return

Celebratio Mathematica

Michael H. Freedman

The topological Poincaré conjecture in dimension 4 (the work of M. H. Freedman)

by Laurent Siebenmann

French to English translation by Min Hoon Kim and Mark Powell

Introduction

At the end of the sum­mer of 1981, in San Diego, M. Freed­man proved that every smooth ho­mo­topy 4-sphere M4 is homeo­morph­ic to S4. Our main goal is to give an ex­pos­i­tion of his proof. In this pa­per, every man­i­fold will be met­ris­able and fi­nite di­men­sion­al. We do not know yet wheth­er such an M4 is al­ways dif­feo­morph­ic to S4. On the oth­er hand, Freed­man proved that every to­po­lo­gic­al ho­mo­topy 4-sphere M4 (without any giv­en smooth struc­ture) is ac­tu­ally homeo­morph­ic to S4 (see be­low).

H. Poin­caré con­jec­tured that every smooth, ho­mo­topy n-sphere Mn is dif­feo­morph­ic to Sn. The first non­trivi­al case, di­men­sion 3, re­mains open (in 1982) des­pite the ef­forts of count­less math­em­aticians. An amus­ing de­tail is that the counter­example of J. H. C. White­head [e1] to his own er­ro­neous proof of this con­jec­ture will play a large role in this lec­ture (see Sec­tion 2).

J. Mil­nor [e4] dis­covered smooth man­i­folds M7 which are homeo­morph­ic to S7 but not dif­feo­morph­ic to S7 (such exot­ic spheres ex­ist in di­men­sion 7 [e13]). There­fore the above Poin­caré con­jec­ture has to be re­vised for di­men­sion 7. S. Smale [e9] es­tab­lished his the­ory of handles to prove that every smooth ho­mo­topy n-sphere is homeo­morph­ic to Sn for n6. His tech­nic­al res­ult, the h-cobor­d­ism the­or­em (see be­low) is more pre­cise. Com­bin­ing this with sur­gery tech­niques of Ker­vaire–Mil­nor [e13] es­tab­lishes the n=5 and 6 cases of the above Poin­caré con­jec­ture. M. New­man ad­ap­ted the en­gulf­ing meth­od of J. Stallings to prove the purely to­po­lo­gic­al ver­sion, that is, every to­po­lo­gic­al ho­mo­topy n-sphere is homeo­morph­ic to Sn if n5. (Smale’s sur­gery meth­od has also been ad­ap­ted to the to­po­lo­gic­al cat­egory [e29].) In sum­mary, the Poin­caré con­jec­ture is es­sen­tially re­solved in di­men­sion 5, is not re­solved in di­men­sion 3, and is par­tially re­solved in di­men­sion 4.

We sketch a proof of Freed­man’s the­or­em which im­plies the to­po­lo­gic­al clas­si­fic­a­tion of smooth, simply con­nec­ted closed 4-man­i­folds and many oth­er res­ults of fun­da­ment­al im­port­ance. Let V and V be two such man­i­folds. Sup­pose that there is an iso­morph­ism Θ:H2(V)H2(V) which pre­serves the in­ter­sec­tion forms. (Note that V is a ho­mo­topy 4-sphere if and only if H2(V)=0.)

The­or­em A: In this situ­ation, Θ is real­ised by a homeo­morph­ism VV.

Proof.  It is not dif­fi­cult to real­ise Θ by a ho­mo­topy equi­val­ence g:VV [e25]. Sur­gery the­ory [e14], [e21] gives a com­pact 5-man­i­fold W with bound­ary W=VV such that the in­clu­sions VW and VW are ho­mo­topy equi­val­ences and such that the re­stric­tion r|V:VV of the re­trac­tion r:WV is ho­mo­top­ic to g. The com­pact tri­ad (W;V,V) is called an h-cobor­d­ism. Smale’s the­ory of handles tries to im­prove a Morse func­tion f:(W;V,V)([0,1];0,1) to ob­tain a situ­ation where f has no crit­ic­al points, that is, f is a smooth sub­mer­sion. Then W is a fibre bundle over [0,1] (a re­mark of Ehresmann) and hence W is dif­feo­morph­ic to V×[0,1]. We are go­ing to find a to­po­lo­gic­al sub­mer­sion f which shows that W is a to­po­lo­gic­al fibra­tion on I (see [e23], Sec­tion 6) so that W is homeo­morph­ic to V×[0,1].  ◻

In par­tic­u­lar, we will prove the simply con­nec­ted, to­po­lo­gic­al 5-di­men­sion­al h-cobor­d­ism the­or­em.

