraic subset o f M w i t h x £ B . The
family (M^laCI) i s l o c a l l y f i n i t e . Hence I (x) := {a 6 I l x E M ^ } i s a
f i n i t e s e t and A x := u(M
(X |a€l \ I ( x ) ) i s a closed l o c a l l y semialge-
b r a i c s u b s e t o f M. We want t o s e p a r a t e x and B by open l o c a l l y semi-
algebraic s e t s . Replacing B by B u A x i f n e c e s s a r y , we may assume

M
that A c B . x F o r every a € I (x) t h e r e e x i s t s an U a £ ^ a ) with x € U a and
U a n B = 0 . We choose s e t s V a € f ( M ^ B) w i t h v
a
n M
a
= u
a (« € I (x) ) , cf.
I, Prop. 3 . 1 1 . Then V := n ( V l a € I ( x ) ) i s an open s e m i a l g e b r a i c n e i g h -

a
bourhood o f x. The c l o s e d s e m i a l g e b r a i c s e t Z := U(U |a€I(x)) c o n t a i n s
V and i s d i s j o i n t from B. Thus x and B a r e s e p a r a t e d by t h e open l o -
cally s e m i a l g e b r a i c s e t s V and M ^ Z . T h i s proves that M i s regular.
Running through similar arguments, we see t h a t M i s s e p a r a t e d i f all
sets M a a r e s e p a r a t e d spaces. q.e.d.
Lemma 1 . 2 . L e t M be a l o c a l l y s e m i a l g e b r a i c space over R and l e t
(M la€I) be a l o c a l l y f i n i t e f a m i l y i n 7(M) w i t h U ( M l a € I ) = M. Then

a
t h e s t r u c t u r e o f M as a l o c a l l y r i n g e d space i s completely determined
by t h e subspaces i n t h e f o l l o w i n g way:
a) A s u b s e t U o f M belongs t o f(M) i f and o n l y i f U fl M a € T(M ) for
every a € I .
b) A f a m i l y (U^|A€A) i n T(M) b e l o n g s to Cov M i f and o n l y i f
(U x fl M a I A € A) € C o v M f o r every a € I .
c) A f u n c t i o n f : U-» R on some U € T(M) belongs t o G> (U) i f and o n l y i f

M
flUflM 6<9 M (U n M ) f o r every <x€I.
Furthermore, i f a l l subspaces M Q a r e paracompact, then M i s a l s o para-
compact.
P r o o f . The statement (c) has a l r e a d y been proved i n I, § 3 (gluing o f
locally s e m i a l g e b r a i c maps, c f . Prop. 3 . 1 6 ) . Notice that also the
" o n l y i f " p a r t s i n t h e statements (a) and (b) a r e t r i v i a l . To prove
(a) c o n s i d e r a s u b s e t U o f M w i t h UflM € T(M ) f o r e v e r y a € I . Then

a a
(MÛ) P l M € T ( M ) a a f o r every a G I . We c o n c l u d e that M ^ U £ T ( M ) since
it i s t h e union o f t h e l o c a l l y f i n i t e f a m i l y ((M ^ U) n M I a € I) o f
closed locally s e m i a l g e b r a i c s u b s e t s . Thus U i s i n d e e d an element o f
f (M) .
Now l e t ( U ^ I X e A ) be a f a m i l y i n T(M) w i t h (U^ fl M IA € A ) € C o v
a M for
a
every a € I . Then, by ( a ) , U := U ( U ^ | A € A ) i s an element o f T(.M) .
C o n s i d e r an open s e m i a l g e b r a i c s u b s e t M' o f M. Only f i n i t e l y many

1
sets meet M'. Every U D M ^ r i M i s c o v e r e d by f i n i t e l y many s e t s
U^. Thus U f l M ' i s a l s o c o v e r e d by f i n i t e l y many s e t s U^. Hence
(U^|A€A) i s an a d m i s s i b l e c o v e r i n g o f U. Statement (b) i s proven.
Assume, f i n a l l y , t h a t a l l spaces M Q a r e paracompact. We want t o show
t h a t M i s a l s o paracompact. M i s s e p a r a t e d by Lemma 1.1. L e t
(M pl3€I )
a a be a l o c a l l y finite covering of by open s e m i a l g e b r a i c
subsets. Every closure i s contained i n the union o f the f i n i t e l y
many s e t s M w i t h M _ PI M _ * 0 . Thus M „ i s s e m i a l g e b r a i c and

ay ay ap ap
(M pla€ 1 , 3 6 1 ^ ) i s a locally
a finite f a m i l y i n r(M) w i t h u n i o n M. Re-
placing (M la€I)
a by ( M ^ I a € 1 , 3 € I ) a we may even assume t h a t a l l spa-
ces M a a r e s e m i a l g e b r a i c . F o r any a € I we i n t r o d u c e t h e f i n i t e sets
of indices
1(a) := { 3 € I|M & (1 M * 0}

a
and t h e open l o c a l l y s e m i a l g e b r a i c neighbourhood
W a := M \ U ( M I 3 € I N . 1 ( a ) )
0
of M-
a Since W Q i s c o v e r e d by t h e f i n i t e l y many s e m i a l g e b r a i c s e t s
with 3 € 1 ( a ) , we s e e t h a t W a i s a c t u a l l y open s e m i a l g e b r a i c . The
family (W l<x€I) i s l o c a l l y
a finite and c o v e r s M. Thus M i s indeed
paracompact, and Lemma 1.2 i s proven.
We a r e ready f o r t h e main r e s u l t o f t h i s section.
Theorem 1.3 ( G l u i n g p r i n c i p l e f o r r e g u l a r paracompact s p a c e s ) . L e t M
be a s e t and l e t ( M l a € I ) a be a f a m i l y o f s u b s e t s o f M w i t h
M = U (M |a€I). Assume t h a t f o r every a G I t h e s e t M i s provided with

a structure C f , C o v , © ) o f a r e g u l a r paracompact l o c a l l y
a a a semialgebraic
space o v e r R. Assume f u r t h e r t h a t f o r any two i n d i c e s a,3G I t h e i n t e r -
section M a D Mp i s a c l o s e d locally semialgebraic subset o f M Q and M^,
i.e. i n o b v i o u s n o t a t i o n , M n M € T ( M l and M fl M € T„ (M ) and t h a t Q n fl
c x p o t o c O t p p p '
the subspace s t r u c t u r e s on M fl M^ w i t h r e s p e c t t o ( ^ ' Q' ^ ) and a

C o v
a a
(Tp,CoVp,0^) a r e e q u a l . Assume f i n a l l y that f o r every a G I t h e f a m i l y
(M„ fl M,. I 3 G I) i n T (M ) i s l o c a l l y f i n i t e with respect to the structure

a p a a
( T , C o v , ( o ) . Then t h e r e
a a a e x i s t s a unique s t r u c t u r e (T(M) ' COV
M ' * £ V ^ °f a
locally semialgebraic space over R on M such that:
i) M GT(M) f o r every a G I .
a
ii) The f a m i l y (M laGI) i s l o c a l l y

a finite.
iii) The subspace s t r u c t u r e o f M a i n M with respect to (T(M),Cov ,© ) M M
coincides with ( T , C o v , © ) f o r every a G I .

a a a
Moreover, t h e space M i s r e g u l a r and paracompact.
Proof. 1) We a l r e a d y know from Lemma 1.2 t h a t t h e r e e x i s t s a t most

C O V o f a
one structure (t(M), M '^ ) M locally semialgebraic space on M
(separated o r not) which f u l f i l l s ( i ) - ( i i i ) and how t h i s structure
may be d e s c r i b e d e x p l i c i t l y . We a l s o know t h a t t h i s s t r u c t u r e - i f
it e x i s t s - i s r e g u l a r and paracompact.
2) We prove e x i s t e n c e first i n the case t h a t I i s f i n i t e and e v e r y
M q with i t s given structure i s a f f i n e semialgebraic. Using induction
on t h e c a r d i n a l i t y o f I we r e t r e a t t o t h e c a s e t h a t I i s a two e l e -
ment s e t {1,2}. L e t A := M^ fl M . The i d e a i s t o f i n d 2 a space s t r u c -
t u r e on M f u l f i l l i n g ( i ) - ( i i i ) by i n j e c t i n g M s u i t a b l y i n t o some
m n
R . We choose s e m i a l g e b r a i c embeddings tp : M a <-> R (a = 1,2) o f M Q
with respect to the given structure (T ,Cov ,<0 ). We e x t e n d cp IA t o

a a a a
:M R n
a semialgebraic map £ ( (a, 3) = (1,2) o r = ( 2 , 1 ) ) . T h i s i s
p o s s i b l e by " T i e t z e ' s e x t e n s i o n theorem" ( I , Th. 4.13 o r [DK^, §1]).

n
cp a and y i e l d a map : M->R . We f i n a l l y choose semialgebraic
1 1
functions h 1 : M -» [-1 O] and h
1 f 2 :M 2 [0,1 ] w i t h h ~ (0) = h " 2 (O) = A
(cf. I , 4.15). These f u n c t i o n s a r e then g l u e d t o form a f u n c t i o n
1 1
h : M - [-1 ,1 ] . We have h ' (0) = A, h ~ ( [ - 1 , 0 ] ) = M^ f and
1
h" ([0,1]) = M . 2 The map
f = (ij^ ,\Jj ,h) : M -* R

2
n
xR n
xR
is injective, and i t s r e s t r i c t i o n s to and M 2 are semialgebraic.
In p a r t i c u l a r , : =f(M^) and N 2 := f ( M ) 2 are semialgebraic subsets

m
of R, m = 2n+ 1 . Thus N := UN^ i s also semialgebraic. We e q u i p N
m
w i t h t h e subspace s t r u c t u r e o f t h e s e m i a l g e b r a i c s t a n d a r d space R . By
construction, (resp. N ) 2 i s the s e t of a l l points (x-j , . . . ,x^) i n
N w i t h x <0 ( r e s p . x > 0) . Thus KL and N~ a r e c l o s e d

c semialgebraic
3
m— m— 1 z
s u b s e t s o f N. The r e s t r i c t i o n f = ( c p ^ x ^ h ^ ) of f i s a semialge-
braic isomorphism o f onto , s i n c e cp^ i s an embedding. (The i n -
v e r s e map i s t h e composite o f t h e n a t u r a l p r o j e c t i o n from onto
cp^ (M^ ) and tp^ 1

: cp^ (M^ ) -» ) . f |M 2 i s an isomorphism o f M 2 onto N . 2
Pulling back t h e space s t r u c t u r e o f N t o M by t h e map f we o b t a i n a
space s t r u c t u r e on M w i t h t h e d e s i r e d p r o p e r t i e s (i)- (iii).
3) We c o n s i d e r the case that I i s i n f i n i t e b u t every is still
affine semialgebraic with i t s given space s t r u c t u r e . Our assumption
that the f a m i l i e s (M flMp[3€I)

a are l o c a l l y f i n i t e inM a means i n
this case that every s e t M a meets o n l y f i n i t e l y many s e t s M^. We
introduce the f i n i t e sets of indices
1(a) : = { 3 € 1 1 M fl M * 0 }
a 3 fl
and the subsets o f M
W := U(M I T € I I (a) )
o f M. Every W a i s the union of the f i n i t e family (W DQ I 3 € I (a) ) . We
equip every W D a with i t s subspace s t r u c t u r e i n M^. A p p l y i n g what
has been p r o v e d i n s t e p 2) t o and t h i s f a m i l y , we see t h a t every
W^ has a unique s t r u c t u r e o f a s e m i a l g e b r a i c space such t h a t t h e s e t s
W n M , 3 € I ( c c ) , with t h e i r given
R space s t r u c t u r e s , a r e c l o s e d semi-
a l g e b r a i c subspaces o f W a and t h a t i n t h i s structure W Q i s affine.
More g e n e r a l l y , we o b t a i n f o r every f i n i t e s u b s e t K o f I a unique
semialgebraic space s t r u c t u r e on t h e s e t W K := U ( W l a € K )
a such that
for every 3 € U (I (a) I a €K) t h e s e t W K nM^ w i t h i t s subspace structure
i n Mp i s a c l o s e d semialgebraic subspace o f W , K and W K i s a f f i n e . By
use o f Lemma 1.2 i t i s e a s i l y seen t h a t f o r any two f i n i t e subsets
KcL o f I t h e space W.. i s an open s e m i a l g e b r a i c subspace o f W . We T
equip M = U W v w i t h t h e space s t r u c t u r e as t h e i n d u c t i v e limit of the

K
K
directed system o f t h e s e spaces W , K c f . I , Lemma 2.2. Then every W K -
in i t s given s t r u c t u r e - i s an open s e m i a l g e b r a i c subspace o f M and
(W la€l)
a i s a locally finite covering o f M by open s e m i a l g e b r a i c sub-
s e t s . Every M a meets o n l y f i n i t e l y many W^ and M a n G JT(W^) by con-
s t r u c t i o n . Thus every M Q i s a closed semialgebraic s u b s e t o f M. More-
over, M a i s - i n i t s given s t r u c t u r e - a c l o s e d s e m i a l g e b r a i c sub^
space o f M. Every W^ meets o n l y f i n i t e l y many M . Hence t h e f a m i l y

Q
(M la€I) i s l o c a l l y
a finite i n M. The theorem i s t h e r e f o r e proved i n
the g i v e n special case.
4) We f i n a l l y come t o t h e g e n e r a l c a s e . We choose f o r e v e r y a € I a
locally finite family (N l3€J(a)) of a f f i n e closed semialgebraic

subsets N n of M which c o v e r s M . T h i s i s possible since M i s regu-
a3 a a ^ a
lar and paracompact (cf. the l a s t step i n the proof o f Lemma 1.2 o r
I, Th. 4.11). U s i n g t h e assumptions o f t h e theorem we s e e t h e f o l l o w -
ing: Every meets a s e t M^ f o r o n l y f i n i t e l y many i n d i c e s y € I .

For any such index t h e i n t e r s e c t i o n N n (IM i s c l o s e d and s e m i a l g e -
2
a3 T
braic i n M^. Thus meets o n l y f i n i t e l y many s e t s o f t h e f a m i l y
( N ^ l y e 1,6 6 J ( Y ) ) . We e q u i p every N with i t s subspace structure
in M a and a p p l y what has been proved i n s t e p 3 to the family
(N | a € I/3 € J (a)). We o b t a i n a space s t r u c t u r e on M which i s e a s i l y

ap
checked t o have a l l t h e d e s i r e d p r o p e r t i e s by use o f Lemma 1.2. T h i s
f i n i s h e s the proof o f t h e theorem.

Theorem 1.3 i s somewhat more s p e c i a l than one would expect i n view of
Lemma 1.2. I f we only assume t h a t the spaces M Q are paracompact ( i n -
stead of r e g u l a r and paracompact), then we cannot g u a r a n t e e t h a t a
locally semialgebraic space s t r u c t u r e on M w i t h the p r o p e r t i e s (i) -
(iii) e x i s t s . But Theorem 1.3 s u f f i c e s f o r a l l a p p l i c a t i o n s we have
i n mind.
Example 1.4. In I, §2 we equipped every l o c a l l y finite simplicial
complex X o v e r R w i t h the s t r u c t u r e of a l o c a l l y semialgebraic space
such t h a t f o r every open simplex a € I (X) the s e t o n X i n i t s "simplex
structure" i s a closed semialgebraic subspace o f X. Theorem 1.3 yields
a new justification t h a t t h i s space s t r u c t u r e e x i s t s on X f i s unique
and i s r e g u l a r and paracompact. T h i s was s t a t e d i n I, §4 (Example 4.8).

§2 - E x i s t e n c e o f completions
In t h i s whole s e c t i o n we assume t h a t M i s a r e g u l a r paracompact space
over R. Our g o a l i s to prove the f o l l o w i n g
Theorem 2.1. There e x i s t s an embedding i : M *-» N of M i n t o a partially
complete r e g u l a r paracompact space N ( i . e . an isomorphism o f M onto
a locally semialgebraic subspace o f N).
R e p l a c i n g N by the c l o s u r e of i(M) i n N we then o b t a i n an embedding i
f o r which i(M) i s a l s o dense i n N. Such an embedding w i l l be c a l l e d a
( r e g u l a r paracompact) c o m p l e t i o n of M. N o t i c e t h a t i t does not make
sense t o look f o r an embedding of M i n t o a complete space N, except
i n the t r i v i a l case that M i s semialgebraic, s i n c e by I, § 5 any com-
p l e t e paracompact space N i s s e m i a l g e b r a i c . "Partially complete" i s
the b e s t we can hope f o r .
Our p r o o f of Theorem 2.1 i s based on §1 and the f o l l o w i n g easy lemma.

n n
We denote by S , more p r e c i s e l y by S ( R ) i f necessary, the standard
n-sphere over R, i . e .
s n
= {(x o x )
n e R n + 1
lx o
2
+ ... + x n
2
=1}.
n
We f u r t h e r denote the n o r t h p o l e (0,...,0,1) o f S by <».
Lemma 2.2. For any open s e m i a l g e b r a i c subset U of M there e x i s t s , for
some n € ]N , a l o c a l l y s e m i a l g e b r a i c map cp : M -* S n
such t h a t cp IU i s an
embedding and cp ^ (<») = MÛ.
Proof. The closure U of U i s again semialgebraic ( I , Prop. 4 . 6 ) . We

1
choose an open s e m i a l g e b r a i c neighbourhood M of U. It suffices to
prove the a s s e r t i o n f o r M' and U i n s t e a d o f M and U. Thus we assume

without l o s s o f g e n e r a l i t y t h a t the space M i s a f f i n e semialgebraic.
11 n
We choose an embedding : M^R o f M i n t o an a f f i n e standard space R.
We f u r t h e r choose a s e m i a l g e b r a i c f u n c t i o n f : M-» R w i t h f ^ (0) = MÛ

n+ n +
(e.g. the e u c l i d e a n d i s t a n c e f u n c t i o n from M \ U ) . Let q : R ^ S ^{«>}
be t h e i n v e r s e o f t h e s t e r e o g r a p h i c p r o j e c t i o n w i t h c e n t e r °° ( c f . I , § 7 ) .
Then
1
fq(ijj(x) , f ( x ) " ) x € U
<P(x) :=
I oo x £ U
n +
is a s e m i a l g e b r a i c map from M into S ^ w i t h cp ^ (<») = MÛ. The restric-
n + 1
tion cp|U : U -• S ^ {00} o f cp i s an embedding. q.e.d.
We now start proving the theorem. We choose a l o c a l l y finite covering
( M l a € I ) o f M by open s e m i a l g e b r a i c s u b s e t s
a and a locally semialgebraic
n
map cp : M-+S(a) o f M
a i n t o a standard sphere S(a) = S ^ f o r each a € I
( M
such t h a t t h e n o r t h p o l e c©^ o f S (a) has the preimage M ^ M Q and P a l a
is an embedding. F o r e v e r y a € I we i n t r o d u c e the f i n i t e s e t s o f i n d i c e s
J(a) := (3 € I|M a fl * 0}, 1(a) := U (J (3) I 3 € J (a) )
I(a)' := I(a) x {a}.
U A := U ( M p I 3 € J ( a ) ) i s an open s e m i a l g e b r a i c neighbourhood of M . Q For
each a € I we choose a l o c a l l y semialgebraic function f : M -> [ 0 , 1 ] with

1 1
f " (D
a = M q and f " ~ (0) = M ^ U ,
a A c f . I, Th. 4 . 1 5 . In the set
Z := TT (S(3)*[0,1j)
361
we c o n s i d e r the f a m i l y o f subsets
N(a) : = TTN(a) (a € I) .
P
3
Here N ( a ) Q = S ( a ) x { 1 } , N(a)^ = {(^,0)} - a one point set - f o r
3GI\I(a), and N(a)^ = S ( & ) x [ o , 1 ] f o r 3 6 1 ( a ) ' . We regard every N(a)
as a complete a f f i n e s e m i a l g e b r a i c space i s o m o r p h i c to the product
(S(a)x{1}) x TT , (S(3)x[0,1]j
3GI(a)
in t h e o b v i o u s way. We want t o endow t h e subset
N := U(N (a) I a € I)
of Z with the s t r u c t u r e of a p a r t i a l l y complete l o c a l l y semialgebraic
space o v e r R. C l e a r l y N(a) f!N(3) * 0 i f and o n l y i f 3 G I (a) , and i n
this case
N(a) n N ( 3 ) = T T N (a, 3)
T
yGI
with N ( a , 3 ) Y = S ( Y ) * { 1 } f o r y = a o r T = 3, N ( a , 3 ) T = S(y)x[o,1] i f
1
y € 1(a)* n i ( 3 ) , and N(a,&) = { (~ ,0) } e l s e . We see t h a t a g i v e n N(a)
y
meets N ( 3 ) o n l y f o r f i n i t e l y many i n d i c e s 3/ and t h a t i n t h i s case
N(a) H N ( 3 ) i s c l o s e d s e m i a l g e b r a i c i n N(oc) and i n N ( 3 ) . A l s o , t h e sub-
space s t r u c t u r e s on N(a) 0 N ( 3 ) w i t h r e s p e c t t o N(a) and N ( 3 ) a r e i d e n -
tical. From Theorem 1.3 we know t h e f o l l o w i n g : There e x i s t s on N a
unique l o c a l l y semialgebraic space s t r u c t u r e such t h a t , f o r e v e r y a € I ,
the space N(a) - w i t h i t s g i v e n structure - i s a closed semialgebraic
subspace o f N, and t h e f a m i l y (N(a)la€I) is locally f i n i t e i n N. We
equip N with t h i s s t r u c t u r e . Then N i s r e g u l a r , paracompact, and p a r -
tially complete. (The space s t r u c t u r e on N can be d e s c r i b e d explicit-
ly, c f . Lemma 1.2.)
We d e f i n e f o r every p o i n t x G M a point i ( x ) := (cp^ (x) , f ^ (x) ) GZ.
If x e M , then (cp ( x ) , f (x) ) € S (a) x { 1} and (<p (x) , f (x) ) = (~ ,0) f o r

ft ft a
a a a p P P
every p £ I s 1(a) . (Notice t h a t 3 G I(a) i f f a G I (3) ) . Thus i (x) G N (a) c N.
We c l a i m t h a t t h e map i : M-»N, x w i ( x ) , i s a l o c a l l y semialgebraic em-
b e d d i n g . We prove t h i s i n five steps:
a) i i s i n j e c t i v e . Indeed, l e t x and y be two d i f f e r e n t p o i n t s o f M.
I f x and y b o t h l i e i n some M , a
x + t
then and cp (y) = » . Thus i n
both c a s e s i ( x ) * i ( y ) .
b) We prove t h a t t h e map i i s l o c a l l y semialgebraic. For every finite

s u b s e t K o f I we i n t r o d u c e t h e s e m i a l g e b r a i c s e t s
M K := U(M la€K) G r(M),

a
N R := U(N(a) |a€K) G ?(N) .
C
We have i ( M ) N « K K We may r e g a r d N R as a s e m i a l g e b r a i c s u b s e t o f t h e
finite product TT (S (a) x [o, 1 ]) , w i t h L := U(I(a) I a € K ) . F o r e v e r y a G K

aGL
the subspace s t r u c t u r e o f N(a) i n N c o i n c i d e s w i t h t h e subspace struc-
t u r e o f N(a) i n t h i s f i n i t e p r o d u c t . We c o n c l u d e from Lemma 1.2 t h a t
the subspace structure of N K i n N c o i n c i d e s w i t h t h e subspace struc-
ture i n t h i s f i n i t e p r o d u c t . We have
IIM k = (vVee - L
Thus i l M K i s a s e m i a l g e b r a i c map from M^ t o , hence a s e m i a l g e b r a i c
map from M K t o N. But M i s t h e i n d u c t i v e l i m i t o f t h e d i r e c t e d system
of spaces (M^Kci, K f i n i t e ) , c f . I, Ex. 2.3. Thus i : M -»N i s l o c a l l y
semialgebraic.
c) F o r every a G I we have i (M) 0 N (a) = i ( M ) . Thus t h e map i i s s e m i a l -

a
gebraic ( c f . I.5.2.b). Indeed, we know a l r e a d y t h a t i ( M ) c N ( a ) . On
the o t h e r hand, i f x i s a p o i n t o f M w i t h i ( x ) G N ( a ) , then f ( x ) = 1» a
hence x6M . Q
d) We c o n c l u d e from c) t h a t L := i(M) i s a l o c a l l y s e m i a l g e b r a i c sub-
set o f N. We want t o prove t h a t , f o r any a G I , t h e s e m i a l g e b r a i c map
i = i IM :M -* N(a)
a a a
i s an isomorphism from M Q t o t h e subspace L D N ( a ) o f N. F o r t h i s i t
s u f f i c e s to know t h a t t h e image i (A) = i ( A ) o f any c l o s e d semialge-
b r a i c subset A of M a i s closed i n N(a) (IL. I d e n t i f y i n g N(a) w i t h t h e
space
(S(a)x{l})x TT , (S(B)x[0,1])
&GI(a)
(as above), we have
Let x € M a be g i v e n w i t h i ( x ) € i ( A ) . We have t o v e r i f y t h a t x € A. G i v e n
any e > 0 i n R, t h e r e e x i s t s some y GA w i t h
II cpîx) - cp (y) II < e, l f p ( x ) - f ( y ) I < e

3 p
f o r every 3 6 1 ( a ) . (II II i s t h e e u c l i d e a n standard norm on t h e ambient
space R ^ ^ n + 1
0 f s(3).) We choose an index yGvJ(a) w i t h x E M ^ . Then
cp^(x) * oo. F o r e s m a l l enough we a l s o have cp^(y)#oo^ i . e . y € M ^ . We con-
clude t h a t cp^(x) l i e s i n t h e c l o s u r e o f cp^. (A fl M^) i n tp^ (M^) . But tp^M^
i s a semialgebraic isomorphism from onto cp^(M^), and A fl i s closed
i n M . Thus indeed xGAOM^cA.
e) P u l l i n g back t h e subspace s t r u c t u r e o f L i n N t o M by t h e b i j e c t i v e
map i : M-*L, we o b t a i n a space s t r u c t u r e on M, which has t h e f o l l o w i n g
properties according t o d ) : Every M Q i s a closed semialgebraic subset
of M and t h e subspace s t r u c t u r e on M a c o i n c i d e s w i t h t h e subspace
structure of M with respect t o M i n t h e o l d s t r u c t u r e . The f a m i l y

a
(N(a) |a€I) is locally finite i n N. Thus (N(a)DLIa€I) i s l o c a l l y finite
i n L, and (M laGI) i s l o c a l l y
a finite i n M with respect t o t h e new
space s t r u c t u r e . We c o n c l u d e from Lemma 1.2 t h a t t h e new space s t r u c -
t u r e on M c o i n c i d e s w i t h t h e o l d one. T h i s means t h a t i i s i n d e e d an
embedding o f M i n t o N. The p r o o f o f Theorem 2.1 i s f i n i s h e d .
Our next g o a l i n t h i s c h a p t e r i s t o prove t h a t e v e r y r e g u l a r paracom-
pact space M can be t r i a n g u l a t e d , more p r e c i s e l y , t h a t t h e r e exists a
strictly locally finite simplicial complex X and a l o c a l l y semialgebraic
isomorphism \p : X-^M. Notice t h a t such a t r i a n g u l a t i o n automatically
gives a c o m p l e t i o n o f M - namely t h e embedding o f M i n t o X by t h e map

1
\p . Thus t r i a n g u l a t i o n s a r e a s p e c i a l s o r t o f c o m p l e t i o n s . Theorem 2.1
should be r e g a r d e d as a f i r s t step towards t h e t r i a n g u l a t i o n theorem.

§3 - Abstract simplicial complexes
B e f o r e we s t a r t our p r o o f t h a t r e g u l a r paracompact spaces can be tri-
angulated i t seems wise to d e v e l o p a c o m b i n a t o r i a l p a t t e r n f o r des-
cribing the s i m p l i c i a l complexes over R i n t r o d u c e d i n I, §2. From now
on t h e s e complexes a r e c a l l e d geometric simplicial complexes whereas
the c o m b i n a t o r i a l o b j e c t s which c o r r e s p o n d t o them w i l l be c a l l e d ab-
stract simplicial complexes. A b s t r a c t s i m p l i c i a l complexes are w i d e l y
used i n c l a s s i c a l t o p o l o g y . N o t i c e , however, t h a t our n o t i o n o f a geo-
metric s i m p l i c i a l complex i s more g e n e r a l than the c l a s s i c a l one (even
for R = 3R ) , and t h e r e f o r e our n o t i o n o f an a b s t r a c t s i m p l i c i a l complex
a l s o has to be more g e n e r a l than the c l a s s i c a l one. In our t h e o r y the
s i m p l i c e s o f an a b s t r a c t s i m p l i c i a l complex (see below) r e p r e s e n t the
open s i m p l i c e s o f the c o r r e s p o n d i n g geometric simplicial complex,
w h i l e i n t h e c l a s s i c a l t h e o r y they r e p r e s e n t the c l o s e d s i m p l i c e s .
T h i s d e v i a t i o n from the c l a s s i c a l concepts seems t o be inavoidable in
s e m i a l g e b r a i c geometry, c f . t h e f i r s t remark f o l l o w i n g D e f i n i t i o n 1
i n I, §2.
Definition 1. a) An abstract simplicial complex K i s a p a i r (E(K),S(K))
c o n s i s t i n g o f a s e t E(K) and a s e t S(K) o f non empty f i n i t e subsets

)
of E(K) such t h a t E(K) i s the union of a l l s € S ( K ) * . We call the ele-
ments of E(K) the v e r t i c e s o f K and the elements of S(K) the simpli-
ces o f K. We do admit the empty complex 0 = (0,0).
b) The c l o s u r e o f an a b s t r a c t s i m p l i c i a l complex K i s the p a i r K =
(E(K),S(K)) w i t h E(K) = E(K) and S(K) the s e t o f a l l non empty subsets
t of a l l s € S(K) . The complex K i s c a l l e d closed i f K = K.
c) A s i m p l i c i a l map a : K -* L between a b s t r a c t s i m p l i c i a l complexes K,L
i s a map a : E (K) ->E(L) such t h a t the image a(s) o f every simplex s of
*) Thus the datum E(K) i s determined by S (K), but i t seems c o n v e n i e n t

to m a i n t a i n t h i s redundancy.
K i s a simplex o f L.
d) A subcomplex o f an a b s t r a c t simplicial complex K i s an a b s t r a c t
simplicial complex L w i t h E (L) c E ( K ) and S (L) c S ( K ) . Note t h a t the
c l o s u r e L o f L i s a subcomplex o f K. For any two subcomplexes and
L 2 o f K we can form the i n t e r s e c t i o n nL 2 and the union L-j U L . 2 The
subcomplex f l L ^ i s determined by S(L^ f l L ) 2 = S (L^) D S ( L ) 2 and the
subcomplex UL 2 i s determined by SCL^ U L ) 2 = S (L^) U S ( L ) . We

2 have
E(L 1 UL )2 = E (L«j ) U E ( L ) b u t c o u l d have E ( L fl L ) c E ( L ) fl E ( L ) .

2 1 2 1 2
Clearly f l L ^ i s a subcomplex o f and L 2 and " c o n t a i n s " ( i . e . has
as subcomplex) every subcomplex o f K c o n t a i n e d in and L . S i m i l a r -

2
ly UL 2 i s t h e s m a l l e s t subcomplex o f K which c o n t a i n s and L . 2
More g e n e r a l l y , we may form the i n t e r s e c t i o n and u n i o n o f a r b i t r a r y
f a m i l i e s o f subcomplexes o f K.
e) The c l o s u r e o f a subcomplex L o f K i n K i s the i n t e r s e c t i o n L fl K
o f the subcomplexes L and K o f K. In p a r t i c u l a r , L i s c a l l e d closed
in K i f L = L fl K.
f) A subcomplex L o f K i s c a l l e d full i n K, i f
S (L) = ( s e S ( K ) l s c E(L) }
(In general S(L) may be s m a l l e r than the r i g h t hand s e t . ) Of c o u r s e
every f u l l subcomplex o f K i s c l o s e d i n K.
g) The complement K^L o f a subcomplex L i n K i s the subcomplex N
of K w i t h S(N) = S(K) ^ S ( L ) . The subcomplex L i s c a l l e d open i n K,
if K^L i s c l o s e d i n K.
For the sake o f b r e v i t y we w i l l use the word "complex" i n s t e a d o f
"simplicial complex" i n t h e remainder o f t h i s s e c t i o n as w e l l as i n
the whole e n t i r e t y o f the next s e c t i o n . N o t i c e t h a t f o r every simplex
s o f an a b s t r a c t complex K the p a i r (s,{s}) i s a subcomplex o f K w i t h
a s i n g l e simplex s. We u s u a l l y i d e n t i f y s w i t h t h i s subcomplex o f K.
The f a c e s o f s are the s i m p l i c e s o f the c l o s u r e s o f the complex s.

These a r e a l l t h e non-empty s u b s e t s t o f t h e s e t s. We o f t e n write
t<s i n s t e a d o f t c s i f t i s a f a c e o f s, and we w r i t e t < s i f t c s
but t * s ("proper f a c e s " o f s) . The s i m p l i c e s o f t h e complex sflK
are c a l l e d t h e " f a c e s o f s i n K".
C l e a r l y a subcomplex L o f K i s c l o s e d i n K i f f o r every simplex s o f
L a l l f a c e s o f s i n K a r e s i m p l i c e s o f L. Thus a subcomplex L o f K i s
open i n K i f every simplex s o f K p o s e s s i n g a t l e a s t one f a c e which i s
a simplex o f L i s i t s e l f a simplex o f L.
A s i m p l i c i a l map a : K -> L between a b s t r a c t complexes K, L can a l s o be
r e g a r d e d as a s i m p l i c i a l map from K t o L. T h i s w i l l be denoted by a
and called the c l o s u r e o f t h e s i m p l i c i a l map a. I f a : K -> L i s a sim-
plicial map and i s a subcomplex o f K, we have the image complex
a(K ) = 1 d e f i n e d by
S(L ) 1 = {a(s) |s € S ( K ) } .
1
Note t h a t E(L^) = a(E(K«j)) . Furthermore, we have f o r every subcomplex

-1
1 1
L o f L the preimage complex a ( L ) = K' d e f i n e d by
S(K') = ( s e S(K) |a(s) € S ( L ' ) } .
1 f
The s e t E(K') may be s m a l l e r than a ~ ( E ( L ) ) . We have a ( K ) = a(K^) 1
1 1 7
but o n l y a"" (L' ) c a ~ (L *) .
Now l e t X = (X,I(X)) be a g e o m e t r i c complex o v e r o u r f i x e d r e a l closed
field R ( c f . I , § 2 ) . X y i e l d s an a b s t r a c t complex K(X) = (E(X),S(X))
as f o l l o w s : E(X) i s t h e s e t o f a l l p o i n t s i n X which a r e v e r t i c e s o f
open s i m p l i c e s a of X (i.e. a€X(X)). In p a r t i c u l a r E(X) = E (X) . A
subset s o f E(X) i s an element o f S(X) i f and o n l y i f s i s the s e t o f
v e r t i c e s o f some a € I ( X ) . We c a l l K(X) t h e a b s t r a c t i o n o f the geome-
tric complex X. C l e a r l y t h e c l o s u r e K(X) o f K(X) i s t h e a b s t r a c t i o n
of the c l o s u r e X o f X. A l s o , t h e subcomplexes Y o f X c o r r e s p o n d uniquely

t o t h e subcomplexes L o f K by a b s t r a c t i o n . Our d e f i n i t i o n above o f t h e
words " c l o s e d i n K", "union", " i n t e r s e c t i o n " , "complement", "open i n
K", f o r subcomplexes o f K a r e c o m p a t i b l e w i t h t h e d e f i n i t i o n s o f t h e s e
words f o r subcomplexes o f g e o m e t r i c complexes i n I, § 2 . F o r example,
a subcomplex Y o f X i s open i n X i f and o n l y i f K(Y) i s open i n K ( X ) .
We complete t h i s p a t t e r n o f d e f i n i t i o n s as f o l l o w s .
Definitions 2. a) A subcomplex Y o f a g e o m e t r i c complex X i s c a l l e d
full i n X, i f K(Y) i s f u l l i n K(X). T h i s means t h a t every open simplex
a o f X whose v e r t i c e s a r e a l l c o n t a i n e d i n Y i s an open simplex o f Y.
b) An a b s t r a c t complex K i s c a l l e d finite i f E(K) i s f i n i t e o r , e q u i -
valently, i f S(K) i s f i n i t e . K i s called locally finite i f every sim-
plex s o f K i s a simplex o f an open f i n i t e subcomplex o f K. T h i s means
t h a t every simplex o f K i s a f a c e o f o n l y f i n i t e l y many s i m p l i c e s o f
K (or o f K ) . F i n a l l y K i s c a l l e d s t r i c t l y locally finite i fK i s l o -
cally f i n i t e . T h i s means t h a t every v e r t e x p o f K i s an element o f
only f i n i t e l y many s i m p l i c e s o f K (or o f K ) .
c) The dimension d i m s o f an a b s t r a c t simplex s i s by d e f i n i t i o n t h e
c a r d i n a l i t y o f s minus one, and t h e dimension dimK o f an a b s t r a c t com-
p l e x K i s the supremum o f the dimensions o f i t s s i m p l i c e s (a non nega-
t i v e i n t e g e r o r «>) . S i m i l a r l y , the dimension dim X o f a g e o m e t r i c com-
plex X i s defined as the supremum o f the a f f i n e dimensions o f i t s open
s i m p l i c e s . The empty complex has dimension -1. N o t i c e that forX l o -
cally finite t h i s dimension c o i n c i d e s w i t h the dimension o f X as a
l o c a l l y semialgebraic space.
Any s i m p l i c i a l map f : X->Y between g e o m e t r i c complexes (cf. Definition
6 i n I, §2) o b v i o u s l y y i e l d s a s i m p l i c i a l map K ( f ) : K(X) -* K(Y) between
the a b s t r a c t i o n s o f X and Y by r e s t r i c t i o n t o t h e s e t o f v e r t i c e s . We
c a l l K(f) the a b s t r a c t i o n o f t h e s i m p l i c i a l map f . The c l o s u r e f : X -* Y
o f f has as a b s t r a c t i o n t h e c l o s u r e o f K ( f ) , i . e . K ( f ) = K ( f ) .
We have e s t a b l i s h e d "the a b s t r a c t i o n f u n c t o r " K from t h e c a t e g o r y o f
g e o m e t r i c complexes and g e o m e t r i c s i m p l i c i a l maps t o t h e c a t e g o r y o f
a b s t r a c t complexes and a b s t r a c t s i m p l i c i a l maps. The f o l l o w i n g propo-
sition i s easily verified.
Proposition 3.1. The a b s t r a c t i o n f u n c t o r K is fully faithful, i.e. for
any two g e o m e t r i c complexes X and Y, t h e n a t u r a l map f»-»K(f) from t h e
set o f morphisms Hom(X,Y) t o Hom(K(X),K(Y)) i s b i j e c t i v e .
Definition 3. A r e a l i z a t i o n o v e r R o f an a b s t r a c t complex i s a geome-
tric complex X over R together with a s i m p l i c i a l isomorphism a : L - ^ K ( X ) .
Proposition 3.2. Every a b s t r a c t complex L admits a r e a l i z a t i o n o v e r R.
Indeed, we o b t a i n a "canonical realization" o f L i n t h e same way as i n

( e ( l ) )
the c l a s s i c a l t h e o r y [Go, p. 3 9 f ] . L e t V = R be t h e f r e e R-module
w i t h b a s i s E ( L ) , and l e t X denote t h e s e t o f a l l open s i m p l i c e s i n V,
which a r e spanned by t h e f i n i t e s e t s s c E ( L ) , s G S ( L ) . Then X i s a
g e o m e t r i c complex o v e r R and t h e i d e n t i t y map from E(L) t o E(X) = E ( L )
i s an a b s t r a c t simplicial isomorphism a : L ^ * K ( X ) .
Notice that i fX i s a realization o f L w i t h t h e isomorphism a : L - ^ K ( X )
then X i s a r e a l i z a t i o n o f L w i t h t h e isomorphism oc from L t o K(X) =
K(X). I n t h e case R = 3R t h e g e o m e t r i c complex X may be r e g a r d e d as a
realization o f L i n t h e c l a s s i c a l sense.
If X^ and X 2 a r e two r e a l i z a t i o n s o f L w i t h isomorphisms : L-^>K(X,j)
and a 2 : L - ^ K ( X ) , then by P r o p o s i t i o n
2 3.1 t h e r e exists a unique sim-
plicial isomorphism f : —> X 2 such t h a t a = K ( f ) o o ^ .

2 In t h i s sense
any two r e a l i z a t i o n s o f L a r e c a n o n i c a l l y i s o m o r p h i c . We u s u a l l y choose
for a given a b s t r a c t complex L a fixed realization X over R and denote

this g e o m e t r i c complex by I L | _ (or s i m p l y | L | , i f the field R i s kept
fixed). U s u a l l y we also identify E ( L ) w i t h the s e t E ( X ) by the reali-
z a t i o n isomorphism a : L - ^ » K ( X ) . We call ILI R with t h i s identification
"the" realization o f L o v e r R. Of course, if i s a subcomplex o f L
then we u s u a l l y choose the subcomplex Y o f X w i t h S ( Y ) = S ( L ) f o r 1
the realization of o v e r R. The isomorphism : -^K(Y) i s obtain
from a by r e s t r i c t i o n . In t h i s sense we have f o r m u l a s like
| L l nL i 2 R = | L l i R n IL I 2 R /
ILT^ = I L I R
ILx L l | R = |L| v| R L l l R .
In p a r t i c u l a r , f o r every s € S ( L ) we denote by IsI R the open simplex a
of I L I ^ whose s e t o f v e r t i c e s i s s.
L
Definition 4. The realization o f an abstract s i m p l i c i a l map y : •* 2
L W T 1
o v e r R i s the unique g e o m e t r i c s i m p l i c i a l map g : |LÎ I R 2 ^ R ^- ^
K(g) = y. We denote t h i s map g by l y l ^ .
P r o p o s i t i o n 3.1 and 3.2 together mean t h a t the f u n c t o r K i s an equi-
valence o f the category o f g e o m e t r i c complexes w i t h the category of
a b s t r a c t complexes. We have j u s t constructed a "quasi-inverse" func-
tor to K, the realization functor || , R q u i t e i n the usual way.
The "canonical realization" o f an a b s t r a c t complex L , which we descri
bed i n the p r o o f o f P r o p o s i t i o n 3.2, i s u s u a l l y a g e o m e t r i c complex
i n a very large vector space. One can o f t e n do better.
Proposition 3.3. I f L i s a countable s t r i c t l y locally finite complex
with dimL < n, then L has a realization as a g e o m e t r i c complex X i n

N +
the vector space R ^ ^ such t h a t X i s a c l o s e d l o c a l l y semialgebraic
2 N + 1
s u b s e t o f the l o c a l l y semialgebraic space (R )
n+
Proof. I t s u f f i c e s to realize L i n R^ ^. T h i s can be done by t h e c l a s -
sical c o n s t r u c t i o n as d e s c r i b e d , f o r example, i n [Sp, Chap. I l l , § 2 ] .
Then X becomes even a c l o s e d l o c a l l y s e m i a l g e b r a i c subset o f M ^ q c for

2 n +
M a closed half-space i n R \ e.g. M = ( x > 0 ) .
n
We f i n a l l y d i s c u s s e x t e n s i o n o f t h e base f i e l d . L e t R be a r e a l closed
overfield o f R. L e t X be a r e a l i z a t i o n o f an a b s t r a c t complex L over R
with t h e isomorphism a : L - ^ K ( X ) . Then t h e geometric complex X(R) over
R, o b t a i n e d from X by base e x t e n s i o n i n t h e obvious way, has t h e same
a b s t r a c t i o n as X, i . e . K(X(R)) = K(X), and thus w i l l be regarded as
realization of L over R. I n s h o r t ,
(3.4) I L | ( R ) = |L|~.
R
L w e t n er e a
A l s o , f o r every a b s t r a c t s i m p l i c i a l map Y : L^ 2 ^âin ~
lization i T l g from IYI r i n an obvious way by base e x t e n s i o n ,
(3.5) | T I = S (lYl )g-

R
If L i s l o c a l l y f i n i t e , then (3.4) can a l s o be r e a d as a c a n o n i c a l
isomorphism between t h e base e x t e n s i o n o f t h e l o c a l l y semialgebraic
space iLl R t o R and t h e l o c a l l y s e m i a l g e b r a i c space lL|g. Similarly,
i f L^ and L^ a r e l o c a l l y f i n i t e , then (3.5) can be r e a d as an e q u a l i t y
between l o c a l l y s e m i a l g e b r a i c maps.
§4 - T r i a n g u l a t i o n o f r e g u l a r paracompact spaces
Definition 1. a) L e t M be a l o c a l l y semialgebraic space over R. A tri-
angulation of M is a locally semialgebraic isomorphism cp : X M from a
strictly locally f i n i t e geometric (simplicial) complex X onto M. (Recall
from I, §2 t h a t X has a n a t u r a l s t r u c t u r e as a l o c a l l y s e m i a l g e b r a i c space.)
b) Let (AÎXEA) be a f a m i l y of l o c a l l y semialgebraic subsets o f M. A
s i m u l t a n e o u s t r i a n g u l a t i o n o f M and the family (A^|A€A) is a triangula-
t i o n cp : X —> M such t h a t cp" (A-^) 1

i s a subcomplex o f X f o r every A€A.
Then, o f c o u r s e , the r e s t r i c t i o n cp| :Y^ A^ is a triangulation of
A^ f o r every A € A.
c) I f cp : X M and \\) : Y M a r e t r i a n g u l a t i o n s o f M, we say t h a t \p r e -
f i n e s cp, and w r i t e cp<^, if i s a simultaneous t r i a n g u l a t i o n of M and
the f a m i l y (cp(a) |a€E(X)) c o n s i s t i n g of the images o f a l l open s i m p l i c e s o f X.
d) We say t h a t two t r i a n g u l a t i o n s cp : X M and : Y ^* M are equivalent
if \\) ^*cp i s a s i m p l i c i a l isomorphism from X t o Y. (N.B. This i s a
stronger c o n d i t i o n than cp<^, ip<cp.)
e) I f two t r i a n g u l a t i o n s cp : X -> M 1 and \p : Y -> M 2 of l o c a l l y semialge-
b r a i c subsets , M 2 o f M a r e g i v e n , and i f A is a locally semialgebraic
subset o f M.j 0M , 2 then we say t h a t cp r e f i n e s on A ( r e s p . cp i s e q u i v a -

-1 -1
l e n t t o \p on A) , i f cp (A) and (A) a r e subcomplexes o f X and Y and
1
the r e s t r i c t i o n cp (A) A o f cp r e f i n e s ( r e s p . i s e q u i v a l e n t to) the
-1 ~
restriction (A) —>A of
Of course, i f M has a triangulation, then M must be r e g u l a r and para-
compact. Our g o a l i n t h i s s e c t i o n i s t o prove the c o n v e r s e : Any regu-
l a r and paracompact space M and any locally finite family (A^|A€A) of
locally semialgebraic subsets o f M has a simultaneous triangulation.
T h i s i s well-known t o be true i f M i s a f f i n e semialgebraic ( c f . I,
Th. 2.13). The general case needs some p r e p a r a t i o n . N o t i c e t h a t i t s u f f i

ces t o prove t h e theorem i n t h e case t h a t t h e r e g u l a r paracompact
space i s p a r t i a l l y complete, s i n c e by Theorem 2.1 any o t h e r regular
paracompact space may be c o n s i d e r e d as a l o c a l l y semialgebraic subset
o f such a space.
Lemma 4.1. ( G l u i n g o f t r i a n g u l a t i o n s f o r p a r t i a l l y complete spaces).
L e t M be a p a r t i a l l y complete space over R, and l e t (M^laEI) be a l o -
cally finite family i n T(M) w i t h M = i H M ^ j a E I ) . Assume t h a t f o r every
a € I a t r i a n g u l a t i o n cp^ : X a —> M a i s given. (Thus, by Lemmas 1.1 and
1.2, M i s regular and paracompact.) Assume f u r t h e r t h a t f o r any two
i n d i c e s a, 3 6 I w i t h M Q n M^* 0 the t r i a n g u l a t i o n s cp^ and cp^ a r e e q u i -
v a l e n t on n . Then t h e r e e x i s t s , up t o e q u i v a l e n c e , a unique tri-
angulation cp : X M such t h a t co i s e q u i v a l e n t t o cp^ on f o r every
a € I.
Proof. We prove t h e e x i s t e n c e o f cp, l e a v i n g t h e easy v e r i f i c a t i o n o f
uniqueness t o t h e r e a d e r . F o r ever^j a € I t h e complex X a i s closed

E S
s i n c e the space M Q i s p a r t i a l l y complete. L e t K Q = ( a f a ) denote the
a b s t r a c t i o n o f X^. We want t o c o n s t r u c t a closed abstract complex
K = (E,S) which i s i n some sense t h e "union" o f the K . We d e f i n e t h e
s e t E o f v e r t i c e s as the q u o t i e n t o f t h e d i s j o i n t u n i o n U ( E l a € I ) by a
the following equivalence r e l a t i o n : L e t p 6 E , q € E ^ . Then p ~ q i f f

Q
cp (p)
Q = tpp (q) . (N.B. T h i s makes sense s i n c e every complex X Q is
c l o s e d . T r a n s i t i v i t y o f t h e r e l a t i o n i s e a s i l y checked.) F o r every
a € I we have a n a t u r a l injection i a : E -> E. We d e f i n e S as the union

Q
o f the s e t s {i (s)ls€S }
a a f o r a l l a E I . C l e a r l y K := (E,S) i s a c l o s e d
a b s t r a c t complex. We c l a i m t h a t K i s l o c a l l y f i n i t e . L e t p € E be g i v e n .
We have t o v e r i f y that p i s a vertex of only f i n i t e l y many s € S .
Choose an index a € I such t h a t p = i ( p ) a a f o r some p Q 6E , Q uniquely
determined by p. L e t s 6 S be g i v e n with. p € s . We have s = i ^ ( s ^ ) f o r
some index 3 6 1 and some s E S . C l e a r l y cp (p ) € M . S i n c e the f a m i l y

(MÎT€I) is locally f i n i t e , o n l y f i n i t e l y many i n d i c e s 3 can o c c u r i n
t h i s manner. A l s o , <P (P ) = ^ 3 ^ 3 ) w i t h
a a € u n i q u e l y determined by
p . We
Q have p^€ s^ f o r every simplex s^ as above. Thus, f o r 3 f i x e d ,
o n l y f i n i t e l y many s i m p l i c e s s^ can o c c u r . We see t h a t K i s indeed l o -
cally finite.
We choose a r e a l i z a t i o n X = i K l o f K. For every a € I the injection

E E m
i : a ^ ay be r e g a r d e d as an a b s t r a c t s i m p l i c i a l map from to K.
The realization I i ^ I : X_ ->X of i i s an isomorphism from X onto a

ex a a a
c l o s e d subcomplex o f X. A l s o , the f a m i l y (Y^laGI) o f subcomplexes
of X i s l o c a l l y f i n i t e and has X as i t s u n i o n . For e v e r y a € I we have
a locally semialgebraic isomorphism
S i n c e the t r i a n g u l a t i o n s cp^ and cp^ a r e e q u i v a l e n t on M^flM^, we have
^ (x)
a = ipp(x) f o r every x n ^ Y ^ . Thus the ^ a glue together to a l o c a l l y
s e m i a l g e b r a i c isomorphism : X-^M. Of c o u r s e \p i s e q u i v a l e n t t o cp^ on
M Q f o r every a 6 I. q.e.d.
As the p r o o f shows, Lemma 4.1 is a trivial statement. We d i d not do
any s e r i o u s work. N e v e r t h e l e s s the lemma l e a d s t o a sometimes v e r y
u s e f u l way t o work w i t h triangulations.
Definition 2. For any non n e g a t i v e i n t e g e r n we denote by A n (or more
p r e c i s e l y by A (R) R i f necessary) the c l o s e d s t a n d a r d n-simplex over R,

e n + 1
i . e . A (R) n = Te ,e^,...,e ] with
Q n 0 '*'-'e n the s t a n d a r d b a s i s o f R .
A simplicial a t l a s of a l o c a l l y s e m i a l g e b r a i c space M i s a f a m i l y of
s e m i a l g e b r a i c maps (<p : A ( ) -*M|a€I) w i t h the f o l l o w i n g

a n a five proper-
ties :
a) Every co J i s i n j e c t i v e and hence an isomorphism

J from A , onto a
x
^ n (a)
c l o s e d semialgebraic subset M of M.
b) The f a m i l y (MglaGl) i s locally finite and c o v e r s M.
_i
c) F o r any two i n d i c e s a, 3 € I w i t h M D * 0 t h e s e t s <P (M fl M^) a a a
_i
and cp (M fl M„) a r e ( c l o s e d ) f a c e s o f A , . and A . respectively
3 3 n(3)
Q / n
a n(a)
and
:
â" 1 ( M
a n M
3 }
" ^ (M
« 0
V
is an a f f i n e isomorphism.
d) M a = M 3 a = 3.
A w e
e) F o r e v e r y a € I and every c l o s e d f a c e T o f n ( j a have ^ ( t ) =
for some Y G I .
Notice that i n t h i s d e f i n i t i o n the l a s t two c o n d i t i o n s d ) , e) can a l -
ways be met i f a) - c) a r e f u l f i l l e d by s i m p l y o m i t t i n g o r adding suit-
a b l e maps cp : ^ ( ) ~* * N o t i c e a l s o t h a t , a c c o r d i n g t o § 1 , the space

a n a
M
s t r u c t u r e on t h e s e t M i s u n i q u e l y determined by a g i v e n a t l a s , and
t h a t M i s n e c e s s a r i l y r e g u l a r paracompact and p a r t i a l l y complete.
Definition 3. Two s i m p l i c i a l atlases

(cp I aGI) and (ty, | AGA) o f M a r e
a A
called e q u i v a l e n t i f t h e r e e x i s t s a b i j e c t i o n K : I —>A such t h a t , f o r
every a G I , cp ( A ( a R a ) ) =
^ K ( a ) ( A
n ( K ( a ) ) ) ' i n
p a r t i c u l a r n(a) = n (K (a) ) ,
-1
and \b . . o cp i s an a f f i n e automorphism o f A , \ .
K (a) a * n (a)
T h i s means t h a t t h e index s e t s I and A may be i d e n t i f i e d i n such a way
t h a t every i s g o t t e n from t h e c o r r e s p o n d i n g cp^ by c o m p o s i t i o n with
an a f f i n e automorphism o f A , \ *
* n (a)
I f a t r i a n g u l a t i o n cp : X M i s given with X a closed l o c a l l y finite
geometric complex, we then o b t a i n a s i m p l i c i a l atlas (cp^: A ^->M | aGI (X))

n
by choosing f o r every a G I (X) an a f f i n e isomorphism \b : A , -^*a, x
a n (o)
n(a) := dim a , and s e t t i n g cp^ = cp«i/^. T h i s s i m p l i c i a l atlas i s deter-
mined by t h e t r i a n g u l a t i o n cp up t o e q u i v a l e n c e , and i s c a l l e d an a t l a s
of t h e t r i a n g u l a t i o n cp.
Conversely, i f a s i m p l i c i a l a t l a s (cp^laGI) o f M i s g i v e n then, apply-
i n g Lemma 4.1 t o M and t h e t r i a n g u l a t i o n s cp^ : A

n ( a ) —• c p ^ A ^ ^ ) , we
obtain
Corollary 4.2. Every s i m p l i c i a l a t l a s (cp lotei) o f a l o c a l l y

a semialge-
b r a i c space M i s e q u i v a l e n t t o an a t l a s o f some t r i a n g u l a t i o n cp : X -»M
and cp i s u n i q u e l y determined, up t o e q u i v a l e n c e , by t h e g i v e n a t l a s .
Thus, f o r p a r t i a l l y complete spaces, s i m p l i c i a l a t l a s e s and t r i a n g u l a -
t i o n s a r e r e a l l y t h e same t h i n g . We s h a l l s w i t c h back and f o r t h between
t h e s e concepts a t w i l l .
We now s t a t e a t e c h n i c a l lemma which w i l l h e l p us t o t r i a n g u l a t e regular
paracompact spaces.
Lemma 4. 3 ( E x t e n s i o n o f a r e f i n e m e n t o f a t r i a n g u l a t i o n ) . L e t cp : X M
be a t r i a n g u l a t i o n o f a p a r t i a l l y complete space M. F u r t h e r , l e t M q be
a closed locally s e m i a l g e b r a i c subset o f M with X Q : = ip ( M ) a (neces-

1
Q
s a r i l y closed) subcomplex o f X, and l e t cp : X M denote t h e t r i a n -

^ ô o o
gulation of M Q o b t a i n e d from cp by r e s t r i c t i o n . F i n a l l y l e t I/J : Q
Y
0
M
0
be a t r i a n g u l a t i o n o f M which r e f i n e s cp . Then t h e r e e x i s t s a t r i a n g u -
^ o ô
lation \p : Y -Û M which r e f i n e s cp and i s e q u i v a l e n t t o on M . Q
P r o o f . We assume w i t h o u t l o s s o f g e n e r a l i t y that X = M, cp = i c l ^ . Thus
M i s a closed locally finite s i m p l i c i a l complex and M Q i s a c l o s e d sub-
complex o f M. F o r any k > 0 we denote by M the r e l a t i v e k-skeleton of
(M,M ) i . e . t h e u n i o n o f M and a l l open (or c l o s e d ) s i m p l i c e s of M of

o o
dimension <k. L e t ( x l ^ a
a J
0 ) ^ e a
s i m p l i c i a l atlas of the triangulation
Y M M
ty : 0 0 ~~* Q °f Q • We want t o extend t h i s a t l a s s u c c e s s i v e l y to simpli-
cial atlases (x la€J )
a
K
for a l l skeleta M k
(k>0, J Q C J°c j 1
c . ..) such
that the r e s u l t i n g t r i a n g u l a t i o n of M r e f i n e s the t a u t o l o g i c a l t r i -

k ~ k k
angulation i d : M —>M f o r every k > 0 . Then (x laGU(J
a lk>0)) becomes
a simplicial a t l a s o f M, which r e f i n e s the t a u t o l o g i c a l triangulation
id : M ^ M.
The atlas (Y la€J°) o f M ° is easily u found. M ° i s the u n i o n o f M and

a o
all v e r t i c e s o f M. Simply add the one p o i n t map x p

:
A Q {e} t o the
family (x l a
a € J
0 ) f o r
every e € E ( M ) ^ E ( M ).
Suppose the f a m i l y ( x l a € J ) has

a a l r e a d y been c o n s t r u c t e d f o r some
k>0. Let denote the s e t o f a l l open s i m p l i c e s i n M ^ M Q o f dimen-

— k
s i o n k + 1 . F o r e v e r y a G E^+i boundary do = o ^ o i s c o n t a i n e d i n M .
k k
Let J ( a ) denote the s e t of a l l i n d i c e s a € J with
A
a a n(a) '
and l e t 6 denote t h e b a r y c e n t e r o f a . F o r every p a i r (a,a) with
a € 21^+1 / a G J ( a ) , we i n t r o d u c e the i n j e c t i v e semialgebraic map
x : A M
o,a n(a)+1
defined by
X , (d-t)u
a a + te n ( 0 ) + 1 ) = (1-t) X a (u) + t a .
A a s t n e A = e e
In o t h e r words, we regard n ( ) -| a + cone over n ( )
a ^ '**"' n(a
0
with vertex e , » and extend the isomorphism Y A

: A , * -^>M i n the
n(a)+1 ^ a n(a) a
obvious way 1
to an isomorphism from A , * ,- onto the cone over M with
* n(a)+1 a
k+1
v e r t e x 6. As a simplicial a t l a s of M we choose the u n i o n o f the
three f a m i l i e s (x laGJ ),
a
k
(A ^ o {6} I a G l k + 1 ), ( x a Ia G I k + 1 ,a G J k
(a) ) .
It is easily checked t h a t t h i s union i s indeed a s i m p l i c i a l atlas of
M q.e.d.
Now we are ready to prove our main theorem.

Theorem 4.4 ( T r i a n g u l a t i o n theorem). L e t M be a r e g u l a r paracompact
space over R and ( A ^ I A G A ) a locally finite family of l o c a l l y semial-
gebraic s u b s e t s o f M. Then t h e r e e x i s t s a simultaneous t r i a n g u l a t i o n
( c f . Def. 1) ip : X ^ M o f M and t h i s family.
Proof. Choosing a c o m p l e t i o n o f M (Th. 2.1) we may assume i n a d d i t i o n
that M i s p a r t i a l l y complete. We may a l s o assume t h a t M i s c o n n e c t e d .
We a l r e a d y know t h a t t h e theorem i s t r u e i f M i s semialgebraic ( I , Th.
2.13). In t h e g e n e r a l case we can c o v e r M by a l o c a l l y finite family
(U Inew) o f open s e m i a l g e b r a i c
3
s u b s e t s w i t h U f)U = 0 i f In-ml > 2,
n ^ n m —
cf. I, C o r . 4.19. We s h r i n k this covering to a covering (V | n € U ) with
V €T(M), V c U ,
n n n c f . I , Th. 4.11. We s h a l l work w i t h t h e l o c a l l y
f i n i t e covering (M |n£ E ) o f M by t h e c l o s e d s e m i a l g e b r a i c subsets
M := V . N o t i c e M DM = 0 i f I n-m I > 2. A l s o , f o r every n € U , the

n n n m —
index s e t A (n) c o n s i s t i n g o f a l l A € A w i t h A, PI M * 0 i s f i n i t e .
A n
U s i n g t h e t r i a n g u l a t i o n theorem f o r a f f i n e semialgebraic spaces we
s u c c e s s i v e l y choose f o r n = 1,2,3,... s i m u l t a n e o u s t r i a n g u l a t i o n s
cp : n — o f and t h e f i n i t e l y many s u b s e t s nM , Mn n n -| r
M
n +
M
n n A
A ( A e
^ ( n ) ) with the a d d i t i o n a l property t h a t <P n+1 r e f i n e s cp n
on M n D M , j . (Read M
n+ Q = 0) . Then u s i n g Lemma 4 .3 we r e f i n e cp t o a n
triangulation y\b : Y M such t h a t \b y i s equivalent t o cp on M DM i

* n n n n ^ n n-1 n
M
and equivalent t o <P i on M
n+ n fl M ^ . (Notice that M ^nM n and D n + 1
are d i s j o i n t closed semialgebraic s u b s e t s o f M^.) Now ^ - | n+ i s equi-

v a l e n t t o ill on M , „ n M . By Lemma 4.1 t h e \b can be g l u e d to a t r i -
y 2 y
n n+1 n n ^
angulation : Y-^»M. By c o n s t r u c t i o n \p ^ (A-^) i s a subcomplex o f Y f o r
every A € A. q.e.d.
§5 - T r i a n g u l a t i o n o f weakly s i m p l i c i a l maps, maximal complexes
Def i n i t i o n 1. L e t f : M -> N be a l o c a l l y s e m i a l g e b r a i c map.
a) A t r i a n g u l a t i o n o f f i s a commutative square
(1)
w i t h X and Y s t r i c t l y locally f i n i t e geometric (simplicial) complexes,
cp and ^ l o c a l l y s e m i a l g e b r a i c isomorphisms, and g a s i m p l i c i a l map.
b) A c o m p l e t i o n o f f i s a commutative square
M* - * M
(2)
w i t h i and j c o m p l e t i o n s o f t h e spaces M,N r e s p e c t i v e l y ( i . e . dense
embeddings i n t o p a r t i a l l y complete spaces M,N, c f . § 2 ) , and f a l o -
cally s e m i a l g e b r a i c map.
N o t i c e t h a t the map f i s p a r t i a l l y p r o p e r . N o t i c e a l s o t h a t every tri-
angulation (1) o f f y i e l d s a c o m p l e t i o n o f f , namely
—1 -1
w i t h i and j d e n o t i n g t h e embeddings v i a cp and ^ i n t o the c l o s u r e s
X,Y o f t h e complexes X, Y and g the c l o s u r e o f the s i m p l i c i a l map g.
We have t h e f o l l o w i n g g e n e r a l f a c t about c o m p l e t i o n s o f maps.
Proposition 5.1. E v e r y l o c a l l y s e m i a l g e b r a i c map f : M -+N between r e g u -

l a r and paracompact spaces M,N can be completed. More p r e c i s e l y , given
a c o m p l e t i o n j : N*->Q, we can f i n d a c o m p l e t i o n (2) o f f which involves
j-
P r o o f . We choose some c o m p l e t i o n K : M«-»T. L e t T ( f ) a M x N be the graph
of f . We c o n s i d e r the n a t u r a l factorization
Hf)
of f v i a the graph. L e t P be the c l o s u r e o f the image o f T ( f ) i n T x Q
under the embedding K x j :MxN-»TxQ, and l e t i ^ : r ( f ) «-*P denote the
c o m p l e t i o n o f T(f) o b t a i n e d from K x j by r e s t r i c t i o n . Let finally
g : P -* Q be the r e s t r i c t i o n o f the n a t u r a l p r o j e c t i o n p r 2 : T x Q -* Q t o
P. We have a commutative diagram
TxQ
The c o m p l e t i o n i := i - * P o f M and the map g fulfill the r e q u i r e m e n t s
of the p r o p o s i t i o n . q.e.d.
The q u e s t i o n whether a g i v e n map can be t r i a n g u l a t e d - and how - is
much more d e l i c a t e . In the p r e s e n t s e c t i o n we d e a l w i t h an easy case,
namely a weakly s i m p l i c i a l map f : X-* Y between strictly locally finite
complexes X and Y ( c f . I , §2, Def. 5 ) . The problem h e r e i s , o f c o u r s e ,
t h a t f does not extend c o n t i n u o u s l y - hence s i m p l i c i a l l y - to a map
from X to Y. Our i d e a i s to t r i a n g u l a t e f by " m o d i f y i n g " the complexes
X and Y i n a somewhat c a n o n i c a l way, such t h a t X and Y remain the same
as l o c a l l y s e m i a l g e b r a i c spaces and f extends s i m p l i c i a l l y to the c l o -
sures of the m o d i f i e d complexes.

