Access the full text.
Sign up today, get DeepDyve free for 14 days.
ON THE STRUCTURE OF CERTAIN SEMI-GROUPS OF SPHERICAL KNOT CLASSES By BARRY MAZUR w L Introduction. The problem of classification of k-sphere knots in r-spheres is the problem of classifying "knot pairs": S= ($1, $2), where $2 is an oriented combinatorial r-sphere, $1 a subcomplex of S 2 (isomorphic to a standard k-sphere), and the pair S is considered equivalent to S' (S.~ S') if there is a combinatorial orientation-preserving homeomorphism of S x onto S' 1 bringing S 2 onto S~. Thus it is the problem of classifying certain relative combinatorial structures. The set of all such, for fixed k and r, will be called Z~, and can be given, in a natural manner, the structure of a semi-group. There is a certain sub-semi-group of Z~ to be singled out -- the semi-group S~ of all pairs S ----- ($t, $2) where $1 is smoothly imbedded in S 2 (locally unknotted). In this paper I shall define a notion of equivalence (which I call .-equivalence) between knot pairs which is (seemingly) weaker than the equivalence defined above. Two knot pairs S and S' are .-equivalent if (again) there is an orientation-preserving homeomorphism ~0 : $2-+S~ bringing $1 onto S~. However q~ is required to be combinatorial (not on all of S~, as before, but) merely on S 2 = $2--(pl, ..., p,), where Pl, 9 -, P,,~ $2, where S 2 is considered as an open infinite complex. Thus .-equivalence neglects some of the combinatorial structure of the pair (S~, $2). The set of all ,-equivalence classes of knot pairs forms a semi-group again, called *Z~. 9 r * r Finally the subsemi-group of smoothly imbedded knots in Z k I call S k. The purpose of this paper is to prove a generalized knot theoretic restatement of lemma 3 in [x]. INVERSE T~EOR~.M: A knot S~ is invertible if and only if it is .-trivial. And in application, derive the following fact concerning the structure of the knot semi-groups: 9 * r There are no inverses m S k. 111 20 BARRY MAZUR w .7. Tet:.~inology. My general use of combinatorial topology terms is as in [~,]. It is clear what is meant by the "usual" or "standard" imbedding of a k-sphere or a k-cell in E ~. Similarly an unknotted sphere or disc in E r means one that may be thrown onto the usual by a combinatorial automorphism of E ~. DEFINITION I. Let M k be a subcomplex (a k-manifold) of E ~. Then M k is locally unknotted at a point m (m~M) if the following condition is met with: i) There is an r-simplex A* drawn about m so that ArnMcSt(m), and A'nM is then a k-cell BkcA r, and OBkcOA ~. 2) There is a combinatorial automorphism of A *, sending B k onto the "standard k-cell in A r''. M is plain locally unknotted if it is locally unknotted at all points. Semi-Groups: All semi-groups to be discussed will be countable, commutative, and possess zero elements. DEFINITION 2. A semi-group F is positive if." X + Y ----- o implies X = o (i.e.if F has no inverses). DEFINITION 3" A minimal base of a semi-group F is a collection J = (ll, ... ) of elements of F such that every element of F is a sum of elements in J, and there is no smaller J' cJ with the same property. DEFINITION 4" A prime element p in the semi-group F is an element for which p=x+y implies either x=o or y=o. Clearly, if a positive semi-group F possesses a minimal base, that minimal base has to be precisely the set of primes in F, and F has the property that every element is expressible as a finite sum of primes. DEFINITION 5" An element x~F is invertible if there is a y~F such that x +y= o. w 3" (*)'h~176176 DEFINITION 6. A (Pl, ..-,P,)-homeomorphism, h:Er--~E" will be an orientation preserving homeomorphism which is combinatorial except at the points pieE r. It is a homeomorphism such that h[ E r- (p~) is a combinatorial map- simplicial with respect to a possibly infinite subdivision of the open complexes involved. When there is no reason to call special attention to the points Pl .... ,P,,, I shall call such: a (.)- homeomorphism. 112 ON THE STRUCTURE OF CERTAIN SEMI-GROUPS OF SPHERICAL KNOT CLASSES 2I DEFINITION 7" Two subcomplexes K, K' cE' will be called ,-equivalent (K.