The­or­em B: Every smooth, com­pact, simply con­nec­ted, 5-di­men­sion­al h-cobor­d­ism (W;V,V) is to­po­lo­gic­ally trivi­al. That is, W is homeo­morph­ic to V×[0,1].

For n6, in­stead of 5, Smale’s h-cobor­d­ism the­or­em gives the stronger con­clu­sion that W is dif­feo­morph­ic to V×[0,1]. In di­men­sion 5, his meth­ods ap­ply, but leav­ing to prove that W is dif­feo­morph­ic to V×[0,1]. The fol­low­ing prob­lem is not yet re­solved.

Re­main­ing smooth prob­lem: (Unresolved in February 1982.)  Let S=S1SkandS=S1Sk be two fam­il­ies of dis­jointly em­bed­ded 2-spheres in a simply con­nec­ted 4-man­i­fold M (in fact f1(a point)) in such a way that the ho­mo­lo­gic­al in­ter­sec­tion num­ber SiSj=±δi,j. Can one re­duce SS to k points of in­ter­sec­tion (smooth and trans­verse) by a smooth iso­topy of S in M?

Sim­il­arly, to ob­tain the fact that W is homeo­morph­ic to V×[0,1], we claim (see [e15] and [e29], Es­say III) that it suf­fices to solve the fol­low­ing prob­lem.

Re­main­ing to­po­lo­gic­al prob­lem: (Resolved here.)  With the data of the smooth prob­lem, re­duce SS to k points by a to­po­lo­gic­al iso­topy of S in M, that is giv­en by an am­bi­ent iso­topy ht, 1t1, of Id|M fix­ing a neigh­bour­hood of k-points of SS.

Whit­ney in­tro­duced a nat­ur­al meth­od for solv­ing these prob­lems. In the mod­el (R2;A,A), (this is a straight line A cut­ting a para­bola A in two points), we can dis­en­gage A from A by a smooth iso­topy with com­pact sup­port (that is, fix­ing a neigh­bour­hood of ). One elim­in­ates thus the two in­ter­sec­tion points. We de­duce that in the sta­bil­ised Whit­ney mod­el, (R4;A+,A+)=(R2×R2;A×0×R,A×R×0), there is an iso­topy with com­pact sup­port that makes the plane A+ dis­joint from the plane A+, de­let­ing the two trans­verse in­ter­sec­tion points between A+ and A+.

We call a smooth (resp. to­po­lo­gic­al) Whit­ney pro­cess, a smooth em­bed­ding (resp. a to­po­lo­gic­al em­bed­ding) of a dis­joint uni­on of cop­ies of the mod­el (R4;A+,A+), whose im­age con­tains SS(k points). Such a pro­ced­ure would clearly give the de­man­ded iso­topy to re­solve the re­main­ing smooth prob­lem (re­spect­ively, the re­main­ing to­po­lo­gic­al prob­lem).

The­or­em C: (Casson–Freedman.)  In this con­text, after a pre­lim­in­ary smooth iso­topy of S in M, (adding in­ter­sec­tion points with S by fin­ger moves, far from SS), the to­po­lo­gic­al Whit­ney pro­cess be­comes pos­sible.

The first step of the proof (1973–1976) is due to A. Cas­son. Let B be a smooth, com­pact 2-disc in the bound­ary com­pon­ent of R2AA. The product B×R2 is an open, em­bed­ded 2-handle (as a closed sub­man­i­fold) in the Whit­ney mod­el, and dis­joint from A+A+. In B×R2, Cas­son con­struc­ted cer­tain open sets H=B×R2Ω with bound­ary H=B×R2, that we call open Cas­son handles. (See Sec­tion 2 for the pre­cise defin­i­tion). We are again un­able (in Feb­ru­ary 1982) to de­cide wheth­er H is dif­feo­morph­ic to B×R2 or not. Re­pla­cing B×R2 by HB×R2 in this Whit­ney mod­el (R4;A+,A+) we have an open set (R4Ω;A+,A+), that we call the Whit­ney–Cas­son mod­el. By a re­mark­able in­fin­ite pro­cess, Cas­son proved the fol­low­ing.