Our procedure w i l l be p u r e l y c o m b i n a t o r i a l . A l l c o n s t r u c t i o n s may be
carried o u t i n t h e c a t e g o r y o f s t r i c t l y l o c a l l y f i n i t e complexes w i t h
simplicial maps as morphisms. A c t u a l l y we s h a l l work i n t h e l a r g e r
category o f a l l geometric simplicial complexes over R w i t h t h e s i m p l i -
cial maps as morphisms. (So, i n g e n e r a l , weakly s i m p l i c i a l maps a r e
not r e g a r d e d as morphisms.) T h i s g e n e r a l i z a t i o n w i l l n o t cause any
a d d i t i o n a l d i f f i c u l t i e s . On t h e o t h e r hand, a geometric simplicial
complex X which i s not l o c a l l y f i n i t e can s t i l l be i n t e r p r e t e d geome-
trically as a subspace o f a "weak p o l y t o p e " , c f . end o f I I I , §6. Thus
our c o m b i n a t o r i a l c o n s i d e r a t i o n s s h o u l d be o f a d d i t i o n a l i n t e r e s t later
on.
When i t seems t o be c o n v e n i e n t we s w i t c h over from t h e c a t e g o r y o f
geometric simplicial complexes over R t o t h e c a t e g o r y o f a b s t r a c t
simplicial complexes, w i t h a b s t r a c t s i m p l i c i a l maps as morphisms. These
c a t e g o r i e s a r e e q u i v a l e n t v i a t h e a b s t r a c t i o n and t h e r e a l i z a t i o n func-
tor, c f . § 3 . Thus i t does n o t matter, i n which c a t e g o r y we work. In
particular, t h e base f i e l d R i s o f no importance. As b e f o r e , we speak
simply o f "complexes" i n s t e a d o f s i m p l i c i a l complexes.
In view o f t h e examples 6.15, 6.17 and p r o p o s i t i o n 6.16 i n Chapter I
it i s n a t u r a l to introduce the f o l l o w i n g terminology.
Definition 2. L e t f : X-*Y be a s i m p l i c i a l map between geometric com-
p l e x e s X and Y.
a) f i s c a l l e d p a r t i a l l y proper i f , f o r every open simplex a€I(X),
the f o l l o w i n g h o l d s : I f x i s an open f a c e o f a w i t h f ( t ) € I ( Y ) , then
x C I ( X ) . The map f i s c a l l e d proper i f , i n a d d i t i o n , f o r every p€I(Y)
there e x i s t o n l y f i n i t e l y many a € E ( X ) w i t h f ( a ) = p . I n o t h e r words
the preimage f 1
( p ) o f any p € I(Y) i s a f i n i t e subcomplex o f X.
b) f i s c a l l e d p a r t i a l l y f i n i t e i f f i s p a r t i a l l y proper and f i s i n -
jective on t h e s e t E(c) o f v e r t i c e s o f every a € I ( X ) . The map f i s

called f i n i t e i f , i n addition, t h e preimage f ( p ) o f every P € I(Y) i s
a f i n i t e complex.
C l e a r l y t h e c l o s u r e f : X-> Y o f any s i m p l i c i a l map f : X Y i s partially
p r o p e r . A l s o , a p a r t i a l l y proper map f : X-» Y i s f i n i t e i f and o n l y i f
all fibres o f f a r e f i n i t e s e t s . I f f i s p a r t i a l l y f i n i t e then a l s o f
is partially f i n i t e . But i t may happen t h a t f i s f i n i t e and f i s n o t
finite. F o r example, l e t X be t h e d i s j o i n t union o f c o u n t a b l y many
closed 1-simplices o fi ( n € H) , and l e t Y be t h e u n i o n o f c o u n t a b l y
many c l o s e d 1 - s i m p l i c e s x ( n € ] N ) , a l l meeting
n i n one common v e r t e x p.
We choose a v e r t e x q i n each a . L e t f be t h e s i m p l i c i a l mao from X

^n n ^
onto Y which maps every a onto T w i t h f ( a ) = p. L e t X denote t h e
c
n n *n
subcomplex X ^ { q ^ l n E E ) o f X and Y denote t h e subcomplex Y ^ {p} o f Y.
The r e s t r i c t i o n f : X -*Y o f f i s f i n i t e ( i n f a c t "dominant", see below)
But f i s n o t f i n i t e .
Definition 2a. A s i m p l i c i a l map a : K L between a b s t r a c t complexes i s
called p a r t i a l l y proper ( r e s p . p r o p e r , r e s p . ...) i f t h e r e a l i z a t i o n

L s
|al : R I K | -» l l
R R i - p a r t i a l l y proper (resp. proper, resp. . . . ) .
N o t i c e t h a t t h e s e p r o p e r t i e s o f a do not depend on t h e c h o i c e o f t h e
base f i e l d R.
We now come t o t h e main d e f i n i t i o n of t h i s section.
Definition 3. A s i m p l i c i a l map f : X Y between g e o m e t r i c complexes i s
called dominant, i f f i s p a r t i a l l y p r o p e r and f i s a b i j e c t i o n from t h
u n d e r l y i n g s e t o f X t o t h e u n d e r l y i n g s e t o f Y. Then, o f c o u r s e , f i s
finite.
Notice that, i n t h e case t h a t X and Y a r e l o c a l l y f i n i t e , t h i s just

means t h a t f i s an isomorphism between the l o c a l l y s e m i a l g e b r a i c spa-
ces X and Y. N o t i c e a l s o t h a t , f o r g e n e r a l complexes X and Y, t h e s e t

1
theoretic inverse f : Y -* X o f a dominant map f : X -> Y i s a weakly sim-
p l i c i a l map. I n some sense every weakly s i m p l i c i a l map can be b u i l t up
-1
from such maps f , as we s h a l l see l a t e r .
I f X i s c l o s e d , then every dominant s i m p l i c i a l map f : X -> Y i s a s i m p l i -
cial isomorphism and, i n p a r t i c u l a r , Y i s c l o s e d .
Definition 3a. A s i m p l i c i a l map a : K L between a b s t r a c t complexes i s
c a l l e d dominant i f i t s r e a l i z a t i o n la I i s dominant.
R
We s t a t e some f o r m a l p r o p e r t i e s o f t h e n o t i o n s i n t r o d u c e d i n t h e l a s t
two definitions, l e a v i n g t h e t r i v i a l p r o o f s t o the r e a d e r .
Remark 5.2. a) I f X i s a subcomplex o f a g e o m e t r i c complex Y then t h e
i n c l u s i o n map i : X«->Y i s p a r t i a l l y proper i f and o n l y i f X i s c l o s e d
i n Y. In t h i s case i i s f i n i t e .
b) L e t f : X-> Y and g : Y -> Z be s i m p l i c i a l maps between geometric com-
p l e x e s . L e t P be any one o f t h e p r o p e r t i e s " p a r t i a l l y proper", "par-
tially finite", "finite".
i) I f f and g b o t h have p r o p e r t y P then g • f has p r o p e r t y P.
ii) I f g o f has P then f has P.
iii) I f g • f has P and f i s s u r j e c t i v e (as map from t h e s e t X t o t h e
set Y ) , then g has P.
I f P i s t h e p r o p e r t y "dominant", then t h e statements i ) and i i i ) a r e
a l s o t r u e . Statement i i ) i s o n l y t r u e i f f i s assumed t o be s u r j e c t i v e .
The reader s h o u l d n o t be t r o u b l e d by t h e f a c t t h a t i n t h e c a t e g o r y o f
paracompact r e g u l a r spaces statement i i i ) becomes f a l s e f o r P = "par-
t i a l l y proper" ( I , 6.3) o r P = " p a r t i a l l y finite" ( I , 7.14). Such

maps cannot always be t r i a n g u l a t e d .
In t h e c a t e g o r y o f a b s t r a c t complexes, hence a l s o i n the c a t e g o r y o f
g e o m e t r i c complexes, t h e r e e x i s t a r b i t r a r y f i b r e p r o d u c t s . Indeed, let
a^ : -+ K and a 2 : K - » K be a b s t r a c t
2 simplicial maps. We look a t the
set
F := E(K-) x E ( K ) = { ( e , e ) G E ( K ) x £ ( K ) l a . (e ) = a
9 1 0
1
1
z
9
1 1
1 9 (e )},
9
l E ( K ) z
w i t h t h e n a t u r a l p r o j e c t i o n s p^ : F - » E ( K . j ) , p 2 : F - > E ( K ) . We d e f i n e a
2
complex L = (E(L),S(L)) as f o l l o w s . S(L) i s t h e s e t o f a l l s u b s e t s s
o f F w i t h p^ (s) e S l K ^ and P ( s ) € S ( K ) , and E ( L ) i s the u n i o n o f a l l

2 2
the s u b s e t s s C S (L) o f F. The r e s t r i c t i o n TT : E(L) -+ E (K^)

1 of p ± isa
s i m p l i c i a l map : L-+K. ( i = 1,2) and t h e diagram
(*)
commutes. I t i s now e a s i l y checked t h a t t h i s diagram i s a p u l l back i n

K X K
the category of a b s t r a c t complexes. Thus L i s t h e f i b r e product i K 2
^Th^ f o l l o w i n g statements a r e e a s i l y verified.
Lemma 5.3. L e t P be any one o f the p r o p e r t i e s "partially proper",
"proper", "partially finite", "finite", "dominant". I f i n t h e p u l l
back diagram (*) t h e map a^ has p r o p e r t y P t h e n TT a l s o has p r o p e r t y P 2
Examples 5.4. a) I f b o t h complexes and a r e c l o s e d then i t i s
easily seen t h a t E ( K x K ) 1 K 2 = E(K^)x £ E(K ). 2
b) (The case t h a t K i s t h e one p o i n t complex.) F o r any two a b s t r a c t
complexes and K 2 the d i r e c t product xK 2 e x i s t s and may be des-
c r i b e d as f o l l o w s : E ( K - x K ) 0 = E(K-) x E ( K ) . A s u b s e t s of E ( K ) x E ( K )
9 1 0
i s a simplex of x K i f and o n l y i f t h e p r o j e c t i o n s p^(s) and P ( s )
2 2
are s i m p l i c e s o f and K 2 respectively. The r e a d e r i s warned t h a t t h e
locally s e m i a l g e b r a i c space I x K 2 1 R usually i s different from t h e
product IK^ I x I 2 ' R

R
K o f t h e l o c a l l
Y s e m i a l g e b r a i c spaces IK^ I R and
|K I .
2 R In f a c t , t h i s a l r e a d y o c c u r s i n t h e case where both and K 2
a r e c l o s e d and f i n i t e c f . [ES, p. 6 6 f f ] . The space I I R x | I R is
then a s t r o n g d e f o r m a t i o n r e t r a c t of |K^xK l .
2 R
Definition 4. An a b s t r a c t complex L i s c a l l e d maximal, i f every domi-

r
nant s i m p l i c i a l map a : L ->L i s a s i m p l i c i a l isomorphism. A maximal
h u l l o f an a b s t r a c t complex K i s a dominant map a : L->K w i t h L a ma-
x i m a l complex. Same t e r m i n o l o g y f o r geometric complexes.
Proposition 5.5. L e t a : L -»K and 3 : P -»K be s i m p l i c i a l maps. Assume
t h a t L i s maximal and 3 i s dominant. Then t h e r e e x i s t s a unique simpli-
c i a l map n : L -* P w i t h 3 • n = a. I f a i s dominant, i . e . a i s a maximal
h u l l o f K, then n i s a l s o dominant. In p a r t i c u l a r , any two maximal
h u l l s o f K are isomorphic.
P r o o f . We c o n s i d e r t h e p u l l - b a c k diagram
Q-
of t h e maps a and 3 i n t h e c a t e g o r y o f a b s t r a c t complexes. S i n c e 3 i s
dominant, y i s dominant (Lemma 5 . 3 ) . S i n c e L i s maximal, y i s an i s o -

1
morphism. Thus n := 6°y i s a s i m p l i c i a l map w i t h a = 3 ° n . Uniqueness
of r\ i s e v i d e n t , say, from t h e p u l l - b a c k p r o p e r t y o f t h e diagram. I f
a i s dominant then 6 i s a l s o dominant by Lemma 5.3. Thus n i s dominant
too. q.e.d.
We now s t a r t o u t t o prove t h e e x i s t e n c e o f t h e maximal h u l l f o r a given
a b s t r a c t complex K. We c o n s t r u c t a new complex K as f o l l o w s . L e t H(K)
denote t h e s e t o f a l l p a i r s (p,s) € E ( K ) x s (K) w i t h p € s . We introduce
on H(K) t h e c o a r s e s t e q u i v a l e n c e r e l a t i o n ~ such that (p,t) ~ (p,s)
whenever p G t ;< s . Thus (p^s^) ~ (p ,s )

2 2 i f and o n l y i f p^ = p 2 and
there e x i s t s a f i n i t e sequence.
s = t t ,t = S
1 o' 1'*-* 2m 2
w i t h p^ G t ^ f o r i = 1,...,2m and
t 0 < t ,t1 1 > t ,t 2 2 < t 3 > t 2 m .
We denote t h e e q u i v a l e n c e c l a s s o f a p a i r (p,s) G H (K) by [ p , s ] , and we
take as s e t o f v e r t i c e s E(K) t h e s e t H ( K ) / ~ o f a l l t h e s e equivalence
c l a s s e s . We f u r t h e r i n t r o d u c e f o r every s G S ( K ) the f i n i t e subset
[s] := { [ p , s ] I p G s )
of E(K) and take as s e t o f s i m p l i c e s o f K
S(K) := {[s] I s G S(K)}.
The pair (E(K),S(K)) i s indeed an a b s t r a c t complex K, and t h e n a t u r a l
map TT = TT : E(K)
k -» E ( K ) , Tf([p,s]) = p, i s a s i m p l i c i a l map from K t o
K. More p r e c i s e l y , TT maps every simplex [ s ] o f K b i j e c t i v e l y onto t h e
corresponding simplex s o f K. I f u i s a f a c e o f some [s] GS(K) w i t h
TT(U) = t G S ( K ) , then u = [ t ] G S ( K ) . Thus TT i s dominant. We s h a l l see
l a t e r t h a t TT : K -> K i s indeed a maximal h u l l o f K.

k
F o r any geometric complex X we denote by X t h e r e a l i z a t i o n o f K (X)"

:
and by p x : X -* X t h e r e a l i z a t i o n o f K (X) ~-»K (X) . From now on a
weakly s i m p l i c i a l map from X t o a geometric complex Y w i l l be denoted
by a d o t t e d arrow X ->Y t o keep i n mind t h a t i t i s not an arrow i n
the c a t e g o r y o f complexes.
Theorem 5.6. F o r any weakly s i m p l i c i a l map f : X «.... -> y t h e r e e x i s t s a

unique s i m p l i c i a l map £ : X -* Y such t h a t the diagram
commutes.
Notice that, i n the case t h a t X and Y a r e s t r i c t l y locally finite, X
and Y a r e a g a i n s t r i c t l y locally finite, and thus Theorem 5.6 yields
a triangulation o f the l o c a l l y s e m i a l g e b r a i c map f.
P r o o f o f Theorem 5.6. L e t K := K(X), L := K ( Y ) . Thus X = |K|, Y = ILI•
As s runs through S (K), Isl runs through the s e t I(X) o f open simpli-
ces o f X. We have a map a : S (K) S (L) , d e f i n e d by f ( | s | ) = | a (s) I .
Notice that, i n g e n e r a l , t h i s map a i s not induced by a map from E(K)
to E ( L ) . But i f t e s ( K ) i s a f a c e o f s then a ( t ) i s a f a c e o f a ( s ) .
Moreover, f y i e l d s by r e s t r i c t i o n an a f f i n e map f g : |s| -» |a(s)|,
which extends c o n t i n u o u s l y t o an a f f i n e map T g : IsI -> I a ( s ) I , and f
coincides with f f c on | t | ( c f . I, §2, Def. 5 ) .
We now d e f i n e a map 3 from E(K) t o E(L) by
3([p,s]) = [f (p),a(s)]s (s € S(K) ,p E E(K) ,p€ s) .
I t i s e a s i l y checked t h a t 3 i s w e l l d e f i n e d . Indeed, if p€t<s, then
f (p)
t = f ( p ) . Clearly
g 3 ( [ s ] ) = [ a ( s ) ] . Thus 3 i s a s i m p l i c i a l map
from K t o L. L e t f : X-+Y denote the r e a l i z a t i o n o f 3 . F o r every s€S(K)
the diagram
commutes. Thus i n d e e d f o p = p o f . The uniqueness statement i n t h e
A Y
theorem i s evident, since p Y i s a bijection from t h e s e t Y t o t h e s e t
Y. q.e.d.
: =
Corollary 5.7. F o r any g e o m e t r i c complex X t h e map q P£ X-* X i s a
simplicial isomorphism.
P r o o f . Apply t h e theorem t o t h e weakly s i m p l i c i a l map g = p x ^ from X
to X. We o b t a i n a s i m p l i c i a l map g from X t o X w i t h q • g = 9 • P x
=
-*-^x*
S i n c e q and g a r e b o t h b i j e c t i v e , as maps between s e t s , a l s o g • q =
id£. q.e.d.
C o r o l l a r y 5.8. F o r any g e o m e t r i c a l complex X t h e complex X i s maximal,
and thus p x : X-> X i s a maximal h u l l o f X.
P r o o f . L e t f : Y -> X be a dominant map. A p p l y i n g Theorem 5.6 t o t h e

-1
weakly s i m p l i c i a l maps f and g := f we o b t a i n a commutative diagram
f $ q
w i t h p := p , q := p~. From t h e uniqueness statement i n t h e same theo-

v
~Y X
rem we deduce that g o f i d ~ and f * g id$. Thus f and g a r e s i m p l i -
cial isomorphisms. By t h e p r e c e d i n g c o r o l l a r y a l s o q i s a s i m p l i c i a l
isomorphism. Thus h := p • { q ^ i s a right inverse of f . Since f i s
a bijection o f s e t s , the s i m p l i c i a l map h i s a l s o a l e f t inverse of
f . Thus f i s a s i m p l i c i a l isomorphism. q.e.d.
Corollary 5.9. E v e r y weakly s i m p l i c i a l map f : X >Y from a maximal
complex X t o an a r b i t r a r y complex Y i s simplicial.

P r o o f . C o n s i d e r t h e commutative diagram
Z f
• Y.
p x i s an isomorphism. Thus f = p y •f •p x i s indeed simplicial.
q.e.d.
§6 - T r i a n g u l a t i o n o f amenable p a r t i a l l y f i n i t e maps
We w i l l o n l y c o n s i d e r r e g u l a r and paracompact l o c a l l y s e m i a l g e b r a i c
spaces u n t i l the end o f t h i s s e c t i o n . Thus from now on - u n l e s s some-
t h i n g e l s e i s s a i d - a "space" w i l l always be a r e g u l a r paracompact
locally s e m i a l g e b r a i c space.
We g a t h e r t o g e t h e r some o b s e r v a t i o n s on p a r t i a l l y finite simplicial
maps which m o s t l y appeared i n §5. A few a r e a l r e a d y c o n t a i n e d i n I , §6.
P r o p o s i t i o n 6.1. L e t f : Y -> X be a p a r t i a l l y finite simplicial map
between s t r i c t l y l o c a l l y f i n i t e geometric complexes. Then f and f : Y->X
- regarded as maps between spaces - have t h e f o l l o w i n g properties:

1
i) ( f ) " ( X ) = Y.
ii) f is partially finite.
iii) Every open simplex T o f Y i s mapped under f i s o m o r p h i c a l l y onto
an open simplex a o f X. A l s o , t h e c l o s u r e x n Y o f x i n Y i s
mapped i s o m o r p h i c a l l y onto a fl X, and x i s mapped under f isomor-
phically onto a.
iv) F o r every open simplex a o f X the connected components o f the
locally s e m i a l g e b r a i c subset f ^ (a) o f Y a r e p r e c i s e l y a l l open
s i m p l i c e s x o f X w i t h f ( x ) = a. In p a r t i c u l a r , f is trivial over
a, i . e . f ^ (a) i s i s o m o r p h i c over a t o a p r o d u c t a x F, w i t h F
a d i s c r e t e space i n our c a s e .
In this p r o p o s i t i o n o n l y the l a s t statement (iv) needs f u r t h e r explana-
tion. N o t i c e t h a t t h e s i m p l i c e s x € l ( Y ) w i t h f ( x ) = a a r e open connected

-1
subcomplexes o f t h e complex f (a) and thus a r e open s e m i a l g e b r a i c sub-
-1
sets of f ( a ) . Moreover they a r e p a i r w i s e disjoint.
We draw t h e f o l l o w i n g consequence from t h e statements ( i i i ) and ( i v )

in this proposition.
Corollary 6.2. L e t X,Y,Z be s t r i c t l y l o c a l l y f i n i t e complexes and l e t
Z * -Y
be a commutative diagram o f l o c a l l y semialgebraic maps w i t h f and g
partially f i n i t e . Hence h i s a l s o p a r t i a l l y f i n i t e (cf. I.6.2.ii).
Assume t h a t f and g a r e s i m p l i c i a l . Then h i s weakly s i m p l i c i a l . Thus,
if i n a d d i t i o n t h e complex Z i s maximal (cf. § 5 ) , t h e map h i s s i m p l i c i a l .
We want t o t r i a n g u l a t e p a r t i a l l y f i n i t e maps. A c c o r d i n g t o P r o p o s i t i o n
6.1.iv a n e c e s s a r y c o n d i t i o n for this i s t h a t t h e maps be "amenable"
in the f o l l o w i n g sense.
Definition 1. A l o c a l l y semialgebraic map f : M-> N i s c a l l e d amenable
if there exists a locally f i n i t e family (NÎXeA) i n T(N) such t h a t N

-1
is the u n i o n o f a l l and f i s t r i v i a l over e v e r y , i.e. f (N^)
is isomorphic over t o a product x w i t h some space F ^ .
Then, by t h e t r i a n g u l a t i o n theorem 4.4, t h e r e exists a triangulation
cp : X-^*N such t h a t f is trivial o v e r t h e image cp(c) o f each open sim-
plex o e i ( x ) .
We quote from [DK , Th. 6.4] ( c f . a l s o

3 [BCR, Chap. 9]) the f o l l o w i n g
important f a c t , which i n the case R = IR i s due t o R. Hardt [H], and

1
t h e r e f o r e w i l l o f t e n be c i t e d as " H a r d t s theorem".
Theorem 6.3. E v e r y a f f i n e semialgebraic map f : M -*N i s amenable.
Given a l o c a l l y f i n i t e family (M^liGI) of l o c a l l y semialgebraic sub-
s e t s o f M t h e r e even e x i s t s a l o c a l l y f i n i t e p a r t i t i o n (N, I X € A ) of N i n t o
s e m i a l g e b r a i c s e t s such t h a t f and a l l t h e r e s t r i c t i o n s f|M^:M^-*N ( i € I )
a r e t r i v i a l o v e r each N^.
Actually t h i s has been proved i n [DK^] o n l y i n t h e c a t e g o r y o f s e m i a l g e -
braic s p a c e s , but t h e g e n e r a l i z a t i o n t o t h e l o c a l l y s e m i a l g e b r a i c s e t -
ting i s immediate. A g a i n , t h e s e t s can be chosen as t h e images o f
t h e open s i m p l i c e s o f a s u i t a b l e t r i a n g u l a t i o n Y-^*N.
Another i m p o r t a n t c l a s s o f amenable maps a r e t h e c o v e r i n g s , t o be
studied later (Chap. V ) . There e x i s t p a r t i a l l y f i n i t e maps which a r e
not amenable. F o r example, t a k e R = 3R and c o n s t r u c t an i n f i n i t e l y
high E i f f e l tower T <z [0,1 ] x [0,1 ] ^ x

l o c over [0,1 ]x [0,1] whose diameter
tends t o z e r o as t h e h e i g h t i n c r e a s e s . The p r o j e c t i o n
p : T -* [0,1 ] x [ o , 1 ] ^ { ^2. \) ) i s p a r t i a l l y
r f i n i t e b u t not amenable. Even
simpler, f o r any l o c a l l y complete b u t n o t complete s e m i a l g e b r a i c
space M t h e i d e n t i t y map from M ^ q c to M i s p a r t i a l l y f i n i t e but not
amenable.
To t r i a n g u l a t e a g i v e n amenable p a r t i a l l y f i n i t e map f : M-N we may
assume t h a t N i s a l r e a d y t r i a n g u l a t e d i n such a way t h a t f i s t r i v i a l
o v e r t h e image o f e v e r y open simplex under t h e t r i a n g u l a t i o n map cp : X N
We may as w e l l assume t h a t N = X and cp i s t h e i d e n t i t y . We thus meet
the following
Problem 6.4. L e t f : M -* X be a p a r t i a l l y f i n i t e map from a space M t o
a strictly locally f i n i t e complex X which i s t r i v i a l over e v e r y open
s i m p l e x o f X. Under what f u r t h e r c o n d i t i o n s on f can M be t r i a n g u l a t e d
i n s u c h a way t h a t f becomes s i m p l i c i a l ? More p r e c i s e l y , when does
t h e r e e x i s t a commutative diagram
X
with Y a s t r i c t l y l o c a l l y f i n i t e complex, g a s i m p l i c i a l map, and i|/ a
l o c a l l y semialgebraic isomorphism?
I f such a diagram e x i s t s then we may r e p l a c e Y by i t s maximal h u l l Y
( c f . §5) and assume as w e l l t h a t Y i s maximal. A c c o r d i n g to Corollary
6.2 any other diagram o f the same s o r t - w i t h f f i x e d - i s i s o m o r p h i c
to the g i v e n one i n an o b v i o u s sense.
We s t a y w i t h a map f : M -*X as d e s c r i b e d i n Problem 6.4. We denote by
I(M) the s e t o f a l l connected components o f a l l s e t s f (a) w i t h a r u n -
ning t h r o u g h the s e t I(X) of open s i m p l i c e s of X. The f a m i l y I(M) is
clearly locally f i n i t e and i s a p a r t i t i o n of M i n t o semialgebraic sets.
For a g i v e n open simplex a o f X every connected component T o f f (a)
i s mapped i s o m o r p h i c a l l y onto a by f . We quote a g e n e r a l lemma, which
will a l s o be u s e f u l i n l a t e r sections. Applying t h i s lemma t o t h e re-
s t r i c t i o n f-»ofiX of f , w i t h T the c l o s u r e of T i n M, we see that f
maps T i s o m o r p h i c a l l y onto a f l X .
Lemma 6.5. L e t TT : L N be a p r o p e r map between s p a c e s . Assume t h a t S i s
a l o c a l l y semialgebraic s u b s e t of N such t h a t the f i b r e s TT ^ (s) o f a l l
points s € S are connected and not empty. Assume f u r t h e r t h a t S i s dense
i n N and TT ^ (S) i s dense i n L. Assume f i n a l l y t h a t every p o i n t x i n N\S
has a fundamental system o f open s e m i a l g e b r a i c neighbourhoods U w i t h
U f l S connected. Then a l l f i b r e s o f TT a r e connected and not empty. In
particular i f TT i s a l s o f i n i t e , TT i s an isomorphism.
We postpone the p r o o f o f t h i s lemma f o r a moment and continue the dis-
c u s s i o n o f the p a r t i t i o n E(M) of M. I f a* € I ( X ) i s an open f a c e of

1 1
a € I ( X ) , and T i s a connected component o f f"" (o) , then := T D f ( a ) 1
_i
i s a connected component o f f ( a ^ ) , s i n c e f maps isomorphically onto
. In p a r t i c u l a r , t h e c l o s u r e T o f any T € I(M) i s a union o f s e t s

1
T € I ( M ) . Thus we may r e g a r d I(M) as a " s e m i a l g e b r a i c s t r a t i f i c a t i o n "
o f M, and we c a l l t h e T € I (M) t h e " s t r a t a " o f M.
We now s t a r t o u t t o prove Lemma 6.5. We need t h e f o l l o w i n g g e n e r a l f a c t .
Sublemma 6.6. L e t f be a p r o p e r s u r j e c t i v e map from a space N t o a
space M. Assume t h a t M as w e l l as a l l f i b r e s o f f a r e c o n n e c t e d . Then
N i s connected.
P r o o f . Suppose t h e r e e x i s t s a p a r t i t i o n N = A^ U A 2 o f N i n t o two d i s -
j o i n t open - hence a l s o c l o s e d - non empty l o c a l l y s e m i a l g e b r a i c sub-
s e t s . S i n c e t h e f i b r e s o f f a r e connected, both A^ and A 2 a r e unions
of f i b r e s . S i n c e f i s s u r j e c t i v e M i s t h e d i s j o i n t u n i o n o f t h e s e t s
B 1 := f(A-j) and B 2 := f ( A ) .2 Since f i s proper the sets B i are closed
and l o c a l l y s e m i a l g e b r a i c i n M. But t h i s c o n t r a d i c t s o u r assumption
t h a t M i s c o n n e c t e d . Thus N i s connected. q.e.d.
Proof o f Lemma 6.5. TT (L) i s c l o s e d and c o n t a i n s S. Thus TT(L) = N. L e t

1
a point x £ N \ S be g i v e n . Suppose TT (X) i s n o t c o n n e c t e d . We choose
1 1
a p a r t i t i o n TT (x) = A^ |J A 2 of n (x) i n t o two d i s j o i n t l o c a l l y semi-
- 1
a l g e b r a i c s u b s e t s , which a r e both c l o s e d i n TT ( X ) and non empty. By
a s t a n d a r d argument f o r p r o p e r maps we f i n d an open s e t U c N i n our

-1
g i v e n fundamental system o f neighbourhoods o f x such t h a t n (U) =
V-j U V 2 w i t h V\ an open l o c a l l y s e m i a l g e b r a i c neighbourhood of A i (i=1,2)
and V 1 nV 2 = 0 . We have
TT 1
(U n S ) = [ v 1 n TT*" ( S ) ] •
1
[ v n TT"" ( S ) ] .
2
1
Both s e t s V. f i n (S) a r e non empty, l o c a l l y s e m i a l g e b r a i c , and open i n

-1 -1
TT ( U D S ) . But, by t h e sublemma, TT ( u n S) i s c o n n e c t e d . T h i s contra-
diction p r o v e s t h a t TT ( X ) must be connected. q.e.d.
We c o n t i n u e s t u d y i n g t h e s t r a t i f i c a t i o n I(M) o f M. L e t K(X) = (E(X),S(X))
be t h e a b s t r a c t i o n o f t h e geometric complex X. We want t o c o n s t r u c t an
a b s t r a c t complex K(X,f) = (E (X, f ),S (X, f) ) which h o p e f u l l y w i l l serve
to t r i a n g u l a t e M i n such a way t h a t t h e s t r a t a become t h e images o f
t h e open s i m p l i c e s o f | K ( X , f ) | .
L e t H(X,f) denote t h e s e t o f p a i r s (p,T) G E ( X ) x i ( M ) w i t h pG T(T) .
( f ( T ) means t h e c l o s u r e o f the open simplex f ( T ) i n the c l o s e d complex
X.1 We i n t r o d u c e t h e c o a r s e s t e q u i v a l e n c e r e l a t i o n ~ i n H(X,f) such
T w n e n e v e r T w o a i r s a n d T
that (p,T>j) ~ ( P ' ^ 2 ^\ ^2' cl7
P (P^/Ti) ^P2' 2^
a r e e q u i v a l e n t i f and o n l y i f = p 2 =: p and t h e r e e x i s t s a sequence
U Q = T ,U ,...,U
1 1 2m = T 2 i n T(M) w i t h p € fTuTT f o r 0 2' U
U
2 c U
3 ^ m - l ^ m '
We denote t h e e q u i v a l e n c e c l a s s o f a p a i r (p,T) € H ( X , f ) by [ p , T ] , and
we take t h e s e t o f a l l these e q u i v a l e n c e c l a s s e s as s e t of v e r t i c e s
E ( X , f ) o f o u r complex K ( X , f ) , E ( X , f ) := H ( X , f ) / - . F o r every T € I(M)
we d e f i n e t h e f i n i t e subset
[T] := { [p, T] | p G fTTT fl E (X) }
of E ( X , f ) f and we d e f i n e t h e s e t o f these s u b s e t s [ T ] , w i t h T running
through I(M), t o be t h e s e t o f s i m p l i c e s S(X,f) o f K ( X , f ) . Then
K(X,f) = ( E ( X , f ) , S ( X , f ) ) i s indeed an a b s t r a c t s i m p l i c i a l complex.
We have a w e l l d e f i n e d map
TT = n ^ x f : E(X,f) - E(X), [p,T] » p,
which i s a s i m p l i c i a l map from K(X,f) to K(X).
Lemma 6.7. TT i s p a r t i a l l y finite ( i n t h e sense o f D e f i n i t i o n s 2 and
2a i n §5) .
P r o o f . L e t [T] be a s i m p l e x o f K ( X , f ) . C l e a r l y TT i s i n j e c t i v e on [T] .
In o r d e r t o prove t h a t TT i s p a r t i a l l y f i n i t e we c o n s i d e r a proper face
t 1 o f [T] w i t h i T ( t ) € S ( X ) . We have t o show t h a t t . , e s ( X , f ) . Now

1 TT([T])
i s t h e a b s t r a c t i o n s o f t h e open simplex a := f (T) o f X, and I TT ( t ^ ) I
i s a face of a with €I(X). There e x i s t s a unique s t r a t u m cT
o f M w i t h f (T^) = a 1 (see above) . We have [ T ] < [ T ] , n(t-,) = 1 nd^]),
hence t-j = [T^ ] € S ( X , f ) . Thus TT i s indeed p a r t i a l l y finite.
We need t h e f o l l o w i n g g e n e r a l assumption on our s t r a t i f i c a t i o n I ( M ) .
(*) I f T^ and T 2 a r e two d i f f e r e n t s t r a t a o f M, then [T^] * [ T ] . 2
L e t us l o o k f o r c a s e s where (*) i s f u l f i l l e d .
Lemma 6 . 8 . Assume t h a t T^ and T 2 a r e two d i f f e r e n t s t r a t a o f M w i t h
f (T-|) = f ( T ) =: a and t h a t a has a t l e a s t one v e r t e x

2 p w i t h p E X and
T 1 0 f " (p) * f
1
2 n f" 1
(p) . Then [T ] * [T ] .
1 2
P r o o f . We c l a i m t h a t [ p , T ] * [ p , T ] . Otherwise t h e r e would e x i s t a
1 2
T # U = T w i t h
sequence o f s t r a t a U Q = l' **' 2m 2
and p € fTTTT f o r 0 2m = T 2 fl f ~ (p) . T h i s c o n t r a d i c t s t h e h y p o t h e s i s o f the lemma. Thus
[p/T-j] * [ p / T ] . S i n c e
2 TT maps both [T^ ] and [ T ] 2 bijectively onto t h e same
simplex s € S ( X ) t h e v e r t e x [p,T.j] cannot be c o n t a i n e d in [T ].2 q.e.d.
From t h i s lemma we see immediately
Corollary 6 . 9 . Condition (*) i s c e r t a i n l y f u l f i l l e d i n the f o l l o w i n g
cases. !
i) The complex X i s c l o s e d and f i s t r i v i a l o v e r each c l o s e d simplex
a o f X.
ii) f is a partially f i n i t e map from t h e space M t o a s t r i c t l y locally
finite complex , and f i s t r i v i a l over each open simplex o f X^ .
The complex X i s t h e b a r y c e n t r i c s u b d i v i s i o n o f X^.
1
Here t h e b a r y c e n t r i c subdivision Y of a geometric complex Y i s d e f i n e d
i n t h e o b v i o u s way: s u b d i v i d e Y i n t h e c l a s s i c a l way, and c o l l e c t a l l

1
open s i m p l i c e s o f Y which a r e c o n t a i n e d i n t h e s e t Y.
A third case i n which condition (*) i s c l e a r l y fulfilled i s the f o l l o w -
i n g : M = X, f = i d . Then t h e complex K(X,f)

x i s n o t h i n g o t h e r than t h e
a b s t r a c t i o n K(X) o f t h e maximal h u l l X o f X d e f i n e d i n §5. T h i s case
l e a d s back t o t h e c o n s i d e r a t i o n s i n §5.
From now on we always assume t h a t c o n d i t i o n (*) i s f u l f i l l e d .
Lemma 6.10, L e t T be a s t r a t u m o f M. Then the s i m p l i c e s u € S ( X , f ) hav-
ing [T] as a f a c e a r e i n one-to-one c o r r e s p o n d e n c e w i t h the s t r a t a
U e I(M) w i t h T c U v i a u = [U] . The complex K(X,f) is locally finite.
P r o o f . L e t t := [T] € S ( X , f ) and a := |n(t) I = f (T) € Z(X) . L e t u € S ( X , f )
be a simplex which has t as a p r o p e r f a c e and p := |TT(U) | £ I(X) . Then
p > a . We have u = [U] w i t h a s t r a t u m U € I(M), u n i q u e l y determined by u
a c c o r d i n g t o assumption (*). S i n c e f ( U ) = p , t h e r e e x i s t s a unique
stratum c U w i t h f ( T ) = a . Now
1 := [T^ ] i s a f a c e o f u w i t h
I TT ( 1 ) | = a = ( n ( t ) |. Thus
1 = t . By assumption (*) we have T^ = T.
The first a s s e r t i o n o f the lemma i s now c l e a r . The second follows
from t h e f a c t that the p a r t i t i o n I(M) o f M i s l o c a l l y finite. Indeed,
f o r a g i v e n T £ I(M) and a g i v e n s e m i a l g e b r a i c neighbourhood W o f T,
t h e r e e x i s t o n l y f i n i t e l y many s t r a t a U w i t h U D W * 0 . A fortiori there

e x i s t o n l y f i n i t e l y many s t r a t a U w i t h U H T * 0 , i . e . T C Z T J . Thus f o r
a given t € S ( X , f ) t h e r e e x i s t o n l y f i n i t e l y many u€S(X,f) w i t h u > t .

q.e.d.
I t may happen t h a t K(X,f) i s n o t s t r i c t l y locally finite.
Example 6.11. We c o n s i d e r t h e " i n f i n i t e f a n " M, d e s c r i b e d i n I , §2
after Definition 4. T h i s i s a l o c a l l y f i n i t e g e o m e t r i c complex which
i s not s t r i c t l y locally finite. L e t p be t h e p o i n t o f M m i s s i n g i n M,
and let (T^li€U) be the f a m i l y o f a l l open 2 - s i m p l i c e s o f M, labelled
i n such a way t h a t "T\ n "T 0 M * 0 i f f I i - j ( <2. We i n t r o d u c e t h e o b v i o u s
" f o l d i n g map" 7 from M t o t h e c l o s e d s t a n d a r d t r i a n g l e A = [e ,e.,,e ],

o 2
obtained by g l u i n g t o g e t h e r t h e a f f i n e maps f T : T T A with f (p) =

i
e Q f o r a l l i , f ^ U ^ n T\) = [ e , e ] Q 2 f o r i even, T^0t _<\ n ? ) = [ e

± i Q f e 1 ]
f o r i odd. L e t f : M -» X := A ^ { e } be t h e r e s t r i c t i o n o f f t o M. T h i s
o
map f i s p a r t i a l l y finite and i s t r i v i a l over each open simplex o f X.
The complex K(X,f) i s n a t u r a l l y i s o m o r p h i c t o the a b s t r a c t i o n of the
geometric complex M. N o t i c e t h a t c o n d i t i o n (*) i s f u l f i l l e d f o r f : M -»X
We denote t h e r e a l i z a t i o n |K(X,f)l o f t h e a b s t r a c t complex K(X,f) by
X ( f ) , and t h e r e a l i z a t i o n o f TT - : K(X,f) -»K(X) by p

v v - o r b r i e f l y by p
Theorem 6.12. (As b e f o r e , we assume t h a t c o n d i t i o n (*) i s f u l f i l l e d ) .
There e x i s t s a unique l o c a l l y s e m i a l g e b r a i c map ^ : X ( f ) -+M, such t h a t
the diagram
commutes, mapping every open simplex I [ T ] I of X(f) into (hence onto)
the c o r r e s p o n d i n g stratum T o f M. T h i s map \p i s a l o c a l l y semialgebraic
isomorphism.
P r o o f . F o r any s t r a t u m T E I ( M ) , t h e map p = p - y i e l d s an a f f i n e i s o -
A ,r
morphism from t h e c l o s u r e T f l X ( f ) o f t h e c o r r e s p o n d i n g open simplex
T := I [ T ] I o f X ( f ) onto f (T) . On the o t h e r hand f y i e l d s a l s o a semi-
algebraic isomorphism from T onto f ( T ) . Thus we have a unique s e m i a l -
gebraic isomorphism
if/ : T n X ( f )
T
with (f|T)°i|; T = p|7 f l X ( f ) . These maps ^ T glue to a l o c a l l y semialge-
b r a i c map from X ( f ) t o M, s i n c e f o r any stratum U c T and c o r r e s p o n -
d i n g open simplex y := I [U] I o f X ( f ) we have T < T and ^ T I Y = ^ l Y / and
s i n c e t h e s e s i m p l i c e s Y a r e a l l t h e open f a c e s o f T (Lemma 6.10). We
have = p, and c l e a r l y i s the only l o c a l l y s e m i a l g e b r a i c map
from X ( f ) t o M w i t h f»\J; = p mapping every open simplex T o f X ( f ) i n t o
the c o r r e s p o n d i n g s t r a t u m T o f M. In t h e same way we o b t a i n a l o c a l l y
s e m i a l g e b r a i c map x M-»X(f) w i t h p«y

: =
f mapping every stratum T o f
M i s o m o r p h i c a l l y onto t h e c o r r e s p o n d i n g open simplex T o f X ( f ) . The
map x i s i n v e r s e t o \\>. Thus ijj i s a l o c a l l y s e m i a l g e b r a i c isomorphism.

q.e.d.
I f the map f i s f i n i t e then i t i s p r e t t y o b v i o u s t h a t the complex X ( f )
is strictly locally f i n i t e . Thus we o b t a i n from Theorem 6.3, 6.12 and
Corollary 6.9 t h e f o l l o w i n g f i n a l r e s u l t f o r f i n i t e maps.
Theorem 6.13. Every f i n i t e map between spaces f : M -* N can be t r i a n g u -
l a t e d . More p r e c i s e l y , i f cp : X N i s a t r i a n g u l a t i o n o f N, such that
f i s t r i v i a l over cp(a) f o r every open simplex a o f X, then t h e r e e x i s t s
a t r i a n g u l a t i o n \p : Y M and a s i m p l i c i a l map g from Y t o t h e b a r y c e n -
tric s u b d i v i s i o n X' o f X such t h a t t h e diagram
1
X >- N
cp
commutes. I f N i s p a r t i a l l y complete and f i s t r i v i a l over t h e image
1
<p(o) o f each c l o s e d simplex a o f X the same i s t r u e f o r X i n s t e a d of X.
In the case R= IR R. Hardt o b t a i n e d a s i m i l a r r e s u l t . By use o f subana-
lytic stratifications he proved t h a t f i n i t e (= p r o p e r l i g h t ) subanaly-
t i c maps may be t r i a n g u l a t e d [H^]. Our approach i s more elementary and
more e x p l i c i t .
C o r o l l a r y 6.14. Every f i n i t e map f : M ->N between spaces can be completed
to a f i n i t e map 1 : M -»N between p a r t i a l l y complete spaces (cf. Defini-
tion 1 i n §5).
It i s a l s o obvious t h a t the complex X ( f ) i s s t r i c t l y l o c a l l y finite i f
the complex X i s c l o s e d . Thus we have
Theorem 6.15. E v e r y amenable p a r t i a l l y f i n i t e map f : M->N between p a r -
tially complete spaces can be t r i a n g u l a t e d . More p r e c i s e l y , the state-
ments i n Theorem 6.13 remain t r u e i n t h i s case.
In g e n e r a l , an amenable p a r t i a l l y f i n i t e map f : M ->N can n o t be comple-
ted to a p a r t i a l l y f i n i t e map i f : M->N between p a r t i a l l y complete spaces.
We w i l l not attempt to g i v e a f o r m a l p r o o f o f t h i s n e g a t i v e f a c t here.
J u s t look a t the " f o l d i n g of the f a n " f : M-» A ^ {e } Q i n Example 6.11.
To o b t a i n a c o m p l e t i o n i n the c a t e g o r y of spaces from the c o m p l e t i o n
T : M -»A i n the c a t e g o r y of geometric complexes, one has t o
"blow up" the p o i n t p of M. Then one g e t s a l o c a l l y s e m i a l g e b r a i c map
g : M -> A between p a r t i a l l y complete spaces which extends f . But the
fibre g ^ ( e ) i s not d i s c r e t e . I t i s i n t u i t i v e l y
Q c l e a r t h a t one cannot
get around a b l o w i n g up p r o c e s s to complete f i n the c a t e g o r y of spaces.
Thus, f o r amenable p a r t i a l l y f i n i t e maps, we have t o be c o n t e n t w i t h
"weak t r i a n g u l a t i o n s " i n general.

Definition 2. A weak t r i a n g u l a t i o n o f a space N i s an isomorphism
cp : X N from a l o c a l l y finite (but not n e c e s s a r i l y s t r i c t l y locally
finite) g e o m e t r i c complex X t o N. A weak t r i a n g u l a t i o n o f a locally
s e m i a l g e b r a i c map f : M ->N between spaces i s a commutative diagram
w i t h weak t r i a n g u l a t i o n s cp and \\> o f N and M r e s p e c t i v e l y and a s i m p l i -
c i a l map g.
Now observe that i n our whole study i n t h i s s e c t i o n we do not need
that X i s s t r i c t l y l o c a l l y f i n i t e , as l o n g as we don't demand t h a t the
complex X ( f ) becomes s t r i c t l y l o c a l l y f i n i t e . Thus we obtain
Theorem 6 . 1 5a. E v e r y amenable p a r t i a l l y f i n i t e map f : M-»N between
spaces can be weakly t r i a n g u l a t e d . More p r e c i s e l y , i f cp : x N is a
weak t r i a n g u l a t i o n of N, such t h a t f i s t r i v i a l over cp(a) f o r every
open simplex a o f X, then t h e r e e x i s t s a weak t r i a n g u l a t i o n : Y —> M

1
and a s i m p l i c i a l map g from Y to the b a r y c e n t r i c subdivision X of X
such t h a t t h e diagram
g f
X' = -N
commutes.
The f o l l o w i n g f i n a l proposition shows t h a t c o n d i t i o n (*) i s indispen-
sable for obtaining (even weak) t r i a n g u l a t i o n s o f p a r t i a l l y finite
maps, and t h a t our whole procedure i s somewhat canonical.
Proposition 6.16. A g a i n l e t f : M -»X be a p a r t i a l l y f i n i t e map of a

space M i n t o a l o c a l l y f i n i t e complex X. Suppose t h e r e e x i s t s a commu-
tative triangle
with Y a l o c a l l y f i n i t e complex, g a s i m p l i c i a l map, and x a

locally
semialgebraic isomorphism. Then the stratification f u l f i l l s condition
(*). Thus we a l s o have the commutative diagram d e s c r i b e d i n Theorem
6.12. There e x i s t s a s i m p l i c i a l dominant map h : X ( f ) -> Y w i t h x ° n =

^
and g°h = p ~. In o t h e r words ( c f . Cor. 6.2), the complex X ( f ) is

A ,r
maximal.
N.B. I f we have a diagram (**) with g o n l y weakly s i m p l i c i a l then, r e -
p l a c i n g Y by i t s maximal h u l l Y, we may go ahead and assume t h a t g i s
simplicial.
Proof. The open s i m p l i c e s T of Y correspond b i j e c t i v e l y w i t h the strata
T of M v i a T = X ( ) , T
c f . Prop. 6.1, i i i and i v . (As s a i d b e f o r e , this
remains t r u e under the c o n d i t i o n t h a t X and Y are locally finite in-
stead of s t r i c t l y l o c a l l y f i n i t e . ) We want to map the a b s t r a c t complex
K(X,f) s i m p l i c i a l l y onto K(Y) by a map n which sends a p o i n t [p,T]eE(X,f)
to the unique p o i n t q € E ( Y ) w i t h q a v e r t e x of X 1
(T) = T and g(q) = p.
To see t h a t t h i s map n : E ( X , f ) -* E(Y) i s well-defined we have t o consi-
der a sequence T q ,T^ , . . . / 2 T

m
i n
£ (M) w i t h T^_<| c ~r\ f o r i odd and
T _
i 1 =>"T\ f o r i even, and a point p G E ( X ) with p € f (T ) i for 0 T 2 < . . . < T 2 M _ 1 > T 2 M ,
and g is simplicial. For any T 6 l ( M ) the map n yields a bijection of
the a b s t r a c t simplex [T] o f K(X,f) to the a b s t r a c t i o n o f the simplex

x" 1
(T) o f Y. Thus n i s a s i m p l i c i a l map from K(X,f) to K (Y) . A l s o i f
1 1
[T-j] = [ T ] f o r two
2 strata T 1 f T 2 o f I (M) , then x " ^ ) = x" (T ) , thus
2
T^ = T .
2 So, condition (*) is fulfilled. Denoting the r e a l i z a t i o n of n
n =
by h, we c l e a r l y have x ° • Multiplying by f on the l e f t we obtain
g»h = p v ... The map h is a bijection from the s e t X ( f ) t o Y. Thus h i s
dominant. q.e.d.
§7 - S t a r s and shells
In t h i s s e c t i o n we study the " s t a r s " and " s h e l l s " of a geometric (sim-
plicial) complex X, and t h e i r behaviour under p a r t i a l l y finite simpli-
c i a l maps. Here we cannot r e l y on the c l a s s i c a l l i t e r a t u r e where s t a r s
have been w i d e l y used s i n c e the e a r l y days o f c o m b i n a t o r i a l t o p o l o g y
s i n c e our complexes a r e more g e n e r a l than the c l a s s i c a l ones. Our con-
siderations in this s e c t i o n a r e u s u a l l y p u r e l y c o m b i n a t o r i a l . Thus we
work - as i n §5 - i n the c a t e g o r y o f a l l g e o m e t r i c complexes.
Let X be a g e o m e t r i c complex over R. We a l r e a d y d e f i n e d the s t a r St (A)

x
of any s u b s e t A o f X i n I , §2. We now generalize this definition to
s u b s e t s A of X.
Definition 1. For any x E X we d e f i n e the s t a r S t ^ ( x ) o f x i n X as the
union o f a l l open s i m p l i c e s a o f X w i t h x € a . More g e n e r a l l y we define
for any s u b s e t A o f X the s t a r S t ( A ) x o f A i n X t o be the u n i o n o f a l l
a € I ( X ) w i t h AO o*0. Of c o u r s e , St^(A) i s the union o f the stars
S t ( x ) o f a l l x € A.
x
St (A)
x i s an open subcomplex o f X and X^St (A)
x i s the l a r g e s t subcom-
p l e x Y o f X w i t h Y fl A = 0 . A l s o S t (A) i s the u n i o n of the s t a r s

x St (x)
x
w i t h x r u n n i n g t h r o u g h the open s i m p l i c e s o f X which meet A. In parti-
cular, f o r any open simplex T of X, we have S t ( x )

x = St (x)x f o r every
X€T. The complex S t ( x ) x i s j u s t the u n i o n o f a l l a € I ( X ) w i t h T<a.
For any x € X and z € S t ( x ) the h a l f - o p e n l i n e segment ]x,z] i s c o n t a i n e d i n

x
S t ^ ( x ) . Thus S t ( x ) x i s indeed a " s t a r - s h a p e d s e t " w i t h c e n t e r x, the
c e n t e r perhaps m i s s i n g .
Suppose p o i n t s x € X and zESt (x)x a r e g i v e n . L e t a denote the open sim-

p l e x o f X which c o n t a i n s z. Assume t h a t z * x . There i s a unique p o i n t
w € do := oô with z € ]x,w[. L e t p denote t h e open simplex o f X which
contains w. Then p < o . On t h e o t h e r hand, x C p , s i n c e o t h e r w i s e z
would be c o n t a i n e d i n p , which i s n o t t r u e . Thus wC S t ^ i x ) .
Definition 2. The b o r d e r e d s t a r S t ( x ) x o f a p o i n t x € X with respect to
X i s defined t o be t h e s e t o f a l l p o i n t s y € X w i t h (1-t)x+ t y € St (x)

x
f o r e v e r y t € ] 0 , 1 [ . The s h e l l S h ( x ) o f x i s d e f i n e d
x as t h e d i f f e r e n c e
set S t ( x ) \ S t ( x ) . This
x x i s t h e s e t o f a l l p o i n t s w a r i s i n g as above.
St (x)\{x}
x i s the s e t of a l l points z = ( l - t ) x - r t w with w€Sh (x)x
and 0 < t < 1. The p o i n t w and t h e parameter t a r e u n i q u e l y determined
by z. N o t i c e that x € S t ( x )
x i f and o n l y i f x 6 X. N o t i c e also that
Sh (x)
x i s empty i f and o n l y i f St (x)x = {x}.
We g i v e a s i m p l e example. L e t X be t h e u n i o n o f an open 2-simplex
and a c l o s e d 2-simplex a 2 such t h a t o^ and o* have p r e c i s e l y one v e r -

2
tex e i n common.
Then St (e)
x = X \ [ e , e ] , and S h ( e ) = J e ^ e ^
3 4 x U [e ,e ].
3 4
Lemma 7.1. S h ( x ) and S t ( x )

x x a r e subcomplexes o f X. More p r e c i s e l y
S h ( x ) i s t h e u n i o n o f a l l open s i m p l i c e s
x p o f X with x C "p, and
pi€I(X), where T denotes t h e open simplex o f X c o n t a i n i n g x, and
px denotes t h e open j o i n o f p and T , i . e . t h e s e t o f a l l p o i n t s
(1-t)u + t v with u € p, v € T and 0 < t < 1. ( c f . a l s o Def. 5 below.
Notice that p fl T = 0 . In case p < T we have p i = x . )

The proof i s very easy. As a consequence o f t h i s lemma we see t h a t
the s h e l l s S h ^ x ^ ) and S h ( x ) a r e equal f o r any two p o i n t s x^ , x

x 2 2 in
the same open simplex T o f X ; thus we a l s o see t h a t S ^ C x ^ ) - S t ( x ) . x 2
We a r e t h e r e f o r e justified i n a l s o d e n o t i n g t h e s e subcomplexes o f X
by S h ( i )
v and S t ( x )v r e s p e c t i v e l y and c a l l i n g them t h e s h e l l and the
A A
bordered s t a r of T with r e s p e c t to X .
Lemma 7.2. The c l o s u r e o f t h e complex St^Cx) - and a l s o o f t h e complex
St (x)
x - i s S t — ( x ) . The c l o s u r e o f t h e complex S h ( x ) i s S h — ( x ) .
x
The proof i s again easy. A s t a r S t ( x ) x i s a f i n i t e complex i f and o n l y
i f S h ( x ) i s a f i n i t e complex. By Lemma 7.2 we a l s o know, t h a t

x St (x) x
is finite i f and o n l y i f St (x)

x i s f i n i t e . Notice that t h i s i s true
f o r every x € X i f and o n l y i f t h e complex X i s l o c a l l y f i n i t e , and
f o r every x £ X i f and o n l y i fX isstrictly locally finite.
D e f i n i t i o n 3. a) A g e o m e t r i c complex Y i s c a l l e d connected i f Y i s n o t
the d i s j o i n t u n i o n o f two subcomplexes Y^,Y which a r e both open

2
(hence a l s o c l o s e d ) and non empty i n Y.
b) I f X i s a g e o m e t r i c complex and x a p o i n t o f X , then t h e connected
component C ( x , X ) o f x i n X i s t h e i n t e r s e c t i o n o f a l l subcomplexes Y
o f X which a r e open and c l o s e d i n X and c o n t a i n x.
Notice that C(x,X) i s i t s e l f an open and c l o s e d subcomplex o f X , and
t h a t t h i s complex i s c o n n e c t e d . Thus C ( x , X ) i s t h e s m a l l e s t open and
c l o s e d subcomplex o f X which c o n t a i n s x. I f two p o i n t s x^ and x 2 of X
lie i n t h e same open simplex T € I ( X ) then C ( x ^ , X ) = C ( x , X ) . Thus we

2
also write C(T,X) f o r this s e t and c a l l C ( T , X ) t h e c o n n e c t e d component
o f T i n X . The complex X i s t h e d i s j o i n t union o f a l l i t s subcomplexes
C(T,X).
Definition 4. Two open s i m p l i c e s and of X are called connectable
i n X i f t h e r e e x i s t s a f i n i t e sequence
p P p =
o = <V 1 2m °2
i n I(X) w i t h
P C P 2 m -
"Connectable i n X" i s an e q u i v a l e n c e r e l a t i o n on t h e s e t I ( X ) . F o r any
T€I(X) t h e u n i o n o f a l l s i m p l i c e s t o which T i s c o n n e c t a b l e in X is
an open, c l o s e d , and connected subcomplex o f X . Thus t h i s union i s t h e
connected component C ( x , X ) o f T.
In t h e case t h a t t h e complex X i s l o c a l l y finite, i t i s evident that
the connected components o f X as a complex a r e t h e same s e t s as t h e
connected components o f X as a l o c a l l y semialgebraic space.
The e q u i v a l e n c e r e l a t i o n needed i n § 5 t o d e f i n e t h e v e r t i c e s o f t h e
maximal h u l l X o f X i s c l o s e l y r e l a t e d t o the r e l a t i o n "connectable i n
St (x)"
x f o r open s i m p l i c e s o f S t ( x ) , x w i t h x r u n n i n g through the v e r -
t i c e s o f X. Thus we a r r i v e a t t h e f o l l o w i n g statement.
P r o p o s i t i o n 7.3. A geometric complex X i s maximal i f and o n l y i f t h e
s t a r S t ( x ) o f every v e r t e x x o f X i s connected.
x
F o r a maximal complex X i t may s t i l l happen t h a t t h e s t a r of some p o i n t
i n X ^ X w h i c h i s n o t a v e r t e x i s not connected.
Example 7.4. L e t X be t h e boundary dA o f the s t a n d a r d three-dimensio-
n a l c l o s e d simplex A = [ e , e ^ , e , e ] . L e t X be t h e complex X \ ] e , e ^ [ .
Q 2 3 Q
Then t h e s t a r s S t ( e . ) , 0 < i < 3 , are a l l connected. Thus X i s maximal.

A 1 — —
But f o r any x € ]e , e [ t h e s t a r S t ( x ) has two connected
1 v components.
O I A
We denote t h e s e t o f connected components o f t h e complex X by T T ( X ) . q
A s i m p l i c i a l map f : X -> Y i n d u c e s a map TT (f) : TT (X) -> TT (Y) which sends
any connected component X ' o f X t o t h e c o n n e c t e d component o f Y which
c o n t a i n s t h e ( o b v i o u s l y ) connected complex f ( X ' ) . We o f t e n w r i t e f *
i n s t e a d o f TT (f) . q
Proposition 7.5. L e t X be any g e o m e t r i c complex and l e t x be a p o i n t
i n X ^ X . Then t h e i n c l u s i o n maps i : S t ( x ) <-» S t ( x ) and j : S h ( x ) <•+

x x x
St (x) yield bijections

v TT ( i ) and TT ( j ) . F o r any connected component
A O O
A o f S h ( x ) t h e connected
x component B := j*(A) o f S t ( x ) x i s the set
of p o i n t s ( 1 - t ) x + t z w i t h z € A and 0 < t < 1, and t h e connected compo-
nent D = B f l S t ( x ) o f S t ( x ) w i t h i * (D) = B i s t h e s e t o f p o i n t s
v v
A A
( 1 - t ) x + t z w i t h z € A and 0 < t < 1.
Proof. This i s i n t u i t i v e l y obvious, since St (x) x i s t h e cone w i t h
base S h ( x ) and v e r t e x x, d e p r i v e d o f i t s v e r t e x , and s i n c e S t ( x ) =

x x
St (x) ^Sh (x).

x x In t h e case t h a t X i s s t r i c t l y locally finite, we can
r e g a r d S t ( x ) as a s e m i a l g e b r a i c space.
x Then t h e above argument is al-
ready a p r o o f o f t h e p r o p o s i t i o n . In g e n e r a l , l e t ( X l a € l ) be t h e d i - a
r e c t e d system o f a l l f i n i t e subcomplexes o f X. We choose some index 3
w i t h x € X ^ . The complexes S t ( x ) , x St (x),x Sh (x) are r e s p e c t i v e l y the

x
unions o f t h e d i r e c t e d systems o f complexes ( S t (x) la € l , a > 3 ) , v
_ a
(St ( x ) l a G l , a > 3 ) and ( S h ( x ) ! a € l , o c > 3 ) r e s p e c t i v e l y . We know t h a t
v v
X X
a " a
the p r o p o s i t i o n i s t r u e f o r x and e v e r y X , a a > 3 . From t h i s t h e p r o -
position follows easily f o r x and X.
I f f : X-»Y i s a s i m p l i c i a l map between geometric complexes, t h e n , f o r
any x € X, we have
f (St (x)) c S t ( f ( x ) ) , f ( S t ( x J c S t ( f ( x ) ) ,
x y x Y
as i s e a s i l y v e r i f i e d . The f o l l o w i n g f a c t s w i l l be v e r y u s e f u l f o r us
i n t h e next s e c t i o n , a l t h o u g h i t i s easy t o prove them.
Proposition 7.6. Assume t h a t t h e s i m p l i c i a l map f : X-»Y i s p a r t i a l l y
finite ( c f . §5, Def. 2 ) . L e t y be a p o i n t o f Y.

1
a) f ( S t ( y ) ) Y i s t h e u n i o n o f t h e s t a r s S t ( x ) , w i t h x running x
1
through f ( y ) . These s t a r s a r e p a i r w i s e d i s j o i n t , open and c l o s e d
in f (St (y)).Y In t h e c a s e s where y € Y , o r X i s maximal and y i s a
v e r t e x o f Y, they a r e t h e connected components o f t h e complex

1
f" (St (y)). y
1
b) R e s t r i c t i n g f to S t ( x ) , x f o r any x € f ( y ) , we o b t a i n a p a r t i a l l y
f i n i t e map from S t ( x ) t o S t ( y ) . x y The preimages o f S t ( y ) and S h ( y )

y y
under t h i s map a r e S t ( x ) and S h ( x ) y y respectively.
P r o o f . a) L e t o be any open simplex i n f " ( S t ( y ) ) . Then y € f ( a ) =

Y
1
f(a), and hence o c o n t a i n s some p o i n t x G f ( y ) . T h i s means t h a t
1
oc=St (x), x and we see t h a t f ( S t ( y ) ) Y i s the union of the s t a r s
1 i
S t ( x ) with x € f
x (y) . Suppose t h a t f o r two p o i n t s x and z i n f (y)
t h e s t a r s S t ( x ) and S t ( z ) a r e not d i s j o i n t . L e t a be an open

x x simplex
c o n t a i n e d i n both s t a r s . Then x € o, z € a, and f ( x ) = f ( z ) = y. But f
is i n j e c t i v e on a ( c f . § 5 ) . Thus x = z. A l l t h e complexes S t ( x ) , x
— 1 -1
x€f (y) , a r e open i n f ( S t ( y ) ) , hence they a r e a l s o c l o s e d i n
Y
1 1 1
f" ( S t ( y ) ) . I f y € Y , then f " (y) c X,
y s i n c e X = f " (Y) by t h e p a r t i a l
— 1
p r o p e r n e s s o f f . Thus a l l s t a r s S t ( x ) w i t h x € f (y) a r e connected.
A
v
1
I f y i s a v e r t e x o f Y, y C Y , t h e n a l l x € f (y) a r e v e r t i c e s o f X n o t
c o n t a i n e d i n X. I f i n a d d i t i o n X i s maximal, then we know from Prop.

1
7.3 t h a t every s t a r S t ( x ) w i t h x € f x (y) i s connected,

1
b) We f i x some x e f (y) and c o n s i d e r t h e s i m p l i c i a l map g from St (x)
x
to S t ( y ) o b t a i n e d from f by r e s t r i c t i o n . T h i s map i s p a r t i a l l y
Y finite
1
since St (x) x i s c l o s e d i n f ( S t ( y ) ) . The c l o s u r e g o f g i s a p a r t i a l -
Y
l y f i n i t e map from S t ( x ) t o S t — ( y ) . S i n c e g i s p a r t i a l l y proper, we

x
know t h a t S t ( x ) i s t h e preimage o f S t ( y ) under g. We s h a l l

x y verify
that St (x) v i s t h e preimage o f S t ( y ) v under g. Then a s s e r t i o n b) o f the
A x
p r o p o s i t i o n w i l l be e v i d e n t . F o r z £ S t ^ ( x ) we have ] z, x [ S t (x) , hence

x
] f ( z ) , y [ c z s t ( y ) , which means t h a t f ( z ) G S t ( y ) .
Y y Conversely, i f some
z E Sty(x) w i t h f ( z ) € S t ( y ) v i s g i v e n , then ] z , x [ c s t - ( x ) and t h e image

A x A
of ] z , x [ under g i s c o n t a i n e d i n S t (y) . Thus y ] z , x [ c S t ( x ) , which

x
means t h a t z € S t ( x ) . v q.e.d.
A
C o r o l l a r y 7.7. L e t f : X -»Y be a dominant s i m p l i c i a l map ( c f . §5, Def.
3) . L e t y € Y.
1
i) By r e s t r i c t i o n , f yields, f o r every x G f ( y ) , dominant maps from
the complexes S t ( x ) x (St (x),Sh (x))

x x onto c l o s e d subcomplexes of
St (y) (St (y),Sh (y)).

y y y
1
ii) For d i f f e r e n t points x ^ , x G f 2 ( y ) t h e complexes f ( S t ( x ^ ) ) and x
1
f(St (x )) x 2 a r e d i s j o i n t . Thus, f o r any x € f (y) t h e complexes
f(St (x)) x (f(St (x)), x f ( S h ( x ) ) ) a r e a l s o open i n S t ( y )

x y (St (y), y
Sh (y)).
y
iii) In t h e case t h a t y € Y and x i s the unique p o i n t i n f (y) t h e
complexes S t ( x ) ( S t ( x ) , S h ( x ) ) a r e mapped by f dominantly

x x x
onto S t ( y ) ,
y (St (y),
y Sh (y)).
y
iv) I f X i s maximal and y i s a v e r t e x o f Y, then t h e complexes

1
f(St (x)) x (f(St (x)), x f ( S h ( x ) ) ) with x running
x through f~ (y)
are t h e connected components o f the complexes S t ( y ) , y (St (y),

y
Sh (y)).
y
P r o o f . A s s e r t i o n i ) i s e v i d e n t from t h e p r e c e d i n g p r o p o s i t i o n as soon
as we know t h a t f i s i n j e c t i v e on S t ( x ) . L e t z^ and z x 2 be p o i n t s i n
S t ( x ) w i t h f ( z ^ ) = f ( z ) = w. F o r any t € ]0,1[ t h e two p o i n t s

x 2
( 1 - t ) z ^ + t x ( i = 1,2) have the same image (1-t)w+tx. Since f i s i n j e c -

z
t i v e on t h e s e t X, these two p o i n t s a r e e q u a l . We conclude t h a t z^ ~ 2'
A s s e r t i o n i i ) can be proved i n t h e same way, s i n c e we a l r e a d y know t h a t
St (x )
x 1 and S t ( x )
x 2 a r e d i s j o i n t , hence t h a t f ( S t ( x ^ ) ) and f ( S t ( x ) ) x x 2
are d i s j o i n t . To prove a s s e r t i o n iii) o b s e r v e t h a t S t ( y ) i s connected,
Y
1
hence f ( S t ( x ) ) x = S t ( y ) f o r the unique p o i n t x i n f ( y ) . S i m i l a r l y
Y
we obtain assertion i v ) from the f a c t t h a t t h e complexes St (x),x
St (x), x S h ( x ) a r e c o n n e c t e d by P r o p o s i t i o n s
x 7.3 and 7.5. q.e.d.
A final remark. The shell Sh (x) x o f an open s i m p l e x x € I(X) (= S h ( x ) f o r x
any x € x) i s a somewhat u n u s u a l concept from t h e view p o i n t o f classical
combinatorial t o p o l o g y , even i f the complex X i s c l o s e d . T o p o l o g i s t s
p r e f e r t o work w i t h the l i n k L k ( x ) x of x i n s t e a d o f the s h e l l (and, o f
c o u r s e , t h e y assume X t o be c l o s e d ) . The r i g h t g e n e r a l i z a t i o n o f the
c l a s s i c a l n o t i o n o f l i n k f o r an a r b i t r a r y simplicial complex X and an
a r b i t r a r y open simplex o f X seems t o be p r o v i d e d by t h e f o l l o w i n g defi-
nition .
D e f i n i t i o n 5. a) Two open s i m p l i c e s x = ]u ,...,u [

Q r and
p = ]v,...,v [ g i n a v e c t o r space over R are j o i n a b l e i f the v e r t i c e s

v
u ,...,u ,
Q r 0 '«-wV s are a f f i n e l y independent. In t h i s case the open
join x p o f x and p i s d e f i n e d as the simplex ]u ,... u ,

Q f r
v
0 '«*-' v
s t-
b) For any x € I (X) the l i n k L k ( x ) x o f x i n X i s the subcomplex o f X
whose open s i m p l i c e s a r e the p € I. (X) such t h a t x and p are j o i n a b l e and
x p € I (X) .
As i s e v i d e n t from Lemma 7.1,
fLk ( )*3x
v T i f x cX,
Sh (x) =
[ [ L k ( x ) * 3 x ] ^ 3x
x if TcXs X,
w i t h L k ( x ) * 3 x d e n o t i n g the j o i n o f L k ( x )
x x and 3x, i . e . the subcom-
p l e x o f X which i s the u n i o n o f L k ( x ) , x 3 x , and a l l open j o i n s pcu w i t h
p r u n n i n g through Z ( L k ( x ) ) x and to r u n n i n g t h r o u g h I ( 9 x ) . Thus Sh (x) x
i s determined by L k ( x )
y and x i n both c a s e s .
§8 - Pure h u l l s o f dense p a i r s
As b e f o r e a "space" i s a r e g u l a r paracompact l o c a l l y s e m i a l g e b r a i c
space and a "complex" i s a geometric simplicial complex.
Definition 1. a) A dense p a i r over R i s a p a i r (P,M) c o n s i s t i n g o f a
space P and a dense subspace M. A morphism f : (P,M) -> (Q,N) between
dense p a i r s i s a l o c a l l y s e m i a l g e b r a i c map f : P -> Q w i t h f (M) c A . N o t i c e
t h a t such a morphism f i s u n i q u e l y determined by i t s r e s t r i c t i o n
f IM : M -> N.
b) A completed space i s a dense p a i r (P,M) w i t h P p a r t i a l l y complete.
In o r d e r t o g a i n a b e t t e r u n d e r s t a n d i g o f t h e v a r i o u s p o s s i b i l i t i e s
f o r c o m p l e t i n g a g i v e n space, we w i l l be c h i e f l y interested i n the f u l l
s u b c a t e g o r y o f t h e c a t e g o r y o f dense p a i r s c o n s i s t i n g o f completed
spaces. But i n t h e f o l l o w i n g i t i s o f t e n more n a t u r a l t o work w i t h
a r b i t r a r y dense p a i r s i n s t e a d o f completed spaces.
Every s t r i c t l y locally f i n i t e complex X may be c o n s i d e r e d as a comple-
t e d space (X,X), and every s i m p l i c i a l map f : X ->Y between such complexes
may be c o n s i d e r e d as a morphism f : (X,X) -» (Y,Y) .
In t h e c a t e g o r y o f dense p a i r s over R t h e r e e x i s t f i b r e products.
Namely, i f f ^ : (P^,M^) -> (Q,N) a r e morphisms ( i = 1,2), then d e f i n e

M
M := M^x^^M^/ the f i b r e p r o d u c t w i t h r e s p e c t t o t h e maps f j J j _ / and
d e f i n e P as t h e c l o s u r e o f M i n - | Q 2 '
p X P t
^ i e
f i b r e product with r e s p e c t
t o t h e maps f ^ . We have n a t u r a l p r o j e c t i o n maps p^ : P -* P^ ( i = 1,2) ,
which c a n be r e g a r d e d as morphisms p.^ : (P,M) (P^Mj . The diagram

(P,M) MP ,M )
0 9
p 2 1 1
f
Pi 2
(P ,M )
1 1 — *(Q,N)
i s a pull-back i n the category o f dense p a i r s . I f P^ and P 2 are
partially complete, then P i s a l s o p a r t i a l l y complete.
Definition 2. A morphism f : (P,M) -» (Q,N) o f dense p a i r s i s s e m i a l g e -
braic, i f the map f : P ->Q i s s e m i a l g e b r a i c (hence f [ M : M -> N i s a l s o
semialgebraic).
The f o l l o w i n g lemma w i l l prove t o be v e r y u s e f u l .
Lemma 8 . 1 . L e t f : (P,M) -> (Q,N) be a morphism o f dense p a i r s such
t h a t f l M : M -* N i s s e m i a l g e b r a i c . Then f i s s e m i a l g e b r a i c .
P r o o f . L e t U € r(Q) be g i v e n . The s e t (f 1M)~ (U fl N) = f ~ (U n N)f1M i s1

1
—1
semialgebraic. Consider a non-empty open subset V o f f (U). Then
1
V n M c f " ( U 0 N ) DM and V n M i s n o t empty. We see t h a t f ~ (U n N) n M 1
-1 -1
i s dense i n f (U). T h e r e f o r e f (U) i s a l s o s e m i a l g e b r a i c (cf. I,
Prop. 4.6).
D e f i n i t i o n 3 . L e t f : (P,M) -» (Q,N) be a morphism between dense p a i r s ,
f i s c a l l e d p a r t i a l l y proper (resp. p r o p e r ) , i f the l o c a l l y semial-
g e b r a i c maps f : P -* Q and M-* N b o t h a r e p a r t i a l l y proper (resp. p r o p e r ) .
1
Notice that t h i s forces M=f~ (N). I f P i s p a r t i a l l y complete then
f : P ->Q i s always p a r t i a l l y p r o p e r . Thus, i n t h i s case, t h e morphism

-i
f i s p a r t i a l l y proper i f and o n l y i f f (N) = M.
D e f i n i t i o n 4. L e t f : (P,M) -+ (Q,N) be a morphism between dense p a i r s ,
a) f i s c a l l e d p a r t i a l l y finite (resp. f i n i t e ) i f t h e maps f : P ->Q and

f|M:M->N both a r e p a r t i a l l y finite (resp. finite).
b) f i s c a l l e d dominant i f the map f : P -»Q i s p a r t i a l l y p r o p e r and the
map f|M: M-»N i s an isomorphism.
Remarks 8.2. a) I f f i s dominant, then we know from Lemma 8.1 t h a t t h e
map f : P -* Q i s s e m i a l g e b r a i c . Thus the map f : P -»Q i s p r o p e r ( c f . I,
6.13), and we c o n c l u d e t h a t t h e morphism f i s p r o p e r . F o r t h e same r e a -
son we know t h a t , i f t h e map f : P->Q i s p a r t i a l l y finite and t h e map
f l M : M -»N i s f i n i t e , then the morphism f i s a l r e a d y finite.
b) A l l t h e p r o p e r t i e s o f morphisms d e s c r i b e d i n D e f i n i t i o n s 2-4 a r e
s t a b l e under c o m p o s i t i o n and p u l l - b a c k s . A l s o , i f f : (P,M) -» (Q,N) and
g : (Q,N) -* (S,L) a r e morphisms o f dense p a i r s and i f g * f has one o f
t h e s e p r o p e r t i e s then f has t h e same p r o p e r t y ( c f . I , 5.5 f o r "proper"
and 1.6.2 f o r " p a r t i a l l y proper"). Notice also that, i f f i s proper,
then f ( P ) = Q i f and o n l y i f f(M) = N. I f f ( P ) = Q and g o f i s p r o p e r
(resp. f i n i t e , resp. dominant) then g i s p r o p e r (resp. f i n i t e , resp.
dominant), c f . 1.5.5. I f Q i s p a r t i a l l y complete, and f ( P ) = Q, and
gof i s p a r t i a l l y proper (resp. p a r t i a l l y finite), then g i s p a r t i a l l y
proper (resp. p a r t i a l l y finite), s i n c e we e a s i l y check t h a t g ^ (L) = N.
c) Any s i m p l i c i a l map f : X ->Y between s t r i c t l y locally finite complexes
y i e l d s a morphism f : (X,X) -+ (Y,Y) o f completed spaces. The s i m p l i c i a l
map f i s p a r t i a l l y proper (resp. p a r t i a l l y finite, p r o p e r , f i n i t e , do-
minant) i n t h e sense o f §5 i f and o n l y i f t h e morphism f o f completed
spaces i s p a r t i a l l y proper (resp. p a r t i a l l y finite, p r o p e r , f i n i t e , do-
minant) . In p a r t i c u l a r , i f t h e morphism f i s dominant, then i t i s a l s o
finite.
Definitions 5. a) A t r i a n g u l a t i o n o f a dense p a i r (P,M) i s an isomor-
phism cp : (Z,X) (P,M) o f dense p a i r s w i t h Z a s t r i c t l y locally finite
complex and X a subcomplex o f Z. In o t h e r words, a t r i a n g u l a t i o n o f
(P,M) i s a simultaneous t r i a n g u l a t i o n o f P and M ( c f . § 4 , Def. 1).

Notice that, i f P i s p a r t i a l l y complete, we must have Z = X ( c f . I , 6.1C)
b) A t r i a n g u l a t i o n o f a morphism f : (P,M) -* (Q,N) o f dense p a i r s i s a
commutative square
(Z,X) *(P,M)
<P
(T,Y) ' (Qf N)
w i t h cp and \p t r i a n g u l a t i o n s o f (P,M) and (Q,N) and g : Z -»T a s i m p l i c i a l
map. (Of c o u r s e , g(X) c Y . )
By §4 e v e r y dense p a i r (P,M) can be t r i a n g u l a t e d . By §6 every finite
morphism f : (P,M) -> (Q,N) can be t r i a n g u l a t e d (Th. 6.13). A dominant
morphism can be t r i a n g u l a t e d i f and o n l y i f i t i s f i n i t e ( c f . Remark
8.2.c).
We want t o study f i n i t e dominant morphisms a l o n g l i n e s s i m i l a r to those
we used when s t u d y i n g s i m p l i c i a l dominant maps between complexes i n §5.
In f a c t t h e r e s u l t s o f §5, as w e l l as those o f §6 and §7, w i l l be quite
useful for this.
D e f i n i t i o n 6. A dense p a i r (Q,N) i s c a l l e d p u r e i f every f i n i t e dominant
morphism (S,L) -> (Q,N) i s an isomorphism. A p u r e h u l l o f a dense p a i r
(P,M) i s a f i n i t e dominant morphism (Q,N) -» (P,M) w i t h (Q,N) pure.
The f o l l o w i n g analogue o f P r o p o s i t i o n 5.5 on maximal complexes can be
proved by t h e same f o r m a l argument.
Proposition 8.3. L e t f : (Q,N) (P,M) and g : (S,L) -> (P,M) be morphisms
between dense p a i r s . Assume t h a t (Q,N) i s pure and g i s f i n i t e and do-
minant. Then t h e r e e x i s t s a unique morphism h : (Q,N) -*> (S,L) w i t h
g»h = f . I f f i s f i n i t e and dominant, i . e . f i s a pure h u l l o f (P,M),

then h i s a l s o f i n i t e and dominant. I n p a r t i c u l a r , any two pure hulls
of (P,M) a r e i s o m o r p h i c .
Our goal i n t h i s section i s t o prove t h e e x i s t e n c e o f pure h u l l s . This
needs some p r e p a r a t i o n .
Definition 7 . L e t (P,M) be a dense p a i r and x a p o i n t o f P. Then t h e

T
set f (W) Q x i s d e f i n e d as t h e p r o j e c t i v e l i m i t o f t h e s e t s n ( U n M) o f
Q
connected components o f U n M, w i t h U r u n n i n g through the d i r e c t e d sy-
stem o f open s e m i a l g e b r a i c neighbourhoods o f x i n P. Of c o u r s e ,
TTQ(M) x i s a one p o i n t s e t f o r x € M.
*)
P r o p o s i t i o n 8.4. L e t (P,M) be a dense p a i r w i t h [TT (M) I = 1

q x for
every x 6 P . I f f : (Q,N) -> (P,M) i s a dominant morphism then a l l f i b r e s
o f t h e p r o p e r map f : Q -+P a r e connected. In p a r t i c u l a r , the p a i r
(P,M) i s pure.
T h i s f o l l o w s from Lemma 6.5 and t h e next e v i d e n t Lemma 8.5 by u s e o f
a t r i a n g u l a t i o n o f (P,M).
Lemma 8.5. I f X i s a s t r i c t l y locally finite complex, then t h e n a t u -
r a l map TT (X) 0 x TT ( S t (x)) i s b i j e c t i v e

q x f o r every x € X. More precisely,
the s e t s := { (1-A)x + A z i z € S t ^ ( x ) } , w i t h A r u n n i n g through ]0,1[,
are a fundamental system o f open s e m i a l g e b r a i c neighbourhoods of xi n
X, and, f o r every A€ ] 0 , 1 [ , t h e map z »-> (1-A) x + Az i s an isomorphism
from S t ( x ) onto x nx .
Theorem 8.6. L e t X^ be a s t r i c t l y locally finite complex. L e t X = X,j
be t h e ( f i r s t ) b a r y c e n t r i c s u b d i v i s i o n o f X-j , and l e t f : Y -> X be t h e
maximal h u l l o f t h e complex X ( c f . § 5 ) . Then In (Y)Î

Q = 1 f o r every
y€Y. Thus the completed space (Y,Y) i s pure ( c f . Prop. 8.4).
*) F o r any s e t E we denote by |E| t h e c a r d i n a l i t y o f E.