~K') if there is a ,-homeomorphism h of E r onto itself bringing K onto K'. (If h is a (pi)- homeomorphism I shall also say K,~K'.) To keep from using too many subscripts, whenever a (,)-equivalence comes up in a subsequent proof, I shall act as if it were a (p)-equivalence for a single point p. This logical gap, used merely as a notation- saving device, can be trivally filled by the reader. I'll say a sphere knot is *-trivial if it is ,-equivalent to the standard sphere. w 4. Knot Addition. There is a standard additive structure that can be put on Z~,, the set of combina- torial k-sphere knots in E r (two k-sphere knots are equivalent if there is an orientation- preserving combinatorial automorphism of E r bringing the one knot onto the other). (For details see [z]). I shall outline the procedure of "adding two knots" So, S1. Separate S o and S 1 by a hyperplane H (possibly after translating one of them). Take a k-simplex A i from each Si, i = o, I. And lead a "tube" from A 0 to A 1 (by "thickening" a polygonal arc joining a point p0eA0 to Pi~Ai, which doesn't intersect the S i except at Ai). Then remove the Ar and replace them by the tube T = S k-1 � I, where one end, S k-1 � o is attached to OA 0 by a combinatorial homeomorphism, and the other S k-a X x is attached to 0 A 1 similarly. The resulting knot is called the sum: S o + S~, and its knot-equivalence class is unique. If one added the point at infinity to E r, to obtain S ~, the hyperplane H would become an unknotted S ~-I c S', separating the knot S O q- S 1 into its components S O and S~. In analytic fashion, then, we can say that a k-sphere knot S c S r is split by an S ~-1 c S ~ if: I) S'-i n S is an unknotted (k-- I)-sphere knot in S. 2) S *-I is unknotted in SL 3) S'-IAS is unknotted in S '-i. Let A o and A 1 be the two complementary components of S*-inS in S, and let B be an unknotted k-disc that S*-lnS bounds in S ~-l. Then S0=AoUB , SI=AluB are knotted spheres again, and clearly S.~ S o -k Sr Thus I'll say: S r-1 splits S into S o q-S 1 ; if E 0 and E i are the complementary regions of S r-1 in S *, I'll refer to S 1 as that <~ part of S ,, lying in E D and similarly for S o. Working in the semi-group Zk, one can be slightly cruder, and say: S ~-1 .-splits S if only ~) and 3) hold. Clearly by [x], every S r-1 is .-trivial in S ~. LEMMA I: If S *-I *-splits S, and So, S 1 are constructed in a manner analogous to the above, then S .~ S o 4- S i. w 5- The Semi-Groups of Spherical Knots. This operation of addition, discussed in the previous section, turns ~ into a commu- tative semi-group with zero. Our object is to study the algebraic structure of the 113 a2 BARRY MAZUR r r * r subsemi-group SkCZ k of locally unknotted k-sphere knots. Let Z k be the semi-group of classes of spherical knots under ,-equivalence. Let G~c Z~ be the maximal subgroup of Z~, that is: the subgroup of invertible knots. INVERSE THEOREM: There is an exact sequence r r * r o--> Gk--~ Sk--> Sk---> 0 r * r (where *S~ is the image of S k in Z~) or, equivalently, a knot in S~ is ,-trivial if and only if it is invertible. w 6. Proof of the Inverse Theorem. a) If S is invertible, then S,~o. The proof is quite as in [x]. Let S+S',-~o. (,) Then consider the knots: S| =S+S'+S+S'+...up| S;,=S'+S+S'+S+...up| (See figure I) and notice: (as was done in detail in [x]) S| cU )o S I ,-~ o (p,0) =s+s" H~ H 2 H3 H4 S S' S S' P~ Fig. i LEMMA 2: There is a (,)-homeomorphism f: E'-+E r such that f: S->S +S'. 114 ON THE STRUCTURE OF CERTAIN SEMI-GROUPS OF SPHERICAL KNOT CLASSES 2 3 PROOF: Let D be the k-cell on which the addition of S to S~0 takes place. Since S'_ N o, we may transform figure I to figure 2 by a (p| g which leaves (p| everything to the left of the hyperplane H z fixed, and sends S' to the" standard k-sphere" to the right of H 1. (See figure 2.) H 1 Sk =9 (S--~ Fig. 2 Then, in figure 2, clearly one can construct an automorphism f' which leaves S fixed and sends D onto g(S')--int D. Take f-~g-qf'g, and f has the properties required, and is a (.)