The­or­em D: (Casson [e38], compare [e33].)  After a pre­lim­in­ary smooth iso­topy of S in M, one can find in (M;S,S) smoothly em­bed­ded, dis­joint Whit­ney–Cas­son mod­els so that the mod­els con­tain all the points of SS ex­cept the k in­ter­sec­tion points.

The the­or­em of Cas­son and Freed­man now fol­lows from the the­or­em that we will dis­cuss.

The­or­em E: (Freedman, 1981.)  Every open Cas­son handle is homeo­morph­ic to B2×R2. There­fore, the Whit­ney mod­el (R4;A+,A+) is homeo­morph­ic to (R4Ω;A+,A+).

The non­com­pact ver­sion of The­or­em B is also im­port­ant.

The­or­em F: Let (W;V,V) be a simply con­nec­ted, prop­er smooth 5-di­men­sion­al h-cobor­d­ism with a fi­nite num­ber of ends and a trivi­al π1-sys­tem at each end. Then W is homeo­morph­ic to V×[0,1].

The dif­fi­cult proof pro­posed by Freed­man (Oc­to­ber 1981) ini­ti­ates the proof of the prop­er s-cobordism the­or­em sketched in [e18], while avoid­ing per­form­ing two Whit­ney pro­cesses, in view of the loss of dif­fer­en­ti­ab­il­ity oc­ca­sioned by The­or­em C.

This gives (com­pare [1] and [e34]) the to­po­lo­gic­al clas­si­fic­a­tion of closed, simply con­nec­ted to­po­lo­gic­al 4-man­i­folds that ad­mit (do they all?) a smooth struc­ture in the com­ple­ment of a point. They are clas­si­fied by their in­ter­sec­tion form on H2, to­geth­er with the Kirby–Sieben­mann ob­struc­tion x [e29]; every un­im­od­u­lar forms over Z is real­ised, as well as every xZ2, ex­cept that for even forms, xσ/8Z2. Every to­po­lo­gic­al 4-man­i­fold V which is ho­mo­topy equi­val­ent to S4 is in this class, be­cause V{point} is con­tract­ible and thus V{point} can be im­mersed in­to R4 (com­pare [e29]).

It also fol­lows (see [1], [e34]) that every smooth ho­mo­logy 3-sphere V (that is, H(V)H(S3)) is the bound­ary of a con­tract­ible to­po­lo­gic­al 4-man­i­fold W.

Report

Mike Freed­man an­nounced his proof of the to­po­lo­gic­al Poin­caré con­jec­ture in Au­gust 1981 at the AMS con­fer­ence at UC­SB where D. Sul­li­van was giv­ing a lec­ture series on Thur­ston’s hy­per­bol­iz­a­tion the­or­em. His ar­gu­ment was very bril­liant, but not yet com­pletely wa­ter­tight.

A large group of ex­perts then for­mu­lated cer­tain ob­jec­tions, which led to the state­ment of the ap­prox­im­a­tion the­or­em (The­or­em 5.1). However, Freed­man already had in his head his trick of rep­lic­a­tion, and in a few days, his im­pos­ing form­al proof was born.

In the mean­time, R. D. Ed­wards had found a mis­take in the shrink­ing ar­gu­ments (see Sec­tion 4) and, be­ing an ex­pert in this meth­od, had re­paired the mis­take even be­fore point­ing it out. (I think that he in­tro­duced in par­tic­u­lar the re­l­at­ive shrink­ing ar­gu­ments.) At the end of Oc­to­ber 1981, Freed­man ex­plained the de­tails of his proof, with charm and pa­tience, at a spe­cial con­fer­ence at Uni­versity of Texas at Aus­tin (the school of R. L. Moore) be­fore an audi­ence of spe­cial­ists, in­clud­ing, in the place of hon­our, Cas­son and RH Bing, cre­at­ors of the two the­or­ies es­sen­tial in the proof.

This pa­per relates the proof giv­en in Texas, with im­prove­ments in de­tail ad­ded in be­hind the scenes. Already in 1981, R. An­cel [2] had cla­ri­fied and im­proved the com­plex­it­ies in book­keep­ing of the ap­prox­im­a­tion the­or­em (The­or­em 5.1). In par­tic­u­lar, he was able to re­duce a hy­po­thes­is of Freed­man de­mand­ing that the preim­ages of the sin­gu­lar point con­sti­tute a null de­com­pos­i­tion, show­ing that S(f) count­able or of di­men­sion 0 [e26] suf­fices. J. Walsh con­trib­uted cer­tain sim­pli­fic­a­tions to the shrink­ing ar­gu­ments (end of Sec­tion 4). W. Eaton sug­ges­ted to me the 4-balls that help to un­der­stand re­l­at­ive shrink­ing (Lemma 4.9 and Pro­pos­i­tion 4.11). I pro­posed a glob­al co­ordin­ate sys­tem of a Cas­son handle. (It was ini­tially ne­ces­sary to em­bed the fron­ti­er of a handle in there.)