Proof. We f i x some x £ X \ X . We want t o prove t h a t S t (y) i s connected y
1
for every y € f (x) . Then we a r e done by Lemma 8.5 above. L e t T denote
the open simplex o f X^ which c o n t a i n s x, and l e t T denote t h e open sim-
p l e x o f X which c o n t a i n s x. The b a r y c e n t e r u o f T i s a v e r t e x o f T,
hence T c S t ^ ( u ) . L e t v ^ , . . . , v f be t h e f i n i t e l y many elements o f f ^ (u).
Since t h e complex Y i s maximal, t h e s t a r s S t ^ v ^ ) , 1 < i < r , a r e connec-
ted (Prop. 7.3). The s e t s V\ := f ( S t ( v ) ) , y i 1 < i < r , a r e t h e connected
components o f S t ( u ) , and f maps every s t a r St.. (v.)

v isomorphically onto
A x 1
V. (Cor. 7.7). The i n c l u s i o n s S t (x) <-+ S t v v (x) and S t ( u ) v St v (u) in-
1 X A-j A A-]
duce b i j e c t i o n s between t h e s e t s o f connected components o f these s t a r s
( c f . Lemma 8.5). Now S t ( x ) x cz S t ( u ) and S t ^ (x) = S t ^ (u) and we con-

x x x
clude t h a t t h e i n c l u s i o n S t ( x ) ^ S t ^ l u ) induces a b i j e c t i o n between

x
the s e t s o f connected components. T h i s means t h a t t h e s e t s V. n S t (x),

1 A
1<i<r, a r e t h e connected components o f S t ( x ) . In p a r t i c u l a r , we see x
that x l i e s i n t h e c l o s u r e V. n S t r ^ u ) o f V . i n S t ^ ( u ) . The preimage

1 A 1 A
1
f (St (u)) x o f S t ( u ) has as connected components Sty(v^),
x 1< i<r,
and f i s a f i n i t e map from t h i s preimage t o S t ^ ( u ) . I t maps every com-
ponent S t — ( v . ) i s o m o r p h i c a l l y onto t h e c l o s u r e V. fl Strr(u) o f V. =

x 1 1 A 1
f ( S t ( v . ) ) i n Strr(u) . Thus x has p r e c i s e l y r preimages y / . . w y . under
x I v
A I r 1 v
f which c a n be numbered such t h a t S t - ( v . ) . The map f sends

Y is
~ (
t ( y
nst (v )
} = f 1 s t
x
( x ) )
y j L
i s o m o r p h i c a l l y onto i t s image, which i s
s t ( x ) n v , n strr(u) = s t ( x ) n v , n s t ( u )
Y v v =st (x)nv.,
v
A 1 A A 1 A A l
a connected s e t . Thus every s t a r S t ( y ^ ) , 1 < i < r , i s y indeed connected.
q.e.d.
Theorem 8.7. Every dense p a i r (P,M) has a pure h u l l .
Proof.We f i r s t c o n s i d e r t h e case t h a t P i s p a r t i a l l y complete. We
choose a t r i a n g u l a t i o n cp : ( X ^ X ^ (P,M) o f (P,M) ( c f . § 4 ) . L e t X = X^
be t h e b a r y c e n t r i c s u b d i v i s i o n o f X^ and f»: Y -> X i t s maximal h u l l . By

Theorem 8.6 t h e morphism f : (Y,Y) -» (X,X) i s a pure h u l l o f (X,X) . Thus
tp«f : (Y,Y) (P,M) i s a pure h u l l o f (P,M). A l s o , by Theorem 8 . 6 ,
| n ( Y ) y | = 1 f o r e v e r y y 6 Y. S i n c e any two pure h u l l s

0 o f (P,M) a r e i s o -
morphic, we know t h e analogous f a c t f o r any o t h e r pure h u l l o f (P,M).
In t h e g e n e r a l c a s e we embed P as a dense subspace i n t o some p a r t i a l l y
complete space P ( c f . § 2 ) . The p a i r (P ,M) has a pure h u l l g : (T,N) ->
(P,M) and I TT^ (N) I = 1 f o r every z £ T. L e t Q be t h e preimage

z o f P under
g, and l e t f : (Q,N) -» (P,M) be t h e morphism o b t a i n e d from g by r e s t r i c -
tion. Clearly f i s f i n i t e and dominant. C o n s i d e r a p o i n t x E Q . Then
U n Q runs through a fundamental system o f neighbourhoods o f x i n Q as U
runs through a fundamental system o f neighbourhoods o f x i n T. Thus
N
0 ( N )
x = n
o ( N )
x' w h e r e n N
§< > x
r
( «sp. T T Q ( N ) ) i s t h e s e t T T ( N )
x q x defined
T
w i t h r e s p e c t t o Q ( r e s p . T) . We know t h a t I TT (N) I = 1 and c o n c l u d e
u x
from P r o p o s i t i o n 8.4 t h a t f i s a pure h u l l o f (P,M). q.e.d.
In t h e c o u r s e o f t h e p r o o f we have c o n s t r u c t e d a pure hull
f : (Q,N) -> (P,M) w i t h | n ( N ) I = 1 f o r e v e r y x € Q . Thus, r e c a l l i n g

N
vj x
P r o p o s i t i o n 8 . 4 , we o b t a i n t h e f o l l o w i n g c o r o l l a r y t o Theorem 8 . 7 .
Corollary 8 . 8 . A dense p a i r (P,M) i s pure i f and o n l y i f |TT (M) I 0 x = 1
for every x G P .
Theorem 8.7 and i t s c o r o l l a r y 8.8 have an analogue i n t h e c l a s s i c a l
theory of cuts i n topology ( c f . [ M ] , [MV]). This w i l l be a p p a r e n t
after a slight change o f t e r m i n o l o g y .
Definition 8. a) A l o c a l l y s e m i a l g e b r a i c s u b s e t A o f a space M i s c a l l -
ed thin i nM i f i t s interior i s empty. I f A i s such a s u b s e t , then we
say t h a t A does n o t c u t t h e space M a t a p o i n t x € A i f | T T ( M ^ A ) I 0 x = 1,
i.e., i f x has a fundamental system o f neighbourhoods U i n M with U ^ A

connected.
b) L e t A be a t h i n s u b s e t o f M. A c u t o f M a l o n g A i s a f i n i t e morphism
f : N •+ M such t h a t f maps N \ f (A) i s o m o r p h i c a l l y onto M\A and

1
B := f (A) i s a t h i n s u b s e t o f N which does not c u t N a t any point
y e B.
Our main r e s u l t s can be e x p r e s s e d i n the f o l l o w i n g terms ( c f . 8.3, 8.7,
8.8) .
Theorem 8.9. L e t A be a t h i n locally s e m i a l g e b r a i c s u b s e t of a space M.
Then t h e r e e x i s t s a c u t f : N-»M o f M a l o n g A. I f g : L-*M is a locally

1
s e m i a l g e b r a i c map and B i s a t h i n s u b s e t o f L such t h a t g (A) c B and B
nowhere c u t s L, t h e n t h e r e e x i s t s a unique locally semialgebraic map
h : L -* N w i t h f • h = g. In p a r t i c u l a r , i f g i s also a cut of M along A,
then h i s an isomorphism.
In the t o p o l o g i c a l t h e o r y o f c u t s t h e " S t e i n f a c t o r i s a t i o n " [MV]
(= " m o n o t o n e - l i g h t factorisation" i n [M]) o f a p r o p e r c o n t i n u o u s map
p l a y s an i m p o r t a n t r o l e . We s h a l l meet the s e m i a l g e b r a i c analogue of
this i n §12.
For l a t e r use we s t a t e a converse to P r o p o s i t i o n 8.4.
Proposition 8.10. L e t f : (Q,N) -* (P,M) be a p r o p e r morphism between
dense p a i r s . Assume t h a t (Q,N) i s pure. Assume f u r t h e r t h a t t h e map
f : Q ->P i s s u r j e c t i v e and has connected f i b r e s . Then (P,M) i s pure.
P r o o f . L e t TT : (P,M) -> (P,M) be a pure h u l l o f (P,M) . By Proposition
8.3 we have a unique morphism ? from (Q,N) to (P,M) w i t h TT«¥ = f . Now ?
i s p r o p e r , hence maps Q onto P. S i n c e f has connected fibres i t follows
t h a t TT has connected f i b r e s . T h i s means t h a t TT i s an isomorphism, i.e.

(P,M) i s pure.
Example 8.11. Even i f the morphism f : (Q,N) -+ (P,M) i s dominant and
has connected f i b r e s , p u r i t y of (P,M) does not n e c e s s a r i l y imply p u r i t y

2
of (Q,N). For example, l e t Q be the c l o s e d u n i t d i s k i n R and let A
1 1
be the c l o s e d t h i n s u b s e t [-•^/-ÎxfO} o f Q. As we shall see in §10
(Prop. 10.4), t h e r e e x i s t s a proper semialgebraic map f :Q P which
c o l l a p s e s A i n t o one p o i n t p and i s an isomorphism from N := QÂ onto
M := P ^ { p } . The morphism f : (Q,N) (P,M) i s dominant and has connected
fibres. (P,M) i s pure, but (Q,N) i s not pure.
We close this s e c t i o n with a very n a t u r a l example o f a dominant
morphism between dense p a i r s .
Example 8.12 (Semialgebraic blowing up). Let (P,M) be a dense p a i r w i t h
A: =PNM c l o s e d i n P. We choose r>2 l o c a l l y s e m i a l g e b r a i c f u n c t i o n s f ^ , . . . , f
on P such t h a t A i s the s e t of common z e r o s of t h e s e f u n c t i o n s . {Actual-
l y A can be w r i t t e n as the zero s e t o f a s i n g l e l o c a l l y s e m i a l g e b r a i c function
[ I , Th. 4.15], but f o r r = 1 the considerations to f o l l o w are trivial.}

n
As usual, l e t TP (R) denote the complete space o f r e a l p o i n t s of the
n - d i m e n s i o n a l p r o j e c t i v e space o v e r R. Let Z be the closed locally

r-1
semialgebraic s u b s e t o f PxJP (R) c o n s i s t i n g of a l l p o i n t s (x,y^:...:y )
with
f (x)y
i j - fj(x) y i = 0
r-1
for 1 <i< j<r- The p r o j e c t i o n p from Pxjp (R) to the first factor
i s a proper semialgebraic map. By r e s t r i c t i o n p y i e l d s an isomorphism
from N:=p ^(M)DN t o M, s i n c e f o r every x€M a t l e a s t one o f the func-
t i o n s f ^ does not v a n i s h . L e t Q denote the c l o s u r e o f N i n Z. Then the
restriction IT : Q -> P i s a l s o a p r o p e r s e m i a l g e b r a i c map, and TT can be
r e g a r d e d as a dominant p r o p e r morphism from the dense p a i r (Q,N) to
(P,M).
We know almost n o t h i n g about s e m i a l g e b r a i c blowing up. The dimension
of t h e f i b r e TT ^ (x) o f a p o i n t x € A i s a measure how much t h e f u n c t i o n s
f^,...,f r a r e "independent" near x. Thus s e m i a l g e b r a i c blowing up seems
to be an i n t e r e s t i n g geometric d e v i c e t o study f i n i t e systems o f
l o c a l l y semialgebraic functions.
Question. I f (P,M) i s pure, under which c o n d i t i o n s on ( f , . . . , f ) i s

1 r
(Q,N) again pure?

§9 - Ends o f spaces, the L O s t r a t i f i c a t i o n
In t h i s s e c t i o n we d e v e l o p a t h e o r y o f "ends" f o r r e g u l a r paracompact
spaces. I t might be a d v i s a b l e f o r the r e a d e r t o pursue t h e l i t e r a t u r e

1
on t h e c l a s s i c a l t h e o r y o f ends i n t o p o l o g y , e.g. F r e u d e n t h a l s origi-
1
n a l papers [ F r ] , [ F r ^ ] o r §1 o f H o p f s paper [Ho]. T h i s w i l l g i v e him
a good f e e l i n g f o r t h e d i f f e r e n t " f l a v o u r " of the semialgebraic theory
as opposed t o t h e c l a s s i c a l t o p o l o g i c a l t h e o r y . I n some sense our t h e o -
ry i s much e a s i e r than the c l a s s i c a l one, s i n c e from t h e view p o i n t o f
g e n e r a l t o p o l o g y our spaces and maps a r e r a t h e r s p e c i a l . So a l o t o f
p a t h o l o g i e s cannot o c c u r . However, i n o t h e r r e s p e c t s our t h e o r y i s more
c o m p l i c a t e d than t h e c l a s s i c a l one. We s h a l l meet new phenomena, i n
particular t h e " c o m p l e x i t y " o f a space ( c f . D e f i n i t i o n s 7 and 8 below),
which a r e u n f a m i l i a r t o t o p o l o g i s t s . The r e a s o n i s t h a t we do not r e -
s t r i c t o u r a t t e n t i o n t o l o c a l l y complete spaces. Thus t h e ends o f a
space M i n some c o m p l e t i o n o f M ( c f . Def. 3 below) a r e n o t n e c e s s a r i l y
complete.
I f M i s s e m i a l g e b r a i c and l o c a l l y complete, then t h e ends o f M can be
regarded as p o i n t s which a r e added t o M t o make M a complete space ( c f .
Th. 9.2 below). In t h i s s p e c i a l case our t h e o r y o f ends i s more o r l e s s
an a d a p t i o n o f F r e u d e n t h a l ' s t h e o r y t o t h e s e m i a l g e b r a i c s e t t i n g . But
already i f M i s l o c a l l y complete and not s e m i a l g e b r a i c , our t h e o r y o f
ends i s d e f i n i t e l y d i s t i n c t from the t o p o l o g i c a l t h e o r y . For example,
if N i s a l o c a l l y complete s e m i a l g e b r a i c space, then t h e space N ^ q c has
no ends a t a l l i n our t h e o r y , s i n c e N ^ Q c i s p a r t i a l l y complete. But t h e
s t r o n g t o p o l o g i e s on N ^ q c and N a r e t h e same. So, i n t h e case R -• 3R ,
N and N ^ Q C have t h e same ends i n t h e t o p o l o g i c a l sense, and t h e r e may
be many o f t h e s e . To g i v e y e t another example, l e t X be a f i n i t e c l o s e d
connected geometric simplicial complex over IR , and l e t p : X -»X be the
u n i v e r s a l c o v e r i n g o f X, c f . Chapter V. Then X i s p a r t i a l l y complete,

and thus has no ends i n our t h e o r y . On the o t h e r hand, the number o f
topological ends of X i s a v e r y i n t e r e s t i n g i n v a r i a n t e(G) o f the f u n -
damental group G of X (e(G) = 0,1,2 o r «>, c f . [Ho], [SW, §5]).
Our t h e o r y of ends needs some p r e p a r a t i o n : As b e f o r e a "space" i s a r e -
g u l a r paracompact l o c a l l y s e m i a l g e b r a i c space o v e r a f i x e d r e a l closed
field R, and a "complex" i s a geometric s i m p l i c i a l complex over R. We
repeat a d e f i n i t i o n from §1.
Definition 1 . A c o m p l e t i o n o f a space M i s a dense embedding cp : M «-> P
of M i n a p a r t i a l l y complete space, i . e . an isomorphism from M onto a
dense subspace cp(M) € T(P) of P.
In t h i s s e c t i o n i t i s sometimes more c o n v e n i e n t t o work w i t h c o m p l e t i o n s
i n s t e a d o f completed spaces, as we d i d i n the l a s t s e c t i o n . Of c o u r s e ,
t h e s e a r e e q u i v a l e n t n o t i o n s . I f cp : M *+ P i s a c o m p l e t i o n then (P,cp(M))
i s a completed space, and i f (M,M) i s a completed space then t h e inclu-
s i o n M «-» M i s a c o m p l e t i o n .
We are m a i n l y i n t e r e s t e d i n the completions of a f i x e d space M. In this
c o n t e x t a morphism from a c o m p l e t i o n cp : M *-> P t o a c o m p l e t i o n i|/ : M -+ Q
i s a l o c a l l y s e m i a l g e b r a i c map f : P -* Q w i t h f«cp = \p. We denote such a
morphism by f : P-^Q. The locally s e m i a l g e b r a i c map f i s automatically
p r o p e r , c f . Remark 8.2a, hence a l s o s u r j e c t i v e . The preimage of Q ^ i|> (M)
i s P \ cp(M) .
There e x i s t s a t most one morphism from cp t o ip. Thus t h e s e t #(M) of iso-
morphism c l a s s e s [cp] of c o m p l e t i o n s cp o f M has a n a t u r a l p a r t i a l o r d e r -
ing: [cp] P-
1 and cp : M *-> P~
? a r e two c o m p l e t i o n s o f M, then t h e r e e x i s t s
T h e
a s m a l l e s t element [cp] i n D(M) w i t h [cp] > [cp^ ], [cp] > U P ] .2 comple-
t i o n cp : M *-> P i s o b t a i n e d as f o l l o w s . Take t h e embedding (cp^ ,cp ) : M2
P x P . L e t P be t h e c l o s u r e o f t h e image o f M i n P^xp,^. Then cp i s the

1 2
r e s t r i c t i o n M <-* P o f (cp^ ,cp ) .2
I f M i s a l r e a d y p a r t i a l l y complete, then JO (M) c o n t a i n s o n l y one element,
cf. I , Cor. 6.10. I f M i s s e m i a l g e b r a i c and l o c a l l y complete - b u t not
complete - then $ (M) has a s m a l l e s t element, namely the one p o i n t

+
c o m p l e t i o n M <-> M , c f . I , Prop. 7.6.
In g e n e r a l , the s e t J) (M) has n e i t h e r a maximal nor a minimal element.
T h i s i s a major d i f f i c u l t y i n u n d e r s t a n d i n g the v a r i o u s c o m p l e t i o n s o f
M. I t seems to be wise t o f o c u s a t t e n t i o n t o t h e "pure c o m p l e t i o n s " o f
M.
Definition 2. A morphism f : Q P from a c o m p l e t i o n : M <-* Q t o a com-
p l e t i o n cp : M «-> P o f M i s c a l l e d f i n i t e , i f t h e f i b r e s o f t h e map f : Q-> P
are discrete, hence f i n i t e , i.e. i ff is a finite s e m i a l g e b r a i c map. A
c o m p l e t i o n cp : M «-> P i s c a l l e d pure i f every f i n i t e morphism from a com-
p l e t i o n o f M t o cp i s an isomorphism. A pure h u l l o f a c o m p l e t i o n
cp : M «-> P i s a f i n i t e morphism 0 ^ P from a pure c o m p l e t i o n \

\
> : M «-> Q t o cp.
These d e f i n i t i o n s a r e a d a p t i o n s o f some n o t i o n s from §8 t o t h e p r e s e n t
needs. F o r example, a c o m p l e t i o n cp : M *-» P i s pure i f and o n l y i f t h e
completed space (P,cp(M)) i s pure. We know from §8 t h a t every c o m p l e t i o n
cp : M P admits a pure h u l l (Th. 8.7) . Any two pure h u l l s a r e isomor-
phic (Prop. 8.3). So, we o f t e n t a l k o f "the" pure h u l l o f cp. More gene-
r a l l y we have t h e f o l l o w i n g p r o p o s i t i o n , which i s a restatement o f Pro-
position 8.3 i n a s p e c i a l c a s e .
P r o p o s i t i o n 9.1. L e t f : Q P be a morphism from a pure c o m p l e t i o n

ij; : M «-> Q t o a c o m p l e t i o n cp : M *-» P. F u r t h e r , l e t g : T P be a f i n i t e
morphism from a c o m p l e t i o n x • M «-> T t o cp. Then t h e r e e x i s t s a unique
morphism h : Q T from i|; t o x w i t h g*h = f .
Definition 3. I f cp : M «-> P i s a c o m p l e t i o n o f t h e space M, then a connec-
t e d component o f t h e space P^cp(M) i s c a l l e d an end o f M i n P (with res-
p e c t t o cp) . The s e t n ( P ^ c p ( M ) )
0 o f ends o f M i n P w i l l u s u a l l y be deno-
t e d by e (M,cp) .
I f M i s s e m i a l g e b r a i c , then P i s s e m i a l g e b r a i c , and e (M,cp) i s a finite
set. In general, e (M,cp) may be i n f i n i t e , b u t i s c o u n t a b l e as long as M
has o n l y c o u n t a b l y many connected components, c f . I , Th. 4.17.
I f M i s s e m i a l g e b r a i c and l o c a l l y complete then M has a pure c o m p l e t i o n
such t h a t t h e ends i n t h i s c o m p l e t i o n a r e p o i n t s . More p r e c i s e l y ,
Theorem 9.2. L e t M be s e m i a l g e b r a i c and l o c a l l y complete. There exists
- up t o isomorphism - a unique pure c o m p l e t i o n cp : M *-> P such t h a t the
ends o f M i n P a r e one p o i n t s e t s . F o r any pure c o m p l e t i o n \p : M «-> Q
t h e r e e x i s t s a (unique) morphism f : Q P from t o cp.
P r o o f . I f M i s a l r e a d y complete, t h e r e i s n o t h i n g t o prove. Assume now
t h a t M i s n o t complete. L e t cp : M «-» M Q
+
be t h e one p o i n t c o m p l e t i o n o f
M ( I , §7) . L e t cp : M <-> P, p : P be a pure h u l l o f c p . The ends o f

Q
M i n P a r e f i n i t e l y many p o i n t s , namely t h e p o i n t s o f t h e f i b r e p ^ («>) .
I f ijj : M «-> Q i s any c o m p l e t i o n o f M then, by I , Prop. 7.6, we have a
unique morphism f Q :Q M +
> t o cp . I f \
from \
\ Q \
> i s pure then, by Prop.
9.1, we have a unique morphism f : Q ^ * P from t o cp w i t h p «f = f ^ . By
Proposition 8.4 t h e f i b r e s o f f a r e c o n n e c t e d . I f t h e ends o f M i n Q
a r e a l s o one p o i n t s e t s , t h e n f must be b i j e c t i v e , hence an isomorphism
(since f i s proper). q.e.d.

We r e t u r n t o an a r b i t r a r y r e g u l a r paracompact space M. E v e r y morphism
f : Q -yf P from a c o m p l e t i o n vp : M «-»• Q t o a c o m p l e t i o n cp : M P induces a
s u r j e c t i o n from t h e s e t e(M,iJ0 = TT (Q \ \p (M) ) t o t h e s e t e (M,cp) = q
TT (P ^ cp(M) ) . We denote
0 t h i s s u r j e c t i o n by KU',CP). I f B i s an end o f M
i n Q, then K , cp) (B) i s t h e end A o f M i n P w i t h A^f(B).
P r o p o s i t i o n 9 . 3 . Assume t h a t the c o m p l e t i o n cp i s pure. Then K(ip,tp) i s
bijective. I f B i s an end o f M i n Q, and A i s the end K(ip,cp) (B) o f M i n
P, 1
t h e n B = f " ( A ) and A = f ( B ) .
P r o o f . f g i v e s by r e s t r i c t i o n a proper s u r j e c t i v e map from Q ^ i|/ (M) to
P^cp(M). By P r o p o s i t i o n 8.4 t h e f i b r e s o f t h i s map a r e c o n n e c t e d . Thus
the c o n n e c t e d components o f Q^4/(M) a r e t h e preimages o f t h e connected
components o f P\cp(M), c f . Sublemma 6.6.
I f cp^ : M «-> P^ and cp : M 2 P 2 a r e two c o m p l e t i o n s and i f cp^ i s pure,
t h e n we o b t a i n a s u r j e c t i o n
:
K(<P<|r<P) 2 ^(M^cp^ -» e(M,tp ) 2
as f o l l o w s . We choose a c o m p l e t i o n ij; : M «-> Q which admits morphisms
f
i :
^"M* iP
^ = 1 2
' ^ t o b o t n
^1 a n d
^2* T n e n
Kdj^cp^) i s a b i j e c t i o n .
We define
1
K(cp ,cp ) := K (ij^,cp ) K (ip,cp )
1 2 2 1 .
The following f a c t s a r e now e a s i l y v e r i f i e d . (To prove t h e f i r s t
statement, compare i|/ w i t h t h e " s m a l l e s t " c o m p l e t i o n cp o f M which ad-
m i t s morphisms t o cp^ and cp , as e x p l a i n e d above.) 2
P r o p o s i t i o n 9.4. The s u r j e c t i o n K(cp^,cp ) does not depend on t h e c h o i c e

2
of \p. I f cp i s a l s o pure, then «(cp^,cp ) i s b i j e c t i v e ,

2 2 t h e i n v e r s e map
being K(CP ,CP ).

2 1 I f cp^ i s a t h i r d c o m p l e t i o n , and both cp and cp a r e 1 2
pure, then
K(CP ,(P )
1 3 = K(cp ,tp )
2 3 K(CP ,CP ).
1 2
For a d e s c r i p t i o n o f the map K(tp^,cp ) which does not use an

2 auxiliary
completion , see Cor. 9.12 below.
We now d e v e l o p an " a b s o l u t e " n o t i o n o f ends which does not r e f e r t o any
c o m p l e t i o n . L e t 3* (M) be the s e t o f a l l p a r t i a l l y complete

c locally semi-
algebraic s u b s e t s A o f M. Every K€ T c (M) i s a c l o s e d s u b s e t of M ( c f . I,
Cor. 6.10). F o r any two elements and K ? of T (M) c the union UK 2 is
a g a i n an element o f 7^(M). In p a r t i c u l a r , the s e t (M), o r d e r e d by i n -
clusion, i s a directed system of subsets of M.
Definition 4 ("absolute" e n d s ) . An end X o f a space M i s an assignment
(MNK)^ o f a connected component (M^K) A o f M^K t o every K € T Q (M)
such t h a t , whenever K c L , (M s K) ^ c o n t a i n s (M^L)^. In o t h e r words, an
end i s an element o f the s e t
c(M) ;= j l i m TTQ ( M S K) .
K€7 (M)
Remark. I t i s not d i f f i c u l t to v e r i f y t h a t the s e t o f ends of a locally
compact and locally connected topological space M i n the sense o f F r e u -
denthal [ F r ] i s t h e p r o j e c t i v e l i m i t o f the s e t s n ( M \ K ) w i t h K Q runn-
ing through the compact s u b s e t s o f M.
We want to e s t a b l i s h f o r every pure c o m p l e t i o n cp : M <-* P a n a t u r a l b i -
j e c t i o n < (cp) from e (M) t o the s e t £(M,cp) o f ends o f M i n P.
Lemma 9.5. L e t cp : M <-+ P be a c o m p l e t i o n o f M. For every K e T (M) the
map cp from
+ T T ( M \ K ) t o TT (P \ cp (K) ) , induced by cp, i s s u r j e c t i v e . I f
q 0
cp i s pure t h e map cp* i s b i j e c t i v e . Moreover, i n t h i s c a s e , the connec-
t e d component cp (A) of Pscp(K) i s the c l o s u r e of cp(A) i n Pscp(K) o f

+ any
A€n (MsK),
n
Proof. We assume w i t h o u t l o s s o f G e n e r a l i t y t h a t M i s a subspace o f P
and cp i s t h e i n c l u s i o n map. cp* i s c l e a r l y surjective, since M ^ Ki s
dense i n P ^ K . More p r e c i s e l y , every connected component C o f P ^ K i s
the u n i o n o f t h e c l o s u r e s : = A n (PNK)
± i n PNK o f a f a m i l y o f compo-
nents ( A l i € I ) o f M^K.
i ( N . B . The f a m i l y (AîGI) i s locally finite.)
Now assume t h a t cp i s pure. We have t o v e r i f y t h a t t h e c l o s u r e s A^ , A 2
i n P ^ K o f any two d i f f e r e n t components A^ and A 2 of M ^ K are d i s j o i n t .
Suppose t o t h e c o n t r a r y that there e x i s t s some p o i n t x € A-^ n A . 2 Since
(P,M) i s pure t h e r e e x i s t s an open neighbourhood U o f x i n P s K such
t h a t UflM i s connected ( c f . C o r . 8.8). UPIM meets b o t h A^ and A . Thus 2
unMcA 1 and U D M c A , a c o n t r a d i c t i o n .
2 q.e.d.
Lemma 9.6. L e t (M,M) be a completed space. F o r any K E T, (M) there exists
some L € T (M) w i t h L D K such t h a t t h e n a t u r a l map from n^(M\M) t o

c u
TTQ(MSL) i s a b i j e c t i o n .
Proof. C h o o s i n g a s i m u l t a n e o u s t r i a n g u l a t i o n o f M, M, and K, we assume
without l o s s o f g e n e r a l i t y that M i s a s t r i c t l y locally finite complex
w i t h c l o s u r e M and t h a t K i s a c l o s e d subcomplex o f M c o n t a i n e d i n M.
In t h e f o l l o w i n g we argue i n a p u r e l y combinatorial way. The assumption
that M i s l o c a l l y f i n i t e w i l l n o t be needed. L e t ( A l a € I ) a be the f a m i l y
of components o f t h e complex M \ M and l e t U be t h e s t a r S t - ( A ) of A„

1
in the barycentric subdivision X = M o f M. The s e t s u* a r e p a i r w i s e
a
d i s j o i n t by t h e f o l l o w i n g g e n e r a l lemma,to be proved later.
Lemma 9.7. L e t Y be a complex and l e t C^ and C 2 be subcomplexes o f Y
with C 1 DC 2 = 0 and C n C1 2 = 0 . Then t h e s t a r s S t , { C ^ ) and S t , ( C ) o f

y y 2
the i n the ( f i r s t ) barycentric s u b d i v i s i o n Y' o f Y a r e d i s j o i n t .

We c o n t i n u e p r o v i n g Lemma 9.6. Every complex i s open i n X and i s
connected, s i n c e i s connected. i s contained i n X^K', s i n c e U Q
is the smallest open subcomplex o f X c o n t a i n i n g A^. L e t L be t h e com-
plement i n X o f t h e open subcomplex U ( U l a € I ) . a This i s a c l o s e d sub-

1
complex o f X w i t h K' <= L c X : = M . Now X s L i s t h e d i s j o i n t u n i o n o f
the U . Thus
Q (u* la€I)
a i s t h e f a m i l y o f components o f t h e complex X ^ L ,
while (A^laGI) i s t h e f a m i l y o f components o f X ^ X . The map n ( X \ X ) -> 0
n (X\L)
Q i sbijective. q.e.d.
We s t i l l have t o prove Lemma 9.7. F o r t h a t we may assume t h a t t h e com-
plex Y i s c l o s e d . L e t be an open simplex o f Y c o n t a i n e d in (i=1,2
We v e r i f y t h a t t h e s t a r s S t ^ , ( o ^ ) and S t ^ , ( c ^ ) a r e d i s j o i n t . Assume on
the c o n t r a r y that there e x i s t s a simplex T o f Y' which i s c o n t a i n e d i n
both s t a r s S t , (o^) and S t , ( a ) . We have a c h a i n

v Y 2 T <
q < ... < x o f r
open s i m p l i c e s of Y with T = ] T , t ^ , . . . , t [ ,
q r where denotes t h e
b a r y c e n t e r o f T^. Then T , T . J , . . . , T
0 a r e p r e c i s e l y a l l open s i m p l i c e s o f
Y which have non-empty i n t e r s e c t i o n w i t h T. Thus = and a 2 =
for s u i t a b l e i n d i c e s i , j € {0, . . . ,r} . But, s i n c e DC 2 and c 1 fl C a r e

2
both empty, n e i t h e r is a face o f a 2 nor i s a 2 a face of . This con-
t r a d i c t i o n proves t h a t indeed S t , ( a ^ ) and S t , ( o )

Y Y 2 are d i s j o i n t .
The lemmas 9.5 and 9.6 imply t h e f o l l o w i n g theorem.
Theorem 9.8. F o r any c o m p l e t i o n cp : M P there exists a natural surjec-
tion
K (cp) : e (M) -» e (M,cp) ,
t o be d e s c r i b e d below e x p l i c i t l y , from t h e s e t e(M) o f ends o f M t o t h e
set e (M,cp) o f ends o f M i n P. I f cp i s pure, then K (cp) i sbijective. If
there e x i s t s a morphism from cp t o another c o m p l e t i o n ij; o f M, then t h e
triangle
e(M,e)
K (CP)
€(M) K (CP , l M
K (IP)
^ e(M^)
commutes. I f cp i s a pure c o m p l e t i o n o f M and \p an a r b i t r a r y comple-
t i o n o f M, then t h i s triangle s t i l l commutes.
P r o o f . We o b t a i n t h e map <(cp) as f o l l o w s . We have a n a t u r a l map
cp* : e(M) = l i m TT (M^K) -* lim_ TT (P s cp (K))
w i t h K r u n n i n g through T (M). By Lemma 9.5 t h i s i s a b i j e c t i o n i f cp
i s p u r e . By Lemma 9.6 we have a n a t u r a l bijection
u:e(M,cp) = TT (Pscp(M))
0 ^ l i m n ( P vtp(K)) .
Q
K(cp) i s d e f i n e d as t h e composite o f cp* w i t h u ~ . The c o m m u t a t i v i t y o f

1
the t r i a n g l e above i s o b v i o u s i n both cases from t h e d e f i n i t i o n s o f
K(CP), K{\\}), and K ( c p , ^ ) . I f ij; i s an a r b i t r a r y completion o f M, t h e n
t h e r e e x i s t s a morphism from a pure c o m p l e t i o n cp o f M t o lp, and we
l e a r n from t h e commutative t r i a n g l e above t h a t K(ip) i s s u r j e c t i v e .
The map K (cp) can be d e s c r i b e d n e a t l y i n terms o f p a t h s , as we shall
now explain.
Definition 5. An i n c o m p l e t e path a : [0,1[ -+ M ( c f . I , § 6 , Def. 2) is
c a l l e d proper i f a i s a proper s e m i a l g e b r a i c map.
Remark 9.9 9 F o r any i n c o m p l e t e path a i n M, the f o l l o w i n g properties
are e q u i v a l e n t .
a) a i s p r o p e r .
b) a ( [ 0 , 1 [ ) i s n o t c o n t a i n e d i n any s e t KCTîn).
c) cx([0,1[) i s c l o s e d i n M, b u t not complete.

Proof. c) b) : I f a ( [ 0 , 1 [ ) were c o n t a i n e d i n some K € (M) , then
ot([0,1[) would be a c l o s e d s e m i a l g e b r a i c s u b s e t o f K, hence would be
complete.
b) => a ) : I t s u f f i c e s t o v e r i f y t h a t , f o r a g i v e n complete semialgebraic

_1
s u b s e t L o f M , t h e preimage a (L) i s a l s o complete ( I , Cor. 6.11). The
set a (L) i s c l o s e d and s e m i a l g e b r a i c i n [0,1[. Thus i t i s a d i s j o i n t
union o f f i n i t e l y many one p o i n t s e t s , c l o s e d i n t e r v a l s , and perhaps
an i n t e r v a l [c,1[ w i t h c € [ 0 , 1 [ . Suppose we had a([c,1[)<=L f o r some
c C [ 0 , 1 [ . Then a ( [ 0 , 1 [ ) would be c o n t a i n e d i n t h e complete s e t
a ( [ 0 , c ] ) UL, i n c o n t r a d i c t i o n t o t h e assumption b ) . Thus a ^ (L) i s a

union o f f i n i t e l y many p o i n t s and c l o s e d i n t e r v a l s . In p a r t i c u l a r ,
-1
a (L) i s complete.
a) c) : S i n c e a i s p r o p e r , L := a ( [ 0 , 1 [ ) i s a closed semialgebraic
subset o f M . But L i s n o t complete, because i t s preimage a ^ (L) =
[0,1[ i s n o t complete.
Definition 6. L e t a be a p r o p e r incomplete path i n M. We d e f i n e an end
X € e(M) as f o l l o w s . We a s s i g n t o every K € T ( M ) t h e connected

c component
(M \ K) ^ o f M s K w h i c h c o n t a i n s a ( [ c , 1 [ ) f o r some c € ] 0 , 1 [ . We c a l l X
the end o f M d e t e r m i n e d by a, and denote t h i s end by e ( a ) .
P r o p o s i t i o n 9.10. I f cp : M «-> P i s a c o m p l e t i o n o f M and a i s a p r o p e r
incomplete path i n M , then t h e map K (cp) : e (M) -» e (M,cp) sends e (a) t o
the connected component e (a,cp) o f P ^ c p ( M ) which c o n t a i n s t h e end p o i n t
cp°a(1) o f t h e c o m p l e t i o n cp«a : [0,1] -> P o f t h e incomplete path cp*a i n P
(cf. I, C o r . 6.9).
T h i s i s p r e t t y e v i d e n t from t h e d e f i n i t i o n o f K (cp) . The p r o p o s i t i o n
g i v e s a v i v i d d e s c r i p t i o n o f t h e map K (cp) as soon as we know t h a t t h e
f o l l o w i n g theorem h o l d s .
Theorem 9.11. For every end X of M there e x i s t s some p r o p e r incomplete
path a i n M w i t h e(a) = X .
P r o o f . We choose a pure c o m p l e t i o n cp : M <-> P and assume w i t h o u t loss of
generality t h a t M i s a subspace of P and cp i s t h e i n c l u s i o n map. Let
A be an end of M and l e t A€ TT (P\M)

0 be the corresponding end K (cp) (A)
of M i n P. We choose a p o i n t x G A and then a path 3 : [0,1] -» P w i t h
3 ([0,1 [ ) c M and 3d) = x. L e t a be t h e proper incomplete path 3 | [0,1 [
i n M. Then x (cp) U(cc)) = A by P r o p o s i t i o n 9.10. But we a l s o have
<(cp)(A) = A. S i n c e K (cp) i s b i j e c t i v e ( c f . Th. 9.8), we have A = e (a) .
q.e.d.
From our d e s c r i p t i o n of the map K (cp) i n terms o f paths we immediately
o b t a i n a d e s c r i p t i o n o f the maps K(CP,IJJ), defined e a r l i e r , i n terms o f
paths as follows.
Corollary 9.12. L e t M «j P and Q be c o m p l e t i o n s o f M. Assume t h a t
e i t h e r t h e r e e x i s t s a morphism from cp t o 4> o r t h a t cp i s p u r e . L e t
A £ e (M,cp) = 7 T ( P s c p ( M ) ) be g i v e n . Then the component B

0 := K (cp, ip) (A) o f
Qsi|i(M) can be d e s c r i b e d as f o l l o w s . L e t a : [0,1] -> P be a path w i t h
a ([0, 1[) c M and a (1 ) G A. L e t 3 : [0, 1 ] -> Q be the unique path i n Q with
3(t) = i j ^ c p " * a (t)

1
for 0 < t < 1 . (N.B. 3 e x i s t s by the path completion
criterion I, Cor. 6 . 9 ) . Then B i s the component o f Q s (M) c o n t a i n i n g
3d ) .
Remark. I t i s p o s s i b l e t o v e r i f y this corollary directly, starting from
the d e f i n i t i o n o f K(CD,IJJ), v i a an a u x i l i a r y pure c o m p l e t i o n x with
[x3>tcp]/[x]>[^jf a n d
not u s i n g the n o t i o n o f an " a b s o l u t e " end. We
leave t h i s as e x e r c i s e to the r e a d e r .
We briefly d i s c u s s the f u n c t o r i a l b e h a v i o r o f the s e t o f ends under par-
tially p r o p e r maps. G i v e n a p a r t i a l l y p r o p e r map f : M ->N between spaces

we o b t a i n a map f * : e (M) -» e (N) as f o l l o w s . L e t A € e (M) be g i v e n . Then
f (X)
+ i s t h e unique end y o f N such t h a t , f o r every L E T (N), (N \ L)
c y
-1
i s t h e component of N ^ L which c o n t a i n s the connected s e t f ( (M ^ f (L) ) ^) .
Notice that t h i s makes sense, s i n c e f ^ (L) £ T c (M) . In t h i s way we ob-
t a i n a f u n c t o r e from the c a t e g o r y of spaces and p a r t i a l l y proper maps
over R t o t h e c a t e g o r y o f s e t s . For example, i f a : [0,1[ -*M i s a proper
i n c o m p l e t e p a t h , then the end e(a) d e f i n e d above i s none o t h e r than the
image o f t h e unique end o f [0,1[ under <x .

+
S i m i l a r l y we have an "end functor" e from t h e c a t e g o r y o f completed
spaces t o t h e c a t e g o r y o f s e t s . {Now i t seems more c o n v e n i e n t t o work
w i t h completed spaces (M,M) than w i t h c o m p l e t i o n s (p : M ^ P.} For any
completed spaces (M,M) the ends o f (M,M) a r e , o f c o u r s e , the ends o f
M i n M w i t h r e s p e c t t o the i n c l u s i o n map cp : M M, i . e . the connected
components o f M^M. We denote the s e t o f ends o f (M,M) by e(M,M). I f
f : (M,M) -> (N,N) i s a p a r t i a l l y proper morphism between completed spaces
( c f . §8, Def. 3) , then f ~ (N) = M, 1

and the r e s t r i c t i o n g : M s M -> N^N
of f induces a map ^ (g)

Q from e (M,M) = TT (M^M)
Q to e(N,N) = TT (N^N).
Q
We denote t h i s map s i m p l y by f * . We f u r t h e r denote the n a t u r a l map K (cp)
from e(M) t o e(M,M) by ^ . Recall that, i n the case t h a t (M,M) is
pure,«^— M j i s b i j e c t i v e . We have a n a t u r a l commutative diagram
t (M) >e(M,M)
K
(M,M)
(*)
e(N)- •e(N,N)
(N,N)
The f o l l o w i n g p r o p o s i t i o n i s an e x t e n s i o n o f our o b s e r v a t i o n s above
on t h e map ic((p,ij;) from e (M,tp) t o z(M,ty) induced by a morphism P Q
between two c o m p l e t i o n s tp : M P and \\> : M Q.
P r o p o s i t i o n 9.13. a) I f f : M-»N i s a p r o p e r s u r j e c t i v e map, then

f * : £ (M) -» £ (N) i s s u r j e c t i v e . If, i n addition, the f i b r e s of f are
connected, then f * i s b i j e c t i v e .
b) I f f : (M,M) -+ (N,N) i s a p r o p e r morphism from a completed space
(M,M) to a pure completed space (N,N), such t h a t f ( M ) = N and t h e f i b r e s
of t h e r e s t r i c t e d map f l M : M -»N a r e connected, then f (M) = N and t h e
f i b r e s o f f : M -* N a r e connected. In t h i s case f * : £ (M,M) £ (N,N) i s
bijective.
P r o o f . I f L runs throuah y (N), then f (L) runs throuah a c o f i n a l sub-

c
set o f the d i r e c t e d s e t (M) , s i n c e f i s p r o p e r . (N.B. P a r t i a l p r o p e r -
ness does not s u f f i c e f o r t h i s . ) f clearly yields a surjection
f * : TT ( M s f ~ (L)) -* TTQ (N s L) f o r every L € y (N) . I f t h e f i b r e s o f f a r e

0
1
connected, then a l l t h e s e maps f * a r e b i j e c t i v e , c f . Sublemma 6.6.
(N.B. A g a i n p a r t i a l p r o p e r n e s s does n o t s u f f i c e forthis conclusion).
Thus. f + : £ (M) -+ £ (N) i s b i j e c t i v e . I n o r d e r t o prove t h e s u r j e c t i v i t y
of t h i s map i n t h e g e n e r a l case we extend f t o a morphism (M,M) -* (N,N)
between completed pairs with (N,N) pure ( c f . Prop. 5.1 and Th. 8.7;
(M,M) c o u l d be chosen pure as w e l l ) . We denote t h i s morphism a g a i n by
f . We a r e i n t h e s i t u a t i o n d e s c r i b e d above b e f o r e P r o p o s i t i o n 9.13.
But now f : M - * N i s p a r t i a l l y proper and s e m i a l g e b r a i c ( c f . Lemma 8.1),
hence p r o p e r . T h i s i m p l i e s t h a t the r e s t r i c t i o n g : MsM -> N s N o f f i s
surjective. Thus, i n t h e diagram (*) above the r i g h t v e r t i c a l arrow
is surjective. K ,— i s b i j e c t i v e and K ,TZ i s surjective. T h i s im-

(N,N) J
(M,M)
plies that f + : £ (M) -> £ (N) i s s u r j e c t i v e . I f t h e f i b r e s o f f : M-> N
are connected then a l s o t h e f i b r e s o f g a r e connected (Lemma 6.5). In
t h i s case the r i g h t v e r t i c a l arrow i n (*) is b i j e c t i v e .
We t u r n t o a d i f f e r e n t type o f q u e s t i o n . Given two pure c o m p l e t i o n s
M -jjj* P and M Q o f M, l e t A G £(M,cp) and B€e(M,ijO be ends o f M i n P
and Q w i t h K(cp,^)(A) = B. A r e t h e r e any g e o m e t r i c p r o p e r t i e s which t h e
spaces A and B have i n common?

We shall g i v e a p o s i t i v e answer f o r two very crude p r o p e r t i e s . The
first one i s s e m i a l g e b r a i c i t y . I t i s e a s i l y seen t h a t A i s s e m i a l g e b r a i c
if and only i f B i s s e m i a l g e b r a i c . A l s o , i f A € e(M) i s the a b s o l u t e end
of M with K(ip) ( A ) = A, then A i s s e m i a l g e b r a i c i f and o n l y i f (M \ K) ^
is semialgebraic f o r some K € 3"MM) . We call such an end A semialqebraic.
The second p r o p e r t y i s "complexity". The complexity c(M) o f a space M
i s a non negative i n t e g e r or «>, which we can a s s o c i a t e to every space
M, cf. Definitions 7 and 8 below, and which measures "how much" the
space f a i l s t o be p a r t i a l l y complete. I t w i l l t u r n out t h a t A and B a l -
ways have t h e same c o m p l e x i t y ( c f . Th. 9.25 below).
Definition 7. Let (P,M) be a completed space.
a) The d e r i v e d sequence ((P^,M )Ik>0)k of (P,M) i s the f o l l o w i n g i n d u c -
t i v e l y d e f i n e d sequence o f completed spaces. (P ,M ) := (P,M).

0 Q is
the subspace P k v. M k of P k and P^ +1 is. the c l o s u r e o f M^ ^ + in P^.
b) I f M i s n o t empty then the c o m p l e x i t y c(P,M) o f (P,M) i s defined
as the supremum o f a l l numbers k > 0 with M k * 0, i . e . c(P,M) i s the high-
e s t index m w i t h M * 0 , m i f i t e x i s t s a n d c(P,M) = «> o t h e r w i s e .

/ The comple
x i t y o f the empty p a i r (0,0) i s d e f i n e d as - 1 .
L e t us take a rough look a t the c o m p l e x i t y c(P,M). Clearly
P^ 3 P. 3 P z>
0 2 *
0
* 1 *
I f U€ T(P) i s an open l o c a l l y s e m i a l g e b r a i c subset of P, then obvious-
l y M^fl U i s dense i n P k n U. Hence we obtain
dimP k n U = dimM k nU>dimM k + 1 DU = dimP k + 1 n U,
p r o v i d e d dim n U i s f i n i t e and M k n U * 0 ( c f . I, Prop. 3 . 2 1 ) . Thus
P k + 1 i s thin i n P R ( c f . § 8 , Def. 8), and
dim P, < dim P - k

x k — x
f o r every x € P^. We see t h a t c(P,M) < dimM . A l s o

c(P ,M ) = 1 + c ( P
r r r + 1 ,M r+ 1 ),
provided i s not empty.
What does i t mean t h a t (P,M) has complexity zero or one? We have
c(P,M) = O i f and only i f M = P. Then M i s p a r t i a l l y complete. Conver-
sely, i f M is partially complete then M i s c l o s e d in P ( I , Cor. 6.10).
hence M = P. Thus c ( P , M ) = O means t h a t M i s p a r t i a l l y complete and
not empty.
Assume now t h a t M i s not p a r t i a l l y complete, i . e . c(P,M) > 1. We know
from I, §7 ( c f . Prop. 7 . 2 ) that M i s l o c a l l y complete i f and only i f
= P s M i s closed i n P, i.e. = P^. We conclude that c(P,M) = 1
if and only i f M is locally complete but not partially complete.
In p a r t i c u l a r the p r o p e r t i e s "c(P,M) = 0" and "c(P,M) = 1" do not de-
pend on the g i v e n c o m p l e t i o n M «-» P of M but o n l y on the space M i t s e l f
This observation can be generalized, c f . Theorem 9 . 1 7 below.
We now t a k e a c l o s e r look a t the d e r i v e d sequence ( ( P , M ^ ) |k>0)

k of a
given completed space (P,M) . The set P^p^ i s contained i n M, and
M= (PNP^ 0 ( P ^ M ^ = (P^p^ u M 2 *)
In p a r t i c u l a r M c M , 2 while is disjoint from M . A l s o M \ M 2 = M^p
i s open i n M, and hence M 2 i s closed i n M. Applying a l l this to (P^,M^
instead of (P,M) we i n d u c t i v e l y see the following:
Remark 9 . 1 4 . M~ => M z> M , z> ... , and z> M z> M z> . . . . Every set
0 * 2 = ^ 4 = * 1 3 =t= 5
0 Q c
2
M 2 J c i s closed i n M = M , and Q every s e t M 2 k + «| i s closed in . The sets
M Q and M 1 are d i s j o i n t . We have, f o r e v e r y k>0,
*) U means " d i s j o i n t u n i o n " i n the s e t t h e o r e t i c a l sense.

M M u p p
k = k + 2 < k" k n- +
We i n t r o d u c e the s u b s e t s
S (P,M) k := P 2 k - P 2 k + 1 (k > 0)
a n d h e n c e i s a
o f M. E v e r y S ( P , M ) i s open i n P k '
k 2 locally complete
s p a c e . We denote t h e c l o s u r e o f S (P,M) i n M by S ( P , M ) . k k
p p i s
Lemma 9.15. a) E v e r y set r ^ r + i dense i n ? « r
b) S (P,M) =
k M . 2 k
c) M 2 k i s t h e d i s j o i n t union o f a l l S (P,M) w i t h 1 > k. 1
P r o o f . We o b s e r v e d already that P r + 1 i s t h i n i n P . T h i s means t h a t

p p P P i n M i s
r ^ r + l i s dense i n P r # The c l o s u r e o f S (P,M) = 2 k ^ 2 k + 1 k
M n ?2 k = M PI M 2 k
M
= 2 k ' -*- s nce M
2 k "*" ^ - Sc o s e <
^ ^ n M # T
^ e
l a s
t assertion
c) i s e v i d e n t from Remark 9.14, as soon as we know t h a t t h e i n t e r s e c -
t i o n o f a l l the sets M 2 k i s empty. But t h i s i s e v i d e n t s i n c e

x £ M a s w e s a w
dim M x 2 k < dim M-2k f o r every
x 2 k ' before, q.e.d.
According t o t h i s lemma the p a r t i t i o n (S (P,M)Ik

k > 0) o f M may be
regarded as a s t r a t i f i c a t i o n o f M - i n a weak sense. We have
s ( P P P
k 1'V = 2k l" 2k+2- +
The c l o s u r e S (P,M) o f S (P,M) i n P i s t h e s e t P^

k k k which i s the u n i o n
of a l l s t r a t a S (P M) of M with 1 f 1 > k and a l l s t r a t a S ^ P ^ M ^ ) o f M <]
with 1 > k. S i m i l a r l y S k ( P ^ ) i s t h e u n i o n o f t h e s t r a t a S-^PjM) o f
M w i t h 1 > k+1 and t h e s t r a t a S ( P , M ) o f M 1 1 1 1 with 1 > k. Thus t h e
stratifications ( S ( P , M ) I k > 0) and ( S ^ P ^ M ^ I k

k > 0) o f M and M 1
fit together to a s t r a t i f i c a t i o n of P = M U .
P r o p o s i t i o n 9.16. L e t (Q,N) be a second completed space w i t h d e r i v e d

sequence ( ( Q , N ) I k > O). L e t f : (P,M) -> (Q,N) be a p r o p e r morphism
k k
w i t h f(M) = N. Then c(P,M) = c(Q,N), and f ( M ) = N

k fef f ( P ) = Q for
K k
every k_>0.
P r o o f . f (P) i s a c l o s e d s e t which c o n t a i n s N, so f ( P ) = Q. We have

1 1
f" (N) = M, and hence f ~ ( N ) 1 = M. 1 We c o n c l u d e that f(M ) = N ,
1 1 which
implies t h a t f (P^) = . Thus f y i e l d s , by r e s t r i c t i o n , a p r o p e r mor-
phism f^ : (P^,M^) -> (Q^,N^) w i t h f^(M^) = . P r o c e e d i n g by i n d u c t i o n
on k, we o b t a i n f (M^.) = N k and f (P^) = Q k f o r every k > 0 . In particu-
lar, N k * 0 i f and o n l y i f M k * 0 . Thus c(0,N) = c(P,M). q.e.d.
Theorem 9 . 1 7 . I f cp : M «-> P and : M *-» Q a r e two c o m p l e t i o n s of a space
M, t h e n c(P,cp(M)) = c(0,^(M)) and
cp" S (P,cp(M) ) = 4 T S ( 0 , i M M ) )
1
k
1
k
for e v e r y k > 0.
P r o o f . We have a t h i r d c o m p l e t i o n x :M T
which admits morphisms
f : (T,x(M)) (P,cp(M)) and g : (T,x(M)) -> (Q,i|>(M)) such t h a t f « x = <P
and g « x = ^. I t s u f f i c e s t o prove the theorem f o r cp and x i n s t e a d o f
cp and \\). The morphism f i s p r o p e r and maps x (M) i s o m o r p h i c a l l y onto
cp(M). We c o n c l u d e from t h e p r e c e d i n g p r o p o s i t i o n that indeed
c(P,cp(M)) = c(T, (M)) X and f ( S ( T , x ( M ) ) k = S ( P , c p ( M ) ) f o r any k > 0 .

k
Since f : x(M)->cp(M) i s b i j e c t i v e , this implies f ( S ( T , x ( M ) ) = S (P,cp(M)),

k k
and then S ( T , x ( M ) ) = f
k S ( P , c p ( M ) ) . T a k i n g preimages under Xr
k
w e o b
~
-1 -1
tain x S ( T , x ( M ) ) = cp
k S ( P , c p ( M ) ) as d e s i r e d .
k q.e.d.
Definition 8 . The c o m p l e x i t y c(M) o f a space M i s t h e number
c(P,cp(M)) f o r any c o m p l e t i o n cp : M *•* P. The L C - s t r a t i f i c a t i o n of M i s
the p a r t i t i o n (S (M) |k>0) o f M i n t o

k l o c a l l y complete subsets
S (M)
k := cp" S (P,cp(M) ) .
1
k
(LC = " l o c a l l y complete".) We c a l l S (M) t h e k - t h LC-stratum
k o f M.
I t s c l o s u r e i n M w i l l be denoted by S ( M ) .
k
The p o i n t i s t h a t , by t h e p r e c e d i n g theorem, t h e c o m p l e x i t y and t h e
L C - s t r a t i f i c a t i o n o f M do n o t depend on t h e c h o i c e o f t h e c o m p l e t i o n
<p o f M.
U s i n g t h e " a b s o l u t e " n o t i o n s o f D e f i n i t i o n 8 we r e - f o r m u l a t e some o f
our o b s e r v a t i o n s on t h e d e r i v e d sequence o f a completed space.
Resumee 9.18. a) The s u b s e t s S (M) o f M a r e p a i r w i s e d i s j o i n t , and

k
d i m S ( M ) < dim M-2k f o r every

x k x x€S (M).
k
b) S (M) i s t h e u n i o n o f t h e s t r a t a S (M) w i t h l > k .

k 1
c) S k + 1 (M) = S (S
Q k + 1 ( M ) ) = S ( S ( M ) ^ S (M) ) f o r every
Q k k k>0.
d) S (S^ (M) ) = S
k k + 1 ( M ) f o r every k > 0 , 1>0.
e) c (M) < dim M .
f) I f t h e r e e x i s t s a h i g h e s t index r w i t h S (M) * 0 , r then
{ 2r
2r+1
i f S (M) i s p a r t i a l l y
else.
I f M i s empty then c(M) = -1. I f a l l S (M) * 0 (hence

r
r
complete
i n particular
dim M = oo) , then c(M) = °°.
g) I f (P,M) i s a non empty completed space, then c (M) = 1 + c ( P ^ M ) .
The c l o s u r e o f S (M) k i n P i s t h e union o f a l l s t r a t a S-^ (M) w i t h l > k
and a l l s t r a t a S-^P^M) w i t h l > k . The c l o s u r e o f S (P^-M) i n P i s t h e k
union o f a l l s t r a t a S (M) w i t h 1 > k+1 and a l l s t r a t a

1 (P ^ M) w i t h l > k
We a l s o f o r m u l a t e P r o p o s i t i o n 9.16 i n a b s o l u t e terms.
Theorem 9.19. L e t f : M -> N be a proper s u r j e c t i v e map from a space M
t o a space N. Then c (M) = c (N) , and f ( S (M) ) = S,(N) f o r every k k>0.

Indeed, by P r o p o s i t i o n 5. 1 our map f can be extended t o a morphism (P ,M) -*
(Q,N) f o r s u i t a b l e completed spaces (P,M) and (N,Q) . T h i s morphism i s p r o p e r
( c f . 1.5.9, 1.6.13, 8.1). Thus P r o p o s i t i o n 9.16 g i v e s the d e s i r e d r e s u l t .
We ask f o r an i n t r i n s i c d e s c r i p t i o n o f the L C - s t r a t a (M) which does
not i n v o l v e any c o m p l e t i o n of M.
P r o p o s i t i o n 9.20. S Q (M) i s the s e t of a l l x G M which have a complete
neighbourhood i n M.
P r o o f . We choose a c o m p l e t i o n cp : M <-* P o f M. We assume w i t h o u t l o s s of
g e n e r a l i t y t h a t M i s a subspace of P and cp i s t h e i n c l u s i o n map. S Q (M)
is, by d e f i n i t i o n , t h e i n t e r i o r M of the s e t M i n P. I f x € M has a
complete neighbourhood i n M, then x £ M by I, P r o p o s i t i o n 7.2. On the
o t h e r hand, M i s a l o c a l l y complete space. Thus e v e r y X € M has a com-
p l e t e neighbourhood i n M, which i s , o f c o u r s e , a complete neighbourhood
i n M. q.e.d.
By t h i s p r o p o s i t i o n the LC-stratum S (M) Q i s the l a r g e s t open locally
complete subspace of M. U s i n g 9.18.C we o b t a i n an i n t r i n s i c - though
c o m p l i c a t e d - d e s c r i p t i o n o f e v e r y s t r a t u m S^(M). As another conse-
quence o f P r o p o s i t i o n 9.20 we obtain
P r o p o s i t i o n 9.21 ( L o c a l n a t u r e o f the L C - s t r a t i f i c a t i o n ) . L e t U be an
open subspace o f M. Then S (U) k = S (M) flU f o r e v e r y

k k>0.
Proof. This i s evident f o r k = 0 from P r o p o s i t i o n 9.20. We then have
s 1 (u) = UNS (U) Q = (MNS (M) ) n U

0 = S 1 (M) n U.
Thus S ( U )
1 i s open i n S ( M ) . We 1 c o n c l u d e , w i t h 9.18.C, t h a t
s 1 (u) = s (s (u))
Q 1 = s (s (M))
Q 1 ns (u)
1 =
= s 1 (M) n s 1 (M) n u = s 1 (M) n u .
Repeating t h i s argument we o b t a i n S (U) k = S (M)k HU f o r e v e r y k. q.e.d.
From t h i s p r o p o s i t i o n and 9.18.f we obtain immediately
Corollary 9.22. a) For every U6 f(M) we have c (U) < c (M) + 1 i f c (M) i s
even, and c(U) < c(M) i f c(M) i s odd.
b) I f (UglocEI) i s any a d m i s s i b l e c o v e r i n g o f M by open l o c a l l y semi-
algebraic s u b s e t s t h e n , f o r e v e r y k > 0,
S (M)
k = U ( S ( U ) |a € I ) ,
k a
and
c(M) = s u p ( c ( U ) |a € I) - 6
a
w i t h 3 = 0 i f c(M) i s odd, and 6 = 0 o r 1 i f c(M) i s even.
Remark 9.23. Starting from 9.22.b we can e s t a b l i s h an LC-stratification
( S ( M ) | k > 0 ) on any
k locally s e m i a l g e b r a i c space M (not n e c e s s a r i l y r e -
g u l a r and paracompact) as f o l l o w s . S (M) k i s the union of a l l S ( U ) f o rk a
any a d m i s s i b l e c o v e r i n g (U^laEI) by open a f f i n e s e m i a l g e b r a i c s u b s e t s . We
f u r t h e r can d e f i n e a c o m p l e x i t y c(M) as i n d i c a t e d i n 9.18.f. I t i s now
p o s s i b l e t o g e n e r a l i z e those f a c t s on the L C - s t r a t i f i c a t i o n and comple-
x i t y which do not r e f e r to a c o m p l e t i o n o f M t o a r b i t r a r y locally semi-
algebraic s p a c e s . We l e a v e the d e t a i l s to the r e a d e r .
N o t a t i o n 9.24. I f cp : M «-> P i s a c o m p l e t i o n of a space M, then f o r any
a b s o l u t e end X € e (M) we denote the c o r r e s p o n d i n g end K (cp) ( X ) o f M i n
P by P ( X ) . Thus (P ( X ) IX € e (M) ) i s the f a m i l y o f a l l ends of M i n P and
P^(p(M) i s the d i r e c t sum o f the spaces P(X).
I f M i s not empty then c l e a r l y c(M) = 1 + sup(c (P(X)) I X 6 e(M)) . I t

t u r n s out t h a t , f o r cp pure, not o n l y the supremum o f the complexities
c ( P ( X ) ) but each number c ( P ( X ) ) itself i s independent o f the c h o i c e
of the c o m p l e t i o n cp o f M.
Theorem 9.25. L e t cp : M «•* P and : M «-+ Q be pure c o m p l e t i o n s o f M. Then,
for any X € e(M), the ends C o f the space P ( X ) i n i t s c l o s u r e P ( X ) i n
P correspond b i j e c t i v e l y t o the ends D o f Q(X) i n i t s c l o s u r e Q(X) i n

-1 -1
Q v i a the r e l a t i o n cp (C) = (D) . (N.B. We have Cctp(M) and D c: \\j (M). )
In particular, c(P(X)) = c ( Q ( X ) ) and |e(P(X))| = U(Q(X))|.
P r o o f . L e t A := P ( X ) and B := Q ( X ) . We may assume t h a t t h e r e e x i s t s a
morphism f : Q P from t o cp ( c f . p r o o f o f Th. 9.17). Thus f y i e l d s
by r e s t r i c t i o n a p r o p e r s u r j e c t i v e map h : B -» A and a l s o a p r o p e r sur-
j e c t i v e map g : B -» A . T h i s a l r e a d y i m p l i e s t h a t c(B) = c ( A ) , c f . Theo-
rem 9.19. We have g ^ (A) = B, and thus o b t a i n a p r o p e r s u r j e c t i v e map
f^ from B s B t o A s A by r e s t r i c t i o n o f f . But B N B C I J J ( M ) and f maps
ip (M) i s o m o r p h i c a l l y onto cp(M) . Thus f^ i s an isomorphism from B s B t o
A s A . The theorem i s now obvious. q.e.d.
With the s i t u a t i o n as i n the p r o o f , we know ( c f . e.g. P r o p o s i t i o n 8.4)
t h a t the f i b r e s o f the p r o p e r map h : B->A a r e c o n n e c t e d . The fibres
of the extended map g : B -> A a r e , o f c o u r s e , a l s o c o n n e c t e d , s i n c e a l l
f i b r e s g ^ (y) w i t h y € A s A a r e o n e - p o i n t s e t s . Thus the morphism
g: (B,B) -+ ( A , A ) i s "shrinking" i n the sense o f t h e f o l l o w i n g defini-
tion
Definition 9. A morphism g : (B,B) -* ( A , A ) between completed spaces i s
called shrinking i f g i s p r o p e r , g(B) = A, hence a l s o g(B) = A all

|
f i b r e s o f g : B-> A a r e connected, and a l l f i b r e s g ^ (y) w i t h y € A s A
are one-points sets. (N.B. Then g g i v e s an isomorphism from B S B onto
A s A by restriction).
We can add t o Theorem 9.25 the f o l l o w i n g
Corollary 9.26. L e t <p : M P and i|/ : M Q be c o m p l e t i o n s o f M which ad-
mit a morphism f : Q -jj*P. Assume t h a t (p i s pure. Then f o r e v e r y A £ e (M)
f y i e l d s by r e s t r i c t i o n a s h r i n k i n g morphism g^: (Q( A) ,Q( A)) -> ( P ( A ) ,P(A))
between t h e completed ends i n Q and P c o r r e s p o n d i n g t o A . The isomor-
phism from Q(A) v Q ( A ) t o P ( A ) \ P ( A ) i n d u c e d by g^ i s the r e s t r i c t i o n
of cpoip"" t o the s e t Q( A) \ Q( A) .
1
§10 - Some proper quotients
We would l i k e t o understand how the ends of a completed space can be
"shrunk" or "expanded". B e f o r e we can a t t a c k t h i s problem we have to
know t h a t s u i t a b l e q u o t i e n t s ( i n p a r t i c u l a r mapping c y l i n d e r s ) e x i s t
i n our c a t e g o r y of spaces. These q u o t i e n t s deserve independent inte-
r e s t . Thus we t r y t o prove t h e i r e x i s t e n c e under the weakest p o s s i b l e
assumptions. They w i l l be s l i g h t l y weaker than n e c e s s a r y f o r t/he a p p l i
c a t i o n s i n §11 and §12.
We work f o r a s h o r t time i n the c a t e g o r y o f a l l (separated) locally
semialgebraic spaces i n s t e a d o f r e g u l a r paracompact spaces. So "space"
now means " s e p a r a t e d l o c a l l y semialgebraic space".
Definition 1. A proper q u o t i e n t o f a space M by an e q u i v a l e n c e rela-

*)
t i o n R on the s e t M i s a proper surjective semialgebraic map p : M -> N
whose f i b r e s are the e q u i v a l e n c e c l a s s e s o f R. We then w r i t e N = M/R
and c a l l p the q u o t i e n t p r o j e c t i o n from M t o M/R.
The definition is justified s i n c e , i n t h i s s i t u a t i o n , N i s indeed a
q u o t i e n t of M i n the c a t e g o r i a l sense, as s t a t e d by the f o l l o w i n g
proposition.
Proposition 10.1. L e t p : M -* N be a proper surjective semialgebraic map
L e t f : M -> L be a l o c a l l y s e m i a l g e b r a i c map which sends every f i b r e of
p t o a p o i n t i n L. Then the unique map f : N -* L w i t h f • p = f i s l o c a l -
ly semialgebraic.
*) R e c a l l t h a t , f o r M paracompact, every proper l o c a l l y s e m i a l g e b r a i c

map p : M -»N i s s e m i a l g e b r a i c ( I , 5.9). Thus our c o n d i t i o n "semi-
a l g e b r a i c " here i s not a s e r i o u s r e s t r i c t i o n .
P r o o f . We choose an a d m i s s i b l e c o v e r i n g (U la€I)
a o f N by open semialge-
_i
b r a i c s u b s e t s . Then (p ( U ) l a 6 I) i s an a d m i s s i b l e
a covering of M
by open s e m i a l g e b r a i c s u b s e t s . We have t o v e r i f y t h a t f | U : U^-^L i s Q
semialgebraic f o r every a € I. R e p l a c i n g N by and M by p ^ (U^) f o r
some f i x e d a, we may assume t h a t M and N a r e s e m i a l g e b r a i c spaces. Re-
p l a c i n g L by i t s s e m i a l g e b r a i c subspace f(M) we a l s o assume t h a t L i s
semialgebraic.
A s u b s e t B o f N i s c l o s e d and s e m i a l g e b r a i c i n N i f and o n l y i f p ( B ) 1
i s c l o s e d and s e m i a l g e b r a i c i n M. From t h i s f a c t one deduces immediate-
l y that f i s continuous. The graph r ( f ) c N x L i s t h e image o f t h e graph
T (f) c M x L under t h e (proper) s e m i a l g e b r a i c map p x i d : MxL N x L . Thus
T(f) i s a semialgebraic subset o f N x L . T h i s f i n i s h e s t h e proof t h a t f
is semialgebraic. q.e.d.
Proposition 1 0 . 2 . L e t p : M-+N be a proper s u r j e c t i v e map. Assume t h a t
M i s r e g u l a r and paracompact. Then N i s a l s o r e g u l a r and paracompact.
Proof. a) We f i r s t prove r e g u l a r i t y . We have t o s e p a r a t e a given closed
locally semialgebraic s e t A i n N from a g i v e n p o i n t x i n N ^ A . S i n c e M
i s r e g u l a r and paracompact t h e r e e x i s t open l o c a l l y semialgebraic sets
U , U
1 2 i n M w i t h p ~ (x) c
1
, p"* (A) c U
1
2 and ^ nU 2 = 0 (I, 4.14) . Since
p i s proper t h e r e e x i s t open l o c a l l y s e m i a l g e b r a i c neighbourhoods ,V" 2
of x and A r e s p e c t i v e l y w i t h U. D p " 1
(V.), namely V i N\p(MÛ ). i
Since p i s s u r j e c t i v e DV 2 = 0 . T h i s proves t h a t N i s r e g u l a r ,
b) We choose a l o c a l l y finite covering (A^laGI) o f M by c l o s e d s e m i a l -
gebraic subsets ( c f . I , Th. 4 . 1 1 ) . Then (p(A ) | a € I ) i s a c o v e r i n g o f
N by c l o s e d s e m i a l g e b r a i c s u b s e t s . We c l a i m t h a t t h i s f a m i l y i s l o c a l l y
finite. The map p i s s e m i a l g e b r a i c (I,5 . 9 ) . F o r any U 6 J'(N) t h e s e t

1
p ( U ) i s s e m i a l g e b r a i c and hence meets A a f o r o n l y f i n i t e l y many a € I .
We c o n c l u d e t h a t U meets p(A ) o n l y f o r f i n i t e l y many a € I .