-homeomorphism. Therefore, by the above lemma, S,~ S + S" = S| ,~o (,) (*) and finally: S,~o (,) which proves (a). and S~o, then S is invertible. b) If S ~ S~ (v) PROOV: First observe that if k = r-- I, invertibility of knots is generally true (by [I]), and so we needn't prove anything. LEMMA 3: If k<r--I, and SeS~,,S,~o for pr then S,~o. (p) PROOF: There is an r-cell A containing S but not p. Then fl A is combinatorial, and by a standard lemma: LEMMA 4: If g : A~A' is a combinatorial homeomorphism of an r-cell AcE r to an r-cell A'C E r, then g can be extended to a combinatorial automorphism of E' (see [2]). Thus, restrictfto A, and extendf[A to a combinatorial automorphism g ofE r. This g yields the equivalence S,~o. Therefore, assume S,.~o, and psS. (p) 115 BARRY MAZUR Fig. 3 Let B be a small r-cell about p, so that C = BnS is in St(p), and hence an unknotted k-cell, by the local unknottedness of S. 0BiaS = 0C and 0C is unknotted in 0B. Letfbe the (p)-homeomorphism taking S onto the standard S *. Fig. 4 Now let D be an unknotted disc, the image of a perturbation off(C) with the properties: i) O(f(e))-=OD; ii) int Dr B ; iii) f(p) r ; iv) the knot K=Do(Sk--f(C)) is still trivial. Thenf -1 takes K to a knot K' =f-l(K), split by 0B into the sum: K'-----S +S' where S is the knot lying in the exterior component of 0B, and S' in the interior. 116 ON THE STRUCTURE OF CERTAIN SEMI-GROUPS OF SPHERICAL KNOT CLASSES 2 5 But K~o, and K'..~K where f(p)r therefore by lemma 3, f(K),~K. t (p) So: S + S' ~f(K)~ K,-~ o, and S' is invertible. S k is a positive semi-group. COROLLARY: * r So we have that S k is precisely S~ ~ modulo units ~>. * r w 7" Infinite Sums in Z k. Let X o i = i, ..., be knots representing the classes )i e E~. Define Y~ X; to be the infinite one point compactified sum of the knots X o in that order (figure 5). X~ X 2 X 3 X 1 9 . . . . .p~ %/-- Fig. 5 As it stands, X = Y- X~ will not represent a knot in X~, because X is not combina- i=1 torially imbedded (at p| ~o DEFINITION 8. ~ Xi= X converges if there is a (p| H : X-+Y, * r where Y is combinatorially imbedded. In that case, the knot class ye E k is uniquely 0o determined by the Xi~ N~, and I shall say ~ Xi =Y. ~ i=1 * r Sk" If ~Xi is in Sk, I'll say that ~)~ converges in * ~ i=l /=1 oo * r THEOREM I. If ~ )4 converges in S~, then it does so finitely. That is, there is i=1 an N such that Xi~o, i>N. * r PROOF: Notice that by the inverse theorem, there are no inverses in S k. oo Let X= ~ X~., and H:X-+Y where Y is a subcomplex of E r and H a i=I (,)-homeomorphism. 117 ~6 BARRY MAZUR Fig. 6 Let B be a ball about p' such that BnY is a disc in St(p'), and by the local unknot- tedness of Y, OB splits Y into two knots, y = yr + yI2) where YI r B is trivial, and Y,-~ Y2. Fig. 7 Now transform the situation by H-< Let B'=H-I(B), and we have that OB' ,-splits X into: X ~ X (t) + X (~I and H yields the ,-equivalences: X (1) ~ Y(1) ~ o X (2) ~.o y(2) Find an i so large that A~cint B'. Then SA~ splits X (*) further: X (1) ~ X(3) + X (4) where X (3t is the part of X (1) lying in A~. But then, by figure 6, X (3) is nothing more than: 118 ON THE STRUCTURE OF CERTAIN SEMI-GROUPS OF SPHERICAL KNOT CLASSES 27 ao X/3/~ Y X s. Passing to equivalence classes in Sk, one has: X (3) + Z (4) = o Z (3) ----- .~.Zi * r (where x the ,-equivalence class of X). But repeated application of the fact that S k has no inverses yields Z~ = o for j >1 i, which proves the theorem. REFERENCES [1] B. MAZUR, On Imbeddings of Spheres, Acta Mathematica (to appear). [2] V. K. A. M. GtrGOENgmM, Piece wise Linear Isotopy, Journal of the London Math. Soc., vol. 46. Refu le r6 novembre I959.
Publications mathématiques de l'IHÉS – Springer Journals
Published: Aug 4, 2007
You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
Copy and paste the desired citation format or use the link below to download a file formatted for EndNote
Access the full text.
Sign up today, get DeepDyve free for 14 days.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.