My ex­pos­i­tion (Janu­ary 1982) does not seem to have changed es­sen­tially from my memor­ies of Texas. Only my con­struc­tion of cor­rect­ive 2-discs (the D(α) of Sec­tion 3.9) de­vi­ates, prob­ably for reas­ons of taste. I am in­debted to A. Mar­in for his broth­erly and in­sight­ful com­ments.

1. Terminology

This ter­min­o­logy is used from now on ex­cept when oth­er­wise in­dic­ated. All spaces ad­mit a met­ric, de­noted gen­er­ally by d. Maps are all con­tinu­ous. The sup­port of a map f:XX is the clos­ure of {xXf(x)x}. The sup­port of a ho­mo­topy, or an iso­topy ft:XX (0t1) is the clos­ure of {xXft(x)x for some t[0,1]}. For a sub­set A, define the clos­ure A¯, the in­teri­or A˚ and the fron­ti­er δA, al­ways with re­spect to the un­der­stood am­bi­ent space (the largest in­volved). If A is a man­i­fold, it is of­ten ne­ces­sary to dis­tin­guish A˚ from its form­al in­teri­or IntA and δA from the form­al bound­ary A.

A de­com­pos­i­tion D of a space X will be a col­lec­tion of com­pact dis­joint sub­sets in X that is USC (up­per semi con­tinu­ous); the quo­tient space X/D is ob­tained by identi­fy­ing each ele­ment of D to a point (see [e27] for a met­ric). The quo­tient map XX/D is closed, which is ex­actly equi­val­ent to the USC prop­erty.

The set of con­nec­ted com­pon­ents of a space X is de­noted by π0(A). If A is com­pact, π0(A) is at the same time a de­com­pos­i­tion of A for which the quo­tient A/π0(A) is a com­pact set of di­men­sion 0 (totally dis­con­tinu­ous), that is iden­ti­fied with π0(A) as a set. If AX, π0(A) gives a de­com­pos­i­tion of X whose quo­tient space is de­noted by X/π0(A). The en­d­point com­pac­ti­fic­a­tion will ap­pear in Sec­tion 2.

The man­i­folds and sub­man­i­folds men­tioned will be (un­less oth­er­wise in­dic­ated) smooth. For man­i­folds, we ad­opt the usu­al con­ven­tion ([e29], Es­say I); in par­tic­u­lar, Rn is the Eu­c­lidean space with the met­ric d(x,y)=|xy|;  Bn={xRn|x|1};  I=[0,1]. A mul­tidisc is a dis­joint uni­on of fi­nitely many discs (each are dif­feo­morph­ic to B2). Sim­il­arly, for mul­ti­handle, etc. The sym­bols , and in­dic­ate a dif­feo­morph­ism, a homeo­morph­ism and a ho­mo­topy equi­val­ence, re­spect­ively.

2. Casson tower and Freedman’s mitosis

We will use two ver­sions B2 and D2 of the stand­ard smooth 2-disc {(x,y)R2x2+y21}. The stand­ard 2-handle is (B2×D2,B2×D2); its at­tach­ing re­gion is B2×D2; its skin + is B2×D2, its core is B2×0. A 2-handle is a pair (H4,H)  diffeomorphic to  (B2×D2,B2×D2). An open 2-handle is a man­i­fold dif­feo­morph­ic to B2×D2˚. For a 2-handle (pos­sibly open), the at­tach­ing re­gion, the skin and the core are defined by a dif­feo­morph­ism with the stand­ard 2-handle (per­haps the open one). In this pa­per, we can al­low ourselves to omit the pre­fix “2-”; handles of in­dex 2 ap­pear rarely. Also, we write D2˚ where we ought strictly to write IntD2.