We now know by Lemma 1 . 2 t h a t N i s i n d e e d paracompact. q.e.d.
From now on a "space" w i l l a g a i n mean a r e g u l a r paracompact locally
s e m i a l g e b r a i c space. By P r o p o s i t i o n 1 0 . 2 we know t h a t a proper quotient
of a space i s a g a i n a space.
In o r d e r t o prove t h e e x i s t e n c e o f p r o p e r q u o t i e n t s the f o l l o w i n g
lemma w i l l be u s e f u l .
Lemma 1 0 . 3 . L e t (M lcc€I) a be a l o c a l l y f i n i t e f a m i l y of l o c a l l y semial-
g e b r a i c s u b s e t s o f a space M which c o v e r s M. Assume t h a t e i t h e r a l l M Q
are c l o s e d or a l l M Q a r e open i n M. L e t R be an e q u i v a l e n c e relation
on t h e s e t M such t h a t e v e r y M a i s a union of equivalence classes. L e t
R denote t h e r e s t r i c t i o n o f t h e e q u i v a l e n c e r e l a t i o n t o M . Assume
a a
M
that, f o r every a € I , t h e p r o p e r quotient a / ^ a exists. Then t h e p r o p e r
q u o t i e n t M/R exists.
N.B. I t i s c l e a r t h a t , i f t h e p r o p e r q u o t i e n t M/R exists, a l l the s e t s
Mg/R^ a r e c l o s e d , r e s p . open, l o c a l l y s e m i a l g e b r a i c s u b s e t s o f M / R , and
t h a t on M / R a a t h e subspace s t r u c t u r e i n M/R i s t h e same as t h e q u o t i e n t
s t r u c t u r e coming from M « A l s o , a (M^/R^laGI) i s a locally finite cover-
ing o f M/R.
P r o o f . We denote the q u o t i e n t M^/R^ by N Q and t h e p r o p e r p r o j e c t i o n map from
^ 3 ^
: = o f N i s
M Q to N a by p . The s u b s e t N ^
a P a
M
a
n M
a l o c a l l y semial-
g e b r a i c and p (N .) = M n M . Q The r e s t r i c t i o n p.:M n M ->N . o f p

f l
^ *a a3 a 3 â3 a 3 a3 *a
i s again proper. T h i s map shows t h e s u b s p a c e N o f N^ t o be t h e p r o -
per q u o t i e n t o f M D M ^ by t h e e q u i v a l e n c e
Q relation R obtained from
R by r e s t r i c t i o n . (Notice that M n M Q consists of f u l l equivalence

c l a s s e s o f R.) S i n c e t h e same h o l d s f o r N ^ we have a c a n o n i c a l commu-
Q
pa
tative triangle
with an isomorphism v _,
^ A
a 3
M a s
We r e g a r d t h e spaces = a /^ a subsets o f t h e s e t N := M/R. Then
N
o = N
r, a s
s e t s , and v rt becomes t h e i d e n t i t y map. Thus t h e subspace
s t r u c t u r e s on N =N with respect to N and N a r e e q u a l . I f every1 M i s

a3 3a 3
n ft rt
a ^ a
c l o s e d i n M then every N ^ i s c l o s e d i n N and N . Then we know from
rt
J
cx3 a 3
Theorem 1.3 t h a t t h e r e e x i s t s on N a u n i q u e space s t r u c t u r e such t h a t
(N^laEI) i s a locally finite family of closed l o c a l l y semialgebraic
s u b s e t s o f N and the s u b s p a c e s t r u c t u r e s on are the given quotient
s t r u c t u r e s o f t h e spaces M . a We e q u i p N w i t h t h i s space s t r u c t u r e .
C l e a r l y the n a t u r a l p r o j e c t i o n p : M->N i s a proper semialgebraic map,
and we a r e done.
N = M n M i s e n
Assume now t h a t every i s open i n M. Then p ^ a ^' a p °P
in M and i n M „ . I t f o l l o w s t h a t N n i s open i n N and i n N„ , s i n c e

a 3 a3 cx 3
p^ and p^ a r e proper and s u r j e c t i v e . On t h e s e t N we have a unique
s t r u c t u r e o f a r i n g e d space o v e r R such t h a t (N^locGI) i s an a d m i s s i b l e
open c o v e r i n g and every N , Q i n i t s g i v e n s t r u c t u r e , i s an open sub-
space o f N. T h i s space N i s l o c a l l y s e m i a l g e b r a i c , b u t not y e t known
to be s e p a r a t e d . Once we have v e r i f i e d t h i s we a r e done: By P r o p o s i t i o n
1.5.7 and Theorem 1.5.9 t h e map p i s p r o p e r and s e m i a l g e b r a i c .
L e t x and y be d i f f e r e n t p o i n t s o f N. We have t o s e p a r a t e them by open
neighbourhoods. We choose i n d i c e s a, 3 € I w i t h x € N and y € N . The f t

-1 -1
fibres p (x) and p (y) a r e complete, hence c l o s e d i n M. S i n c e M i s
r e g u l a r and paracompact we f i n d open s e m i a l g e b r a i c neighbourhoods
-1 -1
cM a and U c M ^ o f p
2 (x) and p (y) w i t h n CJ = 0 . S i n c e t h e maps
2
P a and Pp a r e proper we f i n d open s e m i a l g e b r a i c neighbourhoods cN^
and V c N ^ 2
1
o f x and y w i t h p ~ (V ) c IK. C l e a r l y V
± 1 nV 2 = 0. q.e.d.
We now prove t h e e x i s t e n c e o f proper quotients i n a very s p e c i a l case.
P r o p o s i t i o n 10.4. L e t A be a complete - hence s e m i a l g e b r a i c - subset
o f a space M. There e x i s t s t h e proper q u o t i e n t M/A o f M by t h e e q u i v a -
l e n c e r e l a t i o n which has as e q u i v a l e n c e c l a s s e s t h e s e t A and t h e one
p o i n t s e t s {x} w i t h x € M s A.
P r o o f . We have t o f i n d a proper s u r j e c t i v e map p : M N whose f i b r e s
a r e A and t h e s e t s {x} w i t h x € M ^ A . We choose an embedding M P of M
i n t o a p a r t i a l l y complete space P. I t s u f f i c e s t o f i n d a p r o p e r surjec
t i o n q : P-> L whose f i b r e s a r e A and t h e s e t s {x} w i t h x € P ^ A . Then t h
r e s t r i c t i o n p : M-*q(M) o f q w i l l be a proper q u o t i e n t o f M by t h e o r i -
g i n a l equivalence relation. (N.B. q(M) i s a subspace o f L s i n c e q i s
s e m i a l g e b r a i c ) . Thus we assume w i t h o u t loss of g e n e r a l i t y that the
space M i s p a r t i a l l y complete.
We chosse an open s e m i a l g e b r a i c neighbourhood U o f A i n M. The c l o s u r e
B := U o f U i n M i s complete. I t s u f f i c e s t o prove t h a t t h e p r o p e r quo
t i e n t B/A e x i s t s . Then, a p p l y i n g Lemma 10.3 t o t h e c o v e r i n g o f M by
the two c l o s e d l o c a l l y semialgebraic sets MÛ and B, we see t h a t t h e
proper q u o t i e n t M/A e x i s t s . Now we have reduced the proof t o t h e case
t h a t t h e space M i s complete and (hence) semialgebraic.
+
L e t N denote t h e one p o i n t c o m p l e t i o n (M\A) o f the l o c a l l y complete
s e m i a l g e b r a i c space M \ A ( c f . I , § 7 ) . The i d e n t i t y map M \ A-»M \ A ex-

1
tends t o a p r o p e r map p : M -* N w i t h p (°°) = A ( I , Prop. 7.6), and we
a r e done. q.e.d.
Given a l o c a l l y s e m i a l g e b r a i c map f : X -*Y between spaces X and Y , we
ask whether t h e mapping c y l i n d e r Z(f) exists i n t h e c a t e g o r y o f spaces
In analogy t o t h e t o p o l o g i c a l case we mean by Z ( f ) t h e q u o t i e n t o f t h e
d i r e c t sum
M := ( X x [ o , 1 ] ) U Y
by t h e c o a r s e s t e q u i v a l e n c e r e l a t i o n R w i t h (x,1) ~ f ( x ) f o r every x € X
Suppose t h a t even t h e p r o p e r q u o t i e n t p : M-*Z(f) e x i s t s . S i n c e Y i s
c l o s e d and l o c a l l y s e m i a l g e b r a i c i n M, t h e s e t p ( Y ) i s c l o s e d and l o -
c a l l y s e m i a l g e b r a i c i n Z ( f ) , and p maps Y i s o m o r p h i c a l l y onto p ( Y ) .
I d e n t i f y i n g Y w i t h p ( Y ) we r e g a r d Y as a c l o s e d subspace o f M and a l s o
o f Z ( f ) . Now p maps Xx{1} t o Y and t h e r e s t r i c t i o n XX{1}-»Y of p i s
a g a i n p r o p e r . But t h i s map i s j u s t a copy o f f : X - * Y . Thus f must be
proper. So, i n t h e p r e s e n t framework, where we o n l y l o o k f o r proper
q u o t i e n t s , we have t o assume from the beginning that f i s proper.
Theorem 10.5. L e t f : X - » Y be a p r o p e r map between s p a c e s . Then t h e map
ping c y l i n d e r Z ( f ) e x i s t s as a proper q u o t i e n t o f M := ( X x [ o , 1 ] ) U Y .
P r o o f . L e t ( U l a € J ) be a l o c a l l y
a f i n i t e c o v e r i n g o f Y by open s e m i a l -
g e b r a i c s e t s U^. Suppose t h a t f o r e v e r y a € J t h e mapping c y l i n d e r o f

1
the r e s t r i c t i o n f " (Ug) -*U o f f e x i s t s as a p r o p e r q u o t i e n t . Then i t
a
f o l l o w s from Lemma 10.3 t h a t Z ( f ) e x i s t s as a p r o p e r q u o t i e n t . Thus
we may henceforth, assume t h a t X and Y a r e ( a f f i n e ) s e m i a l g e b r a i c spaces.
L e t f : X - * Y be a c o m p l e t i o n o f f ( c f . Prop. 5.1). Then X = f ~ 1

(Y)
since f i s proper. Suppose t h a t t h e p r o p e r q u o t i e n t

p : Xx[0,1 ] LIY - Z ( f )
exists. L := p ( X x [ 0 , 1 ] U Y ) i s a semialgebraic subset of Z(f) with pre-
image X x [ o , 1 ] U Y . Thus the restriction
p : X x [ 0,1 ] IJ Y -* L
o f p i s a proper map, and t h i s map shows L to be the mapping c y l i n d e r
Z ( f ) o f f . These c o n s i d e r a t i o n s a l l o w us t o r e t r e a t t o the case t h a t
both X and Y a r e complete (affine) semialgebraic spaces.
In o r d e r t o c o n s t r u c t Z ( f ) i n t h i s case we s t a r t w i t h the proper pro-
jection map
cp : X x l -» C(X) := Xxl/Xx{1 }
o f the "cone" C(X) o f X , which e x i s t s by the p r e c e d i n g p r o p o s i t i o n 10.4.
C o n s i d e r the p r o p e r surjective map
* : X x y x l -^>XxIxY p x l d > C(X) x Y /
w i t h a d e n o t i n g the s w i t c h (x,y,t) •* ( x , t , y ) . N o t i c e t h a t two different
points (x^y^t-j) and x

( 2'Y2' 2^ t h a v e t h e s a j n e
iniage under ij; i f and
only i f t 1 = t 2 = 1 and y^ = y « 2 We choose a p o i n t x Q £ X . L e t Z be the
image under o f the c l o s e d s e m i a l g e b r a i c subset
N := ( T ( f ) x i ) u ({X }XYX{1})
Q
of X x Y x I . {As u s u a l , r ( f ) denotes the graph o f f.} Then Z i s a complete
space and the r e s t r i c t i o n x : N

-* z o f
4> i s
proper and surjective. We
have another proper surjective map
n : ( X x l ) LIY -+ N,
d e f i n e d by K(x,t) = (x,f(x),t), n (y) = (x , y , 1 ) . The composite p := x O T T
i s a proper s u r j e c t i v e map from ( X x l ) [JY to Z which shows t h a t Z i s the
mapping c y l i n d e r o f f . q.e.d.
From P r o p o s i t i o n 10.4 and Theorem 10.5 we obtain

Corollary 10.6. L e t f : X -» Y be a s e m i a l g e b r a i c map from a complete
space X t o a space Y (hence f i s p r o p e r ) . Then the mapping cone
C(f) := Z ( f ) / X x { 0 }
e x i s t s as a proper quotient of (Xxi)UY.
We now come t o t h e main r e s u l t o f t h i s s e c t i o n . L e t A be a c l o s e d l o -
cally semialgebraic subset o f a space X and l e t f : A ->Y be a proper
map t o another space Y . We want t o g l u e X t o Y "along A",- i . e . we ask
f o r t h e q u o t i e n t X U ^ Y o f X U Y by t h e c o a r s e s t e q u i v a l e n c e relation
with a~f(a) f o r every a € A .
Theorem 10.7. The proper quotient X U Y of X U Y e x i s t s .

f
N o t i c e t h a t t h i s theorem c o n t a i n s o u r p r e c e d i n g results 10.4, 10.5 and
10.6 as s p e c i a l c a s e s . N o t i c e a l s o t h a t properness of f i s certainly
necessary f o r t h e e x i s t e n c e o f X U Y as a proper
f quotient.
The p r o o f o f Theorem 10.7 i s easy i f A = X . I n t h i s case t h e proper
map TT : X U Y -+ Y , TT(X) = f (x) (x € X) , n(y) = y (y 6 Y) shows t h a t
Y = XU Y. f
We now p r o o f Theorem 10.7 i n t h e case t h a t X i s p a r t i a l l y complete.
Choosing a simultaneous t r i a n g u l a t i o n o f X and A we assume t h a t X i s
a closed l o c a l l y f i n i t e s i m p l i c i a l complex and A i s a c l o s e d subcom-
plex of X .
We f i r s t c o n s i d e r t h e case t h a t X \ A c o n t a i n s o n l y f i n i t e l y many open
s i m p l i c e s . We choose a f i l t r a t i o n
A = X c XQ 1 c X 2 c ... c X r = X
o f X by c l o s e d subcomplexes X i such t h a t X ^ = Ua i with a s i n g l e

x
open simplex CK whose boundary do^ i s c o n t a i n e d i n ^_^•
U s i n g i n d u c t i o n on i , we prove t h a t t h e p r o p e r q u o t i e n t X i UY f exists
for i = 0,1,...,r. We know t h i s from above f o r i = 0. Assume t h a t
0<i<r-1 and t h a t t h e p r o p e r q u o t i e n t map
p : X.UY X.U Y f =: Y ±
e x i s t s . We i d e n t i f y X i + 1 w i t h t h e mapping cone C(cp) of the i n c l u s i o n
map cp : i n t h e o b v i o u s way such t h a t , say, the b a r y c e n t e r o f
a i + 1 i s the vertex o f t h e cone. We a l s o i n t r o d u c e t h e mapping cone
C(q«cp) o f t h e composite o f cp w i t h t h e r e s t r i c t i o n q : X ^ - » Y i o f p. L e t
X : (3a i + 1 x i ) UX. -* C(cp) and y : O a . + 1 x l ) UY. -* C(q«cp) be t h e p r o p e r
quotient p r o j e c t i o n s o f t h e s e mapping cones. Then we have a u n i q u e map
Y from t h e s e t C(cp) U Y t o t h e s e t C(q°tp) such t h a t t h e diagram
(3o i + 1 xl) UX.UY idLJfi »(3a. + 1 xl) UY ±
X U id
C(cp) U Y *C(q.<p)
commutes. S i n c e the other t h r e e maps a r e p r o p e r s e m i a l g e b r a i c surjec-
t i o n s we c o n c l u d e t h a t y i s a l s o a p r o p e r s e m i a l g e b r a i c surjection.
V i a t h i s map r , C(q«>cp) i s t h e q u o t i e n t X ± + 1 U Y o f C (cp) U Y

f = X ± + 1 UY,
as i s e a s i l y checked. Thus our theorem i s proved i n the case that
XÂ contains only f i n i t e l y many open simplices.
Assume now t h a t X ^ A c o n t a i n s infinitely many open s i m p l i c e s . L e t
(X^jXEA) be t h e f a m i l y o f a l l c l o s e d subcomplexes o f X which contain
A and c o n t a i n o n l y f i n i t e l y many s i m p l i c e s o u t s i d e o f A. We o r d e r Ain
such a way t h a t X<y i f f X, c x . Then (X, IAEA) i s a d i r e c t e d system o f

— A y A
partially complete s p a c e s . F o r every X E A we have a p r o p e r q u o t i e n t map
P :X UY
X A -X U Y x f =: M, .
If A < y t h e n M, embeds i n t o M i n a c a n o n i c a l way as a c l o s e d subspace.
— A u
Thus (M^|A€A) i s again a d i r e c t e d system o f p a r t i a l l y complete spaces
with closed embeddings as t r a n s i t i o n maps. I t i s now e a s i l y seen that
the r i n g e d space M := l i m M, i s a p a r t i a l l y complete r e g u l a r and p a r a -

~A~* A
compact l o c a l l y s e m i a l g e b r a i c space. F o r example, r e g a r d M as t h e u n i o n
of the sets w i t h M^cM^ f o r A < y , and a p p l y Theorem 1.3 t o t h e f a m i -
ly consisting o f t h e space Y and t h e spaces p^ ( a ) w i t h A € A , a € I ( X ) ,
ocx^, a X^ f o r u < A , a <t A. The maps p^ f i t together to a proper sur-
j e c t i v e map p : X U Y -> M ( c f . I , Prop. 3.16), and by t h i s map M i s t h e
d e s i r e d q u o t i e n t X U Y o f X(JY. This f i n i s h e s the proof i n the case

f
that X i s p a r t i a l l y complete.
In o r d e r t o prove t h e theorem i n g e n e r a l we quote t h e f o l l o w i n g fact
from [DK ]. 5
Proposition 10.8 [DK^, Prop. 2 . 2 ] . L e t X be a l o c a l l y finite simplicial
complex and A a c l o s e d subcomplex o f X. L e t V denote t h e open star
St^,(A) o f A i n t h e f i r s t barycentric s u b d i v i s i o n X' o f X. Then t h e r e
exists a s e m i a l g e b r a i c r e t r a c t i o n map r : V~>A such t h a t , f o r every
x€VÂ, t h e open l i n e segment ]x,r(x)[ i s c o n t a i n e d i n t h e same open
simplex S c V o f X' as x.
T h i s has been proved i n [DK^] i n t h e c a s e t h a t t h e complex X i s f i n i t e ,
by use o f an e x p l i c i t f o r m u l a f o r r i n terms o f b a r y c e n t r i c coordinates
The same f o r m u l a and p r o o f work i f X i s l o c a l l y finite.
We r e t u r n t o t h e s i t u a t i o n o f Theorem 10.7. Choosing a s i m u l t a n e o u s t r i
a n g u l a t i o n o f X and A we l e a r n from P r o p o s i t i o n 10.8 t h a t t h e r e exists
an open l o c a l l y s e m i a l g e b r a i c neighbourhood U o f A i n X which admits a
s e m i a l g e b r a i c r e t r a c t i o n r t o A. (For t h i s i t s u f f i c e s t o know that
Prop. 10.8 i s t r u e f o r s t r i c t l y locally finite complexes). I f we know

t h a t t h e proper q u o t i e n t UU exists, then we see by an application
o f Lemma 10.3 t o the open subsets X ^ A and U|JY o f XLJY t h a t the pro-
per q u o t i e n t X U Y f e x i s t s . Thus we may r e p l a c e X by U and henceforth
assume t h a t X admits a s e m i a l g e b r a i c r e t r a c t i o n R t o A.
The f u n c t i o n g = f«r : X ->Y i s semialgebraic (perhaps not proper) and
extends f : A-+Y. We choose a completion g :X Y o f g, c f . P r o p o s i t i o n
5.1. We denote the c l o s u r e of A i n X by A and the r e s t r i c t i o n o f g t o
A by f : A-> Y. Now we f o r g e t the f u n c t i o n g. The s o l e purpose o f this
f u n c t i o n was to y i e l d a completion f :A Y o f f w i t h A the c l o s u r e o f
A i n some c o m p l e t i o n X of X.
S i n c e f i s p r o p e r , we have f ^ (Y) = A. As shown above the p r o p e r quo-
tient map
p :XUY^XU jY
e x i s t s . The locally s e m i a l g e b r a i c subset L := p(X|jY) o f X U | Y has the
preimage XLJY under p. Thus the restriction
p I X U Y : XLiY -> L
i s a l s o proper and shows L to be the q u o t i e n t X U ^ Y . T h i s f i n i s h e s the
proof o f Theorem 10.7.

§11 - M o d i f i c a t i o n o f pure ends
Definition 1. A s h r i n k i n g ( r e s p . an expansion) o f a completed space
(A,A) i s a s h r i n k i n g morphism ( c f . Def. 9 a t the end o f §9) g : (A,A)
(B,B) ( r e s p . a s h r i n k i n g morphism h : (C,C) -> (A,A) ) .
I f cp : M <-» P and : M <-> Q a r e pure c o m p l e t i o n s o f a space M then t h e r e
exists a t h i r d pure c o m p l e t i o n x : M
*-* T
which admits morphisms t o cp
and \\>. We know from C o r o l l a r y 9.2 6 t h a t , f o r any end X € e (M) , we can
go from t h e completed end ( P ( X ) , P ( X ) ) i n P t o the completed end
( T T T J , T ( X ) ) i n T b y an e x p e n s i o n and then from ( T T X ) , T ( X ) ) t o (QTX) , Q ( X ) ) by
a s h r i n k i n g . The q u e s t i o n a r i s e s : Which s h r i n k i n g s and expansions o f
completed ends can be r e a l i z e d by morphisms between pure c o m p l e t i o n s ?
For the s h r i n k i n g problem we have a f u l l y satisfactory answer.
Theorem 11.1. L e t cp : M ^ P be a pure c o m p l e t i o n o f M. Given a family
of shrinkings (g. | X € e ( M ) ) , g, : (P ( X ) ,P ( X ) ) -> (B\ ,B. ) , o f the completed

A A A A
ends o f M i n P, there e x i s t s a morphism f : P Q from the c o m p l e t i o n
cp : M <-» P t o a n o t h e r pure c o m p l e t i o n \|; : M *-> Q such t h a t the shrinking
morphisms h^: ( P ( X ) , P ( X ) ) -+ (Q ( X ) , Q (X) ) o b t a i n e d from f a r e i s o m o r p h i c
to the g^ under (P ( X ) , P (X)) {i.e. there e x i s t isomorphisms
u x : (Q(X),Q(X)) (B ,B ) with u ^ h
A X A = g >.
A
P r o o f . We r e g a r d M as a subspace o f P and cp as the i n c l u s i o n map. Let
A =
: P\M = U ( P ( X ) IX € e(M)) .
Then
A = U ( P T x T | X € e(M))
and t h e f a m i l y (P ( X ) I X € e(M)) is locally finite, s i n c e the f a m i l y
( P ( X ) IX € e(M)) is locally finite. Let R(X) denote the e q u i v a l e n c e
r e l a t i o n on P ( X ) whose e q u i v a l e n c e c l a s s e s a r e the f i b r e s o f
g A : P (A) -» . The s e t P (A ) \ P ( A ) i s a u n i o n o f o n e - p o i n t equivalence
c l a s s e s . We c o n c l u d e t h a t , f o r A * y , t h e s e t P (A) D P ( y ) i s a u n i o n o f
o n e - p o i n t e q u i v a l e n c e c l a s s e s w i t h r e s p e c t t o b o t h r e l a t i o n s R ( A ) and
R(y). Thus the e q u i v a l e n c e r e l a t i o n s R ( A ) f i t t o g e t h e r t o an e q u i v a -
l e n c e r e l a t i o n R on A t o which Lemma 10.3 i s a p p l i c a b l e . By t h i s lemma
the p r o p e r q u o t i e n t p r o j e c t i o n h : A -> A/R =: L e x i s t s . F o r every
A € e(M) we have a commutative triangle
PCX)
w i t h h^ := h | P ( A ) , := h ( P ( A ) ) , and u^ an isomorphism. Since h i s
proper, the s e t := h ( P ( A ) ) i s locally s e m i a l g e b r a i c , and t h e s e t
is indeed the c l o s u r e of i n the p a r t i a l l y complete space L. A l -
so, by c o n s t r u c t i o n , u^(C^) = B^. Thus t h e proper p r o j e c t i o n h
yields, f o r every A, a morphism o f completed spaces
h A : ( P T A T , P ( A ) ) -+ (C ,C ),
A A
which i s i s o m o r p h i c t o g^ under (P(A),P(A)). In p a r t i c u l a r , e v e r y h^
i s a s h r i n k i n g morphism.
By t h e main r e s u l t i n t h e l a s t s e c t i o n , Theorem 10.7, we have a p r o -
per p r o j e c t i o n map
p : PLJL - P U L
H =: Q
g l u i n g t h e c l o s e d subspace A o f P t o L by t h e map h. {Here we u s e
Theorem 10.7 o n l y i n t h e easy case where t h e space X i n q u e s t i o n i s
partially complete.} The r e s t r i c t i o n o f p t o P i s s t i l l a proper s u r -
j e c t i o n f : P-*Q. The s e t N := f (M) i s l o c a l l y s e m i a l g e b r a i c and dense

1
i n Q. I t s preimage under f i s M. S i n c e a l l f i b r e s f~~ (y) ,y € N, a r e one-
point sets, the r e s t r i c t i o n f IM : M -*• N i s an isomorphism. Thus

f y i e l d s a completion : M -> Q by restriction to M. Clearly f is a
morphism from the c o m p l e t i o n M *-» P to ip. By P r o p o s i t i o n 8.10, ^ is
pure. The s h r i n k i n g maps o f the morphism f : P Q a r e the morphisms
h^ : ( P ( A ) , p ( A ) ) •+ ( C , C ) from above, which a r e

A A i s o m o r p h i c to the
g i v e n morphisms g^ : ( P ( A ) , P ( A ) ) -> (B^,B ) under A (P(A),P ( A ) ) . Theorem
11.1 i s proved.
By Theorem 11.1 we may shrink the ends P ( A ) of M i n a pure completion
P i n any way we want. In t h i s way we can make them " s i m p l e r " . For
example, l e t A = P ( A ) be a semialgebraic end of M. We triangulate (A,A).
L e t K denote the maximal c l o s e d subcomplex o f A c o n t a i n e d in A with
respect to the first b a r y c e n t r i c s u b d i v i s i o n of t h i s t r i a n g u l a t i o n .
Then K i s a s t r o n g d e f o r m a t i o n r e t r a c t o f A i n the o b v i o u s semialge-
braic sense, c f . Chapter I I I , §1 or [DK^, p. 136]. In p a r t i c u l a r K i s
connected. Applying Theorem 11.1 t o the p r o j e c t i o n (A,A) -> (A/K,A/K)
and the i d e n t i t y morphisms f o r the o t h e r ends o f M, we s h r i n k the end
A to a c o n t r a c t i b l e end A/K. I f the end A was complete (i.e. A = A),
then the new end i s a c t u a l l y a p o i n t , but otherwise there i s no means
f o r s h r i n k i n g A to a p o i n t , s i n c e the complexity c(A) has to be pre-
served ( c f . Th. 9.25).
E x a m p J ^ ^ J J ^ ^ Assume t h a t dimM = 2 and that the end P (A) o f M i n P i s
semialgebraic and not complete. Then P ( A ) is locally complete, since
a priori dimP(A) = 1 and hence c ( P ( A ) ) <_ 1. By the p r o c e d u r e just
described, we can s h r i n k P(A) i n such a way that (P(A),P(A)) becomes

•
i s o m o r p h i c t o the p a i rc (K ,K ) w i t h K the closed "star"
n n n
/ \
/ \
with n > 1 arms [PQ,P^ ], . . . , [ p / P l 0 n
a n <
^ K
n
t n e
°P e n
subcomplex where
the p o i n t s p.j , p , . . . , P
2 N a r e removed. The number n i s an i n v a r i a n t o f
X, namely the number o f ends o f P(X) ( c f . Th. 9.25).
We call an end P(X) o f M i n some c o m p l e t i o n P pure i f t h e dense pair
(P,P\p(X)) i s pure. C l e a r l y the c o m p l e t i o n i s pure i f and o n l y i f
all ends a r e pure.
One may ask whether a l l s t a r s ( K , K ) o c c u r as pure ends o f 2-dimen-

n n
sional semialgebraic spaces. This i s t r i v i a l l y t r u e by the following
general observation.
Example 11.3. L e t (A,A) be a completed space w i t h A c o n n e c t e d . Consi-
der t h e completed space
(P,M) := (Ax[0,1], (Ax[0,1]) ^ (Ax{0})).
The p a i r (P,M) i s pure and M has the unique end A x { 0 } i n P. The comple
ted end i s isomorphic to (A,A).
We now l o o k f o r e x p a n s i o n s o f the ends o f M i n a pure c o m p l e t i o n
cp : M <-> P.
Problem 11.4 (Expansion problem f o r ends). Given a f a m i l y of expan-
sions (g^ | X € e (M) ) ,
g
X : (
^X' X C }
(PTTy,P(A) ) ,
of the completed ends i n P, does there e x i s t a morphism f : Q P from
a pure c o m p l e t i o n : M -» Q t o cp such t h a t every s h r i n k i n g morphism
h^ : (Q(X),Q(X)) -+ ( P ( X ) , P ( X ) ) , o b t a i n e d from f by r e s t r i c t i o n , i s iso-
morphic t o g^ o v e r (P(X),P(X)) { i . e . , such t h a t t h e r e e x i s t s an iso-
morphism v : (C ,C ) (Q(X),Q(X)) w i t h h » v = g }?
We c a l l a morphism f : Q P as d e s c r i b e d here a s o l u t i o n o f the expan-
s i o n problem f o r cp : M P and t h e f a m i l y (g |X€e(M)) .

A
In o r d e r t o s t u d y t h e e x p a n s i o n problem i t s u f f i c e s to consider spaces
w i t h o n l y one end. T h i s i s a consequence of the f o l l o w i n g proposition.
Proposition 11.5. L e t cp : M<->P be a pure c o m p l e t i o n o f the space M and
let (g IX€e(M)) be a f a m i l y o f e x p a n s i o n s o f t h e completed ends o f M i n

A
P, g^ : ( C , C ) -* (P (X) ,P (X)) . We r e g a r d M as a subspace o f P and cp as the

A A
i n c l u s i o n map. The e x p a n s i o n problem i s solvable f o r M^*P and t h e f a m i l y
(g IX€e(M)) i f and o n l y i f , f o r e v e r y X€e(M), the e x p a n s i o n problem i s

A
C
solvable f o r P ^ P ( X ) * P and the e x p a n s i o n g^ o f the unique completed end
of P\P(X) i n P.
Proof. L e t f : Q P be a s o l u t i o n o f the e x p a n s i o n problem f o r M*=-> P and
the family (g^|-A€e(M)) . F i x an a b s o l u t e end X€e(M). A c c o r d i n g t o Theorem
11.1 we may s h r i n k a l l ends Q ( u ) , y * X, back t o P(y) l e a v i n g Q(X) un-
changed, u s i n g the i d e n t i t y as s h r i n k i n g map f o r (Q(X),Q(X)) and t h e
restrictions o f f as s h r i n k i n g maps f o r (Q(y),Q(y)), y 4= X. We then ob-
t a i n a c o m p l e t i o n o f PSP(X) which expands the unique end P(X) i n P i n
the p r e s c r i b e d way. Assume c o n v e r s e l y t h a t , f o r e v e r y XGe(M), a p r o p e r
s u r j e c t i v e map p^ : Q ^ P i s g i v e n which y i e l d s an isomorphism from
Q to P\P(X) and whose r e s t r i c t i o n p^ (P(X))-*P(X) i s isomor-
p h i c o v e r P ( X ) t o the g i v e n map g^ : P (X) . Assume a l s o that
p^ 1
(P ( X ) ) -* i s a pure c o m p l e t i o n . F o r e v e r y f i n i t e s e t S c e (M) we
introduce the f i b r e product o f t h e spaces X G S , over P which

(S)
r e s p e c t t o t h e maps p^. The n a t u r a l map p : Q -+P i s a g a i n a p r o p e r
g
-1
s u r j e c t i o n . Moreover, p y i e l d s an isomorphism from p
g (P\U ( P ^ I X e s ) ) g
-1 (S)
onto P ^ U ( P I X € S ) . The i n c l u s i o n map from p
A (P (P^ I X£S)) t o Q isag
pure c o m p l e t i o n , as i s e a s i l y seen. Over each P^, X e s , p g i s isomorphic
to p^, hence t o g^. F o r f i n i t e s u b s e t s S c= T o f e(M) we have a unique

(T) (S)
proper s u r j e c t i o n p T g :Q ->Q with P ° P
s T s = P - We
T consider
the s e t (!)
( S )
Q := l i m Q
S
w i t h S r u n n i n g through t h e f i n i t e s u b s e t s o f e(M). We want to e q u i p Q
w i t h the s t r u c t u r e o f a p a r t i a l l y complete space such t h a t the n a t u r a l

(S)
surjections q g : Q -* Q are a l l p r o p e r .
T h i s can be done e a s i l y . L e t q = q^ be the n a t u r a l s u r j e c t i o n from Q
to P. We choose a s i m u l t a n e o u s t r i a n g u l a t i o n o f P and M. Then e v e r y
P(A) i s a c l o s e d subcomplex of P. For e v e r y c l o s e d s i m p l e x o i n P,
the preimage q ^ (a) i n Q i s mapped b i j e c t i v e l y to pg^ (a) by q g for S large
enough. We t r a n s f e r the space s t r u c t u r e o f the complete semialgebraic

-1 — -1 — -1
space p s ( a ) t o the s e t q ( a ) by use o f q . For d i f f e r e n t
g simplices
a t h e complete spaces q ^ ( a ) fit t o g e t h e r w e l l . A p p l y i n g Theorem 1.3,
we o b t a i n the s t r u c t u r e o f a p a r t i a l l y complete space on Q such t h a t
e v e r y space q ^ (a), w i t h i t s g i v e n space s t r u c t u r e , i s a subspace o f

-1 —
Q, and the f a m i l y o f a l l q (a) i s l o c a l l y f i n i t e . With t h i s structure,
the map q : Q-»P i s p r o p e r s e m i a l g e b r a i c . I t y i e l d s an isomorphism from
q ^(M) t o M and s o l v e s the expansion problem f o r the g i v e n family
( g I A E c(M)) .
A q.e.d.
The expansion problem seems to be much more d i f f i c u l t than the s h r i n k -
i n g problem. With our p r e s e n t methods we can do almost n o t h i n g . Never-
t h e l e s s , the d i s c u s s i o n up to now makes i t c l e a r t h a t , f o r a l l q u e s t i o n s
about m o d i f i c a t i o n s of ends, we can d e a l w i t h each end o f M separately.
Thus, f o r the r e s t o f t h i s s e c t i o n , we u s u a l l y assume t h a t M has a s i n g l e
a b s o l u t e end, and we speak o f "the" end o f M i n a g i v e n c o m p l e t i o n .
The end A of M i n P can be s i m p l i f i e d by s h r i n k i n g , b u t u s u a l l y the
way i n which A i s embedded i n P, i . e . t h e r e l a t i o n o f A t o M, w i l l be-
come more c o m p l i c a t e d . Thus, f o r many purposes, i t seems t o be impor-

t a n t t o l o o k f o r an e x p a n s i o n o f A which makes t h e embedding as s i m p l e
as p o s s i b l e . We f o c u s a t t e n t i o n on t h e case where t h e end A i s l o c a l l y
complete, which means t h a t M has c o m p l e x i t y < 2 . Then 3A = A ^ A i s c l o s e d
i n P. L e t P := P ^ 3 A .
Definition 2 . The end A o f M i n P i s c a l l e d c o l l a r e d i f t h e r e e x i s t s an
open l o c a l l y s e m i a l g e b r a i c neighbourhood U o f A i n P t o g e t h e r w i t h an
isomorphism a : Ax [ o , 1 ] - S u f 1 P such t h a t a (Ax [ o , 1 [ ) = U and a (a,0) = a
for every a € A .
Of c o u r s e , a c o l l a r e d end i s pure and l o c a l l y complete, hence has com-
plexity <1 .
Theorem 1 1 . 6 . Assume t h a t M has c o m p l e x i t y c(M) < 2 . Then, f o r any
c o m p l e t i o n M <-* P, t h e r e e x i s t s a c o m p l e t i o n M <-> Q and a p r o p e r morphism
f :Q P such t h a t t h e end o f M i n Q i s c o l l a r e d .
To prove t h i s we need t h e f o l l o w i n g g e n e r a l f a c t which has been proved
for locally complete s e m i a l g e b r a i c spaces i n [DK^, § 3 ] . The p r o o f goes
through f o r a r b i t r a r y locally complete ( r e g u l a r , paracompact) spaces
w i t h o u t any s e r i o u s modification.
Theorem 1 1 . 7 [DK^, Th. 2 ] . L e t N be a l o c a l l y complete space and B a
closed l o c a l l y s e m i a l g e b r a i c s u b s e t o f N. Then t h e r e e x i s t s an open
locally s e m i a l g e b r a i c neighbourhood U o f B i n N, a p r o p e r s e m i a l g e b r a i c
map g from 3U := U ^ U t o B, and an isomorphism a : Z (g) U from t h e
mapping c y l i n d e r o f g t o U w i t h a(u,0) = u f o r u € 3U and oc(b) = b f o r
b€B. (We r e g a r d 3Ux[o,1[ and B as subspaces o f Z ( g ) , as usual.)
The proof o f Theorem 1 1 . 7 i n [DK^] g i v e s an e x p l i c i t d e s c r i p t i o n o f U,
g, and a. C h o o s i n g a s i m u l t a n e o u s t r i a n g u l a t i o n of N and B one t a k e s

for U t h e s t a r o f B i n N w i t h r e s p e c t t o t h e second b a r y c e n t r i c s u b d i -
v i s i o n o f t h e t r i a n g u l a t i o n . For g one t a k e s t h e r e s t r i c t i o n t o 3U o f
the r e t r a c t i o n r c o n s t r u c t e d i n the proof of P r o p o s i t i o n 10.8, cf.
[DK^, § 2 ] . F o r a one t a k e s t h e map d e s c r i b e d by t h e f o r m u l a s
oc(p(u,t)) = ( 1 - t ) u + t r ( u ) (u € 3U,t e [0, 1 ]) .
a (b) = b (b € B)
w i t h p : 3Ux [0,1 ] -> Z (g) t h e c a n o n i c a l p r o j e c t i o n .
We s h a l l use Theorem 11.7 o n l y i n t h e case where N i s p a r t i a l l y comple-
t e . We f i r s t prove Theorem 11.6 i n t h e s p e c i a l case t h a t t h e end A : = P^M
of M in P i s partially complete. A p p l y i n g Theorem 11.7 t o t h e space
N := P and t h e s u b s e t B :=.A, we o b t a i n a new c o m p l e t i o n Q o f M by g l u -
ing P ^ U t o O U ) x [ o , 1 ] a l o n g 3U w i t h the o b v i o u s isomorphism 3U3Ux{0}
( c f . Theorem 1 .3 o r Theorem 10.7). Then g l u i n g the i d e n t i t y of P ^ U
w i t h the map ocp : 3Ux [0,1 ] ->U, where p i s t h e p r o j e c t i o n from 3Ux[0,1]
to Z ( g ) , we o b t a i n a p r o p e r map f : Q-» P ( c f . I , Prop. 3.16). f i s an
morphism from t h e c o m p l e t i o n Q o f M t o t h e c o m p l e t i o n P o f M, and t h e
end o f M i n Q i s c o l l a r e d .
Assume now t h a t the end A = P N M o f M i n P i s o n l y l o c a l l y complete.
The s e t 3A = A ^ A i s closed i n P and c o n t a i n e d i n M. We c o n s i d e r t h e
open subspace L := 3A o f M. The space P i s a c o m p l e t i o n o f L, and L
has t h e end A i n P. We know t h a t the c o m p l e t i o n L ^ P o f L i s dominated
by a c o m p l e t i o n L «-> Q' such t h a t t h e end o f L i n Q' i s c o l l a r e d . L e t

f
f :Q -r*P be t h e morphism from t h e second completion of L to the f i r s t
JLi
1
one. L e t C denote t h e preimage o f 3A i n Q and l e t h : C-»3A denote t h e
restriction o f f t o C. Then
Q := Q' U 3A
h
is a c o m p l e t i o n o f M w h i c h dominates t h e c o m p l e t i o n P o f M. The end o f
M i n Q i s c o l l a r e d , as i s e a s i l y seen, and Theorem 11.6 i s p r o v e d .

We shall prove i n the n e x t c h a p t e r ( I I I , §8) t h a t a c o l l a r e d end A o f
M i s , up t o homotopy e q u i v a l e n c e , independent of the chosen completion
P o f M. A l s o , i f M i s an (open) l o c a l l y semialgebraic manifold, then
e v e r y c o l l a r e d end A o f M i s a g a i n a (closed) l o c a l l y semialgebraic
m a n i f o l d . Thus M has a c o m p l e t i o n which i s a l o c a l l y s e m i a l g e b r a i c mani-
f o l d w i t h boundary. We hope to e x p l a i n t h i s i n the second volume o f the
l e c t u r e notes ( c f . [SY, Th. 6.2] f o r R= IR and M s e m i a l g e b r a i c ) .

§12 - The S t e i n factorization o f a s e m i a l g e b r a i c map
We s t a r t w i t h a p r o p e r (hence s e m i a l g e b r a i c ) map f : M -*N between spa-
ces. L e t R ( f ) denote t h e e q u i v a l e n c e r e l a t i o n whose e q u i v a l e n c e c l a s -
ses a r e t h e connected components o f t h e non empty f i b r e s o f f . We ask
whether t h e proper q u o t i e n t M/R(f) e x i s t s . Assume t h i s i s t h e case and
cp : M->M/R(f) i s the q u o t i e n t p r o j e c t i o n . S i n c e f maps every e q u i v a l e n c e
class o f R ( f ) t o a p o i n t we have a factorization
(*)
of f i n t o p r o p e r maps which we c a l l a S t e i n factorization of f(in
analogy t o t h e complex a n a l y t i c Stein factorization). We s h a l l often
speak o f " t h e " S t e i n factorization of f , but a c t u a l l y the quotient
space M/R(f) i s o n l y determined up t o isomorphism.
The second factor TT i n t h e S t e i n factorization (*) o f f has f i n i t e
f i b r e s and i s p r o p e r . Thus TT i s a f i n i t e map. The f i b r e s o f t h e
first factor cp a r e connected, and cp i s s u r j e c t i v e . Assume t h a t
i s any f a c t o r i z a t i o n of f with p f i n i t e , \p s u r j e c t i v e , and t h e f i b r e s
of \\j connected. Then t h e f i b r e s o f a r e t h e connected components o f
t h e f i b r e s o f f , and i s p r o p e r . Thus t h e map shows L t o be t h e
q u o t i e n t o f M by R ( f ) and, up t o isomorphism, the f a c t o r i z a t i o n s (*)
and (**) a r e t h e same.
Theorem 12.1. Every p r o p e r reap f : M-»N between spaces has a S t e i n
factorization.
P r o o f . The s e t f(M) i s c l o s e d and l o c a l l y semialgebraic i n N. Suppose
we have found a S t e i n f a c t o r i z a t i o n g = p«tp o f the map g : M -» f (M) ob-
tained from f by r e s t r i c t i o n of the range space. Then f = Ttotp, w i t h TT
the composite o f p and the i n c l u s i o n f(M) ->N, i s a Stein factorization
o f f . Thus, i n o r d e r to prove t h e theorem, we may assume t h a t f i s s u r -
jective. Let
p : ( M x [ 0 , 1 ]) U N - Z ( f )
be the q u o t i e n t p r o j e c t i o n onto the mapping c y l i n d e r of f . As usual,
we identify t h e subspaces M x [ o , 1 [ and N o f (Mx[0,1])UN with their
images under p, thus r e g a r d i n g them a l s o as subspaces of Z ( f ) . S i n c e
f i s surjective, the r e s t r i c t i o n g : M x [ 0 , 1 ] ->Z(f) o f p i s a l s o s u r -
j e c t i v e . Moreover g i s p r o p e r . We now c o n s i d e r the dense p a i r s
( M x [ 0 , 1 ] , M x [ 0 , 1 [ ) and ( Z ( f ) , M x [ 0 , 1 [ ) . We may r e g a r d g as a morphism
from the f i r s t p a i r t o the second. Consequently we have a unique
f a c t o r i z a t i o n of g
Mx[0,1]
(***)
w i t h h the pure h u l l o f the dense p a i r (Z (f) ,Mx [ 0 , 1 [) (Prop. 8 . 3 ) The map
h is finite and t h e map g has connected fibres (Prop. 8 . 4 ) . Moreover
g i s s u r j e c t i v e and p r o p e r . L e t N := h ^ (N). We have g ^(N) = g ^ (N) =
M x { l } . I d e n t i f y i n g MX{1} w i t h M i n the o b v i o u s way and r e s t r i c t i n g the
diagram (***) to Mx{1} and i t s images N and N we obtain a Stein facto-
r i z a t i o n of f . q.e.d.
As a consequence o f t h i s p r o o f we have a s i d e remark on mapping cylin-
ders .
Corollary 1 2 . 2 . L e t f : M ->N be a p r o p e r s u r j e c t i v e map between s p a c e s .

The dense p a i r (Z(f),Mx[0,1[) i s pure i f and o n l y i f the f i b r e s o f f
are connected.
The S t e i n f a c t o r i z a t i o n can a l s o be c h a r a c t e r i z e d as a "maximal" f a c t o -
r i z a t i o n with t h e second f a c t o r a f i n i t e map.
Proposition 12.3. Assume f : M-*N i s a p r o p e r map and
a r e two f a c t o r i z a t i o n s o f f i n t o l o c a l l y semialgebraic maps w i t h TT and
p finite. Assume a l s o t h a t cp i s s u r j e c t i v e and has c o n n e c t e d fibres,
{i.e. the f i r s t factorization i s Stein. Notice t h a t (p and are auto-
m a t i c a l l y proper.} Then t h e r e e x i s t s a unique l o c a l l y semialgebraic
map h : S -> T, such t h a t p°h = TT and h°cp = . {N.B. h i s p r o p e r , hence s e m i a l a e b r a i c
Proof. I t i s evident t h a t t h e f i b r e s o f >j» a r e u n i o n s o f c o n n e c t e d com-
ponents o f the f i b r e s o f f . Thus t h e r e e x i s t s a unique l o c a l l y semialgebraic
map h : S -* T w i t h h»cp = Composing w i t h p we o b t a i n
p«h»(p = p»\|/ = TT • cp.
Since cp i s s u r j e c t i v e t h i s implies p«h = TT . q.e.d.
We seek a g e n e r a l i z a t i o n o f t h e S t e i n f a c t o r i z a t i o n o f p r o p e r maps f o r
an a r b i t r a r y semialgebraic map f : M-+N, i n t h e sense t h a t Proposition
12.3 remains t r u e . I t w i l l t u r n o u t t h a t such a f a c t o r i z a t i o n f = TT©cp
does e x i s t , and t h a t t h e f i r s t f a c t o r cp i s a "pure" s e m i a l g e b r a i c map.
We f i r s t e x p l a i n what we mean by a pure map and s t a t e some easy o b s e r -
vations about t h e s e maps.
Definition 1. A s e m i a l g e b r a i c map f : M -> N i s c a l l e d pure a t a p o i n t

y€N, i f y has a fundamental system o f open s e m i a l g e b r a i c n e i g h b o u r -
-1
hoods U such t h a t every preimage f (U) i s non-empty and c o n n e c t e d .
The map f i s called pure i f f i s pure a t e v e r y p o i n t o f N.
A pure map has a dense image.
Example 12.4. L e t f : M ->N be an embedding, i . e . an isomorphism from M
onto a subspace o f N. Then f i s pure i f and o n l y i f f(M) i s dense i n N
and the dense p a i r (N,f(M)) i s pure i n t h e sense o f §8.
P r o p o s i t i o n 12.5. F o r e v e r y s e m i a l g e b r a i c map f : M-*N the f o l l o w i n g
are equivalent.
i) f i s pure.
ii) For e v e r y non-empty connected open s e m i a l g e b r a i c s u b s e t U o f N
the preimage f (U) i s non-empty and c o n n e c t e d .
iii) For e v e r y non-empty connected open l o c a l l y s e m i a l g e b r a i c s u b s e t

1
U o f N t h e preimage f (U) i s non-empty and c o n n e c t e d .
P r o o f . The i m p l i c a t i o n s i i i ) => i i ) i ) are t r i v i a l . ( R e c a l l t h a t every
p o i n t of a l o c a l l y s e m i a l g e b r a i c space has a fundamental system o f open
connected s e m i a l g e b r a i c neighbourhoods.) In o r d e r t o prove i ) i i i ) ,

o -1
l e t a non-empty connected s e t U C 7(N) be g i v e n . C e r t a i n l y f (U) 4=0,
-1
s i n c e f has dense image, Suppose f (U) i s t h e d i s j o i n t union of two
open non-empty l o c a l l y s e m i a l g e b r a i c s u b s e t s V-^V^. Then f ( f ^ (U) ) =
U n f(M) i s the u n i o n o f the two s u b s e t s f (V,j) , f (V" ) . They a r e l o c a l l y

2
s e m i a l g e b r a i c s i n c e f i s s e m i a l g e b r a i c . S i n c e U fl f (M) i s dense i n U,
we have
u = (u n f ( v ) )
1 u (u n f ( v ) )
2 .
S i n c e U i s connected, t h e s e t s U n f ( V ^ ) and U n f ( V ) must meet i n some 2
p o i n t y. L e t W be any open s e m i a l g e b r a i c neighbourhood o f y w i t h WcU.

Then
wnf(M) = [ w n f ( v ) ] 1 u [wnf(v )] 2
and n e i t h e r s e t WD f (V^) n o r Wfl f (V ) i s empty. T a k i n g preimages

2 under
f we o b t a i n
f" 1
(w) = [ f ~ (w) n v ^ u [ f ~ (w) n v l .
1 1
2
The two s e t s on t h e r i g h t - h a n d s i d e a r e open, s e m i a l g e b r a i c , non-empty

_1
and d i s j o i n t . Thus f (W) i s n o t connected, and we see t h a t f i s n o t
-1
pure a t y. T h i s c o n t r a d i c t i o n proves t h a t f (U) i s connected.
q.e.d.
-1
Proposition 12.6. A proper map f : M-+N i s pure i f and o n l y i f f (y)
i s connected and non-empty f o r e v e r y y € N .
P r o o f . I f f i s pure, then f has a dense image, hence f i s s u r j e c t i v e .
Assume t h a t a f i b r e f (y) i s t h e union o f two non-empty disjoint
open s e m i a l g e b r a i c s u b s e t s - | f V .
v
2 We may choose d i s j o i n t open semi-
1
a l g e b r a i c subsets U 1 f U 2 o f M w i t h U fl f " (y) = V
± i ( i = 1,2). S i n c e
f i s proper, t h e r e i s an open connected s e m i a l g e b r a i c neighbourhood

1 1
U o f y i n N such t h a t f ~ ( U ) i s c o n t a i n e d i n U U . Thus f ~ ( U ) i s
2
not connected, i n c o n t r a d i c t i o n t o Prop. 12.5.ii.
C o n v e r s e l y assume t h a t f ^(y) i s connected and non-empty f o r e v e r y
y€N. Then the preimage f ^ (U) o f any connected open l o c a l l y semi-
a l g e b r a i c subset U o f N i s connected by Sublemma 6.6.
Lemma 12.7. L e t f : M-*S, g : S-»N be s e m i a l g e b r a i c maps.
a
i) I f f and g a r e pure then g f i s pure.
ii) I f g o f i s pure and f has a dense image then g i s pure.
iii) I f gof i s pure and g i s an embedding, then f i s pure.

P r o o f . i ) The first statement i s evident.
ii) L e t U be a connected open s e m i a l g e b r a i c s e t i n N. The set

-1 -1 -1
f (g (U)) i s connected, hence i t s image g (U) n f(M) under f i s
connected. S i n c e f(M) i s dense, the c l o s u r e o f t h i s set contains

-1 -1
g (U). Thus g (U) i s a l s o connected.
i i i ) We r e g a r d g as an i n c l u s i o n map. L e t U be an open connected semi-
a l g e b r a i c s e t i n S. We choose an open s e m i a l g e b r a i c s e t W i n N
w i t h WDS = U ( c f . I, Prop. 3.14). L e t V be the connected compo-
nent o f W c o n t a i n i n g U. Then a l s o V n S = U. We have
1 1
f" (U) = (g«f)"" (V) .
S i n c e gof i s pure t h i s s e t i s connected. q.e.d.
Definition 2. A pure c o m p l e t i o n of a l o c a l l y semialgebraic map
f : M-*N i s a commutative diaaram
(*)
wi-^h a and 3 pure completions o f M and N.
Remark 12.8. Every l o c a l l y s e m i a l g e b r a i c map f has a pure completion
(*) w i t h p r e s c r i b e d pure completion 3 o f N. I f f i s semialgebraic,
then h i s p r o p e r .
Indeed, g i v e n a pure c o m p l e t i o n 3 o f N, we o b t a i n a diagram (*) with
a a completion o f M by a p p l y i n g P r o p o s i t i o n 5.1. Then we r e p l a c e a by
a pure h u l l a : M*-> P o f a and h by the composite hop w i t h p the finite

p r o j e c t i o n map from P t o P. The second sentence i n t h e remark i s e v i -
dent from Lemma 8.1.
Lemma 12.7 has t h e f o l l o w i n g immediate consequence which g i v e s us an
u n d e r s t a n d i n g o f p u r i t y o f a s e m i a l g e b r a i c map f i n terms o f a pure
completion of f .
Proposition 12.9. L e t
a ^
(*)
be a commutative diagram o f s e m i a l g e b r a i c maps w i t h a and 3 dense em-
beddings .
i) I f f and 3 a r e pure then h i s pure.
ii) I f h and a a r e pure then f i s pure.
iii) I n p a r t i c u l a r , i f (*) i s a pure c o m p l e t i o n o f f , then f i s pure
if and o n l y i f h has connected fibres.
In t h e f o l l o w i n g we do n o t work w i t h c o m p l e t i o n s b u t w i t h "proper ex-
t e n s i o n s " o f s e m i a l g e b r a i c maps.
D e f i n i t i o n 3. A proper e x t e n s i o n o f a s e m i a l g e b r a i c map f : M N is a
commutative triangle
M < - *P
^ \ (
** )
w i t h a a dense embedding and g a proper map.
Remark 12.10. Every s e m i a l g e b r a i c map f : M ->N h a s a p r o p e r e x t e n s i o n
(**) w i t h a pure. We o b t a i n every such p r o p e r e x t e n s i o n from a pure

completion (*) o f f by r e s t r i c t i n g h to t h e preimage o f N under h ( r e -
g a r d i n g a and 3 as i n c l u s i o n maps).
Definition 4. A S t e i n f a c t o r i z a t i o n o f a s e m i a l g e b r a i c map f : M-»N. i s
a commutative triangle
(***)
w i t h cp pure and TT f i n i t e .
Theorem 12.11 . i ) E v e r y s e m i a l g e b r a i c map f : M -> N has a Stein factori-
zation.
ii) If (***) i s a Stein factorization and
i s any f a c t o r i z a t i o n o f f w i t h TT^ f i n i t e , then t h e r e e x i s t s a unique
s e m i a l g e b r a i c map h : S->S^ w i t h h»cp = cp^ and TT^*h = TT.
Proof. i ) We choose a p r o p e r extension (**) o f f w i t h pure embedding
a and a Stein f a c t o r i z a t i o n g = nof o f g.
Then cp := i|;»a i s a pure map and f = TT• cp i s a S t e i n f a c t o r i z a t i o n of f .
ii) We choose a p r o p e r e x t e n s i o n o f the map (cp,cp^) : M - > S x S ^ .
M c 9 * p
( C P , C P ) \ ^
1 ^/[tyrtyj)
SxS,
We then have a commutative diagram o f maps
w i t h ]p pure ( c f . Lemma 1 2 . 7 . i i ) and TT and TT.J f i n i t e . S i n c e a has dense
image, n»ij; = TT^<»^. By P r o p o s i t i o n 12.3 t h e r e e x i s t s a unique semialge
b r a i c map h : S-^S- w i t h T ^ • h = TT and h»ip = ij;-. . Hence h«cp = cp- . q.e.d
According t o t h e second p a r t o f t h i s theorem any two S t e i n factoriza-
t i o n s o f a s e m i a l g e b r a i c map f a r e i s o m o r p h i c . Thus we a r e a l l o w e d t o
t a l k about " t h e " S t e i n f a c t o r i z a t i o n of f .
Corollary 12.12 (Comparison o f S t e i n f a c t o r i z a t i o n s ) . L e t
be a commutative square w i t h a, 3 l o c a l l y s e m i a l g e b r a i c maps and f , g
s e m i a l g e b r a i c maps. Let.
be t h e S t e i n f a c t o r i z a t i o n s o f f and g. Then t h e r e e x i s t s a unique l o -
cally s e m i a l g e b r a i c map y : s T
with X*Y 3°TT
= =
and IJMOC = y*<P»
T h i s can be e a s i l y seen by a p p l y i n g t h e l a s t theorem t o t h e S t e i n fac-
t o r i z a t i o n o f f and t h e f i n i t e map Nx^T->N o b t a i n e d by p u l l i n g back
X : T -+ Q a l o n g 3.
We g i v e an a p p l i c a t i o n o f C o r o l l a r y 12.12.
Theorem 12.13. L e t
be a commutative square ( s o l i d arrows) w i t h f a pure s e m i a l g e b r a i c map,
g a partially f i n i t e map, and a,3 locally s e m i a l g e b r a i c maps. Then t h e r e
e x i s t s a unique l o c a l l y s e m i a l g e b r a i c map y : N -> P w i t h y°f = a and g*Y=3«
Proof. I f M i s s e m i a l g e b r a i c then, r e p l a c i n g P by the s e m i a l g e b r a i c sub-
space a (M) , we may assume t h a t g i s f i n i t e . In t h i s c a s e our c l a i m i s
contained in Corollary 12.12. In t h e g e n e r a l case we choose an admis-
sible covering ( N ^ l i G I ) o f N by open s e m i a l g e b r a i c s u b s e t s . For every

1 1
i € I the s e t f (N^) i s s e m i a l g e b r a i c and the r e s t r i c t i o n f ^ : f (N^)->N^
o f f i s a g a i n pure. Thus we have a unique s e m i a l g e b r a i c map y^ : ISL -> P

l
w i t h g»Yj_ = 3 1 ^ and f ^ = a | f ~ ' ( N ) . Byi the u n i q u e n e s s statement i n
Corollary 12.12 we have y. IN. PIN. = Y - l N . f l N . . Thus t h e y. f i t together

l i "2 J l 3
to a l o c a l l y s e m i a l g e b r a i c map Y : N -+P with the d e s i r e d p r o p e r t i e s .
q.e.d.
In the s p e c i a l case where f i s t h e i n c l u s i o n map L * [ o , 1 [ *-» L x [ o , 1 ] f o r
some space L t h i s theorem means a r e l a t i v e path completion theorem w i t h
parameters f o r p a r t i a l l y f i n i t e maps ( c f . I, Th. 6.8 for relative path
completion without parameters). I t would be interesting t o know some-
t h i n g about r e l a t i v e path c o m p l e t i o n w i t h parameters more g e n e r a l l y f o r
p a r t i a l l y proper maps.
Using the S t e i n f a c t o r i z a t i o n f = n° cp o f a g i v e n semialgebraic map
f : M->N we get a h o l d on the set of a l l points y € N with f pure a t y
(cf. D e f i n i t i o n 1).
Theorem 12.14. For any y € N, t h e map f i s pure a t y i f and only i f
_i
TT (y) c o n s i s t s o f one p o i n t . The set of these points y i s l o c a l l y
semialgebraic.
Proof. S i n c e cp i s pure, f i s pure a t y i f and only i f TT i s pure a t y.
Thus we may r e p l a c e f by TT i n the whole c o n s i d e r a t i o n and assume hence

-1
forth that f i s f i n i t e . If f (y) i s empty then there e x i s t s a neigh-
bourhood o f y which i s d i s j o i n t from the c l o s e d s e t f ( M ) . Thus f i s
not p u r e a t y. If f (y) c o n t a i n s s e v e r a l p o i n t s x^,...,x r then, by
standard arguments, t h e r e e x i s t s an open s e m i a l g e b r a i c neighbourhood
U of y such t h a t f (U) i s the d i s j o i n t union o f open semialgebraic
neighbourhoods V^,...,V r of # ...,x r e s p e c t i v e l y . The preimage f ^ (U
of e v e r y semialgebraic n e i g h b o u r h o o d U' c u of y i s disconnected. Thus
f i s not pure a t y.
Assume now t h a t f ^(y) = {x}. T h e r e e x i s t s a t r i a n g u l a t i o n o f f such
t h a t y i s a v e r t e x o f N, hence x i s a v e r t e x of M (Th. 6.13; we iden-
t i f y M and N with the c o r r e s p o n d i n g simplicial complexes). Now f is a
simplicial f i n i t e map, hence maps e v e r y open simplex of M isomorphical
l y o n t o an open simplex of N. For purely combinatorial reasons the
preimage o f the open s t a r U o f y i n N i s t h e open s t a r V of x i n M. It
then f o l l o w s t h a t f o r every A € ]0,1[ t h e preimage of t h e set (1-A)U+A{
is (1-A)v+A{x}, and we see t h a t f i s pure a t y. The last sentence i n
this theorem i s now evident from H a r d t ' s theorem (Th. 6.3). q.e.d.
We now want to d e s c r i b e the f i b r e s of the pure f a c t o r cp of the S t e i n
factorization f = TT• cp of a g i v e n s e m i a l g e b r a i c map f : M->N. Let Con(f)
denote t h e s e t o f a l l c o n n e c t e d components o f the preimages f ^(U) of
all open s e m i a l g e b r a i c subsets U of N with Ufl f (M) * 0 .
Lemma 12.15. L e t
be a commutative diagram o f s e m i a l g e b r a i c maps w i t h x pure. Then
Con(f) = ( x " (W) i W E C o n ( g ) } .

1
-1 -1 -1
P r o o f . F o r a g i v e n U E Jf(N), t h e preimage f (U) = x <? (U) i s n o t empty
1
if and o n l y i f g (U) i s n o t empty. Assume t h a t we a r e i n t h i s c a s e and
-1 -1
let , . ..,W r be t h e c o n n e c t e d components o f g (U). Then f (U) i s t h e
-1 -1
d i s j o i n t union o f t h e non empty open s e t s x (W-j),..., x (W ) , and t h e s e

r
a r e connected (Prop. 12.5). Thus t h e x 1

(W^) a r e t h e connected compo-
nents of f (U). T h i s g i v e s us t h e c l a i m i n t h e lemma.
As b e f o r e , l e t f : M-> N be a s e m i a l g e b r a i c map. F o r any x € M we denote

1
by C ( f , x ) t h e c o n n e c t e d component o f x i n t h e f i b r e f f (x) through x.
We f u r t h e r denote by D(f,x) t h e i n t e r s e c t i o n o f a l l s e t s V€Con(f)
which contain x. We have
1
C(f,x) c D(f,x) c f~ f(x).
Lemma 12.15 has t h e f o l l o w i n g immediate consequence.
Lemma 12.16. In t h e s i t u a t i o n o f Lemma 12.15,
1
D(f,x) = " ( D ( g , ( x ) ) )
x X
f o r every x € M.
If f i s proper t h e n i t i s p r e t t y o b v i o u s t h a t C ( f , x ) = D(f,x) f o r e v e r y
x € M. In p a r t i c u l a r , i f f i s finite t h e n D ( f , x ) = {x} f o r every x € M .
A p p l y i n g t h i s remark t o t h e f i n i t e factor TT o f a S t e i n factorization
f = TT • cp o f an a r b i t r a r y s e m i a l g e b r a i c map f we o b t a i n t h e d e s i r e d
d e s c r i p t i o n o f t h e f i b r e s o f cp from Lemma 12.16.

Theorem 12.17. F o r every x € M the f i b r e cp cp(x) through x o f t h e pure
f a c t o r cp o f a S t e i n f a c t o r i z a t i o n o f f c o i n c i d e s w i t h t h e s e t D ( f , x ) .
N.B. In p a r t i c u l a r , t h e s e t s D(f,x) a r e unions o f components o f f f ( x ) .
Another a p p l i c a t i o n o f Lemma 12.16 g i v e s us
Proposition 12.18. L e t
a
M P
be a proper e x t e n s i o n o f f w i t h a pure. We r e g a r d a as an i n c l u s i o n map.
Then, f o r e v e r y x £ M
D(f ,x) = C(g,x) n M.
I t may w e l l happen t h a t D(f,x) * C ( f , x ) , i f f i s n o t p r o p e r . F o r example
l e t g : P -* N be a p r o p e r map w i t h connected f i b r e s and l e t f : M-> N be
the r e s t r i c t i o n o f g t o a pure dense l o c a l l y semialgebraic subset M of
P. Then, by P r o p o s i t i o n 12.18, t h e s e t s D(f,x) a r e j u s t t h e f i b r e s o f
f . But some f i b r e s may be n o t connected.
Most o f the r e s u l t s o f t h i s s e c t i o n have w e l l known analogues i n topo-
l o g y , c f . [M], [MV].
§13 - S e m i a l g e b r a i c spreads
We now make a more d e t a i l e d study o f the case where the pure factor
of t h e S t e i n f a c t o r i z a t i o n o f a g i v e n s e m i a l g e b r a i c map f i s an em-
bedding ( i . e . an isomorphism onto i t s image).
Definition 1 . A l o c a l l y s e m i a l g e b r a i c map f : M-*N i s c a l l e d a spread
if t h e f o l l o w i n g h o l d s f o r every x € M: I f U runs through the open semi-
a l g e b r a i c neighbourhoods o f y := f ( x ) then the connected component o f

-1 -1
x in f (U), which we denote by f ^^ ' x runs through a fundamental
system o f neighbourhoods of x. Of c o u r s e , a s p r e a d has d i s c r e t e fibres.
We r e f e r the r e a d e r t o [F] , [M], [M^] f o r the t h e o r y o f spreads i n
topology.
Examples 13.1. i ) E v e r y f i n i t e map i s a spread.
ii) E v e r y embedding i s a spread.
iii) The composite g«f o f two s p r e a d s f : M-* N and g : N -* L i s a s p r e a d .
T h i s i s o b v i o u s , s i n c e f o r any p o i n t x € M and any open s e m i a l g e b r a i c
neighbourhood U o f g»f(x) we have, w i t h y := f ( x ) ,
1 1 1
(*) (g«>f)" ( U ) x = f " (g" ( U ) ) . y x
i v ) A map f : M-*N between spaces i s c a l l e d a local isomorphism i f every
x € M has an open s e m i a l g e b r a i c neighbourhood V such t h a t f ( V ) i s open
and s e m i a l g e b r a i c i n N and f y i e l d s an isomorphism from V onto f ( V ) .
{Notice t h a t here we use the word " l o c a l " i n a weak sense, but c f . Theo-
rem 13.8 below.} C l e a r l y every l o c a l isomorphism i s a spread. In p a r t i -
c u l a r , c o v e r i n g s , t o be d i s c u s s e d i n Chapter V , are spreads. A l s o , f o r
every l o c a l l y complete space M, t h e i d e n t i t y map from M ^ q c to M ( c f . I,
§7) i s a l o c a l isomorphism and hence a s p r e a d .
v) More g e n e r a l l y l e t f : M->N be a map between spaces. Assume t h a t
every x £ M has a neighbourhood U i n M, such t h a t the restriction

f lu : U -* N i s a spread. Then f i s a spread.
To prove this fact, l e t x€M and U be a neighbourhood o f x i n M such
t h a t f l U i s a spread. S i n c e M i s r e g u l a r , t h e s e t ^ := {V€r(M)lxGV,
VcU} i s a fundamental system o f neighbourhoods o f x i n M. F o r e v e r y
V € V we f i n d an open s e m i a l g e b r a i c neighbourhood W o f f ( x ) such t h a t

1 1
(f |U)" (W) c v . S i n c e V c U , the s e t ( f l U ) ~ ( W ) x i s n o t o n l y open and
c l o s e d i n U n f " ( w ) , but a l s o i n f " ( W ) . Hence f "

1 1 1
(W) = (fIU)~ (W) .
1
We see t h a t f i s indeed a spread.
v i ) A p a r t i a l l y f i n i t e map between spaces i s a spread. T h i s f o l l o w s
immediately from i ) and v ) .
vii) I f M i s l o c a l l y complete, then every map f : M-*N with discrete
fibres i s a spread. Indeed, every x € M has a complete s e m i a l g e b r a i c
neighbourhood K. The map f l K : K-*N i s proper and has f i n i t e fibres,
hence i s f i n i t e . Thus f!K i s a spread. By v) f i s a l s o a s p r e a d .
Let f :X Y be a s i m p l i c i a l map between l o c a l l y f i n i t e complexes X,Y.
Then f i s a spread i f and o n l y i f f l a i s i n j e c t i v e f o r every open sim-
p l e x a o f X. Indeed, t h i s c o n d i t i o n i s c e r t a i n l y necessary since a
s p r e a d has d i s c r e t e f i b r e s . Assume now t h a t f l a i s i n j e c t i v e f o r every
a € I ( X ) . As always, we denote t h e c o m p l e t i o n o f f t o a s i m p l i c i a l map
from X t o Y by f . C l e a r l y f la i s i n j e c t i v e f o r every a 6 I ( X ) . L e t x € X
1
be g i v e n and y := f (x). I t i s p r e t t y o b v i o u s t h a t f ( S t ( y ) ) i s the
Y
d i s j o i n t union o f t h e s t a r s S t ^ ( z ) w i t h z running through f ^ (y) ( c f .
b e g i n n i n g o f t h e p r o o f o f Prop. 7.6). Thus S t ( x ) i s the connected

v com-
ponent o f f ^ ( S t ( y ) ) c o n t a i n i n g x. T h i s i m p l i e s t h a t , f o r every
Y
A € ]0,1[ the "shrunken s t a r " (1-A)x + A s t ( x ) i s a connected

x component
of f ~ ( ( 1 - A ) y + A s t ( y ) ) . We see t h a t f i s a spread.
1
y
In t h e f o l l o w i n g we f o c u s a t t e n t i o n on s e m i a l g e b r a i c spreads, i . e .
spreads which a r e s e m i a l g e b r a i c maps. But n o t i c e t h a t f : M-» N i s a

s p r e a d i f and o n l y i f t h e r e s t r i c t i o n f|X o f f t o e v e r y s e m i a l g e b r a i c
subset X o f M i s a spread (cf. 13.1.ii, iii, v ) . Thus o u r r e s u l t s will,
t o some e x t e n t , y i e l d an u n d e r s t a n d i n g of a r b i t r a r y spreads.
Theorem 13.2. L e t f : M-»N be a s e m i a l g e b r a i c map. f i s a spread i f and
o n l y i f t h e pure f a c t o r cp i n the S t e i n f a c t o r i z a t i o n o f f i s an embedding.
Proof. Let M - -S
N
/^^
be a S t e i n f a c t o r i z a t i o n o f f . I f cp i s an embedding then both TT and cp
are s p r e a d s , hence f i s a l s o a s p r e a d .
Assume now t h a t f i s a s p r e a d . Then, f o r every x € M, the s e t D(f,x)
i n t r o d u c e d i n §12 i s t h e one p o i n t s e t {x}. Thus, by Theorem 12.17,
the map cp i s i n j e c t i v e . In o r d e r t o prove t h a t cp i s an embedding i t
suffices to v e r i f y t h a t the b i j e c t i v e map : M->cp(M), o b t a i n e d from
cp by r e s t r i c t i o n o f the range, i s open. L e t x € M, y := cp(x). S i n c e cp

-1 -1
i s pure, cp (TT ( U ) ) i s connected

y f o r every open s e m i a l g e b r a i c
neighbourhood U o f f ( x ) . Thus by f o r m u l a (*) i n 1 3 . 1 . i i i , we have

cp" 1
1
(TT" ( U ) ) = f " ( U ) .
y
1
x
Applying t o t h i s e q u a l i t y we o b t a i n
iMf~ 1
(U) ) x = TT" (U)
1
n cp(M) .
-1
Since f i s a spread, f x runs through a fundamental system o f
neighbourhoods o f x i f U runs through t h e open s e m i a l g e b r a i c n e i g h -
bourhoods o f f ( x ) . Thus i s i n d e e d open. q.e.d.

We look f o r a c r i t e r i o n s a y i n g t h a t a g i v e n s e m i a l g e b r a i c map f : M-*N
has a finite extension ( i n the sense o f §12, Def. 3), i . e . a factori-
zation
(**) M « - *P
w i t h a an embedding and TT f i n i t e . I f such a f a c t o r i z a t i o n (**) exists
then, r e p l a c i n g the embedding a by i t s pure h u l l a : M -> P and TT by nop,
w i t h p the f i n i t e p r o j e c t i o n from P onto P, we obtain a factorization
(**) w i t h a a pure embedding and TT f i n i t e . T h i s i s a S t e i n f a c t o r i z a -
t i o n o f f . Thus a n e c e s s a r y (and o b v i o u s l y s u f f i c i e n t ) condition for
the e x i s t e n c e o f a f i n i t e e x t e n s i o n o f f i s t h a t the pure f a c t o r cp i n
the S t e i n f a c t o r i z a t i o n i s an embedding. In view o f Theorem 13.2 we may
express this result i n the f o l l o w i n g way.
Theorem 13.3. A s e m i a l g e b r a i c map f : M ->N admits a f i n i t e extension i f
and o n l y i f f i s a s p r e a d . In t h i s case we have, up t o isomorphism, a
unique f i n i t e extension (**) where the embedding a i s pure, namely the
S t e i n f a c t o r i z a t i o n of f .
In §6 we have seen t h a t every f i n i t e map has a completion which i s a g a i n
finite (Cor. 6.14). Theorem 13.3 p r o v i d e s us w i t h a new proof of this
result.
Corollary 13.4. L e t f : M-* N be a f i n i t e map and l e t \J> : N «-> Q be a com-
p l e t i o n o f the space N. Then t h e r e e x i s t s a unique completion
c
M
of f w i t h f f i n i t e and cp pure.
Indeed, ip»f i s a s p r e a d and f,cp a r e t h e f a c t o r s o f t h e S t e i n f a c t o r i z a -
t i o n o f T|> • £ .
S e m i a l g e b r a i c spreads can be t r i a n g u l a t e d . Even t h e f o l l o w i n g holds.
Theorem 13.5. L e t f : M -»N be a s e m i a l g e b r a i c s p r e a d . L e t ( M ^ l i € I ) and
( N j I j €J) be l o c a l l y finite families of l o c a l l y semialgebraic subsets
o f M and N r e s p e c t i v e l y . Then t h e r e e x i s t s a t r i a n g u l a t i o n
x se—» M
g f (T)
-1 -1
of f such t h a t every \p (N_.) i s a subcomplex o f Y and every cp (M^)
is a subcomplex o f X.
P r o o f . We choose a f i n i t e e x t e n s i o n o f f ,
c
M *P
and r e g a r d M as a subspace o f P. I f we prove the claim f o r the f i n i t e
map TT and t h e f a m i l i e s (M, (JVLliei)) and ( N j j G J ) o f s u b s e t s o f P and N,
then t h e c l a i m i s a l s o proved i n the o r i g i n a l s e t t i n g . Thus we may

1
assume w i t h o u t loss of generality that f i s f i n i t e . Applying Hardt s
theorem (= Th. 6.3) we choose a simultaneous t r i a n g u l a t i o n ijj : Y -»N o f
N and t h e s e t s N_. such t h a t t h e map f and a l s o every restriction
f |Mj_ : ML -*N i s t r i v i a l over (a) f o r every open simplex a o f Y. Re-
placing \\) by i t s f i r s t barycentric s u b d i v i s i o n we know i n a d d i t i o n
t h a t t h e r e e x i s t s a (unique) triangulation (T) o f f which i n v o l v e s t h e
g i v e n t r i a n g u l a t i o n \p o f Y. ( c f . Th. 6.13). We want t o v e r i f y t h a t , f o r

1
every i € I, t h e s e t X i := cp (M^) i s a subcomplex o f X. But t h i s i s
o b v i o u s : The s e m i a l g e b r a i c map g|X.^ : X - » Y i s t r i v i a l

i over every a€Z(Y).
S i n c e g has d i s c r e t e f i b r e s t h i s means t h a t g maps every connected

component of g ^ (a) f i x , isomorphically onto a . Now g ^ ( a ) has as c o n -
n e c t e d components the open s i m p l i c e s T o f X w i t h g(x) = o, and g maps
every T i s o m o r p h i c a l l y onto a . (This i s a g e n e r a l fact f o r f i n i t e sim-

-1
plicial maps, c f . § 6 ) . Thus t h e c o n n e c t e d components o f g ( a ) n X.
must be some o f t h e s e s i m p l i c e s T . Running through a l l a £ I ( Y ) we see
t h a t X^ i s a u n i o n o f open s i m p l i c e s o f X. q.e.d.
T h i s theorem i m p l i e s a comparison statement f o r open c o v e r i n g s of M
and N i f a semialgebraic s p r e a d f : M -* N i s g i v e n .
Corollary 13.6. A g a i n , l e t f : M ->N be a s e m i a l g e b r a i c spread. L e t
(M li£I) and (N_.|j£J) be l o c a l l y

i finite coverings o f M and N by open
semialgebraic s e t s . Then t h e r e exists a locally finite covering (U^lkEK)
o f N by open s e m i a l g e b r a i c s e t s and f o r e v e r y k € K, a f a m i l y ( V ^ I A€A (k))
c o n s i s t i n g o f c o n n e c t e d components o f f (U^) such t h a t t h e f o l l o w i n g h o l d s :
a) E v e r y i s contained i n some N_. and every N.. i s t h e u n i o n o f f i n i -
t e l y many s e t s U^.
b) E v e r y i s contained i n some and every i s the union of
f i n i t e l y many s e t s .
In t h e s p e c i a l case t h a t f i s f i n i t e the l o c a l l y finite covering
(U^lkGK) by open s e m i a l g e b r a i c s e t s can be chosen i n such a way t h a t
a) and b) h o l d w i t h (V^ IA€A(k))

A the f a m i l y of a l l connected components
1
of f" (U ). k
Proof. We choose a t r i a n g u l a t i o n o f f as i n d i c a t e d i n Theorem 13.5. Re-
placing t h e t r i a n g u l a t i o n s o f M and N by t h e i r b a r y c e n t r i c subdivisions
we may assume t h e f o l l o w i n g : M = X, N = Y w i t h s t r i c t l y locally finite
complexes X and Y. The map f i s s i m p l i c i a l and i t s c l o s u r e f i s f i n i t e .
Every i s an open subcomplex X^ o f X, and e v e r y N.. i s an open subcom-
p l e x Y_. o f Y. Furthermore, every open simplex a <z X ( r e s p . a c X , ) has a t
l e a s t one v e r t e x lying i n X ( r e s p . i n X - ) / and e v e r y open simplex xcY

Y I n
(resp. T C Z Y J ) has a t l e a s t one vertex lying in Y (resp. j ) - this
situation i t is easily verified t h a t the covering (U^lkEK) w i t h
running t h r o u g h the open s t a r s S t ( w ) w i t h w € E ( Y ) flY and

Y the families
(V ^|X€A(k))
K w i t h V^.^ r u n n i n g through the open s t a r s S t ( v ) x with
1
v€f (w) , where u* = S t ( w ) have the d e s i r e d p r o p e r t i e s .
k Y q.e.d.
We defined a s p r e a d by a local condition i n the weak sense ( c f . Def. 1
a b o v e ) . As a consequence o f Theorem 13.5, or i t s C o r o l l a r y 13.6, we now
see that semialgebraic s p r e a d s can be characterized by a local condition
i n the strong sense.
Definition 2. A base of a space N i s a s u b s e t X of f ( N ) such t h a t every
U € J'(N) i s the union of a f i n i t e family in X .
Proposition 13.7. L e t f : M -*N be a semialgebraic map and l e t X be a base
of N. Let Con(f,.£) denote the s e t o f a l l connected components of a l l

-1
sets f (B) w i t h B r u n n i n g through Z . The map f i s a s p r e a d i f and only
if Con(f,<£) i s a base o f M.
P r o o f . I f C o n ( f , / ) i s a base o f M then i t i s e v i d e n t directly from De-
finition 1 t h a t f i s a s p r e a d . Assume now that f i s a spread, and let
X£ T(M) be given. We have to v e r i f y t h a t X i s the union of finitely
many open s e t s i n C o n ( f , £ ) . Choosing an open s e m i a l g e b r a i c s e t U z> f(X)

-1
i n N we may replace N by U, M by f (U), f by its restriction from
-1
f (U) to U and X by {W6/|Wcu) . Thus we may asssume w i t h o u t l o s s of
generality t h a t M and N are semialgebraic. Applyina Corollary 13.6
t o the covering {N} o f N and the covering {M,X} o f M, we obtain a
finite family {U.j,...,U } i n ?"(N) and

r f o r e v e r y u\ a (finite) family
1
(V\ j I j G J (i)) of c o n n e c t e d components o f f " (0\) such t h a t X i s the
union of a l l V\j with i € { 1 , . . . , r } , j€J(i). Choosing f o r every U\
a finite family (B..|X€A(i)) i n ^ w i t h u n i o n U. i t i s evident that

B a n c
X i s t h e union o f some o f t h e components o f t h e preimages f ( i>^ ' ^
these a r e f i n i t e l y many elements o f Con(f,.£). q.e.d.
Using t h e method i n t h e p r o o f o f Theorem 13.5 and C o r o l l a r y 13.6 we can
also v e r i f y that a l o c a l isomorphism, as d e f i n e d i n 13.1.iv, i.e. "in
t h e weak sense", i s a c t u a l l y a l o c a l isomorphism " i n the strong sense".
Theorem 13.8. L e t f : M->N be a l o c a l isomorphism between s p a c e s . Then
there exists a locally finite covering (U^|A€A) o f M by open semialge-
braic s e t s such t h a t , f o r every A € A, f yields an isomorphism o f
onto an open s e m i a l g e b r a i c s e t f(U^) i n N. I f f i s s e m i a l g e b r a i c we
can find such a c o v e r i n g (U^|A€A) with (f(U^)|A€A) also l o c a l l y finite.
P r o o f . We may assume t h a t M and N a r e s e m i a l g e b r a i c . Now r u n t h r o u g h
the proof o f C o r o l l a r y 13.6 a g a i n w i t h (MîeD = {M}, ( N j | j € J ) = {N} .
We choose as t h e s e t s Uy t h e open s t a r s S t ( v ) w i t h v € E ( X ) n X. S i n c e
x
f maps a s m a l l neighbourhood o f v i s o m o r p h i c a l l y onto a s m a l l neigh-
bourhood o f w := f ( v ) i t i s e v i d e n t t h a t f maps S t ( v ) x isomorphically
onto S t ( w ) .
v Thus t h e c o v e r i n g ( S t ^ ( v ) Iv € E(X) n X ) o f X has t h e d e s i r e d
properties. q.e.d.
Theorem 13.8 h a s f i r s t been proved i n f u l l g e n e r a l i t y by Roland Huber
at Regensburg. He a c t u a l l y proved a more g e n e r a l theorem by an elemen-
t a r y d i r e c t method, which i s i n d e p e n d e n t l y i n t e r e s t i n g . We s h a l l pre-
s e n t a j a z z e d - u p v e r s i o n o f h i s theorem and p r o o f i n t h e next s e c t i o n
(Huber worked o n l y i n the semialgebraic setting).

§14 - Huber's theorem on open mappings
A locally s e m i a l g e b r a i c map f : M -*N is called open i f the image o f any
open s e m i a l g e b r a i c s e t i s open. T h i s means t h a t f i s open i n the topo-
logical sense w i t h r e s p e c t t o the s t r o n g t o p o l o g i e s on M and N.
Theorem 14.1 (Huber). L e t f : M-> N be an amenable map ( c f . §6) between
spaces. Assume f u r t h e r t h a t f i s open and every point x€M has an open
semialgebraic neighbourhood V such t h a t f | V : V-+N has connected fibres.
Then t h e r e e x i s t s an a d m i s s i b l e c o v e r i n g (M^|i€I) o f M by open locally
semialgebraic s e t s such t h a t , f o r every i € I, the fibres of flMîIVL-^N
are c o n n e c t e d components o f f i b r e s of f . I f the components of a l l f i b r e s
of f are s e m i a l g e b r a i c then (jyLli€I) can be chosen as a l o c a l l y finite
covering with semialgebraic sets M^.
Examples 14.2. The assumptions o f the theorem are f u l f i l l e d i n the
following cases.
a) f i s a s e m i a l g e b r a i c local isomorphism. In t h i s case the theorem
g i v e s our p r e v i o u s r e s u l t Theorem 13.8 anew.
b) M i s an open l o c a l l y s e m i a l g e b r a i c subset o f the p r o d u c t LxN of
two spaces and f i s the r e s t r i c t i o n to M o f the c a n o n i c a l p r o j e c -
tion LxN -> N.
In o r d e r to prove Theorem 14.1 we choose a s u i t a b l e t r i a n g u l a t i o n of N
and then assume t h a t N i s a l o c a l l y finite simplicial complex Y and f
is trivial over each open simplex a of Y.
Let I(M) denote the s e t o f a l l connected components o f a l l i n v e r s e

1
images f ( a ) o f a l l open s i m p l i c e s a € £(Y) . C l e a r l y I(M) is a locally
finite partition of M i n t o l o c a l l y s e m i a l g e b r a i c s e t s . We use the
n o t a t i o n Z(M) s i n c e - as we shall see - I(M) is a "stratification"

o f M ( i n a m i l d sense) s i m i l a r t o the s t r a t i f i c a t i o n Z(Y) o f Y i n t o
open s i m p l i c e s . The p r o o f w i l l be s i m i l a r in spirit t o the c o n s i d e r a -
tions i n §6 about t r i a n g u l a t i o n s o f p a r t i a l l y finite amenable maps,
where such a " s t r a t i f i c a t i o n " was a l s o s t u d i e d .
We a l r e a d y s t a r t t o c a l l t h e s e t s S € Z(M) t h e s t r a t a o f M. S i n c e f i s
trivial o v e r each a £ Z ( Y ) , f maps every s t r a t u m S onto a whole open
simplex a o f Y and, f o r every y G a , the i n t e r s e c t i o n S n f (y) i s a

1
connected component o f f ( y ) . We d i v i d e t h e p r o o f o f Theorem 14.2
into several steps.
First (and most important) s t e p . I f S^ and S 2 are d i f f e r e n t strata
w i t h f ( S ) = f ( S ) = a , then S J H S ^ = 0 .
1 2
P r o o f . Suppose t h e r e e x i s t s a p o i n t x € S^ H S . Then y := f ( x ) i s a
2
p o i n t i n 3a. We choose an open l o c a l l y s e m i a l g e b r a i c neighbourhood V
o f x such t h a t f | V has connected f i b r e s . Then we choose some Uef(M)
w i t h x £ U < z V . F i n a l l y we choose some open s e m i a l g e b r a i c neighbourhood
W o f f ( x ) i n N w i t h W c f (U) and WD a c o n n e c t e d . This i s possible since
f i s open. L e t Z ( M , a ) denote t h e s e t o f a l l s t r a t a S € Z(M) w i t h
f ( S ) = a , and l e t A denote the s u b s e t o f a l l S £ Z ( M , a ) with
UDf (W) n S * 0 . The s e t A i s f i n i t e s i n c e Z(M) i s l o c a l l y finite.
C l e a r l y both S-j and S a r e elements o f A . F o r e v e r y S € A t h e s e t

2
-1
f (UD f (W) n S) i s open s e m i a l g e b r a i c i n W n a and
(*) wn a = u f ( u n f " (w) n s ) .

1
S€A
This i s c l e a r since f i s t r i v i a l over o, Z ( M , a ) i s t h e s e t o f connected
components o f f 1
( a ) , and f i s open. F o r e v e r y y € W n a we have
f~ 1
(y) n U c f " 1
(y) n V c S
w i t h a unique S € A . Thus the r i g h t hand s i d e o f (*) i s a d i s j o i n t
union o f non empty open s e m i a l g e b r a i c s u b s e t s o f WD a , and t h e r e a r e

at l e a s t two such s e t s . T h i s c o n t r a d i c t s the f a c t t h a t WD a i s connected.
We c o n c l u d e t h a t s T n S~I = 0.
2nd S t e p . F o r every o € I (Y) we have f (a) = f (a).

~~-1 -1 — -1 —
Indeed, f (a) c f ( a ) . Given a p o i n t x € f ( a ) and an open n e i g h b o u r -
_ i
hood V o f x we have f ( V ) D a 4= 0 s i n c e f ( V ) i s open. Thus V D f ( a ) * 0,
and we see t h a t x £ f ( a ) . 1
3rd S t e p . L e t a and T be open s i m p l i c e s o f Y w i t h T e a . Then, f o r every

-1 — -1
stratum S c f ( a ) , the i n t e r s e c t i o n S f l f (T) i s a union o f s t r a t a
-1 -1 —
Tcf ( T ) . Every s t r a t u m T c f (T) l i e s i n the c l o s u r e S o f a unique
stratum S c f 1
(a) .
Proof, f (T) c f (a) = f ( a ) , and f ( a ) i s the d i s j o i n t union ( c f .
step 1) o f the c l o s u r e s S o f the connected components S o f f ^ ( a ) . Thus

-1 -1
f (T) i s t h e d i s j o i n t u n i o n o f t h e c l o s e d s e m i a l g e b r a i c s e t s S (1 f ( T )
w i t h S r u n n i n g through the s t r a t a S c f " ( a ) . T h i s g i v e s both r e s u l t s .

1
q.e.d.
1
Remark. S f! f (x) may c o n s i s t of s e v e r a l s t r a t a as t h e f o l l o w i n g
picture shows.
Y
) = 1 C
Here M i s an open subset i n JJR , f i s the p r o j e c t i o n onto the f i r s t
coordinate a x i s , and Y i s an open i n t e r v a l s u b d i v i d e d by one p o i n t .

F o r e v e r y T 6 Z (M) we d e f i n e the s t a r S t ^ (T) as the u n i o n o f a l l S G Z (M)
w i t h S 3 T. S i n c e t h e s t a r S t ( x ) o f T : = f ( T )
Y c o n s i s t s o f f i n i t e l y many
open s i m p l i c e s , i t i s e v i d e n t from the t h i r d s t e p t h a t St^CT) consists
of f i n i t e l y many s t r a t a . In p a r t i c u l a r , St^CT) i s l o c a l l y s e m i a l g e b r a i c
4th S t e p . S t ( T ) i s open f o r e v e r y T G Z(M).

M
P r o o f . I f S i s a s t r a t u m o f M w i t h S £ T then S i s d i s j o i n t from S t ^ ( T ) .
Thus t h e complement o f S t ( T ) i n M i s the union o f t h e f a m i l y

M
( S I S G Z ( M ) , S4>T). The f a m i l y ( S l S G Z ( M ) ) i s locally finite s i n c e Z (M)
is locally f i n i t e . Thus t h i s union i s c l o s e d . q.e.d.
5th S t e p . F o r e v e r y T G Z(M) the f i b r e s o f f | S t (T) a r e connected
components o f the f i b r e s o f f .
P r o o f . L e t T : = f ( T ) . The image f ( S t ( T ) ) i s a subcomplex o f St^Ci:) which

M
c o n t a i n s T . L e t y € f ( S t ( T ) ) and l e t a denote t h e unique open simplex o f

M
S t ( x ) c o n t a i n i n g y . By Step 3 t h e r e e x i s t s a unique
Y stratum S c f (a)
— -1 -1
w i t h S=>T. Thus f (y) D S t ( T ) = f (y) OS, and t h i s
M i s a c o n n e c t e d com
-1
ponent o f f ( y ) .
6th S t e p . The c o v e r i n g ( S t ( T ) | T G Z(M))

M of M i s admissible. Indeed,
e v e r y UG T(M) c a n be c o v e r e d by f i n i t e l y many s t r a t a . A f o r t i o r i , U can
be c o v e r e d by f i n i t e l y many s t a r s .
If t h e connected components o f t h e f i b r e s o f f a r e s e m i a l g e b r a i c , then
the s t r a t a o f M a r e s e m i a l g e b r a i c , hence a l s o t h e i r s t a r s a r e s e m i a l g e -
b r a i c . Now the c l o s u r e S o f a s t r a t u m S i s the u n i o n o f f i n i t e l y inany
strata, and these a r e p r e c i s e l y a l l T G Z (M) such t h a t S f l S t ( T ) * 0.

M
Thus t h e f a m i l y ( S t ( T ) |T £ Z(M))
M i s locally finite.
A l t o g e t h e r we see t h a t t h e theorem h o l d s w i t h the c o v e r i n g
(M liGI)
i := ( S t ( T ) |T€ I(M)) .
M
We ask f o r t h e r e l a t i o n s between our s t r a t i f i c a t i o n o f M and analogous
stratifications o f s u i t a b l e subspaces o f M. T h i s l e a d s t o a refinement
o f Theorem 14.1.
Theorem 14.3. L e t f : M ->N be a l o c a l l y semialgebraic map between spaces.
Assume t h a t (X(A)|AGA) i s a f a m i l y o f l o c a l l y semialgebraic subsets o f
M such t h a t , f o r every HA, the r e s t r i c t i o n f|X(A) : X(A) N fulfills
the assumptions made about f i n Theorem 14.1. Assume f u r t h e r t h a t t h e
family of sets (f(X(A)) IAEA) i s l o c a l l y finite i n N. (N.B. Every s e t
f(X(A)) i s l o c a l l y semialgebraic s i n c e fIX(A) i s amenable.) Then there
exists, f o r every A€A, a l o c a l l y finite covering (X(A,i) Ii€I(A)) o f
X(A) by open l o c a l l y semialgebraic subsets o f X(A), and, f o r any two
indices A,y€A w i t h X(A) c X ( u ) , a map cp , : 1(A) -»I(u) such t h a t t h e

UA
following holds.
a) The f i b r e s o f f | X ( A , i ) a r e c o n n e c t e d components o f f i b r e s o f f | X ( A ) .
b) If X(A)cX(u) then every s e t X(A,i) i s contained i n X(u,cp , (i)) and

VIA
every set X(u,j)nx(A) i s the union o f a l l X ( A , i ) w i t h X(y,cp . ( i ) ) c
pA
e X(y,j) .
c) I f X(A) c x ( p ) c X ( v ) then N be an open l o c a l l y s e m i a l g e b r a i c map between
spaces. Assume t h a t every x € M has a fundamental system o f open semial-
g e b r a i c neighbourhoods V such t h a t f|V has connected f i b r e s . L e t
( Z ^ | K € K ) be any f a m i l y o f open s e m i a l g e b r a i c subsets o f M such t h a t t h e
family ( f ( Z )|K€K) i s locally finite i n N. Then the assumptions o f t h e
theorem h o l d f o r f and the f a m i l y (X(A)|A6A) o f a l l i n t e r s e c t i o n s o f
f i n i t e l y many s e t s Z . Indeed, t h e f a m i l y (f (X (A ))| AG A) i s again locally

f
f i n i t e . Every X(A) i s open and s e m i a l g e b r a i c i n M. By H a r d t s theorem
(Th. 6.3) every r e s t r i c t i o n f l X ( A ) i s amenable.
Proof o f Theorem 1 4 . 3 . For every AEA we choose a locally finite parti-
tion (A, |K€K(A)) of f(X(A)) i n t o s e m i a l g e b r a i c s e t s such t h a t f|X(A)
is t r i v i a l over each A. . The family (A, I A€ A, K E K (A)) i s l o c a l l y finite

AK AK
i n N. We choose a simultaneous t r i a n g u l a t i o n of N and t h i s f a m i l y . Then
we may assume t h a t N i s a l o c a l l y finite simplicial complex Y such that
f o r every AEA and every oEZ(Y), the map f|X(A) is trivial over or. Thus
we have, f o r every AEA, the s t r a t i f i c a t i o n Z(X(A)) of X(A) c o n s t r u c t e d

- -j
above, whose s t r a t a are the connected components o f the s e t s X(A) n f (
We d e f i n e 1(A) : = Z ( X ( A ) ) a n d , f o r every T £ I ( A ) , we d e f i n e X(A,T) : =
S t x ^ (T) . I f X(A)<=X(u), then every s t r a t u m T E I (A) i s c o n t a i n e d i n a

1
unique stratum T' E I ( y ) and we d e f i n e cp , (T) = T . In t h i s s e t t i n g the
yA
c l a i m s a ) , b ) , c) i n the theorem a r e e v i d e n t . q.e.d.
We now s t a t e a consequence o f Theorem 14.1 f o r the t h e o r y o f S t e i n
factorizations.
Theorem 14.5. We c o n s i d e r the S t e i n factorization
o f a p r o p e r map f . The f o l l o w i n g are e q u i v a l e n t .
i) cp i s open and TT i s a l o c a l isomorphism.
ii) f i s open and every p o i n t x E M has an open s e m i a l g e b r a i c neighbour
hood V such t h a t the f i b r e s of f|V are connected.
P r o o f , i ) => i i ) : T h i s i s the t r i v i a l d i r e c t i o n . Of c o u r s e , f = n°cp i s
open. Given a p o i n t x E M we choose an open s e m i a l g e b r a i c neighbourhood

W o f cp(x) which i s mapped i s o m o r p h i c a l l y onto an open semialgebraic
neighbourhood U o f f ( x ) under n. Then V := (p (W) i s an open semialge-
braic neighbourhood o f x, and t h e f i b r e s of f|V are connected components
of fibres of f.
ii) => i ) : By Theorem 14.1 we have a l o c a l l y f i n i t e c o v e r i n g (M^|i€I) o f
M by open s e m i a l g e b r a i c subsets such t h a t , f o r every i € I, the f i b r e s
of f a r e c o n n e c t e d components o f f i b r e s o f f . S i n c e f i s proper this
i m p l i e s t h a t every c o n s i s t s of f u l l f i b r e s o f cp. The image S^ :=

1
cp(NL) i s a s e m i a l g e b r a i c subset o f S, and = cp (S^) . Again u s i n g that
cp i s p r o p e r , we c o n c l u d e t h a t S ^ S i = cp(MîyL) i s c l o s e d i n S, hence S i
i s open i n S. The f a m i l y (S^Jiei) i s a l o c a l l y f i n i t e c o v e r i n g o f S.
Every i s mapped under TT onto the open s e m i a l g e b r a i c s u b s e t :=
f (1VL ) o f N. We c o n s i d e r t h e r e s t r i c t i o n s cp : i -» S^, cp^ : S.^ -» N , and

i
T
f ^ = r °cp :
i i -» of cp, TT , f r e s p e c t i v e l y . The f i b r e s o f cp^ c o i n c i d e
w i t h the f i b r e s o f f . . Thus TT . i s b i j e c t i v e . S i n c e f . i s open, TT . i s

I l J
I ^ I
a l s o open. We see t h a t TT^ i s an isomorphism from S^ onto 1NL and t h a t
cp^ = TU ^°f^ i s open. Thus TT i s a l o c a l isomorphism and cp i s open.
q.e.d.
The f i n i t e maps which a r e l o c a l isomorphisms are j u s t the f i n i t e
c o v e r i n g s t o be s t u d i e d i n Chapter V. Theorem 14.5 i n d i c a t e s a source
of examples o f f i n i t e coverings.
Chapter I I I - Homotopies
§1 - Some s t r o n g d e f o r m a t i o n r e t r a c t s
In t h i s whole c h a p t e r a space means a r e g u l a r paracompact locally semi-
a l g e b r a i c space over some f i x e d r e a l c l o s e d field R. In p a r t i c u l a r , a
s e m i a l g e b r a i c space i s always assumed t o be a f f i n e . A map between

/
spaces M and N i s i m p l i c i t l y assumed t o be l o c a l l y semialgebraic. I f
we do not mean t h i s we c a l l t h e map " s e t t h e o r e t i c " o r "map between
the s e t s M and N".
In t h i s s e c t i o n we r e c a l l some theorems about t h e e x i s t e n c e o f s t r o n g
d e f o r m a t i o n r e t r a c t i o n s which were proved f o r s e m i a l g e b r a i c spaces i n
the papers [DK^] and [DK^]. The p r o o f s t h e r e g e n e r a l i z e word f o r word
to t h e l o c a l l y s e m i a l g e b r a i c s e t t i n g . We s t a r t w i t h some o b v i o u s de-
finitions.
D e f i n i t i o n s 1. a) L e t f and g be two maps from a space M t o a space N.
A homotopy from f t o g i s a map H : Mx[0,1] N such t h a t H = f and
H 1 = g. Here H t denotes t h e map x w H ( x , t ) from M t o N (t € [ 0 , 1 ] ) . As
before the u n i t i n t e r v a l [0,1] i n R w i l l o f t e n be denoted by I .
b) A subspace A o f M (= l o c a l l y s e m i a l g e b r a i c subset o f M) i s c a l l e d
a r e t r a c t o f M i f t h e r e e x i s t s a map r : M -*A w i t h r|A = i d . N o t i c e A
t h a t then A must be c l o s e d i n M. Any such map r i s c a l l e d a r e t r a c t i o n
from M t o A.
c) A subspace A o f M i s c a l l e d a s t r o n g d e f o r m a t i o n r e t r a c t o f M i f
t h e r e e x i s t s a homotopy H : M*I-+M such t h a t H Q i s t h e i d e n t i t y o f M,
H^ i s a r e t r a c t i o n from M t o A and H(a.,t) = a f o r every a £ A, t € I .
We then c a l l H a s t r o n g d e f o r m a t i o n r e t r a c t i o n from M to A.
Theorem 1 .1 ( c f . Theorem 2.1 and 2.7 i n [ DKc- ]) . L e t A be a c l o s e d sub-

space o f a space M. Then t h e r e e x i s t s a l o c a l l y s e m i a l g e b r a i c open
neighbourhood U o f A i n M and a s e m i a l g e b r a i c s t r o n g d e f o r m a t i o n r e -
t r a c t i o n H : U x [ o , 1 ] ->U from t h e c l o s u r e U o f U i n M t o A such t h a t
restricting H to U x [ o , 1 ] y i e l d s a strong deformation r e t r a c t i o n from
U t o A.
Remark. The map H i s o b t a i n e d from t h e r e t r a c t i o n r used i n I I , Prop.
10.8 i n an easy way: Choose a s i m u l t a n e o u s t r i a n g u l a t i o n o f M and A.
Then t a k e f o r U t h e open s t a r o f A i n M w i t h r e s p e c t t o t h e second
b a r y c e n t r i c s u b d i v i s i o n and put
H(x,t) = (1-t)x+ t r ( x ) ,
cf. [DK , 5 §2].
Proposition 1.2 ( E x t e n s i o n o f maps t o a neighbourhood, [DK^, Prop. 4.1]).
A g a i n l e t A be a c l o s e d subspace o f a space M and l e t U be a n e i g h b o u r -
hood of A w i t h the p r o p e r t i e s c l a i m e d i n Theorem 1.1. Any map f : A-»Z

a r e t w o
i n t o some space Z extends t o a map g : U - » Z . I f and exten-
s i o n s o f f t o U, then t h e r e e x i s t s a homotopy G : U x i -* z w i t h G Q = ,
G = g a n d G A
= f f o r e v e r
1 2 t ' Y t € I . (The same i s t r u e w i t h U r e p l a c e d
by U.)
This i s an easy consequence o f Theorem 1.1. Indeed, l e t H : U x l - » U be a
strong deformation r e t r a c t i o n from U t o A. D e f i n e g := f<»H^ and d e f i n e
G:UxI -» Z as f o l l o w s :
Theorem 1.3 ( c f . Th. 5.1 i n [ D K ] ) . I f A i s a c l o s e d

5 subspace o f a space
M then ( A x i ) u (Mx{0}) i s a s t r o n g d e f o r m a t i o n r e t r a c t o f M x i .
The fact that (Axi) u (Mx{o}) i s a r e t r a c t o f M x i means t h a t the p a i r

(M,A) has the "homotopy e x t e n s i o n property".
Corollary 1.4 (Homotopy e x t e n s i o n theorem). L e t A be a c l o s e d subspace
o f a space M . Given a map g : M -+Z i n t o some space Z and a homotopy
F : Axl Z with F Q = g l A t h e r e e x i s t s a homotopy G : M x i Z with G D = g
and GlAxi = F
Mxl
(Axl) U (Mx{0})
FUg
Indeed, we o b t a i n G by composing F U g w i t h a r e t r a c t i o n from M x i t o
(Axl) U (Mx{0}) .
As b e f o r e , a complex means a geometric s i m p l i c i a l complex. In t h e
following we s h a l l o f t e n work w i t h weak t r i a n g u l a t i o n s (cf. I I , §6)
of spaces instead o f t r i a n g u l a t i o n s , i . e . we s h a l l use l o c a l l y finite
instead of s t r i c t l y locally f i n i t e complexes. T h i s i s a minor p o i n t i n
the t h e o r y s i n c e we u s u a l l y have enough t r i a n g u l a t i o n s a t o u r d i s p o s a l ,
but we f e e l t h a t t h e methods become c l e a r e r i f we use weak t r i a n g u l a -
t i o n s . By the way, i n the p r o o f o f Theorem 1.1 i t was a l r e a d y sufficient
t o take a weak s i m u l t a n e o u s t r i a n g u l a t i o n o f M and A.
D e f i n i t i o n s 2. a) A subcomplex Y o f a complex X i s c a l l e d tame i n X i f ,
f o r e v e r y open simplex a o f X , t h e f o l l o w i n g two c o n d i t i o n s are f u l -
filled.
i) I f a c Y then a t l e a s t one v e r t e x o f a i s a p o i n t o f Y.
ii) I f every v e r t e x o f a l i e s i n Y then a c Y (thus a l s o oflXcY).
b) A s i m u l t a n e o u s weak t r i a n g u l a t i o n c p : X - ^ M o f a space M and a sub-
space A o f M ( i . e . cp i s a weak t r i a n g u l a t i o n o f M and cp (A) 1

isa
subcomplex o f X) i s c a l l e d good on A i f t h e complex cp (A) 1

i s tame
in X.
It i s an easy combinatorial e x e r c i s e to v e r i f y
Remark 1.5. F o r e v e r y subcomplex Y o f a complex X t h e b a r y c e n t r i c sub-
division Y 1
i s tame i n X . Thus, i f cp : X
f
M i s a simultaneous trian-
g u l a t i o n o f a space M and a subspace A, then t h e b a r y c e n t r i c s u b d i v i -
s i o n cp : X' -^M
1
o f cp (which as a map between spaces i s t h e same as cp)
i s good on A and on M.
Definitions 3. a) The c o r e o f a complex X i s t h e subcomplex Z o f X
consisting o f a l l open s i m p l i c e s a € I ( X ) w i t h a c X . I n o t h e r words, Z
is t h e unique maximal subcomplex o f X which i s c l o s e d . We denote t h e
c o r e o f X by co X .
b) The c o r e o f a space M w i t h r e s p e c t t o a weak t r i a n g u l a t i o n cp : X M
is the p a r t i a l l y complete subspace cp(coX) o f M. We denote t h i s subset
o f M by co(M,cp) . I f A i s a s u b s e t o f M such t h a t cp (A) i s a subcomplex

1
Y o f X, i . e . cp i s a s i m u l t a n e o u s t r i a n g u l a t i o n o f X and A, then by t h e
c o r e o f A w i t h r e s p e c t t o cp we mean t h e s e t c o ( A , c p l Y ) . We w i l l write
co(A,cp) instead of co(A,cp|Y).
Proposition 1 .6 ( c f . Prop. 2.5 i n [DK^] and i t s p r o o f ) . L e t cp : X M
be a s i m u l t a n e o u s t r i a n g u l a t i o n o f M and A which i s good on A. L e t U
denote t h e image under cp o f t h e open s t a r S t ^ i c o Y) o f the core o f
Y : = cp (A) . N o t i c e t h a t U D A . There
1
exists a semialgebraic strong
deformation r e t r a c t i o n H : U x i - + u from U t o co(A,cp) whose r e s t r i c t i o n
to A x l i s a s t r o n g d e f o r m a t i o n retraction from A t o co(A,cp) .
The map H i s g i v e n by an e x p l i c i t formula [DK^, p. 26] which i s so
simple t h a t we have space here t o w r i t e i t down. We i d e n t i f y X w i t h M
assuming cp = i d - Then H(x,t)

x = (1-t)x + t r ( x ) (x € U, t 6 I) w i t h a r e -
t r a c t i o n r : U -> co Y d e f i n e d as f o l l o w s . I f a i s an open simplex i n U,
a = ] e , . . . , e [ with
Q n Y for0 0, t Q + ... + = 1). We c a l l r the c a n o n i c a l r e t r a c t i o n from
U t o c o Y . I n c i d e n t a l l y , t h e r e t r a c t i o n used i n I I , Prop. 10.8 and i n
Theorem 1.1 i s d e f i n e d by a g e n e r a l i z a t i o n o f t h e formula (*), cf.
[DK , 5 §2].
Corollary 1.7. L e t M be any space and K be a p a r t i a l l y complete sub-
space o f M. Then t h e r e exists a partially complete subspace L z> K i n M
such t h a t L i s a s t r o n g d e f o r m a t i o n r e t r a c t o f M.
Indeed, choose a s i m u l t a n e o u s t r i a n g u l a t i o n cp : X M o f M and K. Then
the core L = c o ( M , c p ' ) of M with respect t o the b a r y c e n t r i c subdivision
cp' o f cp has the d e s i r e d p r o p e r t i e s (apply Remark 1.5 and Prop. 1.6
w i t h A = M) .
For any complex X which i s tame i n X we denote t h e c a n o n i c a l r e t r a c t i o n
from X to c o X by r v . (Caution: r v i s almost never a s i m p l i c i a l map!)

A A
We s t a t e two o b s e r v a t i o n s on c a n o n i c a l r e t r a c t i o n s f o r l a t e r use.
Proposition 1.8. L e t X be a complex which i s tame i n X and l e t Y be a
subcomplex o f X w i t h Y fl co X czY ( f o r example, Y c l o s e d i n X ) . Then Y
is tame i n Y and co Y = Y fl co X . The c a n o n i c a l retraction r y : Y -> co Y
i s the r e s t r i c t i o n of r v : X -* co X t o Y.
We l e a v e t h e easy p r o o f t o the r e a d e r .
Proposition 1.9. A g a i n l e t X be a complex which i s tame i n X and l e t
Y be a subcomplex o f X. Assume t h a t every oEZ(Y) has a t l e a s t one v e r -
tex c o n t a i n e d i n Y, and f u r t h e r t h a t Y n c o X i s open i n c o X ( f o r
example, Y open i n X ) . Then Y i s tame i n X and t h e r e s t r i c t i o n o f r x to

Y is a retraction from Y t o Y n co X , which i s l i n e a r l y homotopic t o i d v
P r o o f . L e t a = ]e ....,e [ be an open simplex o f X a l l whose v e r t i c e s

o n
e^ a r e p o i n t s i n Y . S i n c e X i s tame i n X we know t h a t o c c o X . S i n c e
e Q E Y n co X and Y n co X i s open i n co X we conclude t h a t o c Y 0 co X
and a f o r t i o r i G C Y . Thus Y i s tame i n X.
Now l e t a = ] e , . . . e [ be an open simplex

Q f n i n Y. We may assume t h a t
e Q € Y, e i € X f o r 0 < i < m, and e^ £ X f o r m < i < n, f o r some m € {0, . . . ,n} .
Then r ( a ) = ] e
x Q / ... e [. f m T h i s i s an open simplex o f c o X contained i n
St ^ (e ) and e_ i s a v e r t e x o f Y fl co X. S i n c e Y n co X i s open i n co X
v
co X o o *
we c o n c l u d e t h a t r ( c O c Y fl co X . The r e m a i n i n g a s s e r t i o n s o f the p r o -
x
p o s i t i o n a r e now obvious. q.e.d.

§2 - S i m p l i c i a l approximations
We consider (r+1)-tupels (M,M,j,...,M ) c o n s i s t i n g o f a space M and sub-
spaces M 1 , . . . ,M of M (r>0). Such a t u p e l i s c a l l e d a system o f spaces
If M i s a locally finite simplicial complex X and NL i s a subcomplex X i
of X (1< i < r ) , then i t i s c a l l e d a system o f complexes.
A map
f : (M,M 1 , . . . ,M ) r -+ (N,N 1 , . . . N )
r r
between such systems o f spaces i s , o f c o u r s e , a ( l o c a l l y semialgebraic)
map f : M-> N w i t h f (M^.) c N . i Such a map f is called an isomorphism i f
f : M -»N i s an isomorphism o f spaces and f (M^) = ISL f o r i = 1 , . . . , r .
Then f induces isomorphisms M^-^>1SL.
A homotopy between two maps f , g from (M,M^,...,M ) t o r (N,N^,...,) i s
a map
H. : ( M x i , M ^ I , . . . ,M xI) - r (N,, . . . , )
with H q = f and = g. I f such a homotopy e x i s t s we call f and g homo-
t o p i c and w r i t e f ^ g . S i m i l a r l y we t r a n s f e r a l l the u s u a l t e r m i n o l o g y
from t o p o l o g i c a l homotopy t h e o r y t o the c a t e g o r y o f spaces over R.
We denote the homotopy c l a s s o f a map f from (M,M^,...,M ) to r
( N , , . . . , N ) by r [ f ] and the s e t o f a l l t h e s e homotopy c l a s s e s by
[(M,M ,...,M ),
1 r (N,N^,...,N )]. The g o a l o f the next t h r e e s e c t i o n s
r
will be t o o b t a i n some i n s i g h t i n t o the g e n e r a l n a t u r e o f these homo-
topy s e t s . (We s h a l l prove two "main theorems"). A major t o o l f o r t h i s
purpose a r e s i m p l i c i a l a p p r o x i m a t i o n s , to be e x p l a i n e d now. We have t o
be a l i t t l e more c a r e f u l than i n the t o p o l o g i c a l t h e o r y s i n c e our sim-
plicial complexes a r e not n e c e s s a r i l y closed.

By a weak t r i a n g u l a t i o n ( r e s p . t r i a n g u l a t i o n ) o f t h e system
(M,M-j , . . . ,M ) we mean a s i m u l t a n e o u s weak t r i a n g u l a t i o n

r (resp. trian-
g u l a t i o n ) o f M,,...,M . T h i s i s n o t h i n g more than an isomorphism
cp : (X,X , . . . X )
1 f r (M,M1 , . . . ,M ) r
with X a l o c a l l y finite (resp. s t r i c t l y locally finite) complex and
X^,...,X r subcomplexes o f X.
Definitions 1. a) A system o f complexes (X,X^,...,X ) i s c a l l e d

r tame
if X i s tame i n X ( c f . §1) and e v e r y X i i s tame i n X\. In p a r t i c u l a r
(r = 0) we c a l l a complex X tame i f X i s tame i n X.
b) A weak t r i a n g u l a t i o n cp : (X, , . . . ,X ) r (M,M , . . .,M ) o f a system

1 r
o f spaces i s c a l l e d good, i f (X,X^,...,X ) i s tame. r
N o t i c e t h a t a weak t r i a n g u l a t i o n cp : (X,X^ , . . . ,X^) (M,M^ , . . . ,M ) i s r
good i f and o n l y i f a l l t h e t r i a n g u l a t i o n s c p : X-^»M, cp[X : ^ i

x
(1 < i < r) are good.
By Remark 1.5 t h e f i r s t b a r y c e n t r i c s u b d i v i s i o n cp : (X ,X^

1 ! 1
X )
r
(M,M^,...,M ) o f any weak t r i a n g u l a t i o n cp : (X,X.j,...,X

r ) — * (M,M
i s good. A l s o , i f t h e subspaces are c l o s e d i n M, then a weak t r i a n -
g u l a t i o n cp : (X,X-j , . . . ,X ) -—> (M,M^ , . . . ,M ) i s good

r r i f and o n l y i f t h e
weak t r i a n g u l a t i o n cp : X — • M i s good ( c f . Prop. 1.8).
Definition 2. The c o r e o f a system o f complexes (X,X^,...,X ) i s t h e r
system (co X ,X^ D co X ,...,X n co X) . r I t w i l l be denoted by
co(X,X^,...,X ). r The i n c l u s i o n map from co(X,X^,...,X ) t o r
(X,X ,...,X^) w i l l be denoted by j

n v {or more p r e c i s e l y by j ,
I J- A Â,A^j,...
U s u a l l y o n l y the c o r e s o f tame systems o f complexes w i l l be c o n s i d e r e d
s i n c e o t h e r w i s e co(X,X^,...,X ) may have l i t t l e

r i n common w i t h
(X,X^,...,X ). We o f t e n w i l l
r need an even s t r o n g e r c o n d i t i o n than tame
ness which guarantees t h a t j x i s a homotopy e q u i v a l e n c e , so t h a t we can
replace ( X , , . . . , X ) by i t s c o r e i n homotopy r considerations.
D e f i n i t i o n 3. A system (X,X ,...,X ) o f complexes i s c a l l e d w e l l

1 r cored
if t h e system i s tame and i f t h e c a n o n i c a l r e t r a c t i o n r x : X-»coX (cf.
§1) maps every X^ i n t o i t s e l f , hence onto X. fl co X . We a l s o denote t h e
induced map from (X,X^,...,X ) t o co(X,X^,...,X^) by r r x (or more precise-
l y by r / v
v v v) and c a l l i t the c a n o n i c a l r e t r a c t i o n o f t h e system
A
( A , A ^ , . . . / R ^
(X,X ,...,x ).
1 r
If (X,X^,...,X ) i s w e l l c o r e d then t h e c a n o n i c a l
r retraction
r : (X,X ,...,X ) -+ c o ( X , X i , . . . f X ) i s a homotopy i n v e r s e o f t h e i n -

1
v
A i r 1
r
e l u s i o n map j ^ . More p r e c i s e l y , r ^ j ^ i s t h e i d e n t i t y o f co(X,X^,...,X ) r
and J ° x
r
x
i s
l i n e a r
l y homotopic t o t h e i d e n t i t y o f (X,X^,...,X ) .
By P r o p o s i t i o n s 1.8 and 1.9 a tame system (X,X^,...,X ) i s w e l l r cored
if, f o r every i € {1,...,r}, e i t h e r X\ n co X c X i o r X^ fl co X i s open i n
coX .
P r o p o s i t i o n 2.1. L e t (X,X-j , . . . ,X ) be a tame system o f complexes. Assume r
t h a t every subcomplex X^ I s l o c a l l y c l o s e d i n X (i.e. X^ i s open i n
X. n x o r , what means t h e same, X. i s c l o s e d i n S t ( X . ) ) . Then v
1 1 A 1
(X,X-j , . . . ,X ) r i s well-cored.
r z c z
Indeed, we know t h a t x ( ) for every subcomplex Z o f X which i s
c l o s e d o r open i n X. The same h o l d s f o r an i n t e r s e c t i o n Z = Z^ n Z o f 2
subcomplexes w i t h Z^ open and Z 2 closed i n X.
D e f i n i t i o n 4. a) L e t f : (M,M^,...,M ) -» (N,N^,.../N ) be a map between r r
systems o f spaces. A s i m p l i c i a l approximation to f i s a t r i p l e (cp,<jj,g)
c o n s i s t i n g o f good weak t r i a n g u l a t i o n s cp : (X,X-j , . . . ,X ) r (M,M^ , . . . , M^) ,

: (Y,Y 1 , . . . ,Y ) r (N,N , . . . ,N ) , and a s i m p l i c i a l
1 r map
g : c o ( X , X , . . . ,X )
1 r co(Y Yf r . . . ,Y ) r
such t h a t , f o r every v e r t e x e o f co(X),
1
i|T «f«<p(St (e)) c : S t ( g ( e ) ) .
x Y
b) I f f : (X,X^ , . . . ,X ) -> (Y,Y^ , . . . ,Y ) i s a map between tame systems

r
of locally finite complexes, then by a s i m p l i c i a l approximation to f
we u s u a l l y mean j u s t a s i m p l i c i a l map g from co (X X«j , . . . ,X ) to f
co (Y,Y^ , . . . ,Y ) such, t h a t t h e t r i p l e
r (cp,ijj,g), w i t h cp and \p t h e i d e n t i t y
maps o f (X,X^,...,X ) and ( Y , Y ^ , . . . , Y ) ,

r r i s a simplicial approximation
to f i n t h e sense above.
Notice that a s i m p l i c i a l map g : co (X,X<j , . . . ,X ) -+ co (Y, Y^ , . . . Y ) i s a f
simplicial approximation t o f : (X,X^,...,X ) -* (Y,Y^,...,Y ) i f and

r
o n l y i f g : co X -> c o Y i s a simplicial approximation t o f : X->Y. In
this c a s e g maps co X. i n t o co Y. and t h e r e s t r i c t i o n g l c o X . : co X. -*
co Y i i s a simplicial approximation to f I X i : X^ -> Y^ (1 co (Z , Z^ , . . . ,

r ) i s a simplicial approximation
to a second map h : (Y,Y^,...,Y ) -* (Z,Z^,...,% ) between tame systems r r
of l o c a l l y f i n i t e complexes, then k»g i s a s i m p l i c i a l a p p r o x i m a t i o n t o h°f.
Lemma 2.2. L e t f : X-* Y be a map between tame l o c a l l y f i n i t e complexes,
and l e t g : co X -» co Y be a s i m p l i c i a l . map. Then g i s a simplicial
approximation t o f i f and o n l y i f , f o r every x € X, t h e f o l l o w i n g con-
dition holds:
(*) f ( x ) € p € Z ( Y ) => g - r ( x ) € p . x
(Recall that r x i s the canonical r e t r a c t i o n from X t o co X) .
P r o o f . L e t x € X be g i v e n . L e t a := ] e , . . . , e [ be t h e open simplex o f
Q n
X c o n t a i n i n g x, and l e t p be the open simplex o f Y c o n t a i n i n g f ( x ) . We
assume t h a t e, e x f o r 0 0, r e c a l l that

X i s tame). Then r ( a ) i s the simplex v ]e . ...,e [ i n co X , and g « r ( o ) Y
A o m A
is the open simplex T spanned by t h e v e r t i c e s g ( e ) , . . . , g ( e ) . On t h e Q m
other hand, f ( x ) i s c o n t a i n e d i n the i n t e r s e c t i o n o f the s e t s
f ( S t ( e ) ) with
x i 0<i<m.
Assume t h a t g i s a s i m p l i c i a l a p p r o x i m a t i o n t o f . Then f ( x ) i s c o n t a i n e d
i n t h e i n t e r s e c t i o n o f the s e t s S t ^ C g ^ ^ ) ) w i t h 0 < i < m , which i s t h e
star S t ( x ) Y o f T i n Y. Thus we a l s o have p c S t ( T ) , i . e . T c " p .Y Since
g«r (x) G T , the c o n d i t i o n

x (*) is verified.
Assume now that condition (*) h o l d s f o r e v e r y x € X and assume t h a t
our point x l i e s in St (e) x f o r some e € X H E ( X ) . Then e i s one o f t h e
v e r t i c e s e , . . . , e . We
o m c o n c l u d e from (*) t h a t T i s a f a c e o f p , which
implies t h a t g(e) i s a v e r t e x of p , i . e . p c S t ( g ( e ) ) . Y Thus
f(St (e)) x cSt (g(e)).

Y q.e.d.
Proposition 2 . 3 . L e t f : (X,X^ , . . . ,X ) r -* (Y,Y>j , . . . ,Y ) be a map from a
well cored system t o a tame system o f l o c a l l y finite complexes. L e t
g : co (X,X^ , . . . ,X ) r co (Y,Y,| , . . . ,Y ) r be a s i m p l i c i a l approximation to
f . Then the maps f and J 3° x Y

0< r
f r o m ( X
' X
1 '• • * ' x
r * t o Y Y
( > i /* ** ' r ^ Y
are l i n e a r l y nomotopic.
P r o o f . We put X Q := X, Y := Y. L e t a p o i n t x € X^ be g i v e n f o r some
k€ {0,...,r}. L e t p be t h e open simplex o f Y^ c o n t a i n i n g f ( x ) . By
condition (*) i n Lemma 2 . 2 , the h a l f open l i n e segment [f(x),g«r (x)[ x
i s contained i n p , hence i n Y^. S i n c e r x maps X^ t o X^D c o X , t h e
point g«r (x) l i e s

x i n Y ^ n coY c Y
k . Thus the c l o s e d line segment
[f (x) , g » r (x) ] i s c o n t a i n e d
x i n Y^, more p r e c i s e l y , i n Y^ fl "p. This
implies the p r o p o s i t i o n . q.e.d.
This p r o p o s i t i o n gives t h e r e a s o n why simplicial approximations are

i n t e r e s t i n g t o us i n hoiuotopy t h e o r y . Under f a v o u r a b l e c o n d i t i o n s i t
tells us t h a t , up t o homotopy, we may r e p l a c e a map between t r i a n g u l a -
ted systems o f spaces by the c o m p o s i t i o n o f a c a n o n i c a l r e t r a c t i o n , a
s i m p l i c i a l a p p r o x i m a t i o n and an i n c l u s i o n . T h i s map i s a much s i m p l e r
map than the o r i g i n a l one.
But we s t i l l have t o show t h a t s i m p l i c i a l a p p r o x i m a t i o n s e x i s t under
fairly g e n e r a l assumptions. We f i r s t s t a t e another o b v i o u s consequence
of Lemma 2.2 which g i v e s us a h i n t as t o how t o c o n s t r u c t s i m p l i c i a l
approximations.
Remark 2.4. L e t g : co (X,X^ , . . . ,X ) co(Y,Y^,...,Y ) be a s i m p l i c i a l

r
approximation t o a map f : (X,X^,...,X ) -> (Y,Y^,...,Y^) between tame

r
systems o f complexes. F o r every v e r t e x e € E ( X ) (IX the p o i n t g(e) i s
a v e r t e x o f the open simplex p o f Y which c o n t a i n s f ( e ) . In p a r t i c u l a r ,
if the r e s t r i c t i o n f I X i :X i -> Y^ i s s i m p l i c i a l f o r some i € { l , . . . , r l ,
then the maps f I co X^ and g l c o X ^ from c o X . t o coY. are e q u a l .
Theorem 2.5 ( E x i s t e n c e o f s i m p l i c i a l a p p r o x i m a t i o n s ) . L e t
f : (M,M^,...,M ) -» (N,N^,...,N ) be a map between systems o f s p a c e s .

r r
For every index i e { 1 , . . . , r } assume t h a t e i t h e r 1YL i s c l o s e d i n M o r
N ± i s l o c a l l y closed i n N. L e t cp : (X,X , . . . ,X ) ^ 1 r (M,M , . . . ,M ) and

1 r
: (Y,Y ,...,Y ) ^
1 r (N,N ,...,N ) be weak t r i a n g u l a t i o n s o f (M,M ,...,M
1 r 1 r
and (N,N ,...,N ) . Assume t h a t

1 (X,X ,...,X ) i s tame and t h a t t h e
1 r
system (Y,Y^,...,Y ) i s t h e f i r s t b a r y c e n t r i c s u b d i v i s i o n o f another
system o f complexes. Assume f i n a l l y t h a t ij; .tp maps every open sim-
plex o f X i n t o an open simplex o f Y. Then t h e r e e x i s t s a s i m p l i c i a l
map
g : co(X,X ,...,X ) ^ co(Y,Y ,...,Y )

1 r 1 r
such t h a t (cp,i|;,g) i s a s i m p l i c i a l a p p r o x i m a t i o n to f .
P r o o f . We assume w i t h o u t l o s s of g e n e r a l i t y t h a t (M,M^ ,...,M^) =
(X,X r . . . ,X ) , (N N
r / 1 , . . . ,N ) r = (Y,Y 1 , . . . ,Y ) r and t h a t <p,i|; a r e the
i d e n t i t y maps. We put X q := X, Y Q := Y. For any t £ L ( Y ) we denote
the i n t e r s e c t i o n of a l l Y ^, 0<i<r, with TCY 1 by Y ( T ) . Since
(Y,Y-j , . . . ,Y ) i s the b a r y c e n t r i c s u b d i v i s i o n o f another system o f com-
p l e x e s the complex Y(x) i s tame i n Y ( T ) .
Recall that, f o r any complex Z, we denote the set of v e r t i c e s of Z by
E(Z) and the a b s t r a c t i o n of Z by K(Z) = (E(Z),S(Z)) (II, § 3 ) . We set
out to d e f i n e an abstract s i m p l i c i a l map y : K(coX) K(coY). For
any a £ E(X) the support s u p p f ( a ) of f(o) i s defined as the unique
simplex T € Z(Y) which c o n t a i n s f(a) (by the last assumption i n the
theorem) . For every v e r t e x eofcoX, i . e . e € X f l K (X) , we choose a
vertex y (e) o f p := suppf(e) which l i e s in Y(p). This i s possible
s i n c e Y(.p) i s tame i n Y ( p ) . In t h i s way we obtain a map
y : E(coX) -» E(co Y) .
We claim t h a t p i s a s i m p l i c i a l map from the abstract complex K ( c o X )
to K ( c o Y ) . Let {e ,...,e } be the set of v e r t i c e s of some open sim-
plex c <z co X . We want to v e r i f y t h a t u ( e ) , . . . , y (e^) Q are the vertices
(possibly with r e p e t i t i o n s ) of some T E I ( Y ) . Then T C C O ( Y ) and we are
done.
Let p denote the s u p p o r t of f (a). We have f ( e ^ € f ( a ) c p f o r e v e r y
i = 0,...,n. Thus the support p^ o f f(e^) i s a face of p and y (e^)
i s a vertex of p i n Y. The points y ( e ) , . . . , y (e )

Q n span a simplex T of
Y. Since Y i s tame i n Y t h i s simplex T lies i n Y and our claim i s
proved.
If e i s a vertex of co (X^.) f o r some k £ { 1 , . . . , r } then f ( e ) £ Y^ and
hence y ( e ) £ E ( Y ) n Y. V = E(coY, ) by d e f i n i t i o n of y. This implies that

y maps t h e subcomplex K(co X^) of K(coX) i n t o K(co Y^) .
The r e a l i z a t i o n g = l y l o f y i s a s i m p l i c i a l map from c o X to coY
which maps co X^ i n t o co Y^ f o r 1 < k < r . We c l a i m t h a t g : co X -+coY
i s a s i m p l i c i a l approximation t o f : X-»Y. We have to v e r i f y t h a t , f o r
a g i v e n v e r t e x e G X n E ( X ) , the s e t f ( S t ( e ) ) x i s contained i n S t ( g ( e ) ) . y
L e t a be an open simplex i n S t ( e ) , and

v l e t p denote the support of
f(a). S i n c e e i s a v e r t e x o f a we have f ( e ) G f ( a ) <=. "p. Thus suppf(e)
i s a f a c e o f p and g(e) = y (e) i s a v e r t e x of p . T h i s means pcSty(g(e))
and implies that f(a) c s t y (g(e)).
We f i n a l l y have t o v e r i f y , f o r a g i v e n index k £ {1,...,r}, t h a t g maps
X^ n co X into Y k n co Y . I f X^ Is c l o s e d i n X then X^ n co X = co X^
( c f . Prop. 1.8), and we a l r e a d y know t h a t g ( c o X ^ ) i s contained i n
co Y^ c Y ^ n c o Y . I f X^ i s not c l o s e d i n X then, by assumption, Y^ n co Y
i s l o c a l l y closed i n coY. T h i s means t h a t Y^ fl co Y = Z

k n U
k w i t h
Z k = Y, n co Y and U v = St (Y n co Y) = S t __(co Y, ) . S i n c e g i s sim-

J>- Jv K CO Y K CO x K
p l i c i a l and maps co X, i n t o co Y, we have g(X, fl coX) e g (St v (co X, ))c
rL K JC CO X iC
S t c o Y = U I t
co Y ^ k^ k* remains to v e r i f y t h a t g ( X f l c o X ) c Z ^ . k Since
f i s continuous, f ( X ^ f l coX) c Z^.. In p a r t i c u l a r , f o r every open simplex

a = ]eo . . . .n, e f i K.
n X, fl co X , the p o i n t _L
f (e. ) l i e i n Z . k Thus supp f1 (e . ) c
Z^., which i m p l i e s t h a t g(e^) = y (e^) E Z^.. S i n c e Z^. i s tame i n co Y the
simplex ]g ( e ) , . . . , g ( e ) [ = g(a)
Q R i s c o n t a i n e d i n Z^.. Thus indeed
g ( X ^ f l coX) c Y^ fl c o Y , and Theorem 2 . 5 i s proved.
N o t i c e t h a t , i n view o f Remark 2.4, our procedure f o r c o n s t r u c t i n g the
s i m p l i c i a l approximation g was the most g e n e r a l one. I t seems d i f f i c u l t
to imagine a n a t u r a l h y p o t h e s i s more g e n e r a l than the l a s t one i n the
theorem under which t h i s c o n s t r u c t i o n works. We d i d not use the assump-
tion that (X, X-j , . . . ,X ) r i s tame i n the p r o o f . But s i n c e good t r i a n g u l a -
t i o n s e x i s t i n abundance and tameness of the system (X,X..,...,X ) i s

essential fora simplicial approximation t o be nomotopic t o t h e o r i g i n a l
map (Prop. 2.3), i t seems t o be r e a s o n a b l e t o demand tameness i n t h e
definition of s i m p l i c i a l approximations, as we d i d .
We now ask t o what e x t e n t a s i m p l i c i a l approximation o f a g i v e n map
between systems o f l o c a l l y f i n i t e complexes i s determined by f .
1 ( X X X
Definition 5 ( c f . [Spa, p. 130])**. L e t g^,<3 2 o' 1 '*•'' r * ^
(Y ,,.../Y
Q ) be two s i m p l i c i a l maps between systems o f complexes.
We say t h a t g^ and g 2 are s t r i c t l y c o n t i g u o u s i f , f o r every oEZ(X^),
t h e v e r t i c e s o f g^(a) and g ( c ) 0 span an open simplex o f Y, (0<i<r).
We s a y t h a t g^ and g 2 a r e contiguous i f there e x i s t s a f i n i t e sequence
h Q = g ,h ,h ,...,h
1 1 2 t = g 2
of simplicial maps from (X ,...,X ) t o ( Y , . . . , Y ) w i t h hu

Q r Q r strictly
c o n t i g u o u s to h.^ ^ f o r i = 0 , . . . , t - 1 .
+ (Contiguity i s the equivalence
relation g e n e r a t e d by s t r i c t c o n t i g u i t y . In t h i s section only strict
contiguity w i l l play a role.)
Notice that s t r i c t l y c o n t i g u o u s maps a r e l i n e a r l y homotopic. Thus con-
t i g u o u s maps a r e homotopic.
Proposition 2.6. Any two s i m p l i c i a l approximations
g ,g1 2 : co(X ,...,X ) ^ co(Y ,...,Y )

Q r Q r
of a g i v e n map f : (X ,...,X ) Q r ( Y , . . . , Y ) between tame l o c a l l y

Q r finite
complexes a r e s t r i c t l y c o n t i g u o u s , p r o v i d e d one o f t h e f o l l o w i n g two
conditions holds:
i) F o r every index k 6 ( 1 , . . . , r ) e i t h e r X^ i s c l o s e d i n X o r Y^ i s Q
locally closed i n Y .
*) S p a n i e r uses " c o n t i g u o u s " f o r our " s t r i c t l y c o n t i g u o u s " and has no

word f o r o u r " c o n t i g u o u s " .
ii) (Y ,...,Y ) i s the b a r y c e n t r i c
Q r s u b d i v i s i o n o f another system o f
complexes.
P r o o f . We p u t X := X , Y = Y . L e t a be an open simplex i n X n c o X f o r
Q k
some k € { 0 , 1 , . . . , k } . We choose a p o i n t x i n a and denote t h e open sim-
plex of Y k containing f ( x ) by p . By Lemma 2 . 2 t h e p o i n t s g^(x) and
g (x)
2 a r e both contained i n "p. This implies t h a t g^ ( a ) U g (o) <=• p fl Y^ n 2
coY. L e t x denote t h e open simplex o f Y spanned by t h e v e r t i c e s o f
g^ ( a ) and g^ ( a ) . Of c o u r s e , TCCOY and g^ ( a ) < T ^
then w i l l be done. I f Y^ I s l o c a l l y closed i n Y , then T C Y ^ since
g (a)czY
1 k and p c Y . k I f X^ i s c l o s e d i n X, then X fl co X k = co X^ ( c f .
Prop. 1 . 8 , b u t t h i s I s t r i v i a l ) . Since a c co X^ and t h e g i are s i m p l i -

c Y F i n a l l i f
c i a l we have g^ (a) U g ( a ) c e o Y^ 2 and then T C C O Y ^ k* Y
(Y ,...,Y ) = ( Z
Q r Q Z R ' ) f o r a system o f complexes (Z ,...,Z ) R
then we have open s i m p l i c e s S < ... < Q of Z q whose b a r y c e n t e r s S i
run t h r o u g h t h e s e t E ( g ^ ( a ) ) U E ( g ( a ) ) o f v e r t i c e s o f g (a) and g ( a ) .

2 1 2
Assume, w i t h o u t l o s s o f g e n e r a l i t y t h a t § t i s a vertex o f g^ ( a ) . Then
g^ (a) cr S fc and g^ (a) ^Y^., hence S^cY^.. T h i s implies that
T = ] S , . . . , S [ c Y^.
Q t Thus i c Y ^ d c o Y i n both c a s e s . q.e.d.
We now have e s t a b l i s h e d a rather s a t i s f a c t o r y theory of s i m p l i c i a l
a p p r o x i m a t i o n s from t h e view p o i n t o f homotopy t h e o r y . L e t
f : (M,M^,...,M ) -» (N,N^,...,N ) be a map between systems o f spaces

r r
and assume t h a t , f o r e v e r y index i E {1,...,r}, e i t h e r i s closed i n
M or 1SL i s l o c a l l y closed i n N. Then Theorem 2 . 5 p r o v i d e s us w i t h many
simplicial a p p r o x i m a t i o n s t o f . Indeed, choose some t r i a n g u l a t i o n
: (Y,Y^,...,? ) r (N,N^,...,N ), which i s p o s s i b l e by t h e t r i a n g u l a -

r
t i o n theorem ( I I , Th. 4.4). L e t : (Y,Y 1 , . . . ,Y ) -+ (N,*^ , . . . ,N ) be

r
~ —1
the f i r s t barycentric s u b d i v i s i o n o f \p. The f a m i l y (f ijj (T) I T € I (Y) )
is locally finite i n M. Thus, a g a i n by t h e t r i a n g u l a t i o n theorem, there

exists a simultaneous t r i a n g u l a t i o n co : ( X , , . . . , ) (M,M.j, . . . ,M )
r
o f M,M^ ... M
r f and t h i s f a m i l y . R e p l a c i n g co by i t s b a r y c e n t r i c sub-
division, i f n e c e s s a r y , we may assume t h a t (X,X^,...,X ) i s tame.

r
A c c o r d i n g t o Theorem 2.5 t h e r e e x i s t s a s i m p l i c i a l map
g : co(X,X r .. . ,X )
r - co (Y, Y , . . . , Y )
1 r
such t h a t (cp,i|;,g) i s a s i m p l i c i a l a p p r o x i m a t i o n t o f and, as the p r o o f
of the theorem shows, g can be found by a c a n o n i c a l procedure. By Pro-
position 2.6 any o t h e r s i m p l i c i a l map rj^ o f t h i s k i n d i s s t r i c t l y
c o n t i g u o u s t o g. F i n a l l y , i f every NL i s l o c a l l y c l o s e d i n M, then,
1
by P r o p o s i t i o n s 2.1 and 2.3, the map ^*j °g°r
v »cp i s homotopic t o f .
§ 3 - The f i r s t main theorem on homotopy s e t s ; mapping s p a c e s .
We consider two systems o f s p a c e s (M, A^ , . . . , A ) , (N,B.j , .. . ,B ) r over R
and a r e a l closed field SDR. E v e r y map (= l o c a l l y semialgebraic map,
cf. §1) f : (M, A-j , . . . ,A ) (N ,B ^ , . . . ,B ) r over R y i e l d s by base e x t e n -
sion ( c f . I , 2.10 and I, 4.9) a map
f s : (M(S) ,A 1 (S) , . . . ,A (S) ) r (N (S) , B 1 (S) , . . . ,B (S) ) r
between systems o f spaces o v e r S. I f g i s a second map from
(M,A , . ..,A ) t o
1 r (N,B , ...,B
1 ) and H : ( M x I , A x l , . . . , A x I )
1 r - (N,B ,... ,B
1
is a homotopy from f t o g t h e n
H s : ( M ( S ) x I ( S ) , . . . , A ( S ) xI(S)) -» (N (S) ,B ^ (S) , . . . ,B (S) )

r r
i s a homotopy from f ^ t o g . g (Notice t h a t I(S) i s j u s t the u n i t inter-
val i n S.) Thus we have a c a n o n i c a l map K : [ f ] »-* Cfgl from t h e homo-
topy s e t ( c f . b e g i n n i n g o f §2) [(M,A^,...,A ),(N,B ,...,B )] t o the

1 r
homotopy s e t [ (M(S),A ( S ) , . . . , A ( S ) ) , (N(S),B ( S ) , . . . , B ( S ) ) ] .

1 r 1 r The
latter s e t w i l l be b r i e f l y denoted by [ (M, A^ , . . . ,A ) , (N,B , . . . ,B r 1 ) ] (S) .
The purpose o f t h i s section i s t o prove the f o l l o w i n g theorem.
Theorem 3 . 1 ( F i r s t main theorem). Assume t h a t the space M i s s e m i a l g e -
braic .
a) The canonical map K : [ (M,A^ , . . . ,A ) , (N,B^ , . . . ,B ) ]

r r
[ ( M , A , . . . , A ) , ( N , B , . . . , B ) ] (S) i s always
1 r 1 r injective.
b) Assume t h a t every A^ is locally closed i n M . Assume f u r t h e r that,
for every i £ { 1 , . . . , r } , either A i i s c l o s e d i n M o r B^ is locally
closed i n N. Then K i s a l s o s u r j e c t i v e .
In o r d e r t o s i m p l i f y the n o t a t i o n we assume from now on t h a t r = 1.
The reader w i l l see e a s i l y t h a t a l l arguments remain v a l i d f o r any
r > 0. We w r i t e A,B instead of A ,B . 1 1

In both p a r t s a) and b) o f t h e theorem we may assume w i t h o u t loss of
g e n e r a l i t y t h a t the space N i s a l s o s e m i a l g e b r a i c . Indeed, t o prove
s u r j e c t i v i t y , we have t o c o n s i d e r a map f : (M (S) , A (S) ) -* (N(S),B(S))
and have t o f i n d a map g : (M,A) (N,B) such t h a t f c* g . The image
f(M(S)) i s a semialgebraic subset o f N(S). Now, every semialgebraic
subset o f N(S) i s c o n t a i n e d i n the base e x t e n s i o n D(S) o f some (closed)
semialgebraic subset D o f N. Choosing such a s e t D w i t h f(M(S)) c D(S)
we may c o n s i d e r f as a map from (M(S),A(S)) t o (D (S) , (B fl D)(S)) . I t s u f f i -
ces t o prove t h e c l a i m f o r t h i s new map. I n o r d e r t o prove injectivity
we have t o c o n s i d e r two maps f , g from (M,A) t o (N,B) and a homotopy
H from f s t o g , and we have t o f i n d a homotopy H from f t o g. Now H

g
takes v a l u e s i n D(S) f o r some s e m i a l g e b r a i c s u b s e t D o f N and a g a i n
we may r e p l a c e (N,B) by (D,B fl D) .
n m
We choose f i x e d embeddings o f M and N i n t o s t a n d a r d spaces R ,R . Now
n m
M i s a semialgebraic s u b s e t o f R , N i s a s e m i a l g e b r a i c subset o f R ,
n m
and M(S), N(S) are s e m i a l g e b r a i c subsets o f S , S , d e f i n e d by the same
polynomial inequalities (read over S) as M and N. The p r o o f w i l l be
based on the f o l l o w i n g study o f " s e m i a l g e b r a i c mapping spaces".
We f i x n a t u r a l numbers d , r , s . L e t G. and F.., 1 < i < r , 1 < j <sbe

I i j — — — —
c o p i e s o f the v e c t o r space P(n,m,d) o f p o l y n o m i a l s over R o f t o t a l de-
gree <d i n t h e v a r i a b l e s X.j , . . . ,X , Y^ ...,Y f . We i n t r o d u c e the R-
v e c t o r space
r s
F := F(n,m,d,r,s) := T T (G. x T T F. .)
1 1 J
i=1 j=l
and the f o l l o w i n g semialgebraic subset L o f MxNxF.
L := L(M,N,d,r,s) := { (x,y, (g^, (f ^ ))) € MxNxF I t h e r e e x i s t s some
i € { 1 , . . . ,r} w i t h g (x,y) = 0 i and f ^ (x,y) > 0 f o r 1 < j < s } .
Let TT:L->F be t h e r e s t r i c t i o n t o L o f the c a n o n i c a l p r o j e c t i o n MxNxF-*F.For

X
every element £ € F we have a s e m i a l g e b r a i c s u b s e t TT ^ (£) o f M N (iden-
t i f y i n g MxNx{£} w i t h M x N , as u s u a l ) . N o t i c e t h a t we o b t a i n a l l s e m i a l -
g e b r a i c s u b s e t s o f MxN i n t h i s way, i f we v a r y d , r , s .
We a r e i n t e r e s t e d i n t h e subset Map(M,N,d,r,s) o f F which c o n s i s t s o f

1
all £ € F such t h a t T T ~ (£) i s t h e graph V(f) o f a s e m i a l g e b r a i c map
f:M-*N. An element £ = (g r(f j))

i i of F l i e s i n Map (M, N, d, r , s) i f
and o n l y i f the f o l l o w i n g two c o n d i t i o n s a r e s a t i s f i e d :
i) F o r every x € M t h e r e e x i s t s a unique y € N w i t h (x,y,£) € L.
ii) F o r every ( x , y ) € MxN w i t h Q o ( x

Q fY '£)
0
G L a n d
every e>0 i n R there
exists a 6>0 i n R such t h a t , f o r every x € M with II x-x II <6
there e x i s t s a point y € N with II y — y II <e

o and (x,y,£) € L .
f
Thus, by T a r s k i s theorem on t h e e l i m i n a t i o n o f q u a n t i f i e r s ,
Map(M,N,d,r,s) i s a s e m i a l g e b r a i c subset o f F.
Definition 1. We denote, f o r any p o i n t £ € Map(M,N,d,r,s), t h e s e m i a l -

1
g e b r a i c map f : M-»N w i t h graph F ( f ) = TT ( £ ) by <£>, o r more p r e c i s e l y
a n < w e
by <£> /
R 3 call g a parameter o f f . The v a l u e <£>(x) o f <£> a t a
point x € M w i l l u s u a l l y simply be w r i t t e n £ (x) . We c a l l these maps
<£> t h e s e m i a l g e b r a i c maps from M t o N o f type ( d , r , s ) , and we call
Map(M,N,d,r,s) t h e parameter space o f maps from M t o N of type (d,r,s).
The base e x t e n s i o n Map (M,N,d,r,s) (S) o f t h e space Map(M,N,d,r,s) i s
the parameter space o f maps from M(S) to N(S) o f type (d,r,s). I f £ i s
a p o i n t o f Map(M,N,d,r,s) then £ i s a l s o a p o i n t o f Map(M(S),N(S),d,r,s
and t h e map < £ > s : M(S) -+ N(S) i s t h e base e x t e n s i o n o f t h e map <£> : M-»N
We now adapt t h i n g s t o t h e s i t u a t i o n i n t h e theorem. We f i r s t prove
the i n j e c t i v i t y o f K . We a r e g i v e n two maps f and g from (M,A) t o (N,B)

a
with f g - 9 / S n d we have t o show t h a t f ^ g. F o r any t r i p l e (d,r,s) o f
n a t u r a l numbers we look a t the parameter space H ( f , g , d , r , s ) o f "homo-

n
t o p i e s o f type (d,r,s) from f t o g, i . e . t h e s e t
H(f,g,d,r,s) := { £ € Map ( M x i ,N, d, r , s) I <£> (Ax I) <= B ,
£(x,0) = f ( x ) , £(x,1) = g(x) f o r e v e r y x € M}.
This s e t i s semialgebraic i n Map(Mxi, N ,d,r,s) as f o l l o w s a g a i n from
T a r s k i ' s theorem. Clearly
H(f,g,d,r,s)(S) = H ( f , g , d , r , s ) .
g g
By o u r assumption on f and g t h e r e e x i s t s a t r i p l e (d,r,s) such that
H ( f , g , d , r , s ) ( S ) i s not empty. But then H ( f , g , d , r , s ) i s a l s o n o t empty.
(N.B. Our t h e o r y o f base e x t e n s i o n , which has t h i s consequence, i s
a l s o an a p p l i c a t i o n o f T a r s k i ' s theorem). T h i s means t h a t t h e r e exists
a homotopy (of type ( d , r , s ) ) from f t o g, and the i n j e c t i v i t y of K i s
proved.
The p r o o f o f t h e s u r j e c t i v i t y o f K w i l l be h a r d e r . I t uses t h e t h e o r y
of s i m p l i c i a l a p p r o x i m a t i o n s from §2. We need a lemma. F o r l a t e r use
i n §4 we s t a t e a more g e n e r a l v e r s i o n o f t h e lemma than needed now. The
application i n t h e p r e s e n t s e c t i o n uses o n l y t h e case where t h e s e t C
below i s empty, i . e . t h e map h can be o m i t t e d .
Lemma 3.2. L e t (M,A^,...,A ) and (N,B^,...,B ) be systems o f s e m i a l g e -

r
b r a i c spaces over R and l e t h : C -> N be a map on a subspace C o f M.
Assume t h e r e e x i s t s an isomorphism f : M(S) -* N(S) such t h a t f |C(S) = h g
f ( A (S) ) = B (S) f o r 1 < i < k w i t h some k € {0, . . . ,r} , and f (A (S)) c B , (S)
± i ±
for k < i < r . Then t h e r e e x i s t s an isomorphism g :M N such that
g l C = h, g (A^) = B i f o r 1 < i < k, and g (A ) c B , f o r k < i < r .