A de­fect X in a handle (H4,H) is a com­pact sub­man­i­fold X of H4H such that:

  1. (X,X+H) is a handle where +H is the skin of the handle (H,H);
  2. (+H,X+H) is (de­gree ±1) dif­feo­morph­ic to the White­head double (B2×S1,i(B2×S1)) il­lus­trated in Fig­ure 1;
  3. in the 4-ball H4 (with roun­ded corners), the core A2 of the handle (X,X+H) is an un­knot­ted disc, that is, (H,A) is dif­feo­morph­ic to (B4,B2).

Figure 1.

Figure 2.

A mul­tide­fect X in a handle (H4,H) is a fi­nite sum and uni­on of de­fects such that for an iden­ti­fic­a­tion (H4,H) with (B2×D2,B2×D2), pro­ject to B2 the same num­ber of dis­joint discs in IntB2. A multi-de­fect X in a handle (H4,H) is a fi­nite, dis­joint uni­on iX(i)=X of 1 de­fects X(i), that, for a suit­able iden­ti­fic­a­tion (H4,H)(B2,B2)×D2, are sent, un­der the pro­jec­tion B2×D2B2, to a dis­joint uni­on of discs in B2. A mul­ti­handle (H4,H4) is a dis­joint, fi­nite sum of handles. A mul­tiple de­fect XH4 in a mul­tiple handle is a com­pact sub­set that gives rise, by in­ter­sec­tion, to a mul­tide­fect in each handle. With this data, we have the fol­low­ing.

Lemma 2.1: The tri­ad (H4X˚;H,δX) de­term­ines H4 and X in the fol­low­ing sense. If X is a mul­tide­fect in a handle (H,H) and θ:(HX˚;H,δX)(HX˚;H,δX) is a dif­feo­morph­ism, there ex­ists a dif­feo­morph­ism Θ:HH ex­tend­ing θ.
Figure 3. Twist knots.

Sketch of proof (see [e38]).  If we at­tach a mul­ti­handle (X,X) to HX˚ along the fron­ti­er δX, in such a way that there ex­ists no ex­ten­sion of θ to a dif­feo­morph­ism Θ:H(HX˚)X=H, we claim that (H,H) is dif­feo­morph­ic to (S3,solid torus) where the sol­id tor­us is tied in a non­trivi­al knot — in fact, a con­nec­ted sum of k non­trivi­al twist knots, 1k|π0(X)|.  ◻

A re­sid­ual de­fect Ω in a handle (H4,H4) is the in­ter­sec­tion of a se­quence X1X1˚X2X2˚X3 of com­pact sub­man­i­folds of H4H4 such that, for all k, (Xk,δXk) is a mul­ti­handle in which Xk+1 is a mul­tide­fect. The se­quence X1X2  is called a Rus­si­an doll of de­fects.

Figure 4.

A Cas­son handle is a pair (H4,H4) such that there ex­ists a handle (H,H) with a re­sid­ual de­fect ΩH and an open smooth em­bed­ding i:HH with im­age HΩ, which in­duces a dif­feo­morph­ism i|:HH. In oth­er words, (H,H) is dif­feo­morph­ic to (HΩ,H).

The data of (H,H), the Rus­si­an doll of de­fects Xi and i:HH, con­sti­tute what we will call a present­a­tion of a Cas­son handle (H,H). We will also de­note Hk=i1(HXk˚)andHk=H. Then, H=kHk. The man­i­fold Hk is called a tower of height k, its stages are Ej=i1(Xj1Xj) for jk. The re­stric­tion of i to Hk will be de­noted ik:HkH.

The skin of (H,H) is +H=i1(+H); moreover, by tak­ing in­ter­sec­tion with +H, we define the skin +Hk of Hk and +Ek of Ek. Sim­il­arly +Xk=Xk+H.

A Cas­son handle (H,H) is nev­er com­pact; we will of­ten en­counter the en­d­point com­pac­ti­fic­a­tion H^ of H. Re­call that the en­d­point com­pac­ti­fic­a­tion M^ of a con­nec­ted, loc­ally con­nec­ted and loc­ally com­pact space M is the Freudenth­al com­pac­ti­fic­a­tion that adds to M the com­pact 0-di­men­sion­al space Ends(M) which is the (pro­ject­ive) lim­it of an in­verse sys­tem {π0(MK)KM such that K is compact}.

By i, H^ is iden­ti­fied with the quo­tient of H4 ob­tained by crush­ing each con­nec­ted com­pon­ent of Ω to a point. (To veri­fy this, note that π0(Ω) with the com­pact to­po­logy is the (pro­ject­ive) lim­it of an in­verse sys­tem {π0(U)U is an open sub­set of H con­tain­ing Ω}.)