±
n m
P r o o f . A g a i n we choose f i x e d embeddings M c R , N c R . F o r any t r i p l e
(d,r,s) o f n a t u r a l numbers we c o n s i d e r t h e s u b s e t Q ( d , r , s ) o f
Map(M,N,d,r,s) c o n s i s t i n g o f a l l p o i n t s £ € Map(M,N,d,r,s) such that
<£> extends h and i s an isomorphism from M onto N which maps A. onto

for 1 < i < k and i n t o B^ f o r k < i < r . A l l these conditions can be
e x p r e s s e d by an e l e m e n t a r y p o l y n o m i a l f o r m u l a i n t h e parameter £ over
R by use o f T a r s K i s tneorem on e l i m i n a t i o n o f q u a n t i r i e r s . Tnus
Q(d,r,s) i s s e m i a l g e b r a i c i n Map(M,N,d,r,s). By t h e assumption o f the
lemma t h e r e e x i s t s a t r i p l e (d,r,s) such t h a t Q ( d , r , s ) ( S ) i s not empty.
We c o n c l u d e t h a t Q(d,r,s) i s n o t empty, which means t h a t we have an i s o -
morphism o v e r R (of type d,r,s) w i t h the d e s i r e d p r o p e r t i e s . q.e.d.
We assume t h a t either A i s closed i n M o r b o t h A and B a r e l o c a l l y
closed i n M and N r e s p e c t i v e l y . We a r e g i v e n a map f : ( M ( S ) , A ( S ) ) -*
( N ( S ) , B ( S ) ) and we have t o f i n d a map g : (M, A) -* (N,B) such t h a t g g is
homotopic t o f .
Choosing t r i a n g u l a t i o n s o f (M,A) and (N,B) and p a s s i n g o v e r t o t h e f i r s t
barycentric subdivisions f o r s a f e t y , we assume w i t h o u t l o s s o f g e n e r a l i -
t y t h a t both (M,A) and (N,B) a r e p a i r s o f f i n i t e complexes over R which
are b a r y c e n t r i c s u b d i v i s i o n s o f o t h e r p a i r s o f complexes. We then choose
a simultaneous t r i a n g u l a t i o n cp : (X,X.j) (M(S),A(S)) o f ( M ( S ) , A ( S ) ) and
the f i n i t e l y many s u b s e t s p ( S ) , w i t h p € I (M) , and f ( T ( S ) ) , w i t h

1
T € I(N) , of M ( S ) .
According to § 2 there e x i s t s a s i m p l i c i a l a p p r o x i m a t i o n u : co(X,X^) ->
co (N (S) ,B (S) ) t o f ' c p ( c f . Th. 2 . 5 ) , and t h e maps f*cp and j ^ ^ ' u ^ r ^
from (X X^)
f t o (N(S),B(S)) a r e ( l i n e a r l y ) homotopic (Prop. 2 . 1 and 2 . 3 ) .
Now ( X , ^ ) i s the r e a l i z a t i o n ( |K| gr |K t ) over S o f a p a i r

1 g (K,K^) o f
finite abstract complexes ( I I , § 3 ) . Thus (X,X-]) i s t h e base e x t e n s i o n
(X(S),X (S)) 1 of the p a i r of f i n i t e complexes ( X ^ ) := ( IK | , | K [ ) . R 1 R
Also u i s t h e base extension u g o f a s i m p l i c i a l map u : co(X,X^) -»co(N,B)
and, o f c o u r s e , r% = ( r ) g , x J ( ) = (J^s*
N S
T h U S
~ S V w i t h
v := J - u - r .
N x
A p p l y i n g Lemma 3 . 2 t o t h e isomorphism cp : ( X ( S ) , X ( S ) ) 1 (M(S),A(S)) we

see t h a t t h e r e e x i s t s a s e m i a l g e b r a i c isomorphism x :
(X,X,j) (M, A)
which maps every open simplex a o f X i n t o t h a t open simplex p of M
whose base e x t e n s i o n p(S) c o n t a i n s 3>(a(S)). The maps cp and x s ™ P

a
every
open simplex a(S) o f X(S) i n t o the same open simplex p{S) o f M(S) . Thus
the maps cp and x s from ( X ( S ) X ( S ) ) to

f 1 (M(S),A(S)) are ( l i n e a r l y ) homoto-
p i c , and we have
f-x s - f°cp ^ Vg.
-1
M u l t i p l y i n g by Xg on the r i g h t we obtain
1
f « (v. " ) . X s
T h i s f i n i s h e s the p r o o f o f Theorem 3.1.
I t would be d e s i r a b l e t o prove a v e r s i o n o f Theorem 3.1 where our
assumption t h a t the space M i s s e m i a l g e b r a i c i s e l i m i n a t e d . We may
assume t h a t M i s connected. Then we have an e x h a u s t i o n o f M by a family
of c l o s e d semialgebraic subsets (M I n€!N) w i t h ,

n say, M c M
n n + 1 and the
covering (M n + 2 ^ M | n > 0 ) of M l o c a l l y
n finite (M Q = 0) . I t i s tempting
t o t r y to prove such a v e r s i o n by a p p l y i n g Theorem 3.1 t o the systems
(M ,A^ n M ,...,A D M
n n r ) and (N,B^,...#B ) and somehow u s i n g t h e homotopy
e x t e n s i o n theorem C o r o l l a r y 1.4.
An a p p l i c a t i o n of C o r o l l a r y 1.4 would be d i f f i c u l t i f some o f the
a r e not c l o s e d i n M. But, a l s o i n the case t h a t every A.^ i s c l o s e d i n
M, our C o r o l l a r y 1.4 seems t o be too weak to admit such a proof. One
needs a s t r o n g e r v e r s i o n of C o r o l l a r y 1.4 where o n l y homotopies are
c o n s i d e r e d which do not move p o i n t s on C o r C(S) for a given closed
s e m i a l g e b r a i c s u b s e t C o f M. T h i s l e a d s t o a study o f " r e l a t i v e " homo-
topy s e t s , c f . the next s e c t i o n §4. We s h a l l prove an e x t e n s i o n of
Theorem 3.1 t o r e l a t i v e homotopy s e t s i n §4, p r o v i d e d the A^ are c l o s e d
i n M. We use essentially the same i d e a s as i n the p r o o f o f Theorem 3.1
but a p p l y them w i t h more t e c h n i c a l e f f o r t . We will then be able to
g e n e r a l i z e the theorem t o the case where M i s no longer semialgebraic.

§4 - R e l a t i v e homotopy sets
Agai^ " c n r i r " e " t v o s^stemr Forces (M A. , . ,A ) 'N F. . . F )
o v e r R. We f i x a map h : C-»N on a c l o s e d subspace C of M with
h(Cf1 A.) c B . ( 1 < i < r ) , and we o n l y c o n s i d e r maps from (M,A.j , . . . ,A )
to (N,B^,...,B ) which r extend h. We c a l l any two such maps f , g homo-
t o p i c r e l a t i v e C, and w r i t e f « g r e l . C , i f t h e r e e x i s t s a homotopy
H : (Mxi,A x i , . . . , A x l )
1 r (N,B , . . . ,B ) w i t h H
1 r q = f, H 1 = g, and
H ( x , t ) = h(x) f o r e v e r y x € C, t 6 I . Such a homotopy H i s c a l l e d con-
s t a n t on C. The s e t o f a l l g : (M,A , . . . ,A ) -» (N,B^ , . . . ,B ) which a r e 1 r r
homotopic t o f r e l a t i v e C w i l l be denoted by [ f ] and t h e s e t o f a l l
t h e s e " r e l a t i v e homotopy c l a s s e s " w i l l be denoted by [(M,A ,...,A^),

1
h
(N B ...,B )] .
f v r
A g a i n l e t S be a r e a l c l o s e d o v e r f i e l d o f R. Then h y i e l d s a map
h c : C(S) -+ N(S) which maps every s e t A. (S) fl C(S) i n t o B.(S) (1 < i < r) .
We denote t h e r e l a t i v e homotopy s e t [(M.(S) ,A (S) , . . . ,A (S)) , 1
h ^
S h
(N(S) ,B (S) , . . . ,B (S))]
1 b r i e f l y by [ (M, A , . . . , A ) , (N,B , . . . , B ) ] (S) .
r 1 r 1 r
C C (S)
v ;
As i n §3 we have a c a n o n i c a l map K : [f ] [f ] from
h
[(M,A ,...,A ), 1 r (N B ,... B )]
r 1 r r into this set.
P r o p o s i t i o n 4.1. I f M i s s e m i a l g e b r a i c , then
h h
K : [ ( M , A . . . ,A ) , (N,B , . . . ,B ) ]
r r 1 r -> [ (M, A-, , . . . , A ) , ( N , B . . . ,B ) ] (S)
r r r
is injective.
T h i s i s c l e a r by an o b v i o u s v a r i a t i o n o f t h e p r o o f o f t h e c o r r e s p o n d i n g
fact i n §3. I n s t e a d o f t h e spaces H ( f , g , d , r , s ) used t h e r e , one uses t h e
parameter spaces H ( f , g , C , d , r , s ) o f homotopies r e l a t i v e C between two
g i v e n maps f , g : (M,A) ^ (N,B) which extend h.
Our goal i n t h i s s e c t i o n i s t o prove
Theorem 4.2 ( F i r s t main theorem, second v e r s i o n ) . Assume t h a t every A.

i s c l o s e d i n M . Then t h e c a n o n i c a l map K from [ ( M , , . . .,A ), r
(N,B 1 ,. . . ,B ) ]r
h
t o [ (M,A , . . . ,A ) ( N , B . . . ,B ) ] ( S ) i s b i j e c t i v e .
1 r f r r
h
As i n §3 we assume h e n c e f o r t h , f o r s i m p l i c i t y , t h a t r = 1 and w r i t e = A,
B.j =B. Suppose we have a l r e a d y proved the s u r j e c t i v i t y o f K . Then we
see t h a t K i s i n j e c t i v e by a p p l y i n g the s u r j e c t i v i t y r e s u l t to the
systems (Mxi,Axi), (N,B) and t h e c l o s e d subset C := ( C x i ) U ( M x 3 l ) o f
Mxi i n s t e a d o f (M,A), (N,B), C. Indeed, l e t f , g : (M,A) ^ (N,B) be two
maps with f|C = g l C = h and l e t F : ( M x i , A x i ) ( S ) (N,B)(S) be a homo-
topy w i t h F ( - , 0 ) = f , F(-,1)
s = g , F(x,t) = h (x)
s g for xGC(S), tei(S)
{We w r i t e (M,A) (S) i n s t e a d o f (M(S),A(S)) etc.} We have a map
H : C := ( C x i ) U (MxO) U (Mx1) N,
d e f i n e d by H(x,t) = h(x) f o r (x,t) E C x i , H(x,0) = f (x) , H ( x , 1 ) = g(x),
such t h a t F extends H . By t h e s u r j e c t i v i t y r e s u l t
g t h e r e e x i s t s a map
G : (Mxi,Axi) (N,B)
with G s homotopic t o F r e l a t i v e C ( S ) . C l e a r l y G ( - , 0 ) = f, G(-,1) = g,
and G(x,t) = h(x) f o r (x,t) G C x i .
Thus i t s u f f i c e s t o prove t h e s u r j e c t i v i t y o f K . Suppose t h i s i s a l -
ready done i n t h e case r = 0 . Then t h e s u r j e c t i v i t y i s a l s o c l e a r f o r
r=1 (and more g e n e r a l l y f o r every r > 0 ) by t h e f o l l o w i n g argument.
Let a map f : (M, A) (S) -+ (N,B) (S) be g i v e n w i t h f IC(S) = h . By the s u r -
jectivity f o r r = 0 t h e r e e x i s t s a homotopy G: ( A x i ) (S) -*B(S) w i t h
G(-,0) = f |A(S) , G(x, t) = h (x) f o r x € (C fl A) (S) , t € I (S) , and G(-, 1) =

s
kg f o r some map k : A -> B. We have k | A D C = h l A n C . Let C 1 : = C U A, and
l e t h^ : C-j ->N be the map w i t h r e s t r i c t i o n s h and k t o C and A r e s p e c -
t i v e l y . G l u i n g G w i t h t h e c o n s t a n t homotopy (x,t) ^ h g ( x ) on ( C x l ) ( S )
we o b t a i n a homotopy H: (C^ x i ) ( s ) -* N (S) . We have H ( - , 0 ) = f l C ^ S ) and
H (x, 1 )=(h ) 1 (x) . Now C^ i s c l o s e d i n M . By t h e homotopy e x t e n s i o n
theorem Cor. 1.4 t h e r e e x i s t s a homotopy F : ( M x i ) (s) -»N(S)

with F(-,0) = f and F l ( C j X l ) (S) = H . F i s c o n s t a n t on C ( S ) , maps
( A x i ) (S) i n t o B ( S ) , and ends up w i t h a map f 1 : (M,A) (S) -> (N,B) (S) such
that iC^ \3) - v'h«|)g. ^ppiyxno, agaxn t n e s a i j e c c:w±t;y or K l o r r - 0
we o b t a i n a homotopy F^ : ( M x i ) (s) -> N(S) which i s c o n s t a n t on (S) ,
s t a r t s with and ends w i t h g g f o r some g : M-»N. The composite homo-
topy F := F*F^ i s c o n s t a n t on C ( S ) , s t a r t s w i t h f and ends w i t h g . g
T h i s proves t h e s u r j e c t i v i t y o f K f o r r = 1 p r o v i d e d we know t h e s u r -
jectivity f o r r = 0. N o t i c e t h a t we used i n an e s s e n t i a l way t h a t A i s
closed i n M.
H e n c e f o r t h we assume t h a t r = 0 . We now e x p l a i n how t h e s u r j e c t i v i t y o f
K can be proved i n g e n e r a l once we know t h a t i t h o l d s f o r M s e m i a l g e -
b r a i c . We may assume t h a t M i s connected. We choose a l o c a l l y finite
covering (X ln€3N) o f M by c l o s e d s e m i a l g e b r a i c s u b s e t s . (We can even
choose a c o v e r i n g w i t h Xfl X = 0 i f I n-m I > 1 , c f . I . 4 . 1 9 and 1. 4 .11 .) We

m n
d e f i n e M := X . U . . . U X and C := M U C .
n i n n n
L e t a map f : M(S) N(S) be g i v e n w i t h f |C(S) = hg. We have t o f i n d a
map g : M ~* N w i t h f ^ g g r e l . C(S) . S i n c e we assume t h a t t h e theorem
h o l d s i n the s e m i a l g e b r a i c c a s e , t h e r e e x i s t s a homotopy
H-j' : ( M x l ) ( S ) - N(S) w i t h
1 (-,0) = f l M ^ S ) , ^ ' c o n s t a n t on ( C O M I C S ) ,
and H-j' (-,1) t h e base e x t e n s i o n t o S o f some map from M^ t o N. By t h e
homotopy e x t e n s i o n theorem 1.4 we then o b t a i n a homotopy
H 1 : ( M x i ) (S) N(S)
w i t h H (-,0) = f , H (x,t) = h ( x ) f o r a l l x € C ( S ) , t € I ( S ) ,
1 1 s and a f i n a l
map f 1 := H^-,1) from M(S) t o N(S) such t h a t f | C 1 1 = ( h ^ g f o r some
map :C 1 -+ N which extends h. ( R e c a l l t h a t C 1 := C U M ^ . ) Now vre r e -
peat the same c o n s t r u c t i o n w i t h f,h,M<j r e p l a c e d by f ^ ,h.j , M 2 and ob-
t a i n a homotopy
H 2 : ( M x i ) (S) N(S)
w i t h H ( - , 0 ) = f , H (-,1) =: f
2 1 2 2 ,H 2 constant on C (S) and f |C = ( h )
1 2 2 2 g
f o r some h 2 :C 2 N extending . By i t e r a t i o n we o b t a i n a sequence o f
maps (h | n £ X O , h :C - » N , w i t h h |C - = h
L , a sequence o f maps (f |n€H),
n n n n n-1 n-1 n
f : M(S) -> N ( S ) , w i t h f IC (S) = (h ) , and a sequence o f homotopies
(H ln€U) , Hn n : ( M x i ) (S) -• N(S) , w i t h H n constant on C ^ R 1 (S) ,
H (-,0) = f _
n n 1 (let C Q = C, f Q = f ) and H (-,1) = f R . The maps h f i t n
together t o a map g : M - > N . We choose i n )0,1[ c R a sequence (s InEW)

n
with s < s , f o r a l l n (say s^ = 1 - 2 ~ ) . Then we d e f i n e a s e t t h e o r e -
1 J
n n+i n
t i c map G : ( M x i ) ( s ) -> N(S) by t h e f o l l o w i n g formulas: G(x,t) = g (x)
if x £ M ( S ) and t > s f o r everv n € U , and

n
1
6<x,t) = H ( x , ( s - s _ ) - ( t - s _ ) )
n n n 1 n 1
if x€M(S), s n - 1 <t<s , n with S Q : = 0. We c l a i m t h a t G i s l o c a l l y semi-
a l g e b r a i c . By t h e g l u i n g p r i n c i p l e f o r maps (I, 3.16) i t s u f f i c e s t o
check f o r every m€ IN t h a t t h e r e s t r i c t i o n G l ( M x i ) ( S ) i s semialge-

m
b r a i c . Now, f o r x € M ( S ) , m t € I (S) , and n > m we have H (x, t ) = f _ ^ (x) = n n
(h ) ( x ) = g ( x ) . Thus G i s s e m i a l g e b r a i c on (M x[ s , 1 ] ) ( S ) . But
m o o c e
m rn
c l e a r l y G i s also semialgebraic on (M x [ o , s ] ) ( S ) . Thus G i s semi-
m m
a l g e b r a i c on ( M ^ x [ 0 , 1 ] ) ( S ) . By c o n s t r u c t i o n G(-,0) = f , G(-,1) = g g
and G(x,t) = h g (x) f o r x e C ( S ) , t € I ( S ) , as d e s i r e d .
I t remains t o prove the s u r j e c t i v i t y o f K i n t h e c a s e where M i s semi-
a l g e b r a i c . We then may a l s o assume t h a t N i s s e m i a l g e b r a i c , c f . the
corresponding argument i n §3. L e t a map f from M(S) t o N(S) be g i v e n
with f|C(S) = hg. We have to f i n d a map g from M t o N w i t h f a g r e l . C(
Since things a r e more c o m p l i c a t e d here than i n t h e p r o o f of s u r j e c t i v i -
t y o f < i n §3 we a r e l e s s audacious than i n §3 and f i r s t reduce t o t h e
case where M i s complete. So assume t h a t t h e theorem h o l d s i n t h e case
t h a t M i s complete. We may then prove t h e theorem f o r M s e m i a l g e b r a i c
as follows.
Choosing a s i m u l t a n e o u s t r i a n g u l a t i o n o f (M,C) and p a s s i n g t o i t s f i r s t
barycentric s u b d i v i s i o n , we assume w i t h o u t l o s s o f g e n e r a l i t y t h a t (M,C)
i s a system o f f i n i t e complexes wnich i s t n e b a r y c e n t r i c s u b d i v i s i o n o f

C =
another system o f complexes. Let (M ' )
0 0 co(M,C) be t h e c o r e o f t h i s
M
system and l e t r denote t h e c a n o n i c a l r e t r a c t i o n from (M,C) t o ( 0 /C ).
Q
S i n c e we assume t h a t t h e theorem h o l d s i n t h e complete case we have a
homotopy F : (M xI) (S) -+ N(S) and a map u : M

Q q N such t h a t F(-,0) =
f |M (S) , F ( - , 1 ) = u , F ( x , t ) = hg (x) f o r a l l (x, t) € (C x i ) (S) . We now

Q s
i n t r o d u c e t h e maps F' : = F -(rxid-jOg : (Mxi) (s) -» N(S) , and u«r : M-*N. We
have F* (-,0) = f - r , g F' (-,1) = Ug<»rg, and F' (x,t) = hg*rg(x) f o r a l l
(x,t) € (Cxi) (S) . We have o b v i o u s homotopies G : (Mxi) (s) -> N(S) from f
to f T g / and K Q : (Cxi) N from h * r t o h, d e f i n e d by G(x,t) :=
f (d-t)x U r g (x)) , K ( x , t ) Q h ( t x + (1-t) r(x)) . We extend K q t o a homo-
topy K: ( M x i ) -> N w i t h K(-,0) = u»r. We d e f i n e g := K ( - , 1 ) . Now we com-

1 2
pose t h e t h r e e homotopies G,F',Kg, s a y a t t = ^ and t = ^, t o form a
s i n g l e homotopy
H : ( M x i ) (S) -> N(S) .
We draw a s c h e m a t i c p i c t u r e o f t h i s homotopy.
C(S) const
M(S) G F'
f f° s
r (u g
Then H (Cxi) (S) = (H

o>S w i t h a homotopy H : Cxi N ^
o
p i c t e d as f o l l o w s :
C G
o const K
o
h h •r h« r 1
Here G Q i s t h e i n v e r s e homotopy t o K . Our homotopy H goes Q from f t o
g s but i s n o t y e t constant on C (S) .This can now be e a s i l y remedied. We
have an o b v i o u s homotopy
Cxlxl
w i t h <X>(-,0) = H , <D(x,t,1) = h(x) f o r a l l (x,t) 6 C x I , and $ c o n s t a n t
on (CxO) U ( C x 1 ) . By t h e homotopy e x t e n s i o n theorem <1> can be extended

S
to a homotopy
¥ : ( M x l x l ) (S) -> N ( S )
with ¥ ( - , 0 ) = H and H' c o n s t a n t on [(MxO) U (Mx1)](s). C o n s i d e r the map
H := ¥ ( - , 1 ) from (Mxi)(S) t o N ( S ) . We have H ( - , 0 ) = f, H ( - , 1 ) = g g and
H ( x , t ) = hg(x) f o r a l l (x,t) € ( C x l ) ( S ) . Thus H s o l v e s our problem.
From now on we assume t h a t M i s complete. We choose a c l o s e d s e m i a l g e -
b r a i c neighbourhood D o f C i n M such t h a t C i s a s t r o n g d e f o r m a t i o n
r e t r a c t o f D ( c f . Th. 1 . 1 ) . Then we choose a map h^ : D - » N which extends
h and a homotopy G : ( D x i ) ( s ) -* N ( S ) from f | D ( S ) to ( h ^ g which i s c o n s t a n t
on C ( S ) ( c f . Prop. 1 . 2 ) . We extend G to a homotopy F : (MxI) (S) N (S) w i t h
F(-,0) = f . Then F ( - , 1 ) i s a map f 1 from M(S) t o N(S') w i t h f | D ( S ) =

1
(h^)g. We now r e p l a c e f by f ^ . Changing n o t a t i o n we assume a g a i n that
f i s a map from M(S) t o N ( S ) , b u t t h a t t h e map h i s d e f i n e d on t h e
c l o s e d neighbourhood D o f C, h : D - » N , and t h a t f |D(S) = h . As b e f o r e g
we want t o deform f r e l a t i v e C ( S ) i n t o a map which i s t h e base e x t e n -
sion g s o f a map g : M-+N. L o o s e l y s p e a k i n g , we have b u i l t a barrier
D^C around C, which w i l l h e l p us t o p r o t e c t f from b e i n g disturbed
w i t h i n C. We choose a c l o s e d s e m i a l g e b r a i c neighbourhood E o f C i n M
which i s c o n t a i n e d i n the i n t e r i o r D o f D ( c f . 1 . 4 . 1 4 ) .
U s i n g s u i t a b l e t r i a n g u l a t i o n s o f (M,C,D,E) and N we assume t h a t (M,C,D,E)
i s a system o f f i n i t e complexes, t h a t N i s a f i n i t e complex which i s t h e
b a r y c e n t r i c s u b d i v i s i o n o f a n o t h e r complex, and t h a t h : D - * N maps e v e r y
open simplex o f D i n t o an open simplex o f N . L e t L denote t h e c l o s e d
subcomplex M ^ D o f M . We choose a s i m u l t a n e o u s t r i a n g u l a t i o n o f L ( S )
and the f i n i t e l y many s u b s e t s p (S) , w i t h p £ I (L) , and L(S) fl f 1

(T(S)) ,
with T G I ( N ) . A p p l y i n g I I , Lemma 4.3 (choosing E ( S ) U M S ) as the sub-

space M q t h e r e ) we then o b t a i n a simultaneous t r i a n g u l a t i o n cp : X-+M(S)
of M ( S ) , and the subsets p ( S ) , f ( T ( S ) ) o f M(S)

1
with p running through
I (M) ana T running througn , sach t n a t cp i s e q u i v a l e n c on ti{6) to
the t a u t o l o g i c a l triangulation i d ^ ^ . At t h i s p o i n t our

g assumption
t h a t M i s complete b e a r s fruits.
Let := cp (C(S)) and X 2 := cp ( E ( S ) ) . Then cp|X 2 is a simplicial iso-
morphism if from X 2 t o E(S) which maps the subcomplex X^ of X 2 onto
C(S). A c c o r d i n g to § 2 ( i . e . e s s e n t i a l l y by classical theory [Spa,
Chap. 3 ] s i n c e M i s complete) t h e r e e x i s t s a s i m p l i c i a l approximation
u : -» N ( S ) t o f°cp, and u i s l i n e a r l y homotopic t o f«!p. As in §3 we
write (X,X^,X ) 2 = ( X , X ^ , X ) ( S ) w i t h a system

2 o f complexes (X,X^,X ) 2
over R. We have $ = with a s i m p l i c i a l isomorphism : (X ,X^)

2 (E,C)
and u = Ug w i t h a s i m p l i c i a l map u : X->N. 3y Lemma 3.2 t h e r e exists
a s e m i a l g e b r a i c isomorphism v : X M which extends :X 2 E and maps
every open simplex a o f X i n t o the open simplex p o f M whose base
extension p(S) c o n t a i n s ! p ( a ( S ) ) . The maps x s

a n d
3> from X ( S ) to M(S)
s
are (linearly) homotopic r e l a t i v e X ( S ) . 2 Thus f°X£ i - homotopic t o
f°tp r e l a t i v e X 2 (S) , and a fortiori
f. X s « f«<P rel X 1 (S)
On the o t h e r hand, v/e have a l i n e a r homotopy from f°cp to Ug. But this
homotopy moves p o i n t s i n X^ ( S ) . To remedy t h i s , we choose a semialge-

1
braic function X : X -> [0, 1 ] w i t h A~ (0) = X 1 and (1 ) = X ^ X 2 (cf.
I. Th. 4.15). Then we d e f i n e a homotopy
H : (Xxl) (S) -» N(S)
by the formula (x 6 X ( S ) , t E I ( S ) )
H(x,t) = (1-tXg(x)) (f o Jp) (x) + t X ( x ) Ug(x) . s
T h i s homotopy i s c o n s t a n t on X^ (S) and s t a r t s w i t h H(-,0) = f«tp. We
have H(-,1) = v q w i t h the f o l l o w i n g s e m i a l g e b r a i c map v : X-*N:

u (x) , x € X\ X
v(x) =
1
(1-A (x)) (h • iJO (x) + A (x) u(x) , x € X . 2
Thus f«cp a v rel. X 1 (S) . A l t o g e t h e r we have
rel X-(S)
S
-1
M u l t i p l y i n g by the isomorphism x from M(S) t o X(S) we obtain
f « (v- " )X
1
s
rel C(S)
as d e s i r e d . T h i s completes the p r o o f o f Theorem 4.2.
Discussion 4.3. For a g i v e n base f i e l d R we denote by LSA(R) the cate-
gory o f ( l o c a l l y semialgebraic r e g u l a r paracompact) spaces and (local-
l y semialgebraic) maps o v e r R. We denote by HLSA(R) the "homotopy
category" of spaces over R. This category has the same o b j e c t s as LSA(R)
but i t s morphisms a r e the homotopy c l a s s e s o f maps over R. If S is a
r e a l c l o s e d o v e r f i e l d o f R then we have the base e x t e n s i o n functor
LSA(R) -+ LSA(S). Every o b j e c t M o f LSA(S) i s i s o m o r p h i c to the image
M(S) o f an o b j e c t M o f LSA(R). Indeed, every t r i a n g u l a t i o n |K|g-^»M
y i e l d s such an o b j e c t , namely M = |K| .

D Our f u n c t o r induces a f u n c t o r
K : HLSA(R) -» HLSA(S) . The main message o f Theorem 4.2 i s t h a t K i s an
equivalence o f c a t e g o r i e s . More g e n e r a l l y , Theorems 3.1 and 4.2 imply
s i m i l a r statements about s u i t a b l e homotopy c a t e g o r i e s of systems o f
spaces.
F i n a l remark 4.4. Theorem 4.2 remains true f o r l o c a l l y f i n i t e systems
(A la€I),
a (BgiaGI) o f subspaces o f M and N i n s t e a d of the systems
(A^,...,A ), r (B^,... B ), with f r every A Q c l o s e d i n M, of course. The
proof runs the same way w i t h more n o t a t i o n a l effort.

§5 - The second main theorem; contiguity classes
We assume now t h a t o u r oase r i e l d R i s the f i e l d ]R o f r e a l numbers.
Let two systems o f spaces (M A-j , . . . ,A ) , (N,B^ , . . . ,B ) over IR be g i v e n

f r
B e s i d e l o c a l l y s e m i a l g e b r a i c maps and homotopies from t h e f i r s t system
to t h e second we a l s o c o n s i d e r c o n t i n u o u s homotopy c l a s s e s o f c o n t i -
nuous maps from (M,A^,...,A ) t o (N,B^,...,B ) i n t h e u s u a l

r r topologi-
c a l s e n s e . We denote t h e s e t o f t h e s e c l a s s e s by [(M,A^,...,A ), r
(N,B^,...,B )] p» r t0 We have an o b v i o u s map
X : [(M A , . . .,A ),(N,B , . . .,B )] -

/ 1 r 1 r
[(M,A , . . . , A ) , ( N , B , . . . , B ) ]
1 r 1 r
which sends t h e l o c a l l y s e m i a l g e b r a i c homotopy c l a s s [f] of a l o c a l l y
s e m i a l g e b r a i c map f t o t h e t o p o l o g i c a l homotopy c l a s s [ f l ^ p °f f •
More g e n e r a l l y , g i v e n a l o c a l l y s e m i a l g e b r a i c map
h : (C,C fl A , . . . ,C 0 A ) - (N,B , . . . ,B )
1 1 r
on a c l o s e d subspace C o f A we have a map X from t h e s e t [(M,A^,...,A ) r
h
(N,B-j , . . . ,B ) ] r of l o c a l l y s e m i a l g e b r a i c homotopy c l a s s e s r e l a t i v e C
of l o c a l l y s e m i a l g e b r a i c maps e x t e n d i n g h t o t h e analogous s e t o f t o p o -
logical relative homotopy c l a s s e s [ (M,A^ , . . . ,A ) , (N,B^ , . . . ,B ) ] * ^ -

r r o p
We want t o prove
Theorem 5.1 (Second main theorem). Assume t h a t every A^ i s c l o s e d i n M.

h
Then the map X : [ (M,A , . . . ,A ) , ( N B , . . . B ) ] 1 r f 1 f r [ (M,A , . . . ,A ) ,
1 r
h
(N B f 1 , . . . ,B ) ]
r t Q p i s bijective.
To a g r e a t e x t e n t t h e p r o o f w i l l f o l l o w t h e same p a t t e r n as the p r o o f
of the f i r s t main theorem 4.2 i n t h e p r e c e d i n g s e c t i o n . By f o u r reduc-
tion s t e p s which a r e f u l l y analogous t o t h e c o r r e s p o n d i n g r e d u c t i o n
s t e p s i n §4 we f i r s t see t h a t i t s u f f i c e s t o prove t h e s u r j e c t i v i t y
of A and then t h a t we may assume r = 0 and t h a t M i s s e m i a l g e b r a i c and
complete. The r o l e o f maps over R i n §4 i s p l a y e d here by t h e l o c a l l y
s e m i a l g e b r a i c maps over 3R and the r o l e o f maps over S i n §4 i s p l a y e d
by t h e continuous maps. I n s t e a d o f the homotopy e x t e n s i o n theorem 1.4
one uses the analogous t o p o l o g i c a l f a c t , which i s t r u e as w e l l . (Recall
the p r o o f o f C o r o l l a r y 1.4). We l e a v e t h e d e t a i l s to the reader.
Now l e t f : M -> N be a c o n t i n u o u s map e x t e n d i n g h. S i n c e we assume t h a t
M i s complete, M i s a compact t o p o l o g i c a l space. Thus f(M) i s compact
as w e l l . Choosing a c o v e r i n g o f N by open s e m i a l g e b r a i c s u b s e t s we can
c o v e r f(M) by f i n i t e l y many o f them. Thus f(M) i s c o n t a i n e d i n a semi-
a l g e b r a i c subspace o f N. R e p l a c i n g N by t h i s subspace we assume hence-
forth that N i s also semialgebraic.
As i n §4 we c o n s t r u c t a " b a r r i e r " around C, i . e . a c l o s e d subspace D
of M w i t h D z> c such t h a t f i s homotopic r e l a t i v e C t o a map f^ : M -» N
whose r e s t r i c t i o n t o D i s s e m i a l g e b r a i c . We r e p l a c e f by f ^ and we
denote the r e s t r i c t i o n f I D - i n s t e a d o f f [ C - by h. Thus h i s now a
s e m i a l g e b r a i c map from D t o N. As i n §4 we choose a c l o s e d semialge-
b r a i c neighbourhood E of C i n M with E c D .
U s i n g t r i a n g u l a t i o n s we assume t h a t (M,C,D,E) i s a system o f f i n i t e
complexes and t h a t N i s a f i n i t e complex which i s t h e f i r s t b a r y c e n -
tric s u b d i v i s i o n o f another complex. Now we meet t h e b i g d i f f e r e n c e
to §4. We cannot r e f i n e the t a u t o l o g i c a l t r i a n g u l a t i o n o f (M,C,D,E)
i n such a way t h a t f maps every open simplex o f M i n t o an open simplex
of N, s i n c e now t h e preimages o f the open s i m p l i c e s o f N under f have
no r e a s o n t o be s e m i a l g e b r a i c . Thus our e x i s t e n c e theorem 2.5 f o r
s i m p l i c i a l approximations i s o f no use h e r e . But s i n c e t h e f i e l d TR i s
a r c h i m e d i a n and t h e t o p o l o g i c a l space M i s compact t h e r e e x i s t s a n a t u -
r a l number n such t h a t every open s t a r o f the n - t h b a r y c e n t r i c s u b d i -

(n) 1
vision M o f M i s c o n t a i n e d i n the preimage f (St^Cp)) o f an open s t a r
St (p)N f o r some v e r t e x p € N f l E ( N ) , c f . [Spa, Chap. 3, §3, Th. 14].
The f a m i l y of these s t a r s i s a ( f i n i t e ) open c o v e r i n g o f N s i n c e N i s
tame . We r e p l a c e (M,C,D,E) by i t s n-th b a r y c e n t r i c s u b d i v i s i o n . Then
we choose f o r e v e r y v e r t e x e € E ( M ) a v e r t e x y(e) 6 N n E(N) with
f ( S t ( e ) ) c S t ( u ( e ) ) . The map
M N u : E(M) -• N f l E ( N ) i s an a b s t r a c t simpli-
c i a l map from K (M) t o K ( c o N ) , c f . [Spa, Chap. 3, § 4 ] . E v i d e n t l y the
realization u = I u I : M -» co(N) i s a s i m p l i c i a l approximation t o f : M-> N
We have a l i n e a r homotopy H from f to u. T h i s homotopy moves p o i n t s i n
C. But s i n c e we have b u i l t a b a r r i e r D^C around C we can modify H to
a homotopy which s t a r t s w i t h f , i s c o n s t a n t on C, and ends up w i t h a
s e m i a l g e b r a i c map v by t h e same d e v i c e as i n §4. T h i s completes the
p r o o f o f the theorem.
Remark 5.2. Theorem 5.1 remains true for l o c a l l y finite systems o f
subspaces instead of f i n i t e systems, as does the f i r s t main theorem 4.2
D i s c u s s i o n 5.3. L e t TOP denote the c a t e g o r y of t o p o l o g i c a l (Hausdorff)
spaces and l e t HTOP denote the homotopy c a t e g o r y o f these spaces. We
have f o r g e t f u l f u n c t o r s LSA(3R) -+ TOP and X : HLSA (]R) -+ HTOP. The main
message o f Theorem 5.1 i s t h a t X i s an e q u i v a l e n c e o f HLSA(IR) with a
full subcategory o f HTOP. As o b j e c t s o f t h i s subcategory we may take
those t o p o l o g i c a l spaces which are homeomorphic t o l o c a l l y finite sim-
plicial complexes o r those spaces which a r e homotopy e q u i v a l e n t t o
such spaces.
In §3 we gave a v e r s i o n o f the f i r s t main theorem which o n l y a p p l i e s t o
s e m i a l g e b r a i c spaces and " a b s o l u t e " homotopy c l a s s e s but t h e r e i s much
s t r o n g e r than t h e second v e r s i o n Theorem 4.2, s i n c e the subspaces of M
and N a r e allowed t o be more g e n e r a l , f o r example open. We do not know
whether an analogous v e r s i o n o f the second main theorem h o l d s . We pose

a subquestion of t h i s problem.
Q u e s t i o n 5.4. Given - say - complete s e m i a l g e b r a i c spaces M,N o v e r IR ,
open subspaces U,V o f M,N, and a c o n t i n u o u s map f : (M,U) -* (N,V), does
t h e r e always e x i s t a s e m i a l g e b r a i c map g : (M,U) -* (N,V) which i s (con-
t i n u o u s l y ) homotopic t o f ?
The f i r s t main theorem, i n v e r s i o n 4.2, and t h e second main theorem 5.1
t o g e t h e r form a p o w e r f u l t o o l t o t r a n s f e r r e s u l t s from topological
(= c l a s s i c a l ) homotopy t h e o r y t o l o c a l l y s e m i a l g e b r a i c spaces o v e r any
r e a l c l o s e d base f i e l d . As a f i r s t example we g i v e a c o m b i n a t o r i a l
d e s c r i p t i o n o f t h e homotopy s e t [X,Y] f o r X a f i n i t e complex and Y
a locally finite complex. Other examples w i l l come up l a t e r (cf. in
particular §6).
From now on t h e base f i e l d R may a g a i n be an a r b i t r a r y r e a l closed
field. L e t K and L be a b s t r a c t locally finite complexes.
Definition ( c f . [Spa, p. 130]). L e t cp,ip : K ^ L be two s i m p l i c i a l maps.
a) tp and \\j a r e s t r i c t l y contiguous i f f o r every simplex s 6 S(K) t h e
union cp(s) u (s) i s a simplex o f L.
b) cp and \p a r e c o n t i g u o u s i f t h e r e e x i s t s a sequence cp = cp ,cp^ , . . . , c p = Q r
o f s i m p l i c i a l maps from K t o L such t h a t <Pj__«] and cp^ a r e s t r i c t l y
c o n t i g u o u s f o r i = 1 , . . . , r . The c o n t i g u i t y c l a s s o f cp i s denoted by
[cp] . The s e t o f c o n t i g u i t y c l a s s e s o f s i m p l i c i a l maps from K t o L
i s denoted by [ K , L ] .
N o t i c e t h a t tp and \p are s t r i c t l y c o n t i g u o u s (resp. c o n t i g u o u s ) p r e c i s e l y i f
the r e a l i z a t i o n s |ip| , | i / ; |
R R are s t r i c t l y contiguous (resp. contiguous)
maps from |K| R t o | L | i n t h e sense o f D e f i n i t i o n

R 5 i n §2.
By s d K we denote the f i r s t b a r y c e n t r i c s u b d i v i s i o n o f K. T h i s i s a
p u r e l y c o m b i n a t o r i a l n o t i o n : The v e r t i c e s o f sdK a r e the s i m p l i c e s o f
K and t h e s i m p l i c e s of sd K a r e the s e t s ( s , s^ , . . . , s^} w i t h s, 6 S ( K )

0
and s n cs 1 a ... c s . As r e a l i z a t i o n IsdK I we can ana w i l l choose tne

D i r K
barycentric subdivision IK|' o f IKI_ i n the geometric sense, a v e r t e x
s of sdK b e i n g r e a l i z e d by the b a r y c e n t e r o f the simplex IsI .

R
From now on we assume t h a t the complex K i s c l o s e d . We w i l l use the
classical theory of s i m p l i c i a l a p p r o x i m a t i o n s , a s developed i n [Spa,3.4],
t r a n s f e r e d to an a r b i t r a r y r e a l c l o s e d base f i e l d R i n the obvious
way. The more c o m p l i c a t e d t h e o r y of s i m p l i c i a l approximations
from §2 may be i g n o r e d here s i n c e K i s c l o s e d . We w i l l consider simpli-
c i a l maps from i t e r a t e d b a r y c e n t r i c s u b d i v i s i o n s sd K o f K t o L. Any
such map has i t s image i n the c o r e o f L. Thus we may always r e p l a c e L
by c o L . We assume h e n c e f o r t h t h a t L i s a l s o closed.
L e t h : sdK -»K be the a b s t r a c t i o n of a s i m p l i c i a l a p p r o x i m a t i o n t o the
i d e n t i t y map o f the space IKl_ = I sd KI_ . We o b t a i n every such map h
by mapping every v e r t e x s o f sd K to an a r b i t r a r i l y chosen v e r t e x o f
s. N o t i c e t h a t , f o r any o t h e r r e a l c l o s e d f i e l d R, the map Ihlg i s
a l s o a s i m p l i c i a l a p p r o x i m a t i o n to t h e i d e n t i t y o f IKIg.
h induces a map n : [K,L] -*• [sd K,L] , [cp] [cp»h.]. T h i s map n does not
depend on the c h o i c e of h. Indeed, i f h^ i s the a b s t r a c t i o n of another
s i m p l i c i a l approximation to i d | | K then h and h^ a r e s t r i c t l y contiguous.
A l s o n([tp]) = [\J;] w i t h \p : sd K L the a b s t r a c t i o n o f any simplicial
approximation [ sd K | -> |L| o f |cp|: |K| -> | L [ .
r s
S i m i l a r l y , g i v e n two i t e r a t e d b a r y c e n t r i c s u b d i v i s i o n s sd K and sd K
with r < s , we o b t a i n a c a n o n i c a l map

r S
n : [sd K,L] - [sd K,L]
r s
S IT
from t h e a b s t r a c t i o n h ^ : sd K -+ sd K o f any s i m p l i c i a l a p p r o x i m a t i o n
s
S r
I s d K I -* I s d K I t o t h e map x>->x. I f r < s < t then n •n = n , . Thus
st rs rt
r
we have a d i r e c t system o f s e t s ([sd K,L],r>0).
If cp,i/; : s d K 5 L a r e two s t r i c t l y
r
c o n t i g u o u s maps then t h e i r realiza-
tions I cp | ,11//! : |K| ^ ILI a r e l i n e a r l y homotopic. Thus, i f cp and \

\
> are
c o n t i g u o u s then |cp| and I iM a r e homotopic. F o r t h a t r e a s o n we have a
well defined map
P r
r
: [ s d K , L ] - [|K|,1L|], [cp] * [ I cp I ] .
If r < s then P ° l s
r
r s
=
P r s i n c e t h e map l n
r s l ' IKI -* IK I i s homotopic
t o t h e i d e n t i t y . Thus t h e maps p^_, r € ] N , y i e l d a map q
r
piling [sd K,L] - [|K| ,|L| ].
R R
Theorem 5.5. I f t h e complexes K and L a r e c l o s e d and K i s f i n i t e then
p i s bijective.
P r o o f . We denote t h e map p more p r e c i s e l y by p . F i r s t assume t h a t

IR
R = IR . The composite o f p with the c a n o n i c a l b i j e c t i o n A from

r
[IKL^, I L I ^ ] to [ I K I ^ , i L l ^ t o p ( c f . Th. 5.1) i s the map from l i i n [sd K,L]
to [IK I ILI 3 R ] t d i s c u s s e d i n [Spa 3 . 5 ]. By Theorem 8 i n [Spa 3.5] t h i s

3R
map i s a b i j e c t i o n . Thus p i s a bijection.
Now we assume t h a t R i s t h e f i e l d R Q o f r e a l a l g e b r a i c numbers. The
triangle
[|K ( ] R ,|L| ] K
r
lim [sd K,L]
[|K| K
,|L| ] K
o o
IR
with K the canonical b i j e c t i o n (Th. 3.1 o r 4.2) commutes. S i n c e p is
R
o
a bijection p i s also a bijection. I f R i s an a r b i t r a r y real closed
RQ R
field then R Q c: R. We compare p with p i n the same way and conclude
that p is bijective. q.e.d.
Remark 5.6. More g e n e r a l l y we obtain i n the same way a d e s c r i p t i o n by
contiguity c l a s s e s o f the homotopy s e t [(IKI,IKI^,...,II),
( IL I , IL.j I , . . . , IL I) ] f o r K a f i n i t e and c l o s e d complex and subcomplexes

K / L w t l : i c n a r e c s e d i n K a n <
i i - l° 3 L r e s p e c t i v e l y . Spanier [loc.cit.]
s t a t e s and p r o v e s the c o r r e s p o n d i n g t o p o l o g i c a l theorem f o r r = 1. If
K i s not finite then the theorem becomes wrong f o r r = 0, cf.
[Spa 3.5.7].
Theorem 5.5 gives a purely combinatorial - albeit complicated - des-
c r i p t i o n o f the semialgebraic homotopy s e t [|K|,|L|]. N o t i c e that this
implies a d e s c r i p t i o n of [M,N] for a semialgebraic space M and an
a r b i t r a r y space N over R, as soon as good t r i a n g u l a t i o n s a : X -Û M and
3 :Y N are given. Simply t a k e f o r K and L the abstractions of coX
and coY and n o t i c e t h a t we have a b i j e c t i o n [f] [r °f°J ]

v X from
[X,Y] to [co X , c o Y ] , with j the i n c l u s i o n coX H X and r y the cano-
n i c a l r e t r a c t i o n Y -> co Y , c f . §1.
Despite i t s easy p r o o f Theorem 5.5 i s quite astonishing i f the base
field R i s non a r c h i m e d i a n . The s u r j e c t i v i t y of p means t h a t f o r every
semialgebraic map f : |K| -* |L| we can f i n d a n a t u r a l number n and a

n
s i m p l i c i a l map cp : s d K -> L such t h a t I cp I i s homotopic to f . In f a c t , u s i n g
compactness arguments and the Lebesgue Lemma one proves f o r R = E an
even s t r o n g e r version [Spa, p. 128]: cp can be chosen i n such a way
that |cp| i s a s i m p l i c i a l a p p r o x i m a t i o n t o f , hence i s l i n e a r l y homoto-
pic to f . But t h i s l a s t statement becomes f a l s e i f the base f i e l d is
non a r c h i m e d i a n as i s shown by the f o l l o w i n g simple example.
Counterexample 5.7. Assume t h a t R i s non a r c h i m e d i a n . We choose a po-

s i t i v e element e i n R which i s s m a l l e r than every p o s i t i v e rational
2
p P P i n R F o r
number. L e t X be the boundary o f a 2-simplex [ 0 ' 1' 2^ *
t n e
any A€ [0,1 ] we denote by P ^ ( A ) point (1-A) P + Ap_. . We c o n s i d e r
i
the map f : X-» X w i t h f ( P ) = P , f ( P Q Q Q 1 (2e)) = P , f (P ) = P

1 1 l 2 (1/2) ,
f(P ) = P 2 2 and f l i n e a r (= a f f i n e ) on each i n t e r v a l [ P , P Q Q 1 (2e)],
[P Q 1 (2e) , P-| ], [P<j ,P ], [ P ,P ] .Assume t h a t t h e r e e x i s t s a n a t u r a l num-

2 2 Q
n
b e r n and a s i m p l i c i a l map g : X^ * -> X from t h e n - t h b a r y c e n t r i c sub-
n
d i v i s i o n X^ * o f X t o X w i t h g l i n e a r l y homotopic t o f . L e t Q :=
n ( n )
P 0 1 ( 2 ~ ) . The p o i n t s P Q 1 U) and P ^ (3c) i n t h e simplex ]P ,Q[ o f X
Q
a r e mapped by f i n t o ] P , P ^ [ and ] P ^ , P [ r e s p e c t i v e l y , w h i l e g maps

Q 2
the whole simplex ]P ,Q[ i n t o

Q ( i n f a c t onto) one open simplex a o f X.
S i n c e t h e images under f o f both p o i n t s P Q 1 ( e ) , P ^ (3e) a r e " l i n e a r l y

Q
connectable" i n X to t h e i r images under g, a must be t h e 0-simplex

P n e a r P
{P^}. T h i s f o r c e s g ( P ) = g(Q) = P . A p o i n t T on ] P ' o ^
Q 1 2 o
i s mapped t o a p o i n t near P^ by g, hence t o a p o i n t i n ] P , P [ U [ P , P [ Q 1 1 2
But then f ( T ) = T i s n o t l i n e a r l y c o n n e c t a b l e t o g ( T ) . T h i s contradic-
t i o n proves t h a t such a s i m p l i c i a l map g does n o t e x i s t .

§6 - Homotopy groups
F o r every p o i n t e d space VM,X ) q o v e r R we d e f i n e homotopy g i o a p s
TT (M,x ),
N o n > 1, i n t h e c l a s s i c a l way [Hu, IV, §2]:
n n
TT (M,x )
N o := [ ( l , 3 l ) , (M,x )]. Q
(We w r i t e ( M , x ) i n s t e a d o f (M,{x }) f o r s i m p l i c i t y . )
Q Q The m u l t i p l i c a -
tion i s g i v e n by
[f]-[g] := [f*g],
n n
the p r o d u c t f * g o f two maps f , g : ( I , 3 I ) ^ ( M , X ) b e i n g d e f i n e d by q
_ r f(2t 1 f t ,...,t ),
2 n 0<t, <\
(f*g) (
1
g(2t -1,t ,...,t ),
l 2 n ^< t 1 < 1
1
The inverse [f] of a class [ f ] i s the c l a s s [h] w i t h
h(t ...,t )
r n := f ( 1 - t , t , . . . , t ) . 1 2 n
For n = 0 we s t i l l define a pointed set TT (M,x ) N o as above (1° = one
point space, 31° = 0 ) . T T ( M , X ) o q t u r n s o u t t o be t h e s e t o f connected
components o f t h e space M, t h e base p o i n t b e i n g t h e component o f x . Q
But, i n g e n e r a l , t h e r e i s no r e a s o n a b l e m u l t i p l i c a t i o n on TT (M,X ).
O O
n n n n
Using a standard isomorphism from t h e p o i n t e d space (I /3I ,3I /3I )
n n N + 1
(cf. I I , §10) t o t h e space (S ,<») , w i t h S the u n i t sphere in R and
n
oo t h e n o r t h p o l e o f S , we may i n t e r p r e t TT (M,x ) as t h e s e t o f homo-
N Q
n
topy c l a s s e s [ (S ,«>) , ( M , X ) ] . Q
n n n
F o r every n> 1 we i d e n t i f y l ^ w i t h t h e subspace l ^ *{0} o f I , and
1 1 1 1 n
we denote t h e c l o s u r e o f S I ^! ^ i n 3 I by J Then we d e f i n e f o r
2
n-1
any n > 1 and any t r i p l e (M,A,x ) , w i t h M a space, A a subspace (= l o -
cally s e m i a l g e b r a i c subset) o f M, x Q a point of A, a pointed set

n n
TT (M,A,x )
N o := [(I ,8l ,J n - 1 ) , (M,A,X )] , Q
the base p o i n t b e i n g the c l a s s of the c o n s t a n t map with value X q . For
n>2 this s e t t u r n s out t o be a group w i t h m u l t i p l i c a t i o n and inverse
g i v e n by the same formulas as above, c f . [Hu, IV, § 3 ] . On the other
hand, T T ( M , A , X )
1 q i s the s e t [ (I, {0,1 } , 1 ) , (M, A,x ) ] of homotopy c l a s s e s
o f s e m i a l g e b r a i c paths i n M which s t a r t i n A and end at X q . In g e n e r a l
we a p p a r e n t l y have no r e a s o n a b l e m u l t i p l i c a t i o n on t h i s s e t . N o t i c e t h a t ,
f o r e v e r y n > 1, n n (M, { X } , x ) q Q = n n (M, x ) Q . We d e f i n e T T (M, { X } , X

Q Q Q ) : =TT (M,X )q Q
n n n n
Using a standard isomorphism ( I / J . , 3I /J .,J - /J - ) (D ,S \°°)
n-1 n-I n-! n—I
n n
with D the c l o s e d u n i t b a l l in R , we may identify ( M , A , X ) with
q
n n 1
the s e t [ ( D , S ,°°) , (M,A,x ) ] . Choosing Q some p o i n t Z q in J ^ we may
X n n
also identify TT^ ( M , A , ) Q w i t h the s e t [ ( l , 3 l , z ) , (M, A , x ) ] s i n c e the Q
n n n n
i n c l u s i o n map from (I ,3l ,z ) o to (I ,8l ,J^_^) i s a homotopy equiva-
l e n c e of t r i p l e s , c f . [DKP, p. 201],
Every map cp : ( M , A , X ) ^ q (N,B,y ) between p o i n t e d p a i r s of spaces

Q over
R induces a map between p o i n t e d s e t s
TT (cp) : T T ( M , A , X )
n N Q -+ n ( N , B , y )
n Q
i n the obvious way, [f] » [cp°f], f o r every n > 1 , and, i n case A = {x },Q
TT
B= {y }'Q a l s o f o r n = 0. The map n (^P) i s a group homomorphism i f n> 2,
and a l s o i f n = 1, A = ^ X
Q K B= {y } • As u s u a l , we
Q o f t e n w r i t e cp + instead
of TT (cp) .
n
Having d e f i n e d the a b s o l u t e and the r e l a t i v e homotopy groups completely
along c l a s s i c a l l i n e s , we can t r a n s f e r s e v e r a l b a s i c r e s u l t s about them
from c l a s s i c a l homotopy t h e o r y t o the p r e s e n t l o c a l l y semialgebraic
setting j u s t by copying c l a s s i c a l p r o o f s . But we have t o a v o i d those
p r o o f s which use path spaces, s i n c e such spaces do not e x i s t i n our
s e t t i n g . For example, we see t h a t the group TT (M,X ) i s a b e l i a n f o r

n>2 by l o o k i n g a t the p i c t u r e on p. 18 o f Whitehead's book [W^ ] (or
p. 125 in [W]), rather than u s i n g the g e n e r a l argument t h e r e which
i n v o l v e s the loop space o f ( M , X ) . The q same p i c t u r e can be used to
prove t h a t TT^ ( M , A , X ) q i s abelian f o r n>3. By this picture i t i s also
evident t h a t we obtain the same m u l t i p l i c a t i o n on Tf (M,A,x )

N o as before
i f we compose homotopies by using a coordinate t ^ , with 1 <i<n instead
of t - . The coordinate t may u a l s o be used i n the case A = {x_} .

i n o
For any pointed p a i r of spaces ( M , A , X ) and q n>1 we d e f i n e a boundary
map
3: T T ( M , A , x ) -
N o n _ (A x )
N 1 / Q
n 1
in the u s u a l way, 3 [ f ] := [f|I ~ ]. C l e a r l y 3 maps the n e u t r a l element
of T T ( M , A , x ) t o the
N o n e u t r a l element o f T T ^ ^ ( A , X ) . I f n > 2 q then 3 i s
a group homomorphlsm. 3 depends f u n c t o r i a l l y on the triple (M,A,x ).
The maps 3 and the i n c l u s i o n s i : (A,x ) (M,X ), q j : ( M , x , x ) *-> ( M , A , X )

o Q Q
yield a long sequence
(6-1) ^ V^V ^ V ' o M X }

-^ (M,A,x )
% o JUÂ^)
- . . . - TT 1 (M,A,x ) Q TI (A,X )^->
Q O TT (M,X ) .
O Q
L o o k i n g a t the arguments i n [Hu, IV, § 7 ] , which are entirely semialge-
b r a i c , we see that t h i s sequence i s exact.
We now discuss - among o t h e r things - the dependence of the absolute
and the r e l a t i v e homotopy groups on the base p o i n t . G i v e n a space M
and two points x Q and x^ we denote the set [ ( 1 , 0 , 1 ) , ( M , x , x ^ ) ] o f homo-

Q
i s a
topy c l a s s e s o f paths i n M from X q t o x^ by TT^ ( M , x , x ) . I f x
Q 1 2 third
point i n M we have a m u l t i p l i c a t i o n
n 1 ( M , x , x ) x TT (M,x
Q 1 1 1 ,x ) 2 -» n 1 (M,X ,X ), Q 2
[u][v] := [ u * v ] , by composing p a t h s i n the u s u a l way,

r U(2t) 0 < t < \
(u*v)(t) = \
L
v(2t-1) ^< t< 1 .
We see, as i n the t o p o l o g i c a l t h e o r y , t h a t the d i s j o i n t union TT(M) of
the s e t s TT^ ( M , X , ) , w i t h q (x ,x^)

Q r u n n i n g through MxM, i s a groupoid,
called the ( s e m i a l g e b r a i c ) fundamental groupoid of M.
Let (E,D,C) be a t r i p l e o f spaces w i t h D ^ C and b o t h D and C closed in

n n
E ( f o r example, E = I , D = 3 i , C = ) . Let (M,A) be a p a i r o f
spaces and l e t x ,x-j be two

Q p o i n t s i n A. Then every path u i n A running
from x Q t o x,| y i e l d s a b i j e c t i o n
: [(E,D,C),(M,A, )] - Xl [ (E, D, C) , (M, A, X ) ]

Q
o f p o i n t e d s e t s as f o l l o w s : Given a map f : (E,D,C) -+ (M,A,x^) we extend
the homotopy Cxi->A, (x,t) u ( t ) , to a homotopy F : ( E x i , D x l , C x i )
(M,A,A) w i t h F(-,1) = f , and we d e f i n e u\f] := [f ] w i t h f := F ( - , 0 ) .

# o o
A p p l y i n g the homotopy e x t e n s i o n theorem 1.4 s e v e r a l t i m e s , we see, as
i n the t o p o l o g i c a l t h e o r y [W, I I , § 1 ] , t h a t the map u i s indeed w e l l -
d e f i n e d and b i j e c t i v e , and that u depends o n l y on the class
[u] € n ( M , x , x ) .
1 Q 1
I f v i s a path i n A from x 1 to a t h i r d p o i n t x 0 then c l e a r l y (u*v)
u °v . Thus the f a m i l y ([ (E, D, C) , (M, A,x) ] | x € A) i s a l o c a l system of
p o i n t e d s e t s on the space A. In o t h e r words, i f we r e g a r d TT(A) as a
c a t e g o r y , whose o b j e c t s a r e the p o i n t s o f A and whose morphisms are
the homotopy c l a s s e s o f p a t h s , then we have a f u n c t o r
x [ (E, D,C) , (M,A,x) ], [u]t->u , from TT(A) to the c a t e g o r y o f p o i n t e d

ft
s e t s . In p a r t i c u l a r , the group (A,x) o p e r a t e s on the s e t [(E,D,C),
(M,A,x)] from the left.
As i n t o p o l o g y one easily verifies the f o l l o w i n g u s e f u l f a c t ( c f . [W,
p. 101]).
Proposition 6.2. I f C i s c o n t r a c t i b l e and A i s connected then the for-
g e t f u l map [(E,D,C),(M,A,x)] -* [(E,D),(M,A)] induces a b i j e c t i o n from
the s e t o f o r b i t s o f TT^ (A,x) i n [(E,D,C),(M,A,x)] t o the s e t
[(E,D),(M,A)].
Notice that, i f C i s complete, we can r e p l a c e the t r i p l e (E,D,C) every-
where above by (E/C,D/C,C/C), as i s u s u a l l y done i n t o p o l o g y . But if C
i s only closed i n E, then the spaces E/C, D/C do not n e c e s s a r i l y exist
i n our s e t t i n g . N e v e r t h e l e s s the c o n s i d e r a t i o n s above go through i n
this generality.
n n
If (E,D,C) = (I ,3l ,J .) and n > 2 then the maps u above are group
n— i — #
n n n
homomorphisms. The same h o l d s i f (E,D,C) ( l , I , 3 I ) , n> 1, M = A.
Thus we have l o c a l systems o f groups (TT^ (M,A,X) I X € A ) f o r n > 2 and
(TT (M,x) Ix € M) f o r n> 1 .
G i v e n a map f : (M,A) -> (N,B) the induced homomorphisms f * : TT(A) -»TT(B)
and f + : TT^ (M, A , X ) -> n n (N,B, f (x) ) t o g e t h e r y i e l d a morphism from the
local system ( ( M , A , x ) I x € A) t o the l o c a l system (TT^ (N,B,y) ly € B) f o r
every n> 1. In o t h e r words, f o r every path u i n A from a p o i n t x Q to a
p o i n t x^ the diagram
n (M,A
n f X l ) .n (M,A,x )
n 0
n ( N , B , f (x.,))
n lfno-*" (N,B,f(x ))
n 0
commutes. S i m i l a r l y a map f : M-*N y i e l d s a morphism from the local
system (TT^ (M,x) Ix € M) to (TT (N,y) | y £ N) f o r every n>0.
Finally, f o r every p a i r (M,A) o f spaces the l o n g e x a c t homotopy
sequences (6.1) o f the t r i p l e s (M,A,x) w i t h x r u n n i n g through A f i t

t o g e t h e r to a " l o c a l system o f e x a c t sequences" on A, i n the o b v i o u s
sense.
Up t o now we transferee! r e s u l t s from t o p o l o g i c a l homotopy t h e o r y to
the locally s e m i a l g e b r a i c t h e o r y simply by c o p y i n g s u i t a b l e classical
p r o o f s . T h i s method alone has a r a t h e r l i m i t e d range. A much more
e f f i c i e n t method i s based on the f o l l o w i n g two theorems, which, up t o
trivial considerations, are s p e c i a l cases o f the f i r s t main theorem
(in one o f i t s v e r s i o n s 3.1 o r 4.2) and the second main theorem 5.1.
Theorem 6.3 ( F i r s t main theorem f o r homotopy g r o u p s ) . L e t R be a r e a l
closed f i e l d c o n t a i n i n g t h e base f i e l d R.
i) F o r every p o i n t e d p a i r of spaces (M,A,x ) over R and f o r n > 2

Q the
natural map
K : n (M,A,x ) - TT (M(R) ,A(R) , X ) ,

n o n Q
d e f i n e d by K ( f ] := [ f g ] / i s an isomorphism o f groups. I f A = { x } , o
t h i s holds f o r n=1. In the remaining c a s e s n = 1 , A * ( x } , and

Q
n = 0, A = { x } , K i s an isomorphism o f p o i n t e d
Q sets.
i i ) L e t M be a space over R. The n a t u r a l map [u] ^ [u~] from TT(M) to
T"F(M(K)) i s an i n j e c t i v e g r o u p o i d homomorphism. The image i s t h e
u n i o n o f a l l s e t s TT^ (M(R) , X , x ^ ) w i t h x , x
q Q 1 € M. Given a path u i n
M from a p o i n t x Q to a p o i n t x^ the diagram
TT (M,x )
n 1 >n (M,x )
n o
*n ( M ( S )
' 1 X }
(u^T \ (R) X )
t Q
commutes f o r every n>0.
i i i ) Similarly, g i v e n a p a i r of spaces (M,A) over R and a path u i n A
from a p o i n t x Q to a p o i n t x^ the diagram

TT (M,A,x )
n 1 (M,A,x )
Q
n (M(R) ,A(R) ,x^)

n ^ T T ( M ( R ) ,A(R) , X )
n Q
R #
commutes f o r e v e r y n > 1.
Theorem 6.4 (Second main theorem f o r homotopy g r o u p s ) . Assume t h a t the
base field R i s the f i e l d 1R of real numbers.
i) G i v e n a p a i r o f p o i n t e d spaces ( M , A , X ) then, f o r every n > 2 ,

q the
obvious map
X : n (M,A,x ) - V * ! ' A , x )
n o o t o p
from t h e n - t h s e m i a l g e b r a i c homotopy group t o t h e n - t h topological
(= c l a s s i c a l ) homotopy group i s an isomorphism o f groups. T h i s i s
a l s o t r u e i f n = 1 and A = (x ) . In the r e m a i n i n g cases n = 1 ,

o
A* { Q} / X a n (
3 n = 0, A = r X i s an isomorphism of pointed sets.
i i ) For e v e r y space M o v e r 3R t h e o b v i o u s map A : TT(M) -» ^ top f r o m
the s e m i a l g e b r a i c t o the t o p o l o g i c a l fundamental g r o u p o i d i s an iso-
morphism o f g r o u p o i d s . T h i s map, t o g e t h e r w i t h t h e maps

X : TT (M,x) -^rc (M,xh f o r xGM form an isomorphism from the s e m i -
c
n n top
algebraic local system (n (M,x)|x€M) to the t o p o l o g i c a l local system
( T T (M,x)
N t Q p l x € M) on M (n>0).
i i i ) Similarly, g i v e n a p a i r o f spaces (M,A) o v e r 3R , we have a n a t u r a l
isomorphism from the s e m i a l g e b r a i c l o c a l system (TT (M,A,X) lx € A) to
the topological local system ( n (M,A,x)

n fx € A) on A (n>1).
Theorem 6.4 tells us t h a t the a b s o l u t e s e m i a l g e b r a i c homotopy groups
of a space M o v e r 3R and the r e l a t i v e s e m i a l g e b r a i c homotopy groups
of a p a i r o f spaces (M,A) o v e r 3R are r e a l l y t h e same o b j e c t s as t h e
c o r r e s p o n d i n g t o p o l o g i c a l homotopy g r o u p s . The message o f Theorem 6.3
i s bound t o be more c o m p l i c a t e d s i n c e , i n g e n e r a l , M(K) and A(K) con-
t a i n many more p o i n t s than M and A. Anyway, e v e r y group TT (M(S) ,x)

n
is i s o m o r p h i c to a group TT^ (M, y) s i n c e t h e r e e x i s t s a path i n M(R) from
x t o some p o i n t y € M. Similarly every group (M(R), A (R), x) i s isomor-
p h i c t o a group TT^ (M, A, y) w i t h y £ A.
By the main theorems 6.3 and 6.4 the homotopy groups o f any triple
(M,A,x ) which i s the base e x t e n s i o n to R o f some t r i p l e

Q (M ,A ,x )
o Q o
over the f i e l d R Q o f r e a l a l g e b r a i c numbers a r e i d e n t i c a l w i t h the
t o p o l o g i c a l homotopy groups of (M (3R) ,A (nR) / X

Q Q Q ) - Indeed,
M ( 3 R ) A { M ) X ]
n (M (R),A (R)
n 0 o / X o ) = VM ,A ,x ) c o o = % < G ' o 'o =
x W e
TT (M (]R) ,A (3R) '
n o Q 0 ) t 0 p- ( w r i t e " = " s i n c e the isomorphisms a r e en-
tirely c a n o n i c a l . ) In p a r t i c u l a r , i f V i s an a l g e b r a i c v a r i e t y over
R and x i s a point c i n V(R ), then

o o o
n (V(R),x )
n 0 = n (V(lR)
n ,x ) Q t o p .
If x i s a point i n V(R) then Ti (V(R),x) i s i s o m o r p h i c t o 7 T ( V ( R ) , x )

n n o
f o r any p o i n t x € V ( R ) which l i e s
o Q i n the connected component o f x i n
V ( R ) . Thus we know the s e m i a l g e b r a i c homotopy groups o f spheres,
Grassmannians, S t i e f e l varieties, e t c . o v e r any real closed field to
the same e x t e n t as we know the t o p o l o g i c a l groups o f t h e s e varieties
o v e r TR .
Now observe t h a t , up to isomorphism, every t r i p l e (M,A,x ) over R i s

Q
the base e x t e n s i o n of a t r i p l e over R . Q Indeed, choosing a triangula-
t i o n of (M,A,x ), we Q have a p a i r o f a b s t r a c t l o c a l l y finite simplicial
complexes (K,L) and a v e r t e x e o f L t o g e t h e r w i t h an isomorphism from
(|K| ,|L| ,e) t oR R (M,A,x ). Then
TT (M,A,x ) « n ( | K | , | L | , e >
n o n R R = TT ( I K I ^ ,| L I ^ , e)
n t Q p .
•For example, i f M i s s e m i a l g e b r a i c , then i t i s e v i d e n t t h a t the funda-
mental group TT (M,x ) w i t h r e s p e c t t o any base p o i n t

Q is finitely pre-
sentable, since t h i s i s known f o r f i n i t e closed s i m p l i c i a l complexes
i n the t o p o l o g i c a l t h e o r y ([Spa, 3 . 7 ] ) . (Choose a good triangulation

(X,a) (M,x ) and r e p l a c e X by i t s core!)
Q
More g e n e r a l l y every system o f spaces ( M , , . . . , A ) over R i s isomor-

r
p h i c t o the base e x t e n s i o n t o R o f a system o v e r R , Q since (M,A-j , . . . ,A ) r
can be t r i a n g u l a t e d . U s i n g t h i s o b s e r v a t i o n and our main theorems -
e i t h e r i n t h e g e n e r a l v e r s i o n s i n §3 - §5 o r the s p e c i a l v e r s i o n s i n
t h i s s e c t i o n - we are a b l e t o t r a n s f e r a l a r g e p a r t o f the r e s u l t s o f
c l a s s i c a l homotopy t h e o r y to t h e l o c a l l y s e m i a l g e b r a i c s e t t i n g . We
illustrate t h i s h e r e w i t h two examples. L a t e r we w i l l l e a v e such matters
t o the r e a d e r .
Our first example i s the homotopy e x c i s i o n theorem o f B l a k e r s and
Massey. We choose a v e r s i o n which ( i n t h e t o p o l o g i c a l s e t t i n g ) can be
found i n [DKP, §15].
Theorem 6.5. L e t A^ and A 2 be open l o c a l l y s e m i a l g e b r a i c subsets o f a
space M o v e r R w i t h M = A-j U A 2 • Let A Q := fl A . 2 Assume t h a t p and q
are n a t u r a l numbers w i t h p + q > 3 and n (A^ , A , x) = 0 f o r 1 < r < p ,

r Q
TT (A ,A ,x) =0
r 2 Q for 1 <r<q, f o r every p o i n t x € A . Then the homomorphism
j * : TT (A ,A ,x) n 2 o n (M A ,x)
n / 1
induced by the i n c l u s i o n j : ( A , A ) *•+ (M,A^) i s an isomorphism

2 Q in d i -
mensions 1 < n < p+q - 2 and an epimorphism i n dimension n = p+q - 2 f o r
every x € A . Q
Proof. Choosing a t r i a n g u l a t i o n of (M,A^,A ) we may 2 assume t h a t
(M,A A ) = r 2 (X(R) Y f 1 (R) ,Y (R) ) 2
f o r some t r i p l e (X,Y.j ,Y ) 2 over R Q w i t h Y^ and Y 2 open i n X. L e t
Y Q := Y^ n Y 2 and l e t i denote the i n c l u s i o n map from Y

( 2' o^ Y t o
(X,Y-j). We have X = Y^ U Y 2 and j = i . R We c o n c l u d e from Theorem 6.3
T T Y Y # x v a n i s n
t h a t , f o r every x € Y , Q the groups n ( Y , Y , x ) and
r 1 Q r ( 2' Q ^
for 1 < r < p and 1<r<q r e s p e c t i v e l y . The sets (IR), Y (IR)
2 a r e open
i n X(IR) and c o v e r X(IR) . A g a i n by Theorem 6.3, the groups
TT (Y (3R) , Y
r n Q (IR) ,x) and T T ( Y ( I R ) ,Y (3R) ,x) v a n i s h f o r 1 < r < p
r 2 Q and
1 <r<q respectively and every xGY (IR) Q (not j u s t x € Y ) . S i n c e the
t o p o l o g i c a l analogue o f Theorem 6.5 i s t r u e [DKP, §15] we conclude
from Theorem 6.4, t h a t , f o r every x € Y (IR) , Q
(i^)* : n ( Y ( I R ) ,Y (3R) ,x)

n 2 0 n (X(3R) ^ ( I R )
n ,x)
is an isomorphism i n dimensions 1 < n < p+q - 2 and an epimorphism i n
dimension n = p + q - 2. Now a p p l y Theorem 6.3 t w i c e . We first see that
the analocrous statement i s t r u e over R and then t h a t i t i s true over

o
R, which i s what we want. In the l a s t argument s p e c i a l cases o f t h e
f o l l o w i n g obvious g e n e r a l f a c t have been used: I f f : (M,A,x ) -> (N,B,y )

Q Q
i s a map between p a i r s of p o i n t e d spaces over R, then f o r every n > 1
and every r e a l c l o s e d o v e r f i e l d R o f R the diagram