We re­mark that H^ is the Al­ex­an­droff com­pac­ti­fic­a­tion by a point, ex­actly when ΩH is con­nec­ted, or if each suc­cess­ive mul­tiple de­fect Xi is a single de­fect. The read­er who feels dis­com­bob­u­lated by all the com­plex­it­ies to come may be in­ter­ested in re­strict­ing them­selves at first to this case, which already con­tains all the geo­met­ric ideas.

H^ has all the loc­al ho­mo­lo­gic­al prop­er­ties of a man­i­fold; it is what we call a ho­mo­logy man­i­fold. But its form­al bound­ary, the clos­ure of H, is not a to­po­lo­gic­al man­i­fold near its ends. For ex­ample, if Ω is con­nec­ted, by defin­i­tion, H (which is homeo­morph­ic to H+Ω) is one of the con­tract­ible 3-man­i­folds of J. H. C. White­head [e1], [e2], with a non­trivi­al π1-sys­tem at in­fin­ity. +ΩHS3 is a White­head com­pactum. In the gen­er­al case, +Ω is called a rami­fied White­head com­pactum. Thus, (HΩ^,H) has no chance of be­ing a to­po­lo­gic­al handle. On the oth­er hand, H(+HΩ) is homeo­morph­ic to B2×R2; this will be the cent­ral res­ult of this pa­per.

The­or­em 2.2: (Freedman, 1981.)  Every open Cas­son handle M is homeo­morph­ic to B2×R2.

The proof of The­or­em 2.2 starts with a res­ult of 1979, when Freed­man was able to con­struct a smooth 4-man­i­fold M without bound­ary which is not homeo­morph­ic to S3×R that is however the im­age of a prop­er map of de­gree ±1, S3×RM (see [1] and [e34]).

A Cas­son tower of height k, or more briefly Ck, is a pair dif­feo­morph­ic to (HXk˚,H) where X1X2 is a Rus­si­an doll of de­fects in a handle (H,H).

The­or­em 2.3: (Mitosis, a finite version.)  Let (H6,H6) be a Cas­son tower C6 of height 6. There is a Cas­son tower C12 of height 12, or (H12,H12), such that
  1. H12=H6.
  2. H12H6IntH6.
  3. H12H6 is con­tained in a dis­joint uni­on of balls in IntH6, one ball for each con­nec­ted com­pon­ent.

Figure 5.

Figure 6.

Con­di­tion (3) is re­lated to the fact that, for each Cas­son tower (Hk,Hk), the man­i­fold Hk can be ex­pressed as a reg­u­lar neigh­bour­hood of a 1-com­plex, com­pare [e38]. Fig­ure 5 shows a schem­at­ic dia­gram of Freed­man which sum­mar­ises The­or­em 2.3.

In Sec­tion 3, Fig­ure 6 will rep­res­ent a C6, and Fig­ure 7 will rep­res­ent a C12, etc. From the point of view of the rep­res­ent­a­tion of corners on the bound­ary, it might be bet­ter to use Fig­ure 8.

Figure 7.

The meth­od of Freed­man [1] (com­pare [e34]) al­lows one to give a proof of The­or­em 2.3. However, it is slightly more de­tailed than the ana­logues in [1], [e34]. We will not cov­er this point in this pa­per (see [e37] for an ex­cel­lent write up of the mi­tos­is the­or­em (fi­nite ver­sion, The­or­em 2.3).

Figure 8.

Re­mark: Every pair (k,2k), k>6, in place of (6,12) gives a state­ment that one can de­duce without too much pain and sor­row that we could use in place of The­or­em 2.3 in what fol­lows.

Since we are go­ing to use The­or­em 2.3 of­ten, it is con­veni­ent to make the fol­low­ing:

Change of Nota­tion 2.4: From now on, we write Hk and Xk in place of H6k+6 and X6k+6, k=0,1,2,. (Also the mean­ing of Ek=HkHk1, ik, etc. is changed.)
The­or­em 2.5: (Mitosis, an infinite version.)  Let (H,H) be a Cas­son handle presen­ted as above, and let k0 be an in­teger. There ex­ists an­oth­er Cas­son handle (H,H)(H,H) sat­is­fy­ing the con­di­tions:
  1. Hk1=Hk1 if k1.
  2. HHk1(IntHk)Hk1.
  3. The clos­ure H of H in H is the en­d­point com­pac­ti­fic­a­tion of H.