TT (M,A,x )
n o TT (N,B,y )
n o
f*
TT (M(R) ,A(R) , X )
n q Ti (N(R) ,B(R) ,y )
n Q
R *
commutes, q.e.d.
We a r e not aware of any proof of Blakers-Massey*s e x c i s i o n theorem i n
t o p o l o g y which c o u l d be t r a n s f e r e d t o the g e n e r a l l o c a l l y s e m i a l g e b r a i c
s e t t i n g . The p r o o f i n [DKP] makes heavy use o f the f a c t t h a t IR i s
archimedian.
Our second example o f knowledge t r a n s f e r from t o p o l o g i c a l to semialge-
b r a i c homotopy t h e o r y i s the f o l l o w i n g fundamental theorem which, i n
the s e t t i n g o f CW-complexes, i s due t o J.H.C. Whitehead, c f . e.g.
[Spa, p. 405].
Theorem 6,6. L e t f : M ->N be a map between connected spaces over R. Let
x be a p o i n t o f M and
c v := f ( x ).
o •* o o
T T
i) Assume t h a t , f o r every r > 0, the map TT (f) from n (M,x ) to
R Q r (N/y ) o
is an isomorphism. Then f i s a homotopy equivalence.
i i ) Given some n a t u r a l number n, assume t h a t T T ^ ( f ) : n ( M , x ) -*Tt (N,y ) R Q r o
is an isomorphism f o r 1 < r < n and an epimorphism f o r r = n. Then, f o r
every space P o v e r R w i t h dim P < n, the map f * : [P,M] -* [P,N] induced by f
i s b i j e c t i v e , w h i l e f o r e v e r y n - d i m e n s i o n a l space P t h i s map is surjective.
We g i v e the p r o o f o f p a r t i ) i n d e t a i l l e a v i n g p a r t i i ) t o the reader.
Choosing good t r i a n g u l a t i o n s o f ( M , x ) and

Q (N,y ) we
Q assume t h a t t h e r e
a r e g i v e n tame complexes X and Y over R Q w i t h M = X(R), N = Y(R) and x /Yo 0
v e r t i c e s o f X and Y l y i n g i n X and Y r e s p e c t i v e l y . Now X and Y are homo-
topy e q u i v a l e n t to t h e i r c o r e s , c f . §1. R e p l a c i n g X and Y by their cores
we assume t h a t the complexes X and Y are c l o s e d . Then X(JR) and Y(IR) a r e
certainly CW-complexes. By Theorem 4.2 (= second v e r s i o n of the first
main theorem) t h e r e e x i s t s a map g : (X,x ) -+ ( Y , y ) such

Q Q that g R is
homotopic t o f : (M,x ) -» (N,y ). We may r e p l a c e f by g_ and assume hence-

O O x\
forth that f = g . R A p p l y i n g Theorem 6.3 ( i . e. a g a i n the main theorem) twice
we see f i r s t t h a t the map TT (g) i s an isomorphism and then t h a t TT (g_.)

r r iK
is an isomorphism f o r every r > 0. Now we know by the second main theorem
6.4 and by the t o p o l o g i c a l Whitehead theorem [Spa, p. 403] that
g : X(3R) -»Y(3R) i s a t o p o l o g i c a l homotopy e q u i v a l e n c e . By the second

IK
main theorem, i n i t s g e n e r a l v e r s i o n 5.1, we know t h a t g__ i s a l o c a l l y
IK
s e m i a l g e b r a i c homotopy e q u i v a l e n c e . Then, a p p l y i n g the f i r s t main theo-
rem t w i c e , we see that f = g R i s a homotopy equivalence.
Example 6.7. L e t M be a locally complete space. Then the c a n o n i c a l
map X H X from M ^ q c to M ( c f . I , §7) i s a homotopy e q u i v a l e n c e . Indeed
the h y p o t h e s i s of p a r t i ) o f the theorem i s s a t i s f i e d for trivial
reasons.
How about a proof of Theorem 6 . 6 by semialgebraic methods? L e t us again
confine a t t e n t i o n to p a r t i ) . This suffices to u n d e r s t a n d the difficul-
t i e s . As i n d i c a t e d i n the proof above we may assume t h a t M and N are
partially complete. I f the space M i s s e m i a l g e b r a i c t h e n we can argue
i n a very c l a s s i c a l way. We form the mapping c y l i n d e r Z ( f ) o f f : MN
(cf. I I , §10; f i s now proper). The map f i s the composite p«j of the
i n c l u s i o n map j : M «-> Z ( f ) and the n a t u r a l p r o j e c t i o n p : Z ( f ) ->N. Now p
i s a homotopy e q u i v a l e n c e . Thus we may replace f by the map j . Hence-
f o r t h we assume t h a t M i s a closed subspace o f the p a r t i a l l y complete
space N and f i s the i n c l u s i o n map. The assumption i n ( i ) means t h a t
the r e l a t i v e homotopy groups TT (N,M,x ) a l l v a n i s h .

r Q
r r-1
S i n c e M i s c o n n e c t e d t h i s means t h a t every map (D ,S ) (N,M) is
homotopic t o a c o n s t a n t map, c f . Prop. 6 . 2 . From t h i s one concludes
i n the u s u a l way (e.g. [DKP, p. 2 1 2 ] ) t h a t every map cp : D r

-* N w i t h
cp(S
r— 1
) c M i s homotopic r e l a t i v e S
T~—t o1 a map \p w i t h
r
i|> (D ) c M (cp can
be "compressed" t o M). Choosing a t r i a n g u l a t i o n o f (N,M) i t is now
easy t o c o n s t r u c t a strong d e f o r m a t i o n r e t r a c t i o n from N t o M working
"simplex by simplex".
T h i s i s an e n t i r e l y c l a s s i c a l proof of part ( i ) o f Whitehead's theorem
for M semialgebraic. Unfortunately the p r o o f does not work i f M i s o n l y
locally semialgebraic, s i n c e then we do not have a mapping c y l i n d e r Z ( f )
a t our d i s p o s a l .
Here a s e r i o u s h a n d i c a p of our homotopy t h e o r y comes i n t o s i g h t . Some
o f the most b a s i c c o n s t r u c t i o n s i n topology are impossible i n the cate-
gory o f - say - partially complete s p a c e s . Even the cone CX = X x i / X x { 1 }
over a p a r t i a l l y complete space X (= mapping c y l i n d e r o f the map from
X to the one p o i n t space) does not exist. Indeed, c h o o s i n g a t r i a n g u -
l a t i o n , we may assume t h a t X i s a c l o s e d l o c a l l y finite complex. Then

CX e x i s t s i n the c a t e g o r y o f s i m p l i c i a l complexes. But CX i s not local-
ly f i n i t e i f the complex X i s not finite.
All the more we cannot c o n s t r u c t the s u s p e n s i o n o f a p a r t i a l l y complete
space which i s not s e m i a l g e b r a i c . Thus s t a b l e homotopy t h e o r y i s o u t o f
bounds. We a l s o do not have the analogues o f many important infinite
d i m e n s i o n a l CW-complexes, as f o r example t h e Eilenberg-MacLane spaces
K ( T T , n ) , a t our disposal.
T h i s makes i t d e s i r a b l e t o admit s u i t a b l e i n d u c t i v e l i m i t s o f complete
s e m i a l g e b r a i c spaces over R which a r e more g e n e r a l than p a r t i a l l y com-
plete l o c a l l y s e m i a l g e b r a i c spaces. The authors developed a full
f l e d g e d t h e o r y of such l i m i t s which we call "weak p o l y t o p e s " .
In the c a t e g o r y o f weak p o l y t o p e s mapping c y l i n d e r s , suspensions,
Eilenberg-MacLance spaces, and a l l that e x i s t . This leads, i n p a r t i c u -
lar, t o a s a t i s f a c t o r y s t a b l e homotopy t h e o r y o v e r any real closed
field R, opening a door t o g e n e r a l cohomology t h e o r i e s over R. And,
of coiarse, w i t h weak p o l y t o p e s a d i r e c t p r o o f o f Theorem 6 . 6 can be
given.
In d e v e l o p i n g the theory o f weak p o l y t o p e s one has t o be careful
t h a t the l i m i t s o f complete spaces do not become too " w i l d " and the
geometry i s d e s t r o y e d . T h i s needs more space and time than is avail-
a b l e now. We have t o d e l a y the p u b l i c a t i o n o f t h i s t h e o r y t o a later
occasion.
§7 - Homology, the Hurewicz theorems
Having developed homotopy t h e o r y t h i s f a r , i t i s n a t u r a l t o ask whether
t h e r e e x i s t s an analogue o f s i n g u l a r homology f o r l o c a l l y semialgebraic
spaces and a Hurewicz homomorphism from the homotopy groups t o the ho-
mology groups. We c e r t a i n l y can d e f i n e s i n g u l a r c h a i n s u s i n g semialge-
braic ( i n s t e a d o f j u s t continuous) maps from the s t a n d a r d simplices

n
A, n>0, to the g i v e n space but the a u t h o r s do not know whether the
homology groups d e f i n e d i n t h i s way have r e a s o n a b l e p r o p e r t i e s .
Nevertheless a s a t i s f a c t o r y homology t h e o r y o f s e m i a l g e b r a i c spaces
has been developed a v o i d i n g the s i n g u l a r approach, i n [D] , [DK^], [D.j ] .
The p r o o f s t h e r e r e a d i l y g e n e r a l i z e to l o c a l l y semialgebraic spaces.
Thus we g i v e here o n l y a s h o r t resume o f the b a s i c s o f t h i s homology
t h e o r y , condensed i n t o two theorems. We r e f e r the reader t o the papers
above f o r the d e t a i l s . [DK^] gives mainly an overview on homology and
cohomology. The m i s s i n g p r o o f s can be found i n [D] and [ D ^ ] , the main
d i f f e r e n c e between t h e s e two a r t i c l e s b e i n g the way the homotopy i n -
v a r i a n c e o f cohomology i s proved. (Cohomology i s e s t a b l i s h e d f i r s t ,
homology l a t e r , i n c o n t r a s t to the s i n g u l a r theory.) For a f i r s t read-
ing we recommend CDK^, §3 - § 5 ] .
All we need to know h e r e from the t h e o r y developed i n t h e s e papers i s
the f o l l o w i n g Theorem 7.1. L e t HLSA(2,R) denote the c a t e g o r y whose ob-
j e c t s are the p a i r s o f spaces (M,A) over a g i v e n r e a l c l o s e d f i e l d R
and whose morphisms a r e the homotopy c l a s s e s of (locally semialgebraic)
maps between such p a i r s . L e t E denote the endomorphism (M,A) -* ( A , 0 )
of t h i s c a t e g o r y . L e t G be an a b e l i a n group.
Theorem 7.1. There e x i s t s a f a m i l -y (h n \n£%) o f c o v a r i a n t f u n c t o r s from

1
HLSA(2,R) t o the c a t e g o r y of a b e l i a n groups and a f a m i l y (3 ln£Zf) o f n a t u r a l

transformations 3 :h -> h «, • E w i t h the f o l l o w i n g properties,
n n n-i
1) F o r e v e r y p a i r o f spaces (M,A) the l o n g sequence
h ( M h ( M A l h ( A )
•••-h <A,0) h ^ i r
n n ^ ) h^Tir n ' > 5 (X,A)
n n-1 ^ ----
i s e x a c t . Here i and j denote the i n c l u s i o n maps ( A , 0 ) -* (M,0) and
(M,0) - (M, A) .
2) I f (M,A) i s a p a i r of spaces and U i s an open l o c a l l y s e m i a l g e b r a i c
subset o f M with U c A then the i n c l u s i o n map j : (M\U,A\U) -* (M, A)
induces, f o r e v e r y n € Z, an isomorphism h (j).
3) F o r the o n e - p o i n t space * we have h ( * , 0 ) n =0 for n*0 and h ( * , 0 )

Q =G.
4) For e v e r y f a m i l y o f spaces (M^laCI) and every n € 2 the n a t u r a l map
from the d i r e c t sum o f the groups n

n ( M
a f 0 ) t o the group
h (IX (M Ia€I),0)
n Q i s an isomorphism.
Let us c a l l a family ( h , 3 l n € Z ) w i t h the p r o p e r t i e s

n n indicated i n the
theorem, t o g e t h e r w i t h a f i x e d isomorphism e : G-^»h (*,0),

o a homology
t h e o r y o v e r R w i t h c o e f f i c i e n t s i n G. Theorem 7.1 s t a t e s t h a t such a
homology t h e o r y e x i s t s f o r every G and R. Then i t i s p r e t t y e v i d e n t
t h a t , up t o isomorphism, t h i s homology t h e o r y ( (h. , 3 InEZ) , e) i s unique.

n
Namely, i f (X,A) i s a tame p a i r o f complexes over R, then one deduces
from t h e f u n c t o r i a l l o n g e x a c t sequences ( c f . p r o p e r t y 1) of the pairs
(X,A) and (co X, co A) t o g e t h e r w i t h the f a c t that co X and co A are
s t r o n g d e f o r m a t i o n r e t r a c t s o f X and A, t h a t the i n c l u s i o n map from
(coX, co A) to (X,A) i n d u c e s an isomorphism
h (coX,
n coA) Ah (X,A)n
for every n € Z. Then i t can be shown as i n t o p o l o g y ( c f . [ E S ] , [Mi])
that h ( c o X ,
n co A) i s i s o m o r p h i c t o the s i m p l i c i a l homology H (K,L;G) n
of the a b s t r a c t s i m p l i c i a l complex (K,L) = (K(co X) ,K (co A ) ) . In parti-
c u l a r h (X,A) = 0
n i f n<0 and i f n > dimX.
H e n c e f o r t h we denote the group h (M,A) by H^(M,A;G) andn call i t the

n-th homology group o f t h e p a i r (M,A) w i t h c o e f f i c i e n t s i n G. We write
H (M,G) i n s t e a d o f H (M,0;G). G i v e n a map f : (M,A)

n n (N,B) we o f t e n de-
note a l l t h e homomorphisms ^ ( f ) by f . n +
From our d i s c u s s i o n o f uniqueness o f homology i t i s e v i d e n t t h a t t h e
groups H (M, A;G)

n obey a u n i v e r s a l c o e f f i c i e n t theorem as common i n t h e
s i m p l i c i a l and s i n g u l a r homology t h e o r i e s , c f . [Spa, p. 222].
Theorem 7.2 ( F i r s t and second main theorem f o r homology).
a) L e t R be a r e a l c l o s e d f i e l d containing R. Then we have, f o r e v e r y
p a i r o f spaces (M,A) and every n € %, an isomorphism
K : H (M,A;G) -*>H (M(R) ,A(R) ;G) ,

n n
f u n c t o r i a l with respect t o (M,A), such t h a t t h e diagram

3 (M,A)
n
H (M,A;G)
n H n - 1 (A G) f
H (M(R),A(R);G)
n 3 ( (8),A(«))
M "n-1,(A(K),G)
commutes.
b) F o r every p a i r o f spaces (M,A) o v e r M and e v e r y n € 2 we have a
functorial isomorphism
A : H (M,A;G) H (M,A;G),
n n top
such t h a t t h e diagram
3^(M,A)
n
H (M,A;G)
n
H _ (A,G)
n 1
H
n ( M
' A
' G
W H
n-1 ( A
' G )
top
3
n ( M
' 'top
A
commmutes. Here H (M,A;G). denotes t h e n - t h s i n g u l a r homology Zi L

n top ' ~
group o f the p a i r o f t o p o l o g i c a l spaces (M,A) w i t h c o e f f i c i e n t s i n
G and 8 denotes the boundary map i n s i n g u l a r homoloav.
n,top
It i s now e v i d e n t t h a t we can t r a n s f e r a g r e a t number o f r e s u l t s from
s i n g u l a r homology t h e o r y t o spaces o v e r any r e a l c l o s e d f i e l d R, c f .
the d i s c u s s i o n f o r homotopy groups i n §6. For a c o n c r e t e example,
namely A l e x a n d e r - P o i n c a r e d u a l i t y , see CDK^, § 5 ] . A more s u b t l e method
for transfering results from s i n g u l a r to s e m i a l g e b r a i c homology
is also discussed i n [DK^] ( c f . §7 - §8 o f t h a t p a p e r ) .
It i s clear that, f o r any space M o v e r R, the group H ( M , 2 ) Q i s the f r e e
a b e l i a n group o v e r t h e s e t T T (M) Q o f connected components o f M, and
H (M,G) i s t h e t e n s o r p r o d u c t o f H ( M , Z ) w i t h G o v e r 2.
Q q
For any n e 2 we d e n o t e by H (M,G) t h e k e r n e l o f the homomorphism from

N
H (M,G) t o H (*,G) i n d u c e d by the map

N n from M t o the one p o i n t space
("reduced homology g r o u p s " ) . I f n * 0 then H (M,G) = H ( M , G ) ,

N N w h i l e we
have a s p l i t exact sequence
0 -+ (M,G) -> H (M,G) -> G -+ 0.
If x Q i s any p o i n t i n M then the n a t u r a l map from R (M,G) t o

Q
H ( M , x ; G ) i s an
Q o isomorphism.
H e n c e f o r t h we denote t h e homology groups H ( M , A ; Z ) , N H ( M , 2 ) , 8t (M,2)

N N
simply by H n (M, A) , H n (M) , ^ ( M ) . We want to compare t h e s e groups w i t h
the homotopy groups (or s e t s ) T T ( M , A , x ) , n ( M , x ) .

N o N Q
n n
For any n > 0 and any r e a l c l o s e d base f i e l d R the group H - ( I , 3 I ) i s
n
n ]
i s o m o r p h i c to 2. We choose a c o h e r e n t system o f g e n e r a t o r s Z € H ( I , 3 I n n
as i n d i c a t e d i n [Spa, p. 388], { Z n maps t o the fundamental c l a s s o f the

n
complete s e m i a l g e b r a i c m a n i f o l d 9I with i t s standard o r i e n t a t i o n under
n n n
the isomorphism 9 from H ( I , 3 I )
n t o H^^^ O l ) p r o v i d e d n>1.}
Given a p a i r (M,A,X ) o f pointed
q spaces we d e f i n e a Hurewicz-map
(n> 1, or n = 0 and A = ( X } ) Q
cp : T T ( M , A , x ) -
N o H (M,A)
N
n n
by cp[f] := f ^ ( Z ) , n f o r e v e r y map f : (I , 3l , ) -+ ( M , A , X ) .
q This map
i s well defined s i n c e f * (Z ) depends o n l y on the homotopy class of f ,

n n
in f a c t o n l y on the f r e e homotopy class of f : ( l , 9 l ) -» (M,A) .
Clearly the map cp depends on the t r i p l e (M,A,x ) Q in a functorial
way. Copying arguments i n [Spa, V I I , §4] we obtain
Theorem 7.3. a) I f n> 2, o r i f n = 1 and A {x },

Q the Hurewicz map
i s a group homomorphism.
b) I f n > 1, t h e square
n (M,A,x )
n Q n ( A x )
n-1 ' o
H (M, A) H ( A )
n ' ' n-1
commutes.
Since the image cp[f] o f an element [ f ] € i T ( M , A , x ) only n o depends on t h e
f r e e homotopy c l a s s of f , i t i s evident that f o r any path u i n A from
a point X q t o a p o i n t x^ t h e t r i a n g l e
u
TT (M,A,x )
n 1 * ^n (M,A,x )
n o
commutes (n> 1 ) . S i m i l a r l y , f o r any p a t h v i n M from a p o i n t x Q to a
p o i n t x^ t h e t r i a n g l e
M X ( M x )
V ' 1> — ^% ' o
commutes ( n > 0 ) . Here we have i d e n t i f i e d b o t h H ( M , x ) and H (M,x.j)
N Q N
with H (M) i n t h e c a n o n i c a l way.
If n>2 t h e n we denote by T T ^ (M,A,X ) the q u o t i e n t o f T T ( M , A , x ) N Q by
the normal subgroup g e n e r a t e d by a l l elements u (x)x with x running

#
through T T ( M , A , x ) and
N Q [u] r u n n i n g through 7 T ^ ( A x ) . The f i r s t
f Q triangle
above t e l l s us, i n t h e c a s e = x , t h a t cp i n d u c e s a group homomorphism

+
tp : T T * ( M , A , X ) q H^ (M, A) (n>2).
Similarly, i f n>1, then we d e n o t e by T T * ( M , X )

q the quotient of T T ( M , x ) N Q
by t h e normal subgroup g e n e r a t e d by a l l elements v (x)x with

#
x £ T T ( M , x ) , [v] € TT-J (M,x ) . N o t i c e t h a t TT^ ( M , X ) i s j u s t the f a c t o r
N Q Q
commutator group o f TT ^ ( M , X ) . The q second t r i a n g l e above shows f o r
x Q = x>| t h a t the Hurewicz map cp i n d u c e s a group homomorphism

cp* : TT* (M,x ) -> H ( M )
N o N (n>1).
Theorem 7.4 ( A b s o l u t e and r e l a t i v e Hurewicz theorem).
a) L e t n > 1 . Assume t h a t n ( M , x )R Q = 0 f o r 0 < r < n - 1 . Then the map cp*
from TT * (M,x ) t o H_ (M) i s an isomorphism,

n o n
b) L e t n > 2 . Assume t h a t M and A a r e connected and t h a t T T ( M , A , x ) = 0 R Q
+
for 1 < r < n - 1 . Then the map cp from T T ^ ( M , A , X ) t o H
q n (M, A) i s an
isomorphism.
N.B. P a r t a) t e l l s us i n t h e c a s e n= 1 t h a t , f o r every c o n n e c t e d space
M, t h e f a c t o r commutator g r o u p o f n ( M , X ) i s c a n o n i c a l l y i s o m o r p h i c
1 q
to H^ (M) . I f n > 2 t h e n , under t h e assumptions t h e r e , T T ^ ( M , X ) = T T ( M , x ) . Q N o
We o b t a i n Theorem 7.4 by s t a r t i n g from t h e well-known Hurewicz theorems
in topology ([Spa, p. 3 9 7 f ] , [W, p. 178]) and u s i n g our main theorems
i n homotopy (Th. 6.3, Th. 6.4) and homology (Th. 7.2) together with the
f o l l o w i n g obvious f a c t s :
1) L e t R be a r e a l c l o s e d f i e l d c o n t a i n i n g R and (M,A,x ) a p a i r of
o
p o i n t e d spaces over R. Then the square
n (M,A,x )
n o •H (M,A)
n
n (M(R),A(R),X )
n Q •H (M(R) ,A(R))
n
commutes.
2) Assume t h a t R = 3R . I d e n t i f y i n g the s e m i a l g e b r a i c homotopy and homo-
l o g y groups w i t h t h e c o r r e s p o n d i n g homotopy and homology groups of
topological spaces, t h e s e m i a l g e b r a i c Hurewicz homomorphism
cp : TT (M,A,x ) - H (M,A)
n o n
is the same map as the t o p o l o g i c a l Hurewicz homomorphism.
Similarly, one can t r a n s f e r o t h e r r e s u l t s about the Hurewicz homomor-
phisms t o spaces over any r e a l closed f i e l d . In p a r t i c u l a r , one can
t r a n s f e r t h e f o l l o w i n g theorem due t o Hopf and Fox ( c f . [W, p.248]):
If n>2 and TT^ (M, x ) Q = 0 for 0<r<n-1, t h e n t h e Hurewicz map from
TT (M,x ) t o H . . (M)
i i s surjective.
n+i o n+1
S e m i a l g e b r a i c homology groups are q u i t e u s e f u l i n our homotopy t h e o r y .
This i s a l r e a d y e v i d e n t by the f o l l o w i n g complement t o Theorem 6.6.
It can be o b t a i n e d by t r a n s f e r from a well-known theorem o f J.H.C.
Whitehead ([Spa, p. 399], [W, p. 181]).
Theorem 7.5. L e t f : M->N be a map between connected spaces o v e r R. L e t
x_ be a p o i n t o f M and y :~ f ( x ). G i v e n a n a t u r a l number n we consi-

o o o
d e r t h e i n d u c e d homomorphisms
TT (f) : n ( M , X ) - T i ( N , y ) , H ( f ) : H
q q Q q o q g (M) ->H (N),
q
for 1 < q < n.

a) I f ^ ( f ) i s b i j e c t i v e for1<q<n and s u r j e c t i v e f o r q = n, t h e n
H^(f) i s bijective f o r 1 <q<n and s u r j e c t i v e f o r q = n.
b) Assume t h a t ( M , X ) = 0 and
Q (N,y ) = 0. Then, c o n v e r s e l y , i f
Q
H^(f) i s bijective for1<q<n and s u r j e c t i v e f o r q = n, t h e homo-
morphism TT ( f ) i s b i j e c t i v e for1 <q<n and s u r j e c t i v e f o r q = n.
I f t h e space M i s s e m i a l g e b r a i c then mapping cylinders are at our d i s -
p o s a l , and Theorem 7.5 i s an easy consequence o f t h e r e l a t i v e Hurewicz
theorem. But t h e theorem h o l d s i n g e n e r a l .

§8 - Homotopy groups o f ends
This section refers to t h e t h e o r y o f ends o f a space M d e v e l o p e d in
II, §9 and §11. G i v e n an a b s o l u t e end X o f M we want t o d e f i n e the
q - t h homotopy group o f X as the p r o j e c t i v e l i m i t o f the homotopy groups
TT ( (M ^ K) ^, x ) w i t h K r u n n i n g through
q the s e t (M) of p a r t i a l l y complete
(locally s e m i a l g e b r a i c ) s u b s e t s o f M and s u i t a b l e base p o i n t s x € (M^K)^
{Recall that (M^K)^ i s t h e connected component o f M ^ K determined by
X, c f . I I , §9, Def. 4}. We cannot choose a common base p o i n t i n a l l the
sets ( M ^ K ) ^ s i n c e they have an empty i n t e r s e c t i o n . The most natural
i d e a seems t o be to choose the base p o i n t s x on some p r o p e r incomplete
path a with e(a) = X ( c f . I I , §9, Def. 6 ) . T h i s l e a d s to t h e following
definition.
L e t a : [0,1[ -*M be a p r o p e r incomplete path and l e t X = e(a) be t h e end
determined by a. We c o n s i d e r the s e t A(a) c o n s i s t i n g o f a l l p a i r s ( K , t )
with K e r ( M ) ,c t € [ 0 , 1 [ , a ( [ t , l [ ) c M v K . We define a p a r t i a l ordering
on A(a) by
(K,t) < (L,u) * » K c L , t<u.
If (K,t) < (L,u) then a y i e l d s , f o r every q> 1, a n a t u r a l homomorphism
P 1 M L
(K^t) " ( ^ ' a ( u ) ) - n (M^K,cx(t)) ,
q Q
namely the map i n d u c e d by t h e i n c l u s i o n ( M ^ L , Q ( U ) ) -> ( M ^ K , a ( u ) ) and the path
s H> ((1-s) t + su) from a ( t ) t o a (u) . With t h e s e t r a n s i t i o n maps p / ^ ' | we U
{j\, u;
have, f o r e v e r y q > 1, an i n v e r s e system (TT (M ^ K, a (t)) I (K, t ) € A (a)) o f groups ,
si
Definition 1 . The q - t h homology group TT (X,a) o f the end X based at a
is the p r o j e c t i v e limit o f t h i s system ( q > 1 ) . In s h o r t ,
n (X,a) = lim TT ( M \ K , a ( t ) ) .
q c
(K,t)€A(a) q
These groups behave f u n c t o r i a l l y with r e s p e c t to p a r t i a l l y p r o p e r maps.

Indeed, i f f : M-* N i s a p a r t i a l l y p r o p e r map, then 3 := f°ot i s a p r o p e r
incomplete path i n N which determines the end f ( A ) ,+ c f . I I , § 9 . Every
index (L,s) G A ( 0 ) yields an index (f ( L ) , s ) €A(oc), and f i n d u c e s , f o r
every q > 1 , a homomorphism
These homomorphisms a r e c o m p a t i b l e w i t h the t r a n s i t i o n maps. Thus we
o b t a i n a unique homomorphism
such t h a t , f o r every (L,s) € A ( 3 ) , the square
TT (f*(A) , 3 )
q
TT (Mvf ' (L),a(s)) ? TT (NXL,3(s))

R
4 * 4
w i t h t h e c a n o n i c a l p r o j e c t i o n s as v e r t i c a l arrows, commutes.
How does t h e group TT (A ,a) depend on t h e c h o i c e o f the i n c o m p l e t e

4
path a " i n " t h e end X? Assume t h a t a and 3 a r e two p r o p e r incomplete
paths i n M and t h a t H: [0, 1 [ *I M i s a p r o p e r homotopy from a to 3/
i . e . H i s a p r o p e r map w i t h H(-,0) = a and H(-,1) = 3. F o r any K G T ^ M )

1
the preimage H ( K ) i s a complete s e m i a l g e b r a i c s u b s e t o f [ 0 , 1 [ * I . Thus
1
H (K) i s c o n t a i n e d i n [ 0 , c ] * I f o r some c € ] 0 , 1 [ . F o r e v e r y t € ] c , 1 [
the path s»->H(t,s) runs i n M^K from t h e p o i n t a ( t ) t o the p o i n t 3 ( t ) .
We see t h a t a and 3 determine t h e same end A o f M. Moreover, f o r every
t € ]c,1 [ and q > 1, we have a homomorphism
If u > t i s a second parameter i n ] c , 1 [ then t h e square
H(u,-) #
TT (M SK , CX (u))
(K,u)
P
(K,t)
TT (M<K,a(t))
commutes, s i n c e t h e v e r t i c a l arrows a r e j u s t t h e maps 6 induced
Tr TT
by t h e paths
T(s) = H ( ( 1 - s ) t + s u , 1 ) , 6(s) = H( ( 1 - s ) t + s u,0)
and s i n c e H maps t h e whole r e c t a n g l e [ t , u ] x i i n t o M ^ K . I t i s now e v i -
dent t h a t t h e homomorphisms H(t,-) f o r t h e v a r i o u s K € T (M) and admit-

# c
t e d t € ]0,1[ a l l f i t t o g e t h e r and y i e l d a w e l l d e f i n e d homomorphism
H : TT ( A , 3 ) -* TT (A,a) .
I f <X> : [ 0 , 1 [ x i x i -> M i s a p r o p e r homotopy r e l a t i v e [0,1[*{0,1} from H t o
a second p r o p e r homotopy H 1
: [0,1[xi -» M from a t o 3 then H = H' as i s
easily verified.
I f G : [ 0 , 1 [ x l -» M i s a p r o p e r homotopy from 3 to a t h i r d p r o p e r incom-
p l e t e path y i n M , then t h e composite homotopy H*G i s a g a i n p r o p e r and
t h e homomorphisms H «G and (H*G) from TT (A,y) t c TT (A,a) a r e e q u a l .

# ft 4t 3 4
—1
Now we s e e by t h e u s u a l argument t h a t the i n v e r s e homotopy H z (t,s) *•*
H ( t , 1 - s ) from 3 t o a, which i s a g a i n p r o p e r , y i e l d s a homomorphism
(H from TT (A,a) t o TT ( A , 3 ) which i s i n v e r s e t o H... In p a r t i c u l a r

tr q q TT
H i s Bn isomorphism.
w
Definition 2. The fundamental g r o u p o i d TT (A) o f the end A o f M i s the c a t e -
gory whose o b j e c t s are the p r o p e r incomplete paths a : [0, 1 [ M w i t h e (a) = A
and whose morphisms from an o b j e c t a t o an o b j e c t 3 are the p r o p e r homotopy
c l a s s e s <H> r e l a t i v e [0,1 [ x {o, 1} o f proper homotopies H from a t o 3 .
N o t i c e t h a t e v e r y morphism i n t h e c a t e g o r y TT(A) i s an isomorphism, t h e
i n v e r s e o f <H> b e i n g t h e c l a s s <H Thus TT(A) i s indeed a g r o u p o i d .
We summarize o u r o b s e r v a t i o n s on the maps as f o l l o w s .
P r o p o s i t i o n 8.1. F o r every end A o f M and every q € ]N t h e assignment
a »•+ TT (A,a), <H> » i s a f u n c t o r from the fundamental g r o u p o i d TT(A)

to the category of groups.
The f o l l o w i n g theorem states in particular t h a t the fundamental groupoid
TT(X) i s connected, i . e . f o r any two o b j e c t s oc,3 i n TT(X) there exists
a morphism from a t o 3 . T h i s implies that, f o r given q > 1 , the groups
TTq(X,a), w i t h a r u n n i n g through t h e p r o p e r i n c o m p l e t e paths " i n " X, are
all isomorphic.
Theorem 8.2. Let a , 3 : [ 0 , 1 [3M be proper i n c o m p l e t e paths i n M d e t e r m i n -
ing the same end X. L e t M c P be a pure c o m p l e t i o n ( c f . I I , §9, Def. 2)
of M, and l e t a,"3 : [ 0 , 1 ] denote the unique e x t e n s i o n s of a, 3 t o
paths i n P. Then t h e r e e x i s t s a homotopy
H: (I, [0,1 [,{1}) xi - (P,M,P^M)
with H(-,0) = a,H(-,1) = 3. The r e s t r i c t i o n of H to [0,1[xi i s a proper
homotopy from a t o 3 .
P r o o f . Once we have c o n s t r u c t e d the homotopy H the l a s t sentence o f the

i
theorem w i l l be o b v i o u s s i n c e H (M) = [ 0 , 1 [ x i . Let A := P(X) be the
connected component o f P^M determined by X ( I I , 9.24). Then x := a ( 1 )
and y := 3 ( 1 ) a r e p o i n t s i n A. We choose a path Y : [ 0 , 1 ] A with y ( 0 ) =
and T ( 1 ) = y. Then we choose a simultaneous t r i a n g u l a t i o n of P and the
s e t s M, y ( [ 0 , 1 ] ) , a ( [ 0 , 1 ] ) , " 3 ( [ 0 , 1 ] ) F and we r e g a r d P as a c l o s e d l o -
cally finite complex w i t h t h e s e s e t s as subcomplexes.
The p o i n t s x and y a r e v e r t i c e s o f P. There exists a f i n i t e sequence
of vertices (x IO<i<n)
i in y ( [ 0 , 1 ] ) such t h a t X Q = X, x n = y and
[x _ ,x ]
i 1 i i s a closed 1-simplex of P contained i n y ( [ 0 , 1 ] ) for l£i<n.
For e v e r y i € { l , . . . , n } we choose an open simplex i n M which has
] x ^ _ ^ , x ^ [ as a f a c e . T h i s i s p o s s i b l e s i n c e M i s dense i n P.
H e n c e f o r t h we denote the s e t o f paths

6 : ( [ 0 , 1 ] , [0,1 [,{1}) - (P,M,PNM)
by F(P,M). F o r any two such paths 6^, 6^ we w r i t e 6^ ^ 6^ i f t h e r e
e x i s t s a homotopy from 5^ t o regarded as maps between t r i p l e s o f
spaces. In t h i s n o t a t i o n o u r c l a i m i s a ^ p.
If z i s a p o i n t o f M and w i s a p o i n t o f P ^ M such t h a t the open line
segment ]z,w[ i s c o n t a i n e d i n some open simplex o f M, then t h e l i n e a r
path from z t o w i s an element o f T(P,M). We denote t h i s path by zw. We
make t h e f o l l o w i n g simple but c r u c i a l o b s e r v a t i o n : I f z^,z^ a r e p o i n t s
i n t h e open s t a r S t ( x ^ ) f o r some i E (0,...,n)

M ( c f . I I , § 7 ) , then
z x
z^x^ ^ 2 i* Indeed, s i n c e the dense p a i r o f spaces (P,M) i s assumed
t o be pure t h e complex M i s c e r t a i n l y maximal ( c f . I I , §5, Def. 4).
Thus any two open s i m p l i c e s i n S t ( x ^ ) a r e c o n n e c t a b l e i n S t ( x ^ ) by

M M
a c h a i n o f open s i m p l i c e s i n t h e sense o f I I , §7, Def. 4 ( c f . I I , Prop.
7.3). I t i s then easy to w r i t e down a p i e c e w i s e l i n e a r homotopy from

z x
z^x i to 2 i*
a x f o r e v e r
In p a r t i c u l a r , o^x^ ~ i + 1 i y i £ { 1 , . . . , n - l } s i n c e the b a r y -
a
centers and o f o\ and i + -j are both lying i n S t ( x ^ ) . Obviously
M
x F O R 1 G
we a l s o have ^ ^ - i - î i (1,...,n).
There e x i s t s a unique v e r t e x p i n t h e one d i m e n s i o n a l subcomplex
a([0,1[) o f P such that Jp,x[ i s an edge o f P. L e t a denote the mid-
point o f ]p,x[. This i s a point in St (x). M It i s easily verified that
a ^ ax. S i m i l a r l y 3 ^ by f o r a s u i t a b l e p o i n t b £ S t ( y ) . M By the o b s e r -
v a t i o n above we have ax ^ cLx and by c* 5 x . Thus a l t o g e t h e r we have

2 r
1 o n n
a 3. q.e.d.
We r e t u r n to the d i s c u s s i o n o f t h e f u n c t o r i a l i t y o f t h e groups TT (A,or).
Let f , g : M 4 N be two p a r t i a l l y proper maps and l e t F : Mxi -*• N be a
partially p r o p e r homotopy from f to g. F o r every p r o p e r incomplete path

a i n M t h e map
H := F« ( a x i d ) : [0, 1 [ x i N
i s a p r o p e r homotopy from f«a t o g»a. Thus, i f a determines t h e end X
of M, then f»a and g«a determine t h e same end u = f*(X) = g ( X ) o f N.

+
Moreover, we have t h e f o l l o w i n g proposition.
P r o p o s i t i o n 8.3. The t r i a n g l e
Ti^(y,f «a)
commutes.
The proof may s a f e l y be l e f t t o t h e reader. This p r o p o s i t i o n implies
the i n v a r i a n c e o f t h e groups TT^(X,a) under p a r t i a l l y p r o p e r homotopy
equivalences.
We want t o compare t h e homotopy group TTg(X,a) w i t h t h e homotopy group
n (P (X) ,tp°a (1) ) f o r a g i v e n pure c o m p l e t i o n cp : M -* P. {Again P(X) de-
notes t h e end o f M i n P determined by X,and cp°a denotes t h e e x t e n s i o n
of q)«a t o a path i n P.) We a r e f o r c e d t o assume t h a t P(X) has c o m p l e x i t y
<1, i . e . i s l o c a l l y complete. N o t i c e that t h i s condition only depends
on X b u t not on t h e chosen pure c o m p l e t i o n tp, c f . 11.9.25.
We r e g a r d <p as an i n c l u s i o n . The s e t P(X) i s open i n i t s c l o s u r e P(X)
i n P. Thus t h e s e t 3P(X) = P (X) \ P (X) i s a member o f T' (M) . I f some

Q
K € T (M)c contains 3P(X) then P(X) i s c l o s e d i n P \ K . By
II, Theorem 11.7, t h e r e e x i s t s a s e t L C T ^ M ) w i t h L=>K such t h a t t h e
c l o s u r e U = U \ 3 P ( X ) i n P ^ 9 P ( X ) o f the connected component U o f P \ L

c o n t a i n i n g P ( A ) i s a "mapping c y l i n d e r n e i g h b o u r h o o d " o f P ( A ) . T h i s
means, t h e r e e x i s t s a p r o p e r map h : 3U -+ P ( A ) , cfu : = 8 P ( A ) , such
t h a t the t r i p l e (U, 8 u , P (A) ) i s i s o m o r p h i c t o (Z (h) , (d\J) *0,P ( A ) ) w i t h
Z(h) the mapping c y l i n d e r o f h. L e t t € [0,1[ be a parameter w i t h
a ( [ t , 1 [ ) c M \ L . Since (U,P(A)) i s isomorphic t o (Z (h) \ ( 8 U ) xO, P (A) )
the n a t u r a l map (q>1)
T T ( P ( A ) ,a(1)) -
q TT (U,a(1)) = TT (P
q L, a (1) )
i s an isomorphism. Composing the i n v e r s e o f t h i s map with
Tt (M^ L,a(t) )
q TT (P^L,a(t)) -^TT
q (P^L,a(1) ) ,
— -1
where t h e l a s t arrow i s ( a l t t / 1 ] ) ^ , we o b t a i n , f o r e v e r y q > 1 , a
natural homomorphism
a
(L,t) :
" ( q
M N k
L a(t) ) -
f TT (P(A),a(1)) .
q
L e t A (a) denote the s e t o f a l l p a i r s

1
( L , t ) € A ( a ) w i t h L h a v i n g the
above p r o p e r t i e s . A ( a ) i s c o f i n a l 1
i n A ( a ) . Thus, f o r e v e r y q > 1 , t h e
group T T ( A , a ) i s the p r o j e c t i v e
q l i m i t o f t h e groups TT (M ^ L, a (t) ) w i t h
(L,t) € A ' ( a ) . Now i t i s easily seen t h a t t h e t r a n s i t i o n maps between
t h e s e groups a r e c o m p a t i b l e w i t h the homomorphisms ^ . Thus the
ofT .v yield a n a t u r a l homomorphism

\±j, V.)
o : T T ( A , a ) -> TT (P (A) ,a(1)) .
q q
We abandon t h e i n t e r p r e t a t i o n o f cp as an i n c l u s i o n map and denote this
homomorphism by
a
cp : T T
q ( A
' a )
~* * ( P ( * ) rW ® d ) ) •
q
7
If : M -> Q i s a second pure c o m p l e t i o n o f M and f : Q P i s a morphism from
to cp t h e n f restricts t o a p r o p e r s u r j e c t i v e map g : Q ( A ) ~» P ( A ) . I t
is easily checked t h a t , f o r every q > 1 , the t r i a n g l e

TT ( Q (X) ,i|/«a(1))
Si
(8.4)
n (P(X) ,<p-a( 1) )
Si
commutes.
We recall t h a t every c o m p l e t i o n o f M i s dominated by a c o m p l e t i o n M«-*P
such t h a t the end P(X) in P i s collared ( I I , Th. 11.6). Assume t h a t our
g i v e n c o m p l e t i o n tp : M ^ P i s o f t h i s s p e c i a l t y p e . Then t h e r e e x i s t s ,
as i s e a s i l y seen, a c o f i n a l subset A " ( a ) o f A ' ( a ) such t h a t , f o r every
(L,t) € A " (a) , the connected component U o f P(X) i n P^(p(L) i s an "open
c o l l a r " o f P(X). T h i s i m p l i e s that o (T .* i s an isomorphism. We con-
elude
P r o p o s i t i o n 8.5. I f the end P(X) of M i n P i s c o l l a r e d , then
a
1.
Now we a r e w e l l p r e p a r e d t o g i v e an a p p l i c a t i o n o f our t h e o r y o f homo-
topy groups o f ends.
Theorem 8.6 (Homotopy i n v a r i a n c e o f c o l l a r e d ends) . L e t <p : M P and
\\) : M ** Q be c o m p l e t i o n s o f M such t h a t both ends P ( A ) and Q ( A ) deter-
mined by a g i v e n a b s o l u t e end A o f M i n P and Q are c o l l a r e d . Then the
spaces P(A) and Q ( A ) a r e homotopy e q u i v a l e n t .
P r o o f . There e x i s t s a completion x • M S dominating cp and \p such that
S(A) i s c o l l a r e d . Thus i t s u f f i c e s to c o n s i d e r t h e c a s e t h a t t h e r e
e x i s t s a morphism f : Q -r* P from \\> t o (p. We choose a p r o p e r incomplete

path a i n M w i t h e (a) = X . In t h e commutative diagram (8.4) b o t h maps
and a r e isomorphisms by P r o p o s i t i o n 8.5. Thus g* i s an i s o m o r -
phism f o r every q >_ 1 . By Whitehead's theorem 6.6 the map g : Q ( X ) P(X)
i s a homotopy e q u i v a l e n c e . q.e.d.
In t h e c o u r s e o f t h i s study i t t u r n e d o u t t h a t i t sometimes i s n a t u r a l
to c o n s i d e r p a r t i a l l y proper homotopy c l a s s e s o f p a r t i a l l y p r o p e r maps.
I t would be d e s i r a b l e t o have analogues of our two main theorems on
homotopy c l a s s e s f o r p a r t i a l l y proper homotopy c l a s s e s . Up t o now we
do n o t know how t o prove such theorems.

Appendix A ( t o C h a p t e r I ) : A b s t r a c t l o c a l l y semialgebraic spaces.
T h i s appendix i s meant o n l y f o r those r e a d e r s who a r e f a m i l i a r with
the n o t i o n o f t h e r e a l spectrum S p e r A o f a commutative r i n g A, c f .
[CR] and t h e l i t e r a t u r e c i t e d t h e r e , and f o r an i n t r o d u c t i o n t o r e a l
spectra also [BCR, Chap. 7] and p a r t s o f the a r t i c l e s [L], [ B r ] , [K].
We r e g a r d h e r e Sper A as a t o p o l o g i c a l space, n e c r l e c t i n g i t s u s u a l
s t r u c t u r e sheaf o f " a b s t r a c t Nash f u n c t i o n s " . In f a c t we w i l l introduce
below a new - much b i g g e r - s t r u c t u r e sheaf on S p e r A , the sheaf o f
"abstract semialgebraic functions".
The appendix i s n o t n e c e s s a r y f o r an u n d e r s t a n d i n g o f the other parts
of t h i s book i n any t e c h n i c a l sense. Our g o a l i s t o e x p l a i n how t h e
category of l o c a l l y semialgebraic spaces over a r e a l c l o s e d f i e l d R
fits i n t o t h e framework o f " a b s t r a c t " s e m i a l g e b r a i c geometry, a now
rapidly developing a r e a . I n p a r t i c u l a r we w i l l see t h a t our l o c a l l y
semialgebraic spaces a r e o b j e c t s e q u i v a l e n t t o s u i t a b l e l o c a l l y ringed
spaces over genuine - a l b e i t n o t H a u s d o r f f - t o p o l o g i c a l spaces. This
view p o i n t l e a d s t o a b e t t e r u n d e r s t a n d i n g o f v a r i o u s examples o f
locally s e m i a l g e b r a i c spaces g i v e n i n Chapter I . We w i l l see t h a t ,
although some o f t h e s e spaces l o o k p a t h o l o g i c a l a t f i r s t glance, they
n e v e r t h e l e s s a r e e q u i v a l e n t t o subspaces o f very honest r e a l spectra
and thus may be regarded as " o c c u r i n g i n n a t u r e " .
L e t V be an a f f i n e a l g e b r a i c v a r i e t y over a r e a l c l o s e d f i e l d R. As
u s u a l we denote by R[V] t h e r i n g o f a l g e b r a i c f u n c t i o n s (= "polynomial
f u n c t i o n s " ) on V and by V(R) t h e s e t o f R - r a t i o n a l p o i n t s o f V. The
s e t V(R) i s a subset o f t h e r e a l spectrum Sper R[V] , and t h e subspace
topology o f V(R) i n Sper R[V] c o i n c i d e s with the usual strong topology
of V(R). There i s a c a n o n i c a l 1-1-correspondence between t h e semialge-
b r a i c subsets o f V(R) and t h e c o n s t r u c t i b l e subsets o f Sper R[V] .

Namely, i f Mcrv(R) i s s e m i a l g e b r a i c , and i f we d e s c r i b e M by equalities
and i n e q u a l i t i e s between f u n c t i o n s i n R[V], then t h e s e e q u a l i t i e s and
i n e q u a l i t i e s a l s o d e f i n e a c o n s t r u c t i b l e subset M o f S p e r R [ V ] . The
set M i s the unique c o n s t r u c t i b l e subset o f Sper R[V] w i t h MDV(R) = M,
cf. [CR, § 5 ] . I f M i s a one p o i n t set {x}, then M = M.
Now we consider a fixed s e m i a l g e b r a i c subset M o f V ( R ) . A s e m i a l g e b r a i c
subset X o f M i s open ( r e s p . c l o s e d ) i n M i f and o n l y i f X i s open
(resp. c l o s e d ) i n M [CR, § 5 ] . Moreover t h e open c o n s t r u c t i b l e subsets
of M are p r e c i s e l y t h o s e open s u b s e t s which are quasicompact. Thus the
f a m i l y T(M) o f open s e m i a l g e b r a i c subsets and then a l s o the f a m i l y 2T(M)
of a l l semialgebraic subsets of M i s completely determined by the topo-
logy o f M.
The p o i n t s o f M can be i n t e r p r e t e d as the u l t r a f l i t e r s o f the Boolean
lattice <T(M) i n a c a n o n i c a l way, c f . [Br, p. 260], [CaC, § 1 ] . Then the
c o n s t r u c t i b l e subset X of M corresponding to a s e m i a l g e b r a i c subset X
of M i s j u s t the s e t of a l l u l t r a f i l t e r s F with X € F . In particular
(X = {x}) a p o i n t x o f M i s i d e n t i f i e d w i t h the p r i n c i p a l ultrafilter
F x generated by x.
The open quasicompact s u b s e t s of M are a b a s i s o f t h e t o p o l o g y o f H
[CR, §5] and, s i n c e t h e space M i t s e l f i s quasicompact, every c o v e r i n g
of M by open s u b s e t s has a f i n i t e s u b c o v e r i n g . Thus i t i s e v i d e n t t h a t
t h e r e i s a 1-1-correspondence F «-» F between the sheaves F on the semi-
a l g e b r a i c space M and t h e sheaves F on the t o p o l o g i c a l space M, given
by the formula F(U) = F(U) w i t h U running through £(M) . In p a r t i c u l a r
the sheaf (9^ o f s e m i a l g e b r a i c f u n c t i o n s on M c o r r e s p o n d s to a sheaf of
r i n g s £> M on M. The pair (M,©^) i s a l o c a l l y r i n g e d space i n the u s u a l
t o p o l o g i c a l sense. I t i s an example o f an a b s t r a c t s e m i a l g e b r a i c space,
to be d e f i n e d now.
L e t A be an a r b i t r a r y commutative ring (with 1) and K be a c o n s t r u c t -
ible s u b s e t o f S p e r A . R e c e n t l y t h e r e has been i n t r o d u c e d a sheaf £> K
o f " a b s t r a c t s e m i a l g e b r a i c f u n c t i o n s " on K, c f . [B^, § 3 ] , [ S c h ] , [D^,
§ 1 ] . The s e c t i o n s o f t h i s sheaf may be c o n s i d e r e d as " f u n c t i o n s " which
are c o n t i n u o u s and whose graphs a r e c l o s e d and c o n s t r u c t i b l e , c f . [D^,
§ 1 ] . In t h e case A = R[V], K = M from above,the s h e a f (0 coincides with
2^. So t h e s e f u n c t i o n s a r e the a b s t r a c t analogues o f the s e m i a l g e b r a i c
f u n c t i o n s on s e m i a l g e b r a i c spaces over r e a l c l o s e d f i e l d s [D^, 1.9].
We call a l o c a l l y r i n g e d space (K,0 ) as above a s e m i a l g e b r a i c

K subspace
o f S p e r A . I f one d e s i r e s to e s t a b l i s h an " a b s t r a c t " semialgebraic geo-
metry, as people now do, then i t i s q u i t e n a t u r a l t o s t u d y locally
r i n g e d spaces which l o c a l l y l o o k l i k e s e m i a l g e b r a i c subspaces o f r e a l
spectra.
Definitions 1 [D , 2 I , § 1 ] . a) An a b s t r a c t a f f i n e s e m i a l g e b r a i c space i s
a l o c a l l y r i n g e d space which i s i s o m o r p h i c t o a s e m i a l g e b r a i c subspace
of a r e a l spectrum.
b) An a b s t r a c t l o c a l l y s e m i a l g e b r a i c space i s a l o c a l l y r i n g e d space
(X,(9 ) which has an open c o v e r i n g

X ( X l a € I ) such t h a t ,
a f o r every a € I ,
the space (X.(9 |X V ) i s an a b s t r a c t a f f i n e s e m i a l g e b r a i c space and the

0» A Of
intersection X fl n i s quasicompact f o r any two indices a,0€I. {We
are o n l y interested i n "quasiseparated" spaces.) I f , i n a d d i t i o n , the
t o p o l o g i c a l space X i s quasicompact then (X,(0 ) i s c a l l e d an

X abstract
s e m i a l g e b r a i c s p a c e . S i n c e the X a above a r e quasicompact, this just
means t h a t the i n d e x s e t I can be chosen finite.
c) A l o c a l l y s e m i a l g e b r a i c map between a b s t r a c t locally semialgebraic
spaces (X,(9 ) and

X (Y,(9 ) i s a morphism
V (f,<7) i n t h e c a t e g o r y o f locally
r i n g e d spaces, i . e . f : X-»Y i s a c o n t i n u o u s map and J : f * © -> 0^

Y is a
homomorphism o f sheaves o f r i n g s such t h a t , f o r e v e r y x € X, t h e homo-
morphism : ® Y f ( x ) -*© x x is local. I f the s p a c e s (X,© )x and (Y,^ ) y

a r e s e m i a l g e b r a i c then, o f course,we c a l l such a morphism ( f m o r e
briefly a s e m i a l g e b r a i c map.
The l o c a l l y s e m i a l g e b r a i c spaces and maps o v e r a r e a l c l o s e d field,
as d e f i n e d i n I , §1 w i l l h e n c e f o r t h o f t e n be c a l l e d g e o m e t r i c locally
s e m i a l g e b r a i c spaces and maps i n o r d e r to d i s t i n g u i s h them from t h e
a b s t r a c t spaces and maps d e f i n e d now. The r e a d e r w i l l p e r c e i v e the
c l o s e analogy between the d e f i n i t i o n s here and i n I, § 1 .
Abstract locally s e m i a l g e b r a i c spaces have f i r s t been d e f i n e d by
N . Schwartz [Sch]. He c a l l e d these spaces "real closed spaces".
We g i v e two examples o f a b s t r a c t l o c a l l y s e m i a l g e b r a i c maps.
Examples A1 . i ) L e t cp : A-> B be a r i n g homomorphism. cp i n d u c e s a con-
t i n u o u s map f = Sper cp : Sper B -> Sper A , c f . [CR] . L e t (0 and

oper A
e n o t e t n e
^Sper B ^ sheaves o f a b s t r a c t s e m i a l g e b r a i c f u n c t i o n s on
SperA and S p e r B . "Composition w i t h f " y i e l d s a homomorphism

f o f s n e a v e s o f
^SperA ~* * ^ S p e r B r i n g s , c f . [D^, 1 . 8 ] , hence a homo-
morphism <T: *^sper A f

~* ^ S p e r B * T n i s
homomorphism J i s l o c a l on the
s t a l k s . Thus (f,2) i s a s e m i a l g e b r a i c map between t h e s e m i a l g e b r a i c

spaces (Sper A ,<° ) and (Sper B ,(9 ) . We denote t h i s morphism
oper A oper a
by Sper cp .
ii) Let ( f , J ) : (M,0^) -* ( N , O ) n be a semialgebraic map ( i n the sense
o f I , § 1 ) between s e m i a l g e b r a i c subspaces (M,(9 ) and (N,<C> ) o f a f f i n e

M N
v a r i e t i e s V and W over some r e a l c l o s e d f i e l d R. We c o n s i d e r the semi-
a l g e b r a i c subspaces (M (D^)
f f (Sf,©) o f Sper R[V]
N and Sper R[W] corres-
ponding t o (M,(9 ) and (U,®^) . f y i e l d s a map f from M t o Sf as f o l l o w s :

M
I f F i s a p o i n t o f H, i . e . an u l t r a f i l t e r i n r ( M ) , then ^ ( F ) i s t h e
filter i n 2T(N) generated by t h e s e t s f ( X ) w i t h X r u n n i n g through F. I t
t u r n s o u t t h a t f ( F ) i s a g a i n an u l t r a f i l t e r [Br, p. 2 6 1 ] . I t i s e a s i l y
checked t h a t t h i s map f : M-*N i s continuous. F o r any X € T(N) we have
f (X) = f (X) . F o r any sheaf T on the s e m i a l g e b r a i c space N we have

r
f * ( 3 ) ~ = f * (T) . The homomorphism ^ : f * (£> ) -*^ corresponds N M t o a homo-
morphism J : f * o f sheaves o f r i n g s over M which t u r n s o u t t o
be l o c a l on t h e s t a l k s . Thus o u r s e m i a l g e b r a i c map (f,J) : (M,(9 ) -> (N,<9^)
y i e l d s an a b s t r a c t s e m i a l g e b r a i c map (f ,2b : (M,©^) -> (N,??^) .
It i s e a s i l y seen t h a t themorphisms (f ,Z) i n t h i s l a s t example A . l . i i
are p r e c i s e l y a l l a b s t r a c t s e m i a l g e b r a i c maps from (M,(0jf) t o (N,0~) .
T h i s i m p l i e s t h a t we have an isomorphism (M,C? ) •* (M,Oyj) , (f ,^)

M (f,J)
from t h e c a t e g o r y o f a f f i n e s e m i a l g e b r a i c spaces over R t o a f u l l sub-
c a t e g o r y o f the c a t e g o r y o f a b s t r a c t a f f i n e s e m i a l g e b r a i c s p a c e s . I t
is an easy matter t o prove t h a t t h i s e q u i v a l e n c e extends i n a u n i q u e
way t o an isomorphism (M,(9 ) •+ (M,0~) ,

M (f,J) ( f f r o m the category
of a l l l o c a l l y s e m i a l g e b r a i c spaces over R t o a f u l l subcategory o f the
category of a b s t r a c t l o c a l l y s e m i a l g e b r a i c spaces. We c a l l (M,0g) t h e
a b s t r a c t i o n o f t h e l o c a l l y s e m i a l g e b r a i c space (M,©^) and (f,30 the
abstraction of the l o c a l l y s e m i a l g e b r a i c map (f ,<7).
Many o f the f a c t s about (M,©~) f o r M a f f i n e s e m i a l g e b r a i c , m o s t l y men-
t i o n e d above, g e n e r a l i z e r e a d i l y ( i n a s u i t a b l e way) t o t h e case t h a t
M i s locally s e m i a l g e b r a i c . The p o i n t s o f M can be i n t e r p r e t e d as
those u l t r a f i l t e r s o f t h e Boolean l a t t i c e 3"(M) o f l o c a l l y semialgebraic
s u b s e t s o f M which c o n t a i n s e m i a l g e b r a i c s e t s . The p o i n t s o f M a r e i d e n -
t i f i e d w i t h the p r i n c i p a l u l t r a f i l t e r s , and thus M i s a s u b s e t o f M i n
a n a t u r a l way. i s a sheaf o f R-algebras s i n c e (J> i s a sheaf o f R-

M
a l g e b r a s . The p o i n t s o f M a r e p r e c i s e l y those x € M where t h e r e s i d u e
class f i e l d k(x) o f <9, c o i n c i d e s w i t h R. F o r t h e o t h e r p o i n t s x € M

X
M, x
the f i e l d s k(x) a r e r e a l c l o s e d f i e l d s c o n t a i n i n g R. The p o i n t s o f M
are c l o s e d i n M, b u t u s u a l l y M c o n t a i n s a l s o o t h e r c l o s e d p o i n t s .
The subspace t o p o l o g y o f M i n M i s j u s t the s t r o n g t o p o l o g y ( c f . I , §3)
on M. F o r e v e r y open s e m i a l g e b r a i c s u b s e t U o f M t h e r e e x i s t s a u n i q u e
quasicompact open s u b s e t U o f M w i t h UOM = U. More g e n e r a l l y , f o r
every open l o c a l l y s e m i a l g e b r a i c subset U of M there e x i s t s a unique
retrocompact open s u b s e t U o f M w i t h UDM = U. (A s u b s e t Z o f a topolo-
g i c a l space X i s c a l l e d r e t r o c o m p a c t i n X i f Z fl V i s quasicompact f o r
every quasicompact open s u b s e t V o f X [EGA I * . 0 . 2 . 3 ] ) . In t h i s way we
o b t a i n a b i j e c t i o n U »-» U from t h e s e t f(M) t o the s e t o f a l l r e t r o c o m -
p a c t open s u b s e t s o f M, which maps t(M) onto the s e t o f a l l quasicompact
open s u b s e t s o f M. The b i j e c t i o n U U preserves f i n i t e intersections
and unions (in fact "locally finite" i n t e r s e c t i o n s and u n i o n s ) , The
points of U are p r e c i s e l y the u l t r a f i l t e r s F € M w i t h U€F.
As f o r any a b s t r a c t l o c a l l y s e m i a l g e b r a i c space the quasicompact open
subsets o f M a r e a b a s i s o f t h e t o p o l o g y of M. Thus i t i s a g a i n c l e a r
t h a t we have an isomorphism F •* F from the c a t e g o r y o f sheaves F on the
g e n e r a l i z e d t o p o l o g i c a l space M ( c f . I , §1, Def. 1) t o t h e c a t e g o r y o f

= C7
sheaves on the t o p o l o g i c a l space M. Of c o u r s e , ® M g*
A sheaf Cf on M i s - more o r l e s s - determined by i t s f a m i l y of s t a l k s
(ÎxCM). But t h e r e a r e sheaves o f a b e l i a n groups 7 on M which a r e not

D
zero w h i l e a l l s t a l k s 3^, x € M, a r e z e r o , c f . t 2 ' Example 1.1.7], Thus
i f you want t o argue s t a l k by s t a l k , as i t i s o f t e n c o n v e n i e n t i n sheaf
t h e o r y , then u s u a l l y you have t o work on M i n s t e a d o f M ( c f . [D.j ],
[ D ] f o r some s t r i k i n g
2 examples).
If ( M l a € I ) i s the f a m i l y o f c o n n e c t e d components of M
a ( I , §3) t h e n , as
i s e a s i l y seen, ( K ^ l a E I ) i s t h e f a m i l y o f connected components o f t h e
t o p o l o g i c a l space H .
Let M denote the t o p o l o g i c a l subspace o f H c o n s i s t i n g o f the c l o s e d

p o i n t s o f M. There i s a c l o s e c o n n e c t i o n between some o f t h e p r o p e r t i e s
of (M,0 ) s t u d i e d i n I, §4-§7 and c e r t a i n

M standard properties of the
t o p o l o g i c a l space (which may h o l d o r n o t h o l d ) . We g i v e examples
of t h i s connection. The statements a r e a l l p r o v e n i n [D2, Chap. I , §4,
§ 5 ] . As i s u s u a l i n topology the p r o p e r t i e s "compact", "locally compact",
"paracompact" a r e u n d e r s t o o d t o i n c l u d e "Hausdorff".
m a x
Examples A.2. i ) I f (M,0 ) i s r e g u l a r then M M i s H a u s d o r f f . The con-
verse i s true i f (M,t9 ) i s assumed t o be t a u t

M [ D / 1.4.1].
2
1113
ii) (M,©^) i s r e g u l a r and t a u t i f and o n l y i fM ^ i s locally compact
[D , 2 1.4.5].
m a x
iii) (M,(9 ) i s a f f i n e s e m i a l g e b r a i c
M i f and o n l y i fM i s compact. This
i s an easy consequence o f (ii).
iv) (M,(9 ) i s r e g u l a r and paracompact i f and o n l y

M i f i s paracom-
pact ( i n the t o p o l o g i c a l sense).
v) (M,(i9 ) i s connected i f and o n l y

M i fK m a x
i s connected. In general,
if (M lcxei)a i s t h e f a m i l y o f c o n n e c t e d components o f M, then
M nM a
m a x
= M™ a x
since S i s closed i n M. Thus (M™ a x
|a€I) i s the
f a m i l y o f connected components o f M m a x
.
These examples show t h a t H m a x

i s a w e l l behaved H a u s d o r f f space under
mild c o n d i t i o n s on (M,G? ). M (N.B. These a r e o n l y c o n d i t i o n s on t h e f a m i -
lies o f s u b s e t s t(M), CT(M) o f t h e s e t M ) . The examples a l s o g i v e new
evidence that not "separated" but "regular" i s the reasonable separa-
t i o n property f o r geometric l o c a l l y semialgebraic spaces.
How c l o s e i s t h e c o n n e c t i o n between M and i t s subspace M m a x

? This sub-
space c o n t a i n s t h e s e t M . F o r every non empty open s e m i a l g e b r a i c sub-
s e t U o f M , t h e open subset U of M contains points of i ? n a x

, namely t h e
p o i n t s o f U. Thus M m a x
i s dense i n M. By t h e same argument M m a x
is
even "very dense" i n M i n t h e f o l l o w i n g s e n s e : The a b s t r a c t i o n X o f any

m a x
non empty s e m i a l g e b r a i c subset X o f M meets t h e s e t M .
Assume t h a t (M,©^) i s r e g u l a r . Then every s e m i a l g e b r a i c subset U o f M
is a f f i n e . This implies t h a t the s p e c i a l i z a t i o n s o f a p o i n t x G M form
a chain, s i n c e - as i s w e l l known [CR, p. 32] - the s p e c i a l i z a t i o n s
of a p o i n t in a real spectrum form a c h a i n . (A p o i n t y s p e c i a l i z e s x
if y€TxT [EGA I * , 0 . 2 . 1 ] ) . In p a r t i c u l a r , t h e r e i s a unique closed

m a x
point r(x) G M which s p e c i a l i z e s x. I t t u r n s o u t t h a t the map
r : M-^M max
i s c o n t i n u o u s i f and o n l y i f (M,© ) i s a l s o t a u t
M [D2, 1.4.3].
m a x
Thus f o r (M,© ) t a u t and r e g u l a r the
M space H i s a r e t r a c t o f M.
m a x
Since M i s o f t e n such an honest t o p o l o g i c a l space and i s very dense
i n M, why not replace the a b s t r a c t i o n (M,(9g) o f (M,(9^) by the locally

m a x m a x
ringed space (M , i*(9g) , w i t h i : M «^ M the i n c l u s i o n ? Doing this
we may run i n t o sheaf t h e o r e t i c d i f f i c u l t i e s s i n c e the i n t e r s e c t i o n

m a x
o f an open quasicompact subset o f M w i t h M i s u s u a l l y not quasicom-
m a x
pact. Thus we have t o c o n s i d e r in M "more" open c o v e r i n g s than i n M.
But there i s an even more s e r i o u s o b s t r u c t i o n : I f ( f , J ) i s the abstrac-
tion of a l o c a l l y semialgebraic map ( f , J ) : (M,(9^) -> (N,<9 ) , then N
m a x m a x
usually f will not map M into N . Thus the t o p o l o g i c a l space
does not depend i n a naive f u n c t o r i a l way on (M,(9^) . T h i s makes

rvmax , ~
M l e s s u s e f u l than M.
H e n c e f o r t h we u s u a l l y omit the s t r u c t u r e sheaf (9^ i n the notation (X,O )x
o f an a b s t r a c t l o c a l l y semialgebraic space and the second component ST
i n the n o t a t i o n (f ,<?") o f a l o c a l l y semialgebraic map, as we d i d w i t h
g e o m e t r i c s p a c e s and maps i n most p a r t s o f t h i s book. I n p a r t i c u l a r ,
f o r any commutative r i n g A, the n o t a t i o n SperA now means the real
spectrum o f A equipped w i t h the sheaf o f a b s t r a c t s e m i a l g e b r a i c func-
tions ( i n s t e a d o f the sheaf o f a b s t r a c t Nash f u n c t i o n s , although Nash
functions are v e r y important as soon as we q u i t the p u r e l y semialgebraic

s e t t i n g ) . We a l s o say more b r i e f l y "space" and "map" i n s t e a d o f " l o c a l -
ly s e m i a l g e b r a i c space" and " l o c a l l y s e m i a l g e b r a i c map".
Schwartz has proved i n [SchJ, among many o t h e r t h i n g s t h a t , f o r any

X X
two a b s t r a c t maps f^ : -* Y and f ^ : X2~>Y, the f i b r e p r o d u c t X^ Y 2
e x i s t s i n t h e c a t e g o r y o f a b s t r a c t spaces. We g i v e a d e s c r i p t i o n o f
the f i b r e product i n three s p e c i a l cases.
Examples A. 3. i ) L e t <p : C -»A and : C -* B be r i n g homomorphisms. Then
the f i b r e p r o d u c t o f t h e spaces S p e r A and S p e r B over S p e r C with
r e s p e c t t o t h e s e m i a l g e b r a i c maps Sper cp and Sper \p ( c f . Example A . 1 . i )
i s t h e space Sper A ® ^B, t h e f i b r e product p r o j e c t i o n s from Sper A® ^B
to SperA and S p e r B b e i n g the maps induced by t h e c a n o n i c a l r i n g homo-
morphism A -> A ® B and B -* A ® ^.B.

C
i i ) Let
M — - Mn
1 *1
be a c a r t e s i a n square i n t h e c a t e g o r y o f geometric spaces over a r e a l
closed f i e l d R ( c f . I, Prop. 3.5). Then
P1
M •RL
M 'N
"1
i s a c a r t e s i a n square i n t h e c a t e g o r y o f a b s t r a c t l o c a l l y s e m i a l g e b r a i c
spaces. In s h o r t ,
(M H X
1 "N«2'~ = 1 N*V
iii) L e t M be a g e o m e t r i c space over R and l e t S be a r e a l c l o s e d field

c o n t a i n i n g R. Then the f i b r e p r o d u c t o f M and Sper S with respect to
the o b v i o u s maps M -+ Sper R and Sper S Sper R i s the abstraction
M(S) o f t h e space M(S) o v e r S o b t a i n e d from M by base f i e l d extension
(I, 2.10). In s h o r t ,
K x S e r S =
Sper R P '
The image o f an u l t r a f i l t e r F G M(S) under the p r o j e c t i o n from M(S) to
M i s the u l t r a f i l t e r GGM c o n s i s t i n g o f a l l XG 7(M) w i t h X(S) G F.
As t h e s e examples show i t i s sometimes more s a t i s f y i n g from a f o r m a l
p o i n t o f view t o work i n the c a t e g o r y of a b s t r a c t l o c a l l y semialge-
b r a i c spaces i n s t e a d of the bunch of c a t e g o r i e s o f g e o m e t r i c spaces
over t h e v a r i o u s r e a l c l o s e d f i e l d s . We have a l s o i n d i c a t e d above t h a t
sheaf t h e o r y can be b e t t e r done over a b s t r a c t spaces i n s t e a d o f geome-
tric s p a c e s . At t h i s p o i n t the r e a d e r may suspect t h a t t h e p r e s e n t
a u t h o r s have w r i t t e n t h e wrong book! Why not abandon g e o m e t r i c spaces
a l t o g e t h e r and work o n l y i n the a b s t r a c t category?
The a b s t r a c t i o n M o f a g e o m e t r i c space M c o n t a i n s - i f dimM >0 - many
more p o i n t s than M. While t h i s f a c t makes M more amenable than M f o r
f o r m a l c o n s i d e r a t i o n s and f o r s h e a f t h e o r y , i t makes M u s u a l l y more
difficult than M as soon as c o n c r e t e geometric problems have t o be
solved.
For example, a homotopy between two geometric maps f ,g : M^N over R i s
a locally s e m i a l g e b r a i c map F : Mxi ->N. I t l i v e s on the c a r t e s i a n p r o -
duct o f t h e s e t M and the u n i t i n t e r v a l I i n R and fulfills the r e q u i r e
raents F(x,0) = f (x) and F(x,1) = g(x) f o r x G M (cf. I l l , § 1 ) . The ab-
s t r a c t i o n F o f F l i v e s on the f i b r e p r o d u c t Mx_ _ Y ( c f ; Example

Sper R
A.3.ii). Although SperR has o n l y one p o i n t t h i s f i b r e product contains
many more p o i n t s than the c a r t e s i a n p r o d u c t o f the t o p o l o g i c a l spaces M and I

m a x m a x
Even (M* I) i s b i g g e r than t h e c a r t e s i a n p r o d u c t o f M and
S]per R
r m a x
. Thus t h e g e o m e t r i c homotopy r e l a t i o n c o r r e s p o n d s t o a r e l a t i o n
between a b s t r a c t maps which i s more c o m p l i c a t e d than t o p o l o g i c a l homo-
topy. I t seems t h a t t h i s r e l a t i o n b e a r s much more meaning f o r t h e ab-
s t r a c t maps than t o p o l o g i c a l homotopy, s i n c e t h e r e s u l t s i n Chapter
III, i n p a r t i c u l a r t h e two main theorems i n i t s v a r i o u s v e r s i o n s
(III.3.1, 4.2, 5.1, 6.3, 6.4), g i v e overwhelming evidence t h a t we
have d e f i n e d homotopy between g e o m e t r i c maps i n the r i g h t way.
We f e e l t h a t most p r o o f s i n t h i s book cannot be s i m p l i f i e d by p a s s i n g
from t h e geometric spaces t o t h e i r a b s t r a c t i o n s . I t i s u s u a l l y hope-
l e s s t o u s e t h e s t a n d a r d theorems o f g e n e r a l and a l g e b r a i c topology
i n o r d e r t o prove o u r r e s u l t s , and t h i s n o t o n l y f o r t h e r e a s o n that
the a b s t r a c t spaces a r e almost never H a u s d o r f f .
Q u i t e on t h e c o n t r a r y , we b e l i e v e t h a t our geometric t h e o r y w i l l serve
as a sound and n e c e s s a r y base t o e x p l o r e a b s t r a c t spaces. Certainly
abstract spaces w i l l become extremely important i n s e m i a l g e b r a i c geo-
metry, a s schemes have become important i n a l g e b r a i c geometry. But,
to the. b e s t o f o u r knowledge, the main method a v a i l a b l e up t o now t o
o b t a i n deeper r e s u l t s about a b s t r a c t spaces i s t o reduce t h e problems
to the geometric case.
N o t i c e , f o r example, t h a t the r e a l spectrum o f a f i n i t e l y generated
commutative r i n g i s t h e a b s t r a c t i o n o f a g e o m e t r i c space o v e r t h e f i e l d
R Q o f r e a l a l g e b r a i c numbers. I f A i s an a r b i t r a r y commutative ring
then a problem i n SperA which i n v o l v e s o n l y f i n i t e l y many elements
of A can sometimes be reduced t o S p e r A for A a suitable finitely

o o
generated s u b r i n g o f A and thus reduced t o a geometric problem. A good
e x e r c i s e i n t h i s d i r e c t i o n i s t o prove T i e t z e ' s e x t e n s i o n theorem f o r
s e m i a l g e b r a i c f u n c t i o n s on an a b s t r a c t a f f i n e s e m i a l g e b r a i c space
starting from t h e known e x t e n s i o n theorem §4] ( c f . 1.4.15) i n
the geometric case.
We c l o s e t h i s appendix w i t h an " e x p l a n a t i o n " o f Example 1.2.9 and i t s
subexamples 2.6, 2.8 i n terms o f a b s t r a c t spaces.
Example A . 4 . L e t M be a g e o m e t r i c space and l e t D be an a r b i t r a r y open
subset of (the u n d e r l y i n g t o p o l o g i c a l space of) M. Then D, equipped
w i t h t h e s t r u c t u r e sheaf SÎD, i s an a b s t r a c t locally semialgebraic
space. We want t o d e s c r i b e D i n g e o m e t r i c terms. We choose a c o v e r i n g
(D loc€I) o f D by quasicompact
a open s e t s . L e t := D^HM. Then
• <-->
X € J"(M) and D = X . We now are i n t h e s i t u a t i o n o f Example 1.2.9.
a a a
L e t X be t h e i n d u c t i v e l i m i t , c o n s i d e r e d t h e r e , o f the open s e m i a l g e -
b r a i c subspaces X := U (X lotGJ) w i t h J r u n n i n g through t h e f i n i t e
T
j ot
s u b s e t s o f I . We have
Xj = Dj := U(D |a€J).
a
Obviously (D,£> lD) M i s the i n d u c t i v e l i m i t o f the spaces (D S^\Dj).

Jf
It i s now pretty e v i d e n t t h a t the a b s t r a c t i o n X o f the g e o m e t r i c space
X i s just the open subspace D o f M. Of c o u r s e i t does not matter w i t h
which covering (D^laGI) o f D by quasicompact open s u b s e t s we have
started. I f U i s any quasicompact open s u b s e t o f D then U c D j f o r some
finite s u b s e t J o f I, hence U f l M c X j . Thus a d d i n g U fl M t o the f a m i l y
(X l<x€I) l e a d s t o the same g e o m e t r i c space

a X.
Conversely, i f (X lot€I) i s any f a m i l y o f open s e m i a l g e b r a i c s u b s e t s o f

a
M and X i s the i n d u c t i v e l i m i t o f the open subspaces X j o f M, as con-
sidered i n Example 1.2.9, then the a b s t r a c t i o n X o f X i s the open sub-
space D := U(X l<x€I)

a o f M. Thus the a b s t r a c t i o n s o f the spaces con-
structed i n Example 1.2.9 are p r e c i s e l y a l l open subspaces o f M.

Subexample A. 5. i ) Assume t h a t M is locally complete. Let t (M)
Q denote
the s e t o f a l l complete s e m i a l g e b r a i c s u b s e t s o f M. We choose as f a m i l y
(X |aei)
Q t h e f a m i l y o f a l l open s e m i a l g e b r a i c s u b s e t s U o f M w i t h
x = M
U G ) T ( M ) . T h i s l e a d s t o the space
c ^ o c c o n s i d e r e d i n I , §7, and,
for M a f f i n e , i n Example 1.2.6. The p o i n t s of X a r e p r e c i s e l y a l l ultra-
filters F i n T(M) w i t h F fl (M) * 0 . Intuitively, the " p o i n t s a t infinity"
of M are those u l t r a f i l t e r s F £ M which c o n t a i n e v e r y s e t M^K with
K€T (M) c ( c f . our t h e o r y o f ends i n Chapter I I and n o t i c e t h a t an ultra-
filter F € M which c o n t a i n s a p a r t i a l l y complete s u b s e t o f M a l s o con-
t a i n s a complete s u b s e t of M, since F contains semialgebraic subsets).
The set of points at i n f i n i t y i s c l o s e d i n M, and M ^ q c i s obtained
from M by omitting this set.
ii) F o r c l a r i t y we denote g e o m e t r i c and a b s t r a c t spaces i n the o l d way
as p a i r s involving the s t r u c t u r e s h e a f . As b e f o r e , l e t {M O^) be a geo- r
m e t r i c space. L e t X be any open s u b s e t o f M i n the s t r o n g t o p o l o g y .
Let (D |a€I) be the f a m i l y o f a l l quasicompact open s u b s e t s of M w i t h
D nMcX
a and l e t D := U(D la£I). T h i s s e t D i s the l a r g e s t open subset
of M w i t h D 0 M = X. L e t X : = D f l M , hence D
a a Q = X .
Q The open subspace
(D,0^j|D) o f (M,©yj) i s the a b s t r a c t i o n o f a g e o m e t r i c space (X,(D ) w i t h

X
x x
u n d e r l y i n g s e t X which i s t h e i n d u c t i v e l i m i t o f the spaces ( /0^' ^*
a a
This i s j u s t the space "induced on X" by the space (M,(9 ) , asM consi-
dered i n Example 1.2.7, s i n c e the X a are p r e c i s e l y a l l semialgebraic
open s u b s e t s o f M which a r e c o n t a i n e d i n X.
In o r d e r t o u n d e r s t a n d the message of a l l t h e s e examples the reader

n
s h o u l d choose f o r M an o l d f r i e n d , say M = R . Then M = Sper R[T^ , . . . ,T^] .
The open s u b s e t s D of Sper R[T^ , . . . ,T ] , c o n s i d e r e d as subspaces are
a b s t r a c t i o n s o f geometric spaces X which are r e g u l a r by 1.4.2 but
n e v e r t h e l e s s may be q u i t e c o m p l i c a t e d . I n p a r t i c u l a r paracompactness
can fail f o r such a space X. F o r example, X = M-^ oc i s not paracompact
if R does not c o n t a i n a sequence o f p o s i t i v e elements converging

t o zero, even f o r n = 1 (1.4.20). The example 1.3.15 shows t h a t
SperRfT^T^] contains an open subset D such t h a t X i s not t a u t and
a f o r t i o r i n o t paracompact, even f o r R= IR .
Appendix B (to Chapter I ) : C o n s e r v a t i o n o f some p r o p e r t i e s of spaces
and maps under e x t e n s i o n o f the base field.
In t h e f o l l o w i n g S i s a r e a l c l o s e d f i e l d c o n t a i n i n g o u r base f i e l d R.
Contrary t o t h e usage i n Chapter I I I and a l a r g e p a r t o f Chapter I I a
"space" o v e r R now means an a r b i t r a r y locally semialgebraic separated
space o v e r R. L a t e r we s h a l l r e t u r n t o the c o n v e n t i o n t h a t every space
i s assumed t o be r e g u l a r and paracompact.
Theorem B. 1 . L e t M be a space over R.
i) M(S) i s s e m i a l g e b r a i c i f and o n l y i f M i s s e m i a l g e b r a i c .
ii) M(S) i s a f f i n e s e m i a l g e b r a i c i f and o n l y i f M i s a f f i n e semialge-
braic.
i i i ) M(S) i s paracompact i f and o n l y i f M i s paracompact.
i v ) M(S) i s r e g u l a r and paracompact i f and o n l y i f M i s r e g u l a r and
paracompact.
Proof, i ) L e t (M |a€I) be an a d m i s s i b l e c o v e r i n g o f M by a f f i n e open
semialgebraic s u b s e t s . Then (M^(S)laGI) i s an a d m i s s i b l e c o v e r i n g o f
M(S) by open a f f i n e s e m i a l g e b r a i c s . I f M i s s e m i a l g e b r a i c then there
exists a finite subset J o f I w i t h M = U ( M l a € J ) .

a Then M(S) =
U (M (S) I a€J) , hence a l s o M(S) i s s e m i a l g e b r a i c . C o n v e r s e l y , i f M(S) i s
s e m i a l g e b r a i c , then there e x i s t s a finite subset J o f I with M(S) =
U (M (S) I <x€J) . I n t e r s e c t i n g w i t h M we see t h a t M = U ( M l a € J ) ,

a and we
conclude that M i s semialgebraic.
i i ) We may a l r e a d y assume t h a t M i s s e m i a l g e b r a i c . I f M i s a f f i n e
then, o f course, M(S) i s a f f i n e . Assume now t h a t M(S) i s a f f i n e . We
want t o prove t h a t M i s a f f i n e . We choose a c o v e r i n g (M^M < i < r ) o f
M by f i n i t e l y many open a f f i n e s e m i a l g e b r a i c s and i n d u c t on r . The
case r = 1 i s t r i v i a l . Once we have s e t t l e d t h e case r = 2, then we a r e
done: We know by i n d u c t i o n t h a t U : = M- U ... U M i s a f f i n e and then

t h a t M = UUM i s affine. Thus we assume h e n c e f o r t h t h a t r = 2.
r
We choose isomorphisms (p. : N. -^»M. from s e m i a l a e b r a i c s u b s e t s N. of

n
1 1 1 - 1
i -1
s t a n d a r d spaces R to M ± ( i = 1 ,2) . L e t N ^ 2 := cp (M^M^
1 and N 2 1 :=
cp 2 ^ (M^ n M ) . 2 These are open s e m i a l g e b r a i c subsets of and N 2 respec-
t i v e l y . L e t f denote the isomorphism x »-* cp 2 °co^ (x) from N ^ 2 onto .
S i n c e M(S) i s a f f i n e t h e r e e x i s t isomorphisms g^ from N^(S) onto semi-

n
a l g e b r a i c s u b s e t s X. o f S ( i = 1,2) f o r some n € E , such that
(g ' 2l
2
N (
S ) ) 0 f
s =
?1 | N
-i2 ( S )
(namely, t h i s means an isomorphism from M(S) onto U X 2 which maps
M^(S) onto X\ ) . U s i n g T a r s k i ' s p r i n c i p l e we know t h a t t h e r e e x i s t iso-

n
morphisms g^ from onto s u i t a b l e s e m i a l g e b r a i c s u b s e t s X^ of R with
W e
(g IN ^)•f=g^|2'
2 2 then have an isomorphism from M onto X^ U X 2 and
we see t h a t M i s a f f i n e . For the argument u s i n g Tarski's p r i n c i p l e the
r e a d e r may consult §3 o f Chapter I I I . As i n t h a t s e c t i o n one constructs
a "parameter space" L whose p o i n t s r e p r e s e n t the p o l y n o m i a l data o f a

n
d e s c r i p t i o n of pairs (g^,g ) 2 o f isomorphisms g^ : N ^ - ^ X ^ c R with
N o f a f i x e d
(g 2 |N -J ) °f = g^ I 1 2
2 "type". I f the type i s chosen r i g h t then
L(S) c o n t a i n s a p o i n t which d e s c r i b e s (g<],g )* Thus L(S) i s not empty.

2
By T a r s k i ' s p r i n c i p l e L i s not empty, and we are done.
iii) and i v ) . I f M i s paracompact ( r e g u l a r and paracompact) then we
know from I , §4 t h a t M(S) i s paracompact ( r e s p . r e g u l a r and paracompact)
(cf. 1.4.4, 4.9). Assume now t h a t M(S) i s paracompact and, w i t h o u t l o s s
of generality, t h a t M i s c o n n e c t e d . Then M(S) i s c o n n e c t e d . By 1.4.17,
M(S) i s L i n d e l o f . We choose an a d m i s s i b l e c o v e r i n g (U ln€U)

n of M(S) by
open s e m i a l g e b r a i c s e t s . F o r e v e r y n G I we choose an open semialgebraic
set V n i n M with U c V ( S ) . n n T h i s i s p o s s i b l e by the v e r y d e f i n i t i o n o f
the space M(S) ( I . Ex. 2.10). I f W i s any open s e m i a l g e b r a i c subset o f
M t h e n W(S) i s c o v e r e d by f i n i t e l y many s e t s U , hence a l s o by finitely
many s e t s V ( S ) . T h i s i m p l i e s t h a t W i s c o v e r e d by f i n i t e l y many s e t s
V, and we c o n c l u d e t h a t (V lne:N)
n i s an a d m i s s i b l e covering of M
(cf. I, § 1 ) . Thus M i s L i n d e l o f .
We c l a i m t h a t f o r e v e r y U € HM) the c l o s u r e U i s a g a i n semialgebraic.
By I , Prop. 4.18, t h i s w i l l imply t h a t M i s paracompact. By I, Prop.
3.22.C we know t h a t U(S) = U ( S ) . This i s a semialgebraic set, since
M(S) i s paracompact ( I , Prop. 4.6). We conclude, by ( i ) , t h a t U i s semi-
a l g e b r a i c . Thus M i s indeed paracompact.
Assume t h a t , i n a d d i t i o n , M(S) i s r e g u l a r . L e t U be an open s e m i a l g e -
b r a i c s u b s e t of M. Then U(S) i s open s e m i a l g e b r a i c i n M(S). Since
M(S) i s r e g u l a r t h i s i m p l i e s t h a t U(S) i s a f f i n e hence, by ( i i ) , t h a t
U i s affine. We c o n c l u d e t h a t M i s r e g u l a r ( c f . I , Prop. 4.7). T h i s
finishes the proof o f the theorem.
As b e f o r e we a r e o n l y i n t e r e s t e d i n r e g u l a r paracompact spaces. Asser-
tion ( i v ) i n Theorem B.1 g i v e s us t h e j u s t i f i c a t i o n t o admit o n l y such space
Thus, from nom on, a "space" a g a i n means a r e g u l a r paracompact l o c a l l y s e m i a l
g e b r a i c space. A "map" between spaces means a l o c a l l y s e m i a l g e b r a i c map.
P r o p o s i t i o n B.2. L e t M be a space over R.
i) M i s partially complete i f and o n l y i f M(S) i s p a r t i a l l y complete.
ii) M i s complete i f and o n l y i f M(S) i s complete.
Proof. We choose an isomorphism cp : IKI^-^M, w i t h K a s t r i c t l y locally
f i n i t e abstract simplicial complex ( I I , Th. 4.4). Then (p i s an isomor-

g
phism from |Kl s onto M(S). Now M i s p a r t i a l l y complete (complete) i f
and only i f K i s closed ( c l o s e d and f i n i t e ) and t h i s a l s o means t h a t
M(S) j.s p a r t i a l l y complete (complete). q.e.d.
We d i g r e s s f o r a s h o r t time i n t o the c o n t e n t s o f Chapter I I , §9. We
choose a c o m p l e t i o n j : M ^ P o f M. By t h e j u s t proved p r o p o s i t i o n t h e
space P(S) i s p a r t i a l l y complete, and s i n c e j(M) i s dense i n P a l s o
j (M) (S) = j (M(S)) i s dense i n P(S) ( c f . I , Prop. 3.22). Thus j g : M(S) «->
P(S) i s a completion of M(S). We r e g a r d M as a subspace o f P v i a j ,
hence a l s o M(S) as a subspace o f P ( S ) . L e t ((P ,M ) I k > 0) denote the k k
d e r i v e d sequence o f (P,M) ( I I , §9, Def. 7 ) . I t i s now e v i d e n t t h a t
( ( P ( S ) ,M (S)) I k > 0)
k k i s the d e r i v e d sequence o f (P(S),M(S)). T h i s implies
the f o l l o w i n g r e s u l t .
C o r o l l a r y B. 3. M and M(S) have t h e same c o m p l e x i t y . I f (S (M)|k>0)

k
is the L C - s t r a t i f i c a t i o n o f M ( c f . I I , §9, Def. 8) then (S (M)(S)Ik>0)

k
is the L C - s t r a t i f i c a t i o n o f M(S). I n p a r t i c u l a r , M(S) i s l o c a l l y com-
plete ( i . e . has c o m p l e x i t y <1) i f and o n l y i f M i s l o c a l l y complete.
Remark. L o c a l l y complete spaces a r e r e g u l a r , and a space M i s l o c a l l y complete
i f and o n l y i f every open ( a f f i n e ) s e m i a l g e b r a i c s u b s e t U o f M i s l o c a l l y com-
p l e t e . T h i s shows t h a t the l a s t statement i n C o r o l l a r y B. 3 remains t r u e w i t h -
out the assumption t h a t M i s r e g u l a r and paracompact. ( T h i s remark w i l l be
needed i n [ D ] . ) 2
Now we a r e ready t o d i s c u s s some p r o p e r t i e s of maps.
Theorem B. 4. L e t f : M-* N be a map between spaces over R.
i) f g i s s e m i a l g e b r a i c i f and o n l y i f f i s s e m i a l g e b r a i c .
ii) f g i s p a r t i a l l y proper i f and o n l y i f f i s p a r t i a l l y proper.
iii) f g i s proper i f and o n l y i f f i s p r o p e r .
iv) fg i s partially f i n i t e i f and o n l y i f f i s p a r t i a l l y f i n i t e .
v) f g i s f i n i t e i f and o n l y i f f i s f i n i t e .
Proof, i ) We choose an a d m i s s i b l e c o v e r i n g (N^laGI) o f N by open semi-
a l g e b r a i c s . Then ( N ( S ) | a € I ) i s an a d m i s s i b l e c o v e r i n g o f N(S)
a by open
semialgebraics.The map f g (resp. f ) i s s e m i a l g e b r a i c i f f , f o r every

1 1 1
a€I, t h e space f " (N (S)) = f " (N ) (S) ( r e s p . f"" (N )) i s s e m i a l g e b r a i c .
o Ot Ot 0(
The a s s e r t i o n now f o l l o w s from Theorem B . 1 . i . (Here we d i d n o t need
the assumption t h a t M and N a r e r e g u l a r and paracompact.)
i i ) We choose a completion
M < ^ P
of f ( c f . I I , Prop. 5.1). Then tp and ^ g g are completions of M(S) and
N(S) as shown above. Thus
M(S) < *P(S)
f
S
N(S) < r -Q(S)
i s a c o m p l e t i o n o f f g . Without l o s s of g e n e r a l i t y we r e g a r d tp and ,
hence a l s o cpg and \p^, as i n c l u s i o n maps, f i s p a r t i a l l y proper i f and
only i f g (N) =M and f g i s p a r t i a l l y proper i f and o n l y i f g g (N(S))

-1 -1
M(S). These c o n d i t i o n s a r e e q u i v a l e n t , s i n c e g g (N(S)) = g (N) (S) .
Assertion ( i i i ) i s now e v i d e n t s i n c e a map i s proper i f and o n l y i f
it i s p a r t i a l l y p r o p e r and s e m i a l g e b r a i c .
Next we prove ( v ) . A map is finite i f and o n l y i f i t i s proper and has
finite fibres. T a k i n g i n t o account ( i i i ) we may a l r e a d y assume t h a t f
and (hence) f g i s p r o p e r . Of c o u r s e , i f f g has finite fibres then a
f o r t i o r i f has finite fibres. Assume now t h a t f i s f i n i t e . By Hardt's
theorem ( I I , 6.3) there e x i s t s a l o c a l l y finite partition (N^laGI) of
N i n t o s e m i a l g e b r a i c s e t s such t h a t , f o r every a€I, the restriction
f (N ) -> N o f f i s a t r i v i a l map with a f i n i t e f i b r e F . Then

a cx a
(N (S)|aGI) is a partition of N(S) i n t o s e m i a l g e b r a i c s e t s , and f c is
cx *
over each ^^(S) a t r i v i a l map w i t h the f i b r e F (S) =F . Thus f
rt rt q has
finite fibres.
I t remains to prove ( i v ) . We choose a l o c a l l y finite c o v e r i n g (A^laEI)
of M by c l o s e d s e m i a l g e b r a i c s . Then (A (S)|a€I)
a is a locally finite
c o v e r i n g of M(S) by c l o s e d s e m i a l g e b r a i c s . The map f g is partially
finite i f and o n l y i f , f o r every index ccGI, the r e s t r i c t i o n f lA

c (S)
(flA )(S)
a i s a f i n i t e map. T h i s means, as proved b e f o r e , that, for
every a, the map f lA a is finite, hence t h a t f i s p a r t i a l l y finite.

References
[A] M. A r t i n , "Grothendieck t o p o l o g i e s " . Seminar Harvard
University 1962.
[SGA 4] M. A r t i n , A. Grothendieck, J . L . V e r d i e r , "Theorie des

topos e t cohomologie e t a l e des schemas". S p r i n g e r L e c t u r e
Notes Math. 269 (1972).
[BCR] J • Bochnak, M. Coste, M.F. Roy, "Geometrie a l g e b r i q u e

reelle" i n preparation f o r Springer Verlag
[Bo] N. B o u r b a k i , " T o p o l o g i e g e n e r a l e " , C h a p i t r e s 1-4.

N o u v e l l e e d i t i o n . Hermann, P a r i s 1971.
[Br] L. Brocker, R e a l s p e c t r a and d i s t r i b u t i o n s o f s i g n a t u r e s .

In: "Geometrie a l g e b r i q u e r e e l l e e t formes q u a d r a t i q u e s " ,
Ed. J . L . C o l l i o t - T h e l e n e e t a l . , L e c t u r e Notes Math. 959,
249-272, S p r i n g e r V e r l a g 1982.
[B] G.W. B r u m f i e l , " P a r t i a l l y o r d e r e d r i n g s and s e m i a l g e b r a i c

geometry". London Math. Soc. L e c t u r e Notes 37, Cambridge
U n i v . P r e s s 1979.
[B-] G.W. B r u m f i e l , W i t t Rings and K-theory. Rocky Mountain

J . Math. 14, 733-765 (1984).
[CaC] M. C a r r a l , M. Coste, Normal s p e c t r a l spaces and t h e i r

d i m e n s i o n s . J . Pure Appl. A l g . 30, 227-235 (1983).
[CKLR] M.D. C h o i , M. Knebusch, T.Y. Lam. B. Reznick, T r a n s v e r s a l

z e r o s and p o s i t i v e s e m i - d e f i n i t e forms. I n : "Geometrie
a l g e b r i q u e r e e l l e e t formes q u a d r a t i q u e s " , Ed. J . L .
C o l l i o t - T h e l e n e e t a l . , L e c t u r e Notes Math. 959, 273-298,
S p r i n g e r V e r l a g 1982.
[CR] M. Coste, M.F. Roy, La t o p o l o g i e du s p e c t r e r e e l .

Contemporary Math. 8, Amer. Math. Soc. 1982, 27-59.
[D] H. P e l f s , Kohomologie a f f i n e r s e m i a l g e b r a i s c h e r Raume.

D i s s e r t a t i o n , Regensburg 1981.
[D-] H. P e l f s , The homotopy axiom i n s e m i a l g e b r a i c cohomology.

J. r e i n e angew. Math. 355, 108-123 (1985).
[P ] 2 H. P e l f s , Sheaf theory and Borel-Moore homology on

l o c a l l y s e m i a l g e b r a i c spaces. H a b i l . s c h r i f t , Regensburg
1984.
[DK] H. P e l f s , M. Knebusch, S e m i a l g e b r a i c t o p o l o g y over a r e a l

c l o s e d f i e l d . In: "Ordered f i e l d s and r e a l a l g e b r a i c
geometry", Ed. P.W. Pubois, T. Recio, Contemporary Math.
8, Amer. Math. Soc. 1982, 61-78.
[PK.j],CPK ] 2 H. P e l f s , M. Knebusch, S e m i a l g e b r a i c topology over a r e a l

c l o s e d f i e l d I : Path and components i n the s e t o f r a t i o n a l
p o i n t s o f an a l g e b r a i c v a r i e t y , Math. Z. 177, 107-129
(1981); I I : B a s i c t h e o r y o f s e m i a l g e b r a i c spaces, i b i d.
1978, 175-213 (1981).
[PK^] H. P e l f s , M. Knebusch, On the homology of a l g e b r a i c v a r i e -

t i e s over r e a l c l o s e d f i e l d s . J . r e i n e angew. Math. 335,
122-163 (1981).
[DK ] 4 H. P e l f s , M. Knebusch, Zur T h e o r i e der s e m i a l g e b r a i s c h e n
Wege und I n t e r v a l l e uber einem r e e l l a b g e s c h l o s s e n e n
Korper. I n : "Geometrie a l g e b r i q u e r e e l l e e t formes qua-
d r a t i q u e s " , Ed. J . L . C o l l i o t - T h e l e n e e t a l . , L e c t u r e
Notes Math. 959, 299-323, S p r i n g e r V e r l a g 1982.
[DK^] H. P e l f s , M. Knebusch, S e p a r a t i o n , r e t r a c t i o n s and homo-

topy e x t e n s i o n i n s e m i a l g e b r a i c spaces. P a c i f i c J . Math.
114, 47-71 (1984).
[DK^] H. P e l f s , M. Knebusch, An i n t r o d u c t i o n to l o c a l l y semi-

a l g e b r a i c spaces. Rocky Mountain J . Math. 14, 9 4 5-9 63
(1984).
[DK ] 7 H. P e l f s , M. Knebusch, B a s i c homotopy t h e o r y o f l o c a l l y

s e m i a l g e b r a i c spaces. Rocky Mountain J . Math. 14, 913-918
(1984).
[DKP] T. torn P i e c k , K.H. Kamps, P. Puppe, " H o m o t o p i e t h e o r i e " .

L e c t u r e Notes Math. 157, S p r i n g e r V e r l a g 1970.
[Du] P.W. Pubois, R e a l a l g e b r a i c c u r v e s . T e c h n i c a l Report No.

227, Univ. New Mexico, Albuquerque 1971.
[ES] S. E i l e n b e r g , N. S t e e n r o d , "Foundations o f a l g e b r a i c
t o p o l o g y " . P r i n c e t o n U n i v . Press 1952.
[F] R.H.Fox, C o v e r i n g spaces w i t h s i n g u l a r i t i e s . I n : "Alge-

b r a i c geometry and t o p o l o g y " , Sympos. honour of S. L e f -
s c h e t z , Ed. R.H. Fox e t a l . , P r i n c e t o n U n i v . P r e s s 1957.
[Fr] H. F r e u d e n t h a l , Ober d i e Enden t o p o l o g i s c h e r Raume und

Gruppen, Math. Z. 33,692-713 (1931).
[Fr^] H. F r e u d e n t h a l , Neuaufbau der E n d e n t h e o r i e . Annals Math.

43, 261-279 (1942). _ [ G i ] p 3 1 g
[Go] R. Godement, " T o p o l o g i e a l g e b r i q u e e t t h e o r i e des fais-

ceaux". Hermann, P a r i s 1964.
[EGA I*] A. G r o t h e n d i e c k , J.A. Pieudonne, "Elements de geometrie

a l g e b r i q u e I " . G r u n d l e h r e n math. Wiss. 166, S p r i n g e r
V e r l a g 1971.
[EGA II] A. G r o t h e n d i e c k , J . Pieudonne, "Elements de geometrie

a l g e b r i q u e I I . P u b l . Math. IHES No. 8 (1961).
[H] R.M. Hardt, S e m i - a l g e b r a i c l o c a l t r i v i a l i t y i n semi-

a l g e b r a i c mappings. Amer. J . Math. 102, 291-302 (1980).
[H-j ] R.M. Hardt, T r i a n g u l a t i o n o f s u b a n a l y t i c s e t s and proper

l i g h t s u b a n a l y t i c maps. Invent, math. 38, 207-217 (1977).
[Hi] H. H i r o n a k a , T r i a n g u l a t i o n o f a l g e b r a i c s e t s . A l g e b r a i c
Geometry A r e a t a 1974, P r o c . Sympos. Pure Math. 29,
165-185, Amer. Math. Soc. 1975.
[Ho] H. Hopf, Enden o f f e n e r Raume und u n e n d l i c h e d i s k o n t i n u -

i e r l i c h e Gruppen. Comment. Math. H e l v . 16, 81-100 (1943).
[Hu] S.T. Hu, "Homotopy t h e o r y " . Academic P r e s s 19 59.
[Ke] J . L . K e l l e y , "General Topology", van Nostrand, Princeton

1955.
[K] M. Knebusch, An i n v i t a t i o n to r e a l s p e c t r a . Canad. Math.

Soc. Conf. Proc. 4, 51-105 (1984).
[K^] M. Knebusch, I s o a l g e b r a i c geometry: F i r s t s t e p s . Seminaire

de T h e o r i e des Nombres ( D e l a n g e - P i s o t - P o i t o u ) P a r i s
1980-81 (Ed. M.J. B e r t i n ) , 127-141, P r o g r e s s i n Math. 22
B i r k h a u s e r Boston 19 82.
[KW] M. Knebusch, M.J. Wright, Bewertungen mit r e e l l e r Hense-

l i s i e r u n g . J . r e i n e angew. Math. 286/287, 314-321 (1976).
[L] T.Y. Lam, "The t h e o r y o f o r d e r e d f i e l d s " . Ring theory

and a l g e b r a I I I (Ed. B. McDonald), L e c t u r e Notes Pure
A p p l . Math. 55, 1-152, Dekker 1980.
[L* ] T.Y. Lam, An i n t r o d u c t i o n to r e a l a l g e b r a . Rocky Moun-

t a i n J . Math. 14, 767-814 (1984).
[Lo] S. L o j a s i e w i c z , T r i a n g u l a t i o n of s e m i - a n a l y t i c s e t s .
Ann. S c u o l a Norm Sup. P i s a (3) 18, 449-474 (1964).
[MV] J . M a l g o i r e , C. V o i s i n , F a c t o r i s a t i o n de S t e i n t o p o l o -
gique; decoupe. Topology 20, 191-207 (1981), c f . a l s o :
C a h i e r s Math. 15, U n i v . " * M o n t p e l l i e r 1979.
[M] E. M i c h a e l , Cuts. A c t a Math. 111, 1-36 (1964).
[M^ ] E. M i c h a e l , Completing a spread ( i n the sense of R.H.

Fox) w i t h o u t l o c a l connectedness. Indag. Math. 25,
629-633 (1963).
[Mi] J . M i l n o r , On a x i o m a t i c homology t h e o r y . P a c i f i c J . Math.

12, 337-341 (1962).
[P] A. P r e s t e l , " L e c t u r e s on f o r m a l l y r e a l f i e l d s " . Mono-

g r a f i a s Mat. 22, I n s t . Mat. Pura A p l . , Rio de J a n e i r o
1975, and L e c t u r e Notes Math. 1093, S p r i n g e r V e r l a g 1984.
[R] R. Robson, Embedding s e m i a l g e b r a i c spaces. Math. Z. 183,

365-370 (1983).
[Ry] M.F. Roy, F a i s c e a u s t r u c t u r a l sur l e s p e c t r e r e e l e t fonc

t i o n s de Nash. I n : "Geometrie a l g e b r i q u e r e e l l e e t formes
q u a d r a t i q u e s " , Ed. J . L . C o l l i o t - T h e l e n e e t a l . , L e c t u r e
Notes Math. 959, 406-432, S p r i n g e r V e r l a g 1982.
[Sch] N. Schwartz, Real c l o s e d spaces. H a b i l . s c h r i f t Munchen

1984.
[SW] P. S c o t t , C . T . C . W a l l , T o p o l o g i c a l methods i n group t h e o r y .

I n : "Homological group t h e o r y " , Ed. C.T.C. W a l l , London
Math. Soc. L e c t u r e Notes 36, Cambridge Univ. P r e s s 1979,
137-203.
[SY] M. S h i o t a , M. Y o k o i , T r i a n g u l a t i o n s o f s u b a n a l y t i c s e t s
and l o c a l l y s u b a n a l y t i c m a n i f o l d s . Trans. Amer. Math.
Soc. 286, 727-750 (1984) .
[Spa] E.H. S p a n i e r , " A l g e b r a i c Topology". McGraw H i l l , New
York 1966.
[W] G.W. Whitehead, "Elements o f homotopy t h e o r y " . Graduate

Texts Math. 61, S p r i n g e r V e r l a g 1978.
[W ] 1 G.W. Whitehead, "Homotopy t h e o r y " . M.I.T. P r e s s 1966 .

G i
t ] B. Giesecke, S i m p l i z i a l e Zerlegung a b z a h l b a r e r analyti-
scher Raurr.e, Math. Z. 83, 1 77-2 1 3" ( 1 964) .
T(M) 1 S (M),
k S^(M) 171
T(M) 27 c(P,M) 169
T(M) 57, 63 c(M) 172
T (M)
c
161 Q P 157
M*
r(M) 4,5 P (A) 175
r(M> 30f M 296, 298f

n a x
r(M) 57, 63 ?f 300f
r (M)
c
81 F 296
r (M)
c
81 ? 298f
«
C O V 1 X 33
M
Cov (U) M 2 X 21, 33
(M,0 ) M 3 I(X) 21
0 (0)
M
6 St (A) x 22, 138
6 St (x) x 139
M(R) 19 Sh (x) x 139
19 Sh ( ) x T 140
dim M 37 St (T) x 140
dim^M 37 Lk (x) x 145
Z (M)
n 38 K(X),K(f) 101
I > > 38 E(K) 99

MxN 14 S(K) 99
Mx N s 29 |K|,|K| R 104
M, 15f, 81 IT!, l Y l 104
R
loc
PM 83 X,K 120
f 84 T 111, 163
loc
+
M 78 120
PX'^K
M/fc 178 sd K , s d K r
260f
M/A 182 co X , co (A,tp) 229
2>(M) 158 co(X,X ,...,X 1 r ) 233
S (P,M), S (P,M) 171
x' ( x , x . . .,x )
k k
r r 2 3 c
v r
j
x' j
(x,x,,.. .,x ) r
233 TT (M,A,x ) N o t o p
271
r(f) 30 rr (cp) n 266
f
R 19 n (A,a)
n
286
supp(f) 49 TT(A) 268
supp f (a) 238 TT(A) 288
f* 142, 161 , 164 H (M,G)

n
280
266, 269, 287
u
* 268 H (M,A;G)
N 280
H
#
288 H ( M A G )
280
n ' ' top
XU Y f 185 H (M,G)
N
281
Z(f) 183 H (M),ff (M)

n N 281
C(X) 184 H (M,A)

N
281
C(f) 185 TT*(M,X ) O

283
Con(f) 208 TT^(M,A,X ) Q 283
C(x,X) , C ( i , x ) 140 282
C(f,x) 209 cp*,cp 283
D(f,x) 209 f-g 232
X(f) 138 f ^g relC 249
p
x,f 138 [f ] 232
c (M) 161 [f ] C
249
c (M,cp) 159 []
257
to P
eta) 165 f *g 265, 268
K (CP) 163f TOP, HTOP 259
K ( i j j f CP) 160 LSA(R), HLSA(R) 256
K
(M,M) 167 HLSA(2,R) 278
K 243, 270, 280 Sper A 295
X 257, 271, 280 Sper cp 298
TT (X),TT (f) 142 13

0 0
U
TT (M) 0 36 U 170
\> ( M )
x 152 CO 78, 94
n (M,x )n Q 265 H denotes the s e t {1,2,
TT (M,A,x ) n o 266 of n a t u r a l numbers.

U := U U {0}
o
3o 1 11
3U 178
8 :m n (M,A,x ) Q ^ _ (A,x )
n 1 o 267
3 : H n (M, A; G) - H _^ (A, G)
n 1 2 80
[ (M„A r .,A ) , ( N , B
r r , B ) ] 232
r
n
[ (M A i r r . . . A ) , ( N , B . . . ,B ) ]
f r r r 249
Glossary
absolute end 161
homotopy group 2 65
patn c o m p l e t i o n criterion 68
abstract semialgebraic space 297
a f f i n e s e m i a l g e b r a i c space 297
locally semialgebraic space 297
simplicial complex 99
abstraction of a l o c a l l y semialgebraic space 299
of a geometric simplicial complex 101
o f a s i m p l i c i a l map 102
admissible covering 1
affine s e m i a l g e b r a i c space [DI^, § 7 ] , 42
s e m i a l g e b r a i c map 54
amenable 125
a t l a s o f a t r i a n g u l a t i o n 109f
barycentric subdivision 131, 260f
base o f a space 217
base f i e l d e x t e n s i o n v i i , 19f, [DK^, §4]
blowing up (semialgebraic) 154
bordered star 139
c a n o n i c a l r e t r a c t i o n 230, 234
closed 32
closed simplicial complex 21, 99
subcomplex 22, 100
closure of a simplex 20
of a s i m p l i c i a l complex 21, 99
of a s i m p l i c i a l map 24, 101
of a subcomplex 100
c o l l a r e d end 195
complement o f a subcomplex 100
complete [DK , 2 § 9 ] , 57
completed space 146
completion of a space 94, 157
of a map 113
"complex" 100, 228
complexity 169
cone 184
connectable 141
connected 35f, 140
connected component 36, 140
constant (homotopy) 249
contiguous 240, 260
continuous 32
core of a complex 229
of a system o f complexes 2 33
covering (= Uberlagerung) i v , 225
c u r v e s e l e c t i o n lemma [DK , 2 §12], [DK ,

4 §1]
cut 152, 153
dense 32
pair 146
derived sequence 169
dimension [DK , 2 § 8 ] , 37, 102
direct product o f spaces 14f
sum o f spaces 13f
discrete 36
end 161
determined by a p r o p e r incomplete p a t h 165
i n a completion 159
equivalent simplicial atlases 109
triangulations 106
expansion o f a completed space 189
problem f o r ends 192
e x t e n s i o n of the base f i e l d v i i , 19f, [DK^, §4]
face 100
f i b r e p r o d u c t 29, [DK ,2 § 7 ] , 118
finite e x t e n s i o n o f a s e m i a l g e b r a i c map 214
map 72, 116
morphism between dense p a i r s 147
morphism between completed spaces 158
simplicial complex 22, 102
f i r s t main theorem on homotopy s e t s 243, 249f
f o r homotopy groups 270
full subcomplex 100, 102
fundamental g r o u p o i d 26 8
of an end 288
geometric generic f i b r e i
locally s e m i a l g e b r a i c space 298
s i m p l i c i a l complex 21, 99
g e n e r a l i z e d t o p o l o g i c a l space 1
generic fibre i
gluing principle f o r maps 34
f o r spaces 89
good on A 228
good (weak) t r i a n g u l a t i o n 233
Grothendieck topology i i i , 2
H a r d t ' s theorem [H], 125
homology groups 2 80
t h e o r y over R 279
biomotopic 232
relative C 249
Inomotopy 226
e x c i s i o n theorem 273
e x t e n s i o n theorem 228
groups 265ff
groups o f an end 286
i n v a r i a n c e o f c o l l a r e d ends 293
Hurewicz map 282
theorems 2 83
induced space on an open s u b s e t 17
i n d u c t i v e l i m i t o f spaces 11ff
incomplete p a t h 66
i n t e r s e c t i o n o f a b s t r a c t subcomplexes 100
join 145
joinable 145
LC-stratification 172
LC-stratum 173
Lindelof 50
link 145
local dimension 37
isomorphism 211
l o c a l l y complete 15, 75
locally finite family of sets 6
simplicial complex 22, 102
locally semialgebraic function 6
map 7, 297
space 4, 2 97
subset 27
subspace 28
locally t r i v i a l map 71
"map" 226, 303
mapping cone 185
cylinder 183
maximal complex 119
hull o f a complex 119
microbial 44
morphism between r i n g e d spaces over R 3
dense p a i r s 146
completions 157
one-point completion 78
open 1, 32
join 139, 145
map 219
locally s e m i a l g e b r a i c subset 32
s e m i a l g e b r a i c s u b s e t 4, 32
simplex 20
subcomplex 22, 103
subspace 4
paracompact 43
parameter o f a map 24 5
space 245
partially complete 63
finite 72, 115, 147
p r o p e r 63, 115, 147
p a r t i t i o n o f u n i t y 47, 49, [DK^, §1]
p a t h 35, [DK ,
2 §10], [DK^]
criterion for continuity 36
completion c r i t e r i o n 67, 68
component 35, [DK ,

2 §10, §11], [DK 4
proper [DK , 2 § 9 ] , 57, 115, 147
e x t e n s i o n o f a s e m i a l g e b r a i c map 204
homotopy 2 87
incomplete path 164
quotient 178
pure c o m p l e t i o n of a map 203
c o m p l e t i o n o f a space 158
dense p a i r 149
end 192
hull 149,158
map 200f
of dimension n 37
quotient projection 178
r e a l spectrum 295
realization o f an a b s t r a c t complex 103
a b s t r a c t s i m p l i c i a l map 104
reduced homology groups 281
refinement of a t r i a n g u l a t i o n 106
regular 42
relative homotopy s e t 2 49
homotopy group 266
path completion c r i t e r i o n 67
retract, retraction 226
retrocompact 300
r i n g e d space over R 3
Robson's embedding theorem [R], 42
second main theorem on homotopy s e t s 257
f o r homotopy groups 271
s e m i a l g e b r a i c map [DK , 2 §7], i i i , 7, 9, 54, 298
morphism between dense p a i r s 147

s e m i a l g e b r a i c s e t [DK ,
2 §6]
space [DK , 2 §7], i i i , 4, 297
s u b s e t 30
topology iiif
separated 42
simplex 20, 99
simplicial approximation 234
atlas 108
complex 21, 99
map 24, 99
simultaneous triangulation 26, 106
sheaf 2
shell 139
shrinking morphism between completed spaces 176
of a completed space 189
of a covering 4 7
"space" 42, 75, 124, 226, 303
s p e c i a l i z a t i o n of a point 302
spread 211
s t a r 22, 138
Stein factorization 198, 205
stereographic projection 79
strictly c o n t i g u o u s 240, 260
c o n t i n u o u s 32
locally finite 22, 102
strong d e f o r m a t i o n r e t r a c t ( i o n ) 2 26
t o p o l o g y 31
subcomplex 22, 100
subspace 28, 297
support of a function 4 9
of the image o f a simplex 2 38

system o f complexes 232
of spaces 232
tame i n X 228
system o f complexes 233
taut 82
D K
Tietze's e x t e n s i o n theorem 49, t 5 ' §4]
thin 152
triangulation o f a dense p a i r 146
map 113
morphism o f dense p a i r s 149
space 26, 106
system o f spaces 233
triangulation theorem 112
t r i v i a l map 125
type 2 45
ultrafilter ii, 296
union o f a b s t r a c t subcomplexes 100
universal covering i v
v e r t e x 20, 99
weakly l o c a l l y t r i v i a l map 71
s i m p l i c i a l map 23
weak p o l y t o p e 27 7
weak t r i a n g u l a t i o n of a map 135
space 135
system o f spaces 253
well cored 234
Whitehead's theorem 284f
Univ.-ribliothek
Regensbwo
Vol. 101034: J. M<usiela!k, Orlicz Spaces and Modular Spaces. V, Vol. 1062: J. Jost, Harmonic Maps Between Surfaces. X, 133 pages.
222 papages.;. 1S98c3. 1984.
Vol. 101035: Trne Mathematics and Physics of Disordered Media. Vol. 1063: Orienting Polymers. Proceedings, 1983. Edited by
Proceeeeding:js, 19J83. Edited by B. D. Hughes and B.W. Ninham. J.L. Ericksen. VII, 166 pages. 1984.
VII, 43'/32 pa^gess. 11983.
Vol. 1064: Probability Measures on Groups VII. 'roceedings, 1983.
Vol. 105036: CCoimbiinatornal Mathematics X. Proceedings, 1982. Edited Edited by H. Heyer. X, 588 pages. 1984.
by LR.R.A. CCasssee. XI, 419 pages. 1983.
Vol. 1065: A. Cuyt, Pade Approximants for Operators: Theory and
Vol. 102037: Worn-liinear Partial Differential Operators and Quantization Applications. IX, 138 pages. 1984.
Proceceduress. Proceedings, 1981. Edited by S.I. Andersson and
Vol. 1066: Numerical Analysis. Proceedings, 1£83. Edited by D.F.
H.-D. L Doebonejr. Wll, 334 pages. 1983.
Griffiths. XI, 275 pages. 1984.
Vol. 101038: F. Boirceux, G. Van den Bossche, Algebra in a Localic
Vol. 1067: Yasuo Okuyama, Absolute Summabiliy of Fourier Series
Topos )S withi Afppllications to Ring Theory. IX, 240 pages. 1983.
and Orthogonal Series. VI, 118 pages. 1984.
Vol. 101039: /Anaalyttic Functions, Blazejewko 1982. Proceedings. Edited Vol. 1068: Number Theory, Noordwijkerhout 1)83. Proceedings.
by J. L Lawryynoiwucz. X, 4 9 4 pages. 1983 Edited by H. Jager. V, 296 pages. 1984.
Vol. 101040: A. Gcood, Local Analysis of Selberg's Trace Formula. Ill, Vol. 1069: M. Kreck, Bordism of Diffeomorphisms ind Related Topics.
128 pi pages... 1S98S3. Ill, 144 pages. 1984.
Vol. 1C1041: Lie3 Giroup Representations II. Proceedings 1982-1983. Vol. 1070: Interpolation Spaces and Allied "bpics in Analysis.
Editeced by R. Hierb, S. Kudla, R. Lipsman and J. Rosenberg. IX, Proceedings, 1983. Edited by M. CwikelandJ. Pestre. Ill, 239 pages.
340 p pagess. 1198*4. 1984.
Vol. 101042: /A. CGutt, K. D. Schmidt, Amarts and Set Function Processes.
Vol. 1071: Pade Approximation and its Applications, Bad Honnef
Ill, 25i58 pa^gess. 11983. 1983. Prodeedings. Edited by H. Werner anc H.J. Bunger. VI,
Vol. 1C1043: Limeaar and Complex Analysis Problem Book. Edited by 264 pages. 1984.
V.R H;Havin,, S. V. IHrus6ev and N.K. Nikol'skii. XVIII, 721 pages. 1984. Vol. 1072: F. Rothe, Global Solutions of Reaction-Diffusion
Vol. 1C1044: E. <Ge»keler, Discretization Methods for Stable Initial Value Systems. V, 216 pages. 1984.
P r o b l e m s . Villi, 2201 pages. 1984.
Vol. 1073: Graph Theory, Singapore 1983. Proceedings. Edited by
Vol. 1 1045:: Differential Geometry. Proceedings, 1982. Edited by K. M. Koh and H. P. Yap. XIII, 335 pages. 1984.
A.M. fl. Navesirai. VMM, 194 pages. 1984.
Vol. 1074: E. W. Stredulinsky, Weighted Inequaliies and Degenerate
Vol. 1C1046: / Alcgebjraic K-Theory, Number Theory, Geometry and Analy- Elliptic Partial Differential Equations. Ill, 143 pag«s. 1984.
sis. P Proceeediingis, 1982. Edited by A. Bak. IX, 464 pages. 1984.
Vol. 1075: H. Majima, Asymptotic Analysis for Inferrable Connections
Vol. 1 1047:: Flluicd Dynamics. Seminar, 1982. Edited by H. Beirao da with Irregular Singular Points. IX, 159 pages. 1934.
Veigaga. VII,, 19)3 rpages. 1984.
Vol. 1076: Infinite-Dimensional Systems. Proceeiings, 1983. Edited
Vol. 1. 1048:: Kiinertic Theories and the Boltzmann Equation. Seminar, by F. Kappel and W. Schappacher. VII, 278 pag*s. 1984.
1981J1. Edittedl by/ C. Cercignani. VII, 248 pages. 1984.
Vol. 1077: Lie Group Representations III. Proceidings, 1982-1983.
Vol. 1. 1049:: BJ. Icochum, C6nes autopolaires et algebres de Jordan. Edited by R. Herb, R. Johnson, R. Lipsman, J. Fosenberg. XI, 454
VI, 2'247 piageas. 1984. pages. 1984.
Vol. 1'. 1050:: A. Prcestel, P. Roquette, Formally p-adic Fields. V, 167 pages.
Vol. 1078: A.J.E.M. Janssen, P. van der Steen, Integration Theory.
1984H.
V, 224 pages. 1984.
Vol. 1.1051: AUgelbraic Topology, Aarhus 1982. Proceedings. Edited by
I. Maladsem arnd IB. Oliver. X, 665 pages. 1984. Vol. 1079: W. Ruppert. Compact Semitopological Semigroups: An
Intrinsic Theory. V, 260 pages. 1984
Vol. 1. 1052!: Nurmber Theory. Seminar, 1982. Edited by D.V. Chud-
Vol. 1080: Probability Theory on Vector Spacis III. Proceedings,
novsvsky, G3.V/. Qhudnovsky, H. Cohn and M.B. Nathanson. V, 309
1983. Edited by D. Szynal and A. Weron. V, 373 pages. 1984.
pageges. 19*841.
Vol. I. 10533: P . Hilton, Nilpotente Gruppen und nilpotente Raume. Vol. 1081: D. Benson, Modular Representation "^heory: New Trends
V, 2S221 paigets. H984. and Methods. XI, 231 pages. 1984.
Vol. 1082: C.-G. Schmidt, Arithmetik AbelscherVarietaten mit kom-
Vol. 1.1054;: V. Tmomee, Galerkin Finite Element Methods for Parabolic
Prohoblemss. Wll, 237 pages. 1984. plexer Multiplikation. X, 96 Seiten. 1984.
Vol. I. 10555: Quantum Probability and Applications to the Quantum Vol. 1083: D. Bump, Automorphic Forms on GL (J.IR). XI, 184 pages.
Theceory oif lirrewersible Processes. Proceedings, 1982. Edited by 1984.
L A<Accarctii, /A. Frigerio and V. Gorini. VI, 411 pages. 1984. Vol. 1084: D. Kletzing, Structure and Representitions of Q-Groups.
VI, 290 pages. 1984.
Vol. >l. 10565: Ailgejbraic Geometry. Bucharest 1982. Proceedings, 1982.
Editiited by/ L. B&descu and D. Popescu. VII, 380 pages. 1984. Vol. 1085: G.K. Immink, Asymptotics of Analytc Difference Equa-
Vol. "1.1057:: Biifurrcation Theory and Applications. Seminar, 1983. Edited tions. V, 134 pages. 1984.
by L L. Salwadlorii. VII, 233 pages. 1984. Vol. 1086: Sensitivity of Functionals with Applicaions to Engineering
Vol.il. 105*8: B. Aulbach, Continuous and Discrete Dynamics near Sciences. Proceedings, 1983. Edited by V. Konkov. V, 130 pages.
Maraniioldis o f Equilibria. IX, 142 pages. 1984. 1984
Vol.>l. 10599: Serminaire de Probability XVIII, 1982/83. Proceedings. Vol. 1087: W. Narkiewicz, Uniform Distribute of Sequences of
Editlite parr J. Aztema et M. Yor. IV, 518 pages. 1984. Integers in Residue Classes. VIII, 125 pages. H84.
Vol.)l. 1060): Tropology. Proceedings, 1982. Edited by L. D. Faddeev and Vol. 1088: A.V. Kakosyan, L B . Klebanov, J.A. Melamed, Charac-
A. A A. Mall'cew. V/l, 389 pages. 1984. terization of Distributions by the Method of Inensively Monotone
Vol.il. 10611: Serminaire de Theorie du Potentiel. Paris, No. 7. Pro- Operators. X, 175 pages. 1984.
c e e d i n g s . D)ire<cteurs: M. Brelot, G. Choquet et J. Deny. Redacteurs: Vol. 1089: Measure Theory, Oberwolfach 1983. Froceedings. Edited
F. HHirsch-, et G. Mokobodzki. IV, 281 pages. 1984. by D. Kolzow and D. Maharam-Stone. XIII, 327 >ages. 1984.
Vol. 1090: Differential Geometry of Submanifolds. Proceedings, Vol. 11,7: D.J. Aldojs, J.A. Ibregirnov, J. Jacod, Fcole d'Et* de
1984. Edited by K. Kenmotsu. VI, 132 pages. 1984. Probabilites de Saint-Flour XIII - 1983. Edite par R L Hennequin.
IX, 409 pages. 1985.
Vol. 1091: Multifunctions and Integrands. Proceedings, 1983. Edited
by G. Salinetti. V, 234 pages. 1984. Vol. 1118: Grossissements de filtrations: exemples et applications.
Seminaire, 1982/83. Edit6 par Th. Jeulin et M. Yor. V, 315 pages. 1985.
Vol. 1092: Complete Intersections. Seminar, 1983. Edited by S. Greco
and R. Strano. VII, 299 pages. 1984. Vol. 1119: Recent Mathematical Methods in Dynamic Programming.
Proceedings, 1984. Edited by I. Capuzzo Dolcetta, W.H. Fleming
Vol. 1093: A. Prestel, Lectures on Formally Real Fields. XI, 125 pages. and T.Zolezzi. VI, 202 pages. 1985.
1984.
Vol. 1120: K. Jarosz. Perturbations of Banach Algebras. V, 118 pages.
Vol. 1094: Analyse Complexe. Proceedings, i983. Ldite par E. Amar, 1985.
R. Gay et Nguyen Thanh Van. IX, 184 pages. 1984.
Vol. 1121: Si ngularities and Constructive M ethods for Their Treatment
Vol. 1095: Stochastic Analysis and Applications. Proceedings, 1983. Proceedings, 1983. Edited by P. Grisvard, W. Wendland and J.R.
Edited by A. Truman and D. Williams. V, 199 pages. 1984. Whiteman. IX, 346 pages. 1985.
Vol. 1096: Theorie du Potentiel. Proceedings, 1983. Edite par G. Vol. 1122: Number Theory. Proceedings, 1984. Edited by K. A<ladi.
Mokobodzki et D. Pinchon. IX, 601 pages. 1984. VII, 217 pages. 1985.
Vol. 1097: R.M. Dudley, H. Kunita, R Ledrappier, Ecole d'Ete de Vol. 1123: Seminaire de Probabilites XIX1983/84. Proceedings. Edite
Probabilites de Saint-Flour XII - 1982. Edite par P L Hennequin. X, par J. Azema et M. Yor. IV, 504 pages. 1985.
396 pages. 1984.
Vol. 1124: Algebraic Geometry, Sitges (Barcelona) 1983. Proceed-
Vol. 1098: Groups - Korea 1983. Proceedings. Edited by A.C. Kim ings. Edited by E Casas-Alvero, G.E Welters and S. Xambb-
and B.H. Neumann. VII, 183 pages. 1984. Descamps. XI, 416 pages. 1985.
Vol. 1125: Dynamical Systems and Bifurcations. Proceedings, 1984.
Vol. 1099: C. M. Ringel, Tame Algebras and Integral Quadratic Forms.
XIII, 376 pages. 1984. Edited by B.L.J. Braaksma, H.W. Broer and F. Takens. V, 129 pages.
1985.
Vol. 1100: V. Ivrii, Precise Spectral Asymptotics for Elliptic Operators Vol. 1126: Algebraic and Geometric Topology. Proceedings, 1983.
Acting in Fiberings over Manifolds with Boundary. V, 237 pages.
Edited by A. Ranicki, N.Levitt and FQuinn. V, 523 pages. 1985.
1984.
Vol. 1127: Numerical Methods in Fluid Dynamics. Edited byR Brezzi.
Vol. 1101: V. Cossart, J. Giraud, U. Orbanz, Resolution of Surface VII, 333 pages. 1985.
Singularities. VII, 132 pages. 1984. Vol. 1128: J.EIschner, Singular Ordinary Differential Operators and
Vol. 1102: A. Verona, Stratified Mappings - Structure and Triangu- Pseudodifferential Equations. 200 pages. 1985.
lability. IX, 160 pages. 1984. Vol. 1129: Numerical Analysis, Lancaster 1984. Proceedings. Edited
Vol. 1103: Models and Sets. Proceedings, Logic Colloquium, 1983, by PR. Turner. XIV, 179 pages. 1985.
Part I. Edited by G. H. Muller and M. M. Richter. VIII, 484 pages. 1984. Vol. 1130: Methods in Mathematical Logic. Proceedings, 1983. Edited
by C.A. Di Prisco. VII, 407 pages. 1985.
Vol. 1104: Computation and Proof Theory. Proceedings, Logic
Vol. 1131: K. Sundaresan, S. Swaminathan, Geometry and Nonlinear
Colloquium, 1983, Part II. Edited by M. M. Richter, E Borger, W. Ober-
Analysis in Banach Spaces. Ill, 116 pages. 1985.
schelp, B. Schinzel and W. Thomas. VIII, 475 pages. 1984.
Vol. 1132: Operator Algebras and their Connections with Topology
Vol. 1105: Rational Approximation and Interpolation. Proceedings,
and Ergodic Theory. Proceedings, 1983. Edited by H.Araki, C.C.
1983. Edited by PR. Graves-Morris, E.B. Saff and R.S. Varga.
XII, 528 pages. 1984. Moore, §.Stratila and C. Voiculescu. VI, 594 pages. 1985.
Vol. 1133: K. C. Kiwiel, Methods of Descentfor NondifferentiableOpti-
Vol. 1106: C.T. Chong, Techniques of Admissible Recursion Theory. mization. VI, 362 pages. 1985.
IX, 214 pages. 1984.
Vol. 1134: G. P. Galdi, S. Rionero, Weighted Energy Methods in Fluid
Vol. 1107: Nonlinear Analysis and Optimization. Proceedings, 1982. Dynamics and Elasticity. VII, 126 pages. 1985.
Edited by C. Vinti. V, 224 pages. 1984.
Vol. 1135: Number Theory. Seminar, 1983-1984. Edited by D.V.
Vol. 1108: Global Analysis - Studies and Applications I. Edited by Chudnovsky, G. V. Chudnovsky, H. Cohn and M. B. Nathanson. V, 283
Yu.G. Borisovich and Yu.E. Gliklikh. V, 301 pages. 1934.
pages. 1985.
Vol. 1109: Stochastic Aspects of Classical and Quantum Systems. Vol. 1136: Quantum Probability and Applications II. Proceedings,
Proceedings, 1983. Edited by S. Albeverio, P. Combe and M. Sirugue- 1984. Edited by L. Accardi and W. von Waldenfels. VI, 534 pages.
Collin. IX, 227 pages. 1985. 1985.
Vol. 1137: Xiao G., Surfaces fibrees en courbes de genre deux. IX, 103
Vol. 1110: R. Jajte, Strong Limit Theorems in Non-Commutative
pages. 1985.
Probability. VI, 152 pages. 1985.
Vol. 1138: A. Ocneanu, Actions of Discrete Amenable Groups on von
Vol. 1111: Arbeitstagung Bonn 1984 Proceedings. Edited by F. Hirze- Neumann Algebras. V, 115 pages. 1985.
bruch, J. Schwermer and S. Suter. V, 481 pages. 1985. Vol. 1139: Differential Geometric Methods in Mathenatical Physics.
Vol. 1112: Products of Conjugacy Classes in Groups. Edited by Proceedings, 1983. Edited by H. D. Doebner and J. D. Hennig. VI, 337
Z. Arad and M. Herzog. V, 244 pages. 1985. pages. 1985.
Vol. 1140: S. Donkin, Rational Representations of Alcebraic Groups.
Vol. 1113: P. Antosik, C. Swartz, Matrix Methods in Analysis. IV, VII, 254 pages. 1985.
114 pages. 1985. Vol. 1141: Recursion Theory Week. Proceedings, 1984. Edited by
H.-D. Ebbinghaus, G.H. Muller and G.E. Sacks. IX, 413 pages. 1985.
Vol. 1114: Zahlentheoretische Analysis. Seminar. Herausgegeben
von E Hlawka. V, 157 Seiten. 1985. Vol. 1142: Orders and their Applications. Proceedings, 1984. Edited
by I. Reiner and K. W. Roggenkamp. X, 306 pages. 1935.
Vol. 1115: J. Moulin Ollagnier, Ergodic Theory and Statistical Mechan-
Vol. 1143: A. Krieg, Modular Forms on Half-Spaces of Quaternions.
ics. VI, 147 pages. 1985.
XIII, 203 pages. 1985.
Vol. 1116: S. Stolz, Hochzusammenhangende Mannigfaltigkeiten Vol. 1144: Knot Theory and Manifolds. Proceedings, 983. Edited by
und ihre Rander. XXIII, 134 Seiten. 1985. D. Rolfsen. V, 163 pages. 1985.
Vol. 1145: G . Winkleer, C h o q u e t O r d e r a n d S i m p l i c e s . V I , 143 p a g e s . Vol. ' 3 : H: H'elfslfs, M . K n u s c s c h o c a l l y Semialobra>raic S p e s . XVI,
1985. 3 2 9 pes.;s. 35. 5.
Vol. 1146: S e m i n a i r e ; d ' A l g e b n e Paul Dubreiil en Marie-Paule Malliavin.

Proceedings, 1 9 8 3 - 1 9 8 4 . Edite par M.-P. Malliavin. IV, 4 2 0 p a g e s .
1985.
iVol. 1147: |M. W s c h e s b o r , S u r f a c e s Aleatoirejs. V I I , 111 p a g e s . 1 9 8 5 .
Vol. 1148: M a r k A . Kcon, Probability D i s t r i b u t i o n s in Q u a n t u m Statistical

Mechanics. V, 121 paages. 1 9 8 5 .
Vol. 1149: Universal A l g e b r a a n d Lattice TheoiTy. P r o c e e d i n g s , 1984.

Edited by S . D. C o m o e r . V I , 2 8 2 p a g e s . 1985>.
Vol. 1150: B. Kawohhl, R e a r r a n g e m e n t s a n d C c n v e x i t y of Level S e t s in

PDE. V, 1 3 6 p a g e s . 1985.
Vol 1151: Ordinary a n d Partial Differential E q u a t i o n s . P r o c e e d i n g s ,

1984. Edited by B . D . \ S l e e m a n a n d R.J. Jarvus. X I V , 3 5 7 p a g e s . 1 9 8 5 .
Vol. 1152: H. W i d o D m , A s y m p t o t i c E x p a n s i o n s for Pseudodifferential

O p e r a t o r s on B o u n d d e d D o m a i n s . V, 150 p a g e s . 1985.
Vol. 1153: Probabilityy in B a n a c h S p a c e s V. P r o c e e d i n g s , 1984. Edited

by A. B e c k , R. Duddley, M . H a h n , J. K u e l b s a n d M . M a r c u s . V I , 4 5 7
pages. 1 9 8 5 .
Vol. 1154: D . S . Naaidu, A . K . Rao, S i n g u l a r P e r t u b a t i o n Analysis of

Discrete C o n t r o l Sysstems. IX, 1 9 5 p a g e s . 1 9 8 5 .
Vol. 1155: Stability P r o b l e m s for Stochasitic M o d e l s . P r o c e e d i n g s ,

1984. Edited by V / . V . Kalashnikov a n d V . M . Zolotarev. V I , 4 4 7
pages. 1985.
Vol. 1156: Global [Differential G e o m e t r y a n d G l o b a l Analysis 1984.

Proceedings, 1984. Edited by D. F e r u s , R..B. G a r d n e r , S. H e l g a s o n
and U. S i m o n . V, 33S9 p a g e s . 1 9 8 5 .
Vol. 1157: H. Levine., Classifying Immersions i n t o I R over S t a b l e M a p s

4
of 3-Manifolds i n t o I F R . V, 1 6 3 pages. 1 9 8 5 .
2
Vol. 1153: Stochastinc P r o c e s s e s - M a t h e m a t i c s and Physics. P r o c e e -

dings, 1 9 8 4 . E d i t e d ; by S. Albeverio, P h . B l a n c h a r d a n d L. Streit. V I ,
2 3 0 p a g e s . 1985.
Vol. 1159: Schrodincger O p e r a t o r s , C o m o 1 9 8 4 . Seminar. Edited by S .

Graffi. VIII, 2 7 2 pagees. 1 9 8 5 .
; Vol. 1160: J.-C. v a m der M e e r , The H a m i l t o n i a n H o p f Bifurcation. V I ,

? 115 pages. 1985.
\ Vol. 1 1 6 1 : Harmonwc M a p p i n g s a n d M i n i m a l I m m e r s i o n s , Montecatini

; 1984. Edited by E. (Giusti. V I I , 2 8 5 p a g e s . 1 9 8 5 .
I Vol. 1162: S . J . L . w a n Eijndhoven, J. d e G r a a f , Trajectory Spaces,

| Generalized Functicons a n d U n b o u n d e d O p e r a t o r s . IV, 2 7 2 pages.
I 1985.
t Vol. 1163: Iteration Theory a n d its F u n c t i o n a l Equations. P r o c e e d i n g s ,
\ 1984. Edited by R. ILiedl, L. Reich and G y . T a r g o n s k i . VIII, 2 3 1 p a g e s .
I 1985.
s Voi. 1164: M. M e i s c h i a r i , J . H . R a w n s l e y , S . S a l a m o n , G e o m e t r y
I Seminar "Luigi B i a m c h i " II - 1984. Edited by E. Vesentini. V I , 2 2 4
: pages. 1985.
[
Vol. 1165: Seminarr on Deformations. P r o c e e d i n g s , 1 9 8 2 / 8 4 . Edited

by J.-Lawrynowicz. IX, 3 3 1 p a g e s . 1985.
Vol. 1166: Banach S p a c e s . Proceedings, 1 9 8 4 . Edited by N. Kalton

and E. Saab. VI, 1999 p a g e s . 1985.
Vol. 1167: Geomettry and Topology. P r o c e e d i n g s , 1 9 8 5 . Edited by J.

Alexander and J. Hsarer. V I , 2 9 2 pages. 1 9 8 5 .
Vol. 1168: S . S . Agiaian, H a d a m a r d M a t r i c e s a n d their Applications. Ill,

2 2 7 pages. 1985.
Vol. 1169: W . A . Liight, E . W . Cheney, A p p r o x i m a t i o n Theory in Tensor

Product Spaces. V I I , 157 p a g e s . 1985.
Vol. 1170: B . S . T h i o m s o n , Real Functions. V I I , 2 2 9 p a g e s . 1 9 8 5 .
Vol. 1 1 7 1 : Polyncimes Orthogonaux et A p p l i c a t i o n s . P r o c e e d i n g s ,

1984. Edite par C.. Brezinski, A. Draux, A . P . M a g n u s , P. Maroni et A.
Ronveaux. XXXVII, 5 8 4 p a g e s . 1985.
Vol. 1172: Algebraaic Topology, Gottingen 1 9 8 4 . P r o c e e d i n g s , 1984.

Edited by L. Smith.. VI, 2 0 9 pages. 1985.

Lecture Notes in Mathematics: For Information About Vols. 1-980 Please Contact Your Bookseller or Springer-Verlag

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture Notes in Mathematics: For Information About Vols. 1-980 Please Contact Your Bookseller or Springer-Verlag

Uploaded by

Copyright:

Available Formats

Lecture Notes in Mathematics

Vol 1008 A l g e b r a i c G e o m e t r y . P r o c e e d i n g s , 1981. E d i t e d by J. Dol-

Vol. 9 8 1 : V a l u e D i s t r i b u t i o n Theory. P r o c e e d i n g s , 1981. E d i t e d by Vo\. :

Vol. 9 8 2 : Stability P r o b l e m s for S t o c h a s t i c M o d e l s . P r o c e e d i n g s , Vrjl 1010 J E. D i e s , C h a i n e s d e M a r k o v sur les p e r m u t a t i o n s . IX,

Vol. 9 8 9 : A. B. M i n g a r e l l i , Volterra-Stieltjes Integral E q u a t i o n s a n d Vol. 1016: A l g e b r a i c G e o m e t r y . P r o c e e d i n g s , 1 9 8 2 . E d i t e d by M.

Vol. 9 9 0 : P r o b a b i l i t y in B a n a c h S p a c e s IV. P r o c e e d i n g s , 1 9 8 2 . Vol 1017: Equadiff 8 2 . P r o c e e d i n g s , 1 9 8 2 . Edited by H . W . K n o b l o c h

Vol. 9 9 1 : B a n a c h S p a c e T h e o r y a n d its A p p l i c a t i o n s . P r o c e e d i n g s , Vol 1018: G r a p h Theory, t a g o w 1981. P r o c e e d i n g s , 1981. E d i t e d by

Vol. 9 9 6 : Invariant Theory. P r o c e e d i n g s , 1 9 8 2 . Edited by F. G h e r a r - Vol 1023: S. M c A d a m , A s y m p t o t i c P r i m e D i v i s o r s . IX, 118 p a g e s .

Vol. 9 9 8 : R e c e n t D e v e l o p m e n t s in t h e A l g e b r a i c , A n a l y t i c a l , a n d Vol 10215: D. Tanre, H o m o t o p i e R a t i o n n e l l e : M o d e l e s d e C h e n ,

Ph B i a n c h a r d a n d L. Streit. IX, 213 p a g e s . 1 9 8 3 .

continued on page 335

Locally Semialgebraic Spaces

Mathematics Subject Classification (1980): 1 4 G 3 0 , 54E99, 5 5 Q 0 5 , 5 7 R 0 5

ISBN 3-540-16060-4 Springer-Verlag Berlin Heidelberg New York Tokyo

The primary occupation o f r e a l a l g e b r a i c geometry, o r b e t t e r "semialge-

braic geometry", i s t o s t u d y the s e t of s o l u t i o n s of a f i n i t e system of

polynomial i n e q u a l i t i e s i n a f i n i t e number o f v a r i a b l e s over the field

3R o f r e a l numbers. One wants t o do t h i s i n a conceptual way, not always

m e n t i o n i n g the p o l y n o m i a l d a t a , s i m i l a r l y as i n a l g e b r a i c geometry, say

o v e r €, where one most o f t e n a v o i d s working e x p l i c i t e l y w i t h the

systems o f p o l y n o m i a l e q u a l i t i e s (and n o n - e q u a l i t i e s f * 0) involved.

But a semialgebraic geometry which d e s e r v e s i t s name s h o u l d be able to

work - a t l e a s t - o v e r an a r b i t r a r y r e a l c l o s e d f i e l d R instead of the

f i e l d JR . Such f i e l d s are u s e f u l and even u n a v o i d a b l e i n semialgebraic

geometry f o r much the same r e a s o n as a l g e b r a i c a l l y c l o s e d f i e l d s of

; o v e r <C, as soon as one t r i e s to avoid transcendental techniques or even

' f : V->W be an a l g e b r a i c map between i r r e d u c i b l e v a r i e t i e s over 1R . T h i s

y i e l d s , by restriction, a c o n t i n u o u s map f _ : V(]R) -+WCIR) between the

s e t s o f r e a l p o i n t s . We assume t h a t W(3R) i s Z a r i s k i dense i n W which

.means t h a t W(3R) contains non s i n g u l a r points or, e q u i v a l e n t l y , that the

[ f u n c t i o n f i e l d 3R(W) i s formally r e a l . The generic f i b r e X of f, i . e .

\ i s an a l g e b r a i c scheme o v e r the function f i e l d 3R(W) of W, which con-

\ tains a l o t of information about f and f T O . But i t may be too difficult

[ t o study X, s i n c e the field IR(W) i s u s u a l l y very complicated. In alge-

| b r a i c geometry one often replaces X by the a l g e b r a i c v a r i e t y X obtained

I from X by e x t e n s i o n o f the base f i e l d ]R(W) t o the a l g e b r a i c c l o s u r e C

o f 3E(W) . I t i s much e a s i e r t o s t u d y the "geometric g e n e r i c fibre" X

about f from X. But i n s e m i a l g e b r a i c geometry t h i s p r o c e d u r e i s not ad-

visable, s i n c e most r e a l phenomena i n X w i l l be d e s t r o y e d i n X. Instead

o f X one s h o u l d study t h e v a r i e t i e s X , Q o b t a i n e d from X by base exten-

s i o n from IR(W) to the r e a l c l o s u r e s R a o f IR(W) with r e s p e c t to the

v a r i o u s o r d e r i n g s a o f the f u n c t i o n f i e l d IR(W) , and the s e t s o f ratio-

a r e "as near as p o s s i b l e " t o C and n e v e r t h e l e s s we may hope to d e t e c t

some o f t h e r e a l phenomena o f X, and u l t i m a t e l y of f , i n t h e s e t s

The v a r i e t y X i s the p r o j e c t i v e l i m i t o f the schemes f 1

U r u n n i n g through t h e Z a r i s k i - o p e n s u b s e t s o f W, s i n c e these U are the

p r o j e c t i v e l i m i t of the f i b r e products V ^ u , w i t h r e s p e c t t o the etale

morphisms cp : U -»W from a r b i t r a r y v a r i e t i e s U over IR to W (U * 0 , but

U(IR) may be empty), s i n c e t h e s e morphisms cp a r e the e t a l e neighbour-

hoods o f n. How about the X ? a An o r d e r i n g a of IR(W) corresponds uni-

quely t o an u l t r a f i l t e r F i n t h e Boolean lattice )T(W ( IR ) o f semialge-

braic s u b s e t s o f W(IR) such t h a t every A €F has a non empty i n t e r i o r A

i n the strong topology (= c l a s s i c a l t o p o l o g y on W(IR)), which means

t h a t A i s Z a r i s k i dense i n W, c f . [B, 8.11],[Br, § 4 ] . (A r a t i o n a l func-

and p o s i t i v e on some s e t A C F ) . I t t u r n s o u t t h a t X a i s the projective

l i m i t o f t h e f i b r e p r o d u c t s Vx^u w i t h r e s p e c t t o those e t a l e morphisms

cp : U ->W such t h a t cp(U(IR)) €F. (N.B. cp(U(IR)) i s semialgebraic.) This

is due t o the f a c t t h a t R a can be i n t e r p r e t e d as the u n i o n o f t h e rings

o f Nash f u n c t i o n s & (U) W on t h e v a r i o u s smooth open s e t s U €F, c f . [Ry].

Much more can be s a i d about a geometric i n t e r p r e t a t i o n over IR o f the

tative rings i n v e n t e d by M. C o s t e and M.F. Coste-Roy provide exactly

the r i g h t language t o u n d e r s t a n d a l l t h i s , c f . [CR], [Ry], and t h e

[L, § 4 , § 7 ] , [Br, §3, § 4 ] , [ K ] , [BCR, Chap. 7 ] .