This in­fin­ite ver­sion, The­or­em 2.5, fol­lows from the fi­nite ver­sion, The­or­em 2.3, by an in­fin­ite re­pe­ti­tion. One suf­fi­ciently shrinks balls giv­en by The­or­em 2.3 to en­sure the con­di­tion (3) of The­or­em 2.5.

3. Architecture of topological coordinates

The am­bi­tious con­struc­tion to come ap­plies the mi­tos­is the­or­em (The­or­em 2.5) and ele­ment­ary geo­metry, to con­vert The­or­em 2.2, that every open Cas­son handle is homeo­morph­ic to B2×R2, to two the­or­ems on ap­prox­im­a­tion by homeo­morph­isms. For Cas­son handles, we will use the ter­min­o­logy of Sec­tion 2, un­der the mod­i­fied form in Change of Nota­tion 2.4 (by a rein­dex­ing).

The open Cas­son handle M will be iden­ti­fied with N+N where (N,N) is a Cas­son handle (not open). Let N^ be the en­d­point com­pac­ti­fic­a­tion of N. Sub­tract­ing from N the (to­po­lo­gic­al) in­teri­or of a col­lar neigh­bour­hood of +N in N, very pinched to­wards the ends of N, we ob­tain a Cas­son handle (H,H)(M,M)(N,N) whose clos­ure in N^ is the en­d­point com­pac­ti­fic­a­tion H^ of H. We fix a present­a­tion of (H,H).

We will con­struct a rami­fied sys­tem of Cas­son handles in (N,N), that, in some way, ex­plores its in­teri­or.

Figure 9.

3.1. Construction
For each fi­nite se­quence (a1,,ak) in {0,1} (fi­nite dy­ad­ic se­quence), we can define a presen­ted Cas­son handle (H(a1,,ak),H) con­tained in (H,H), whose present­a­tion con­sists of an em­bed­ding i(a1,,ak):H(a1,,ak)B2×D2, and a Rus­si­an doll of de­fects Xi(a1,,ak), in the stand­ard handle B2×D2 such that (for (1)–(5), see the right fig­ure of Fig­ure 10):
  1. H=H() (case k=0) as a presen­ted Cas­son handle.
  2. H(a1,,ak,1)=H(a1,,ak).
  3. Hk(a1,,ak,0)=Hk(a1,,ak) (re­call that Hk are sets of 6-stages).
  4. The clos­ure H(a1,,ak,0)inH^ is an en­d­point com­pac­ti­fic­a­tion of H(a1,,ak,0).
  5. H(a1,,ak,0)Hk(a1,,ak)H˚k+1(a1,,ak)Hk(a1,,ak).
  6. ik(a1,,ak,0)=ik(a1,,ak), so Xk(a1,,ak,0)=Xk(a1,,ak).
  7. The in­ter­sec­tion of Xk+1(a1,,ak,0) and Xk+1(a1,,ak) is empty, and their uni­on is a mul­tiple de­fect in Xk(a1,,ak).
  8. (Without Change of Nota­tion 2.4) We also re­quire a co­her­ence con­di­tion on the total Rus­si­an doll as­sumed by (7), that is to say {Xk}, where Xk=Xk(a1,,ak). To for­mu­late it, we mo­ment­ar­ily sus­pend the rein­dex­ing con­ven­tion (Change of Nota­tion 2.4) and write Tk=+Xk. The con­di­tion is that there ex­ists an in­ter­val JD2 such that, for all tJ, the me­ri­di­on­al disc Bt=B2×t of the sol­id tor­us B2×D meets the mul­tiple sol­id tori Tk ideally, in the sense that each con­nec­ted com­pon­ent of BtTk is a me­ri­di­on­al disc of Tk, that meets Tk+1 in an ideal fash­ion il­lus­trated in the left fig­ure of Fig­ure 10.

Figure 10.

Ex­e­cu­tion of Con­struc­tion 3.1 (by in­duc­tion on k).  We start with H()=H. Hav­ing defined a presen­ted handle for every se­quence of length k, we define them for every se­quence (a1,,ak,1) by (2). Next, we define H(a1,,ak,0) by the mi­tos­is the­or­em (in­fin­ite ver­sion, The­or­em  2.5). This as­sures that con­di­tions (3), (4) and (5) are met. It re­mains to define the present­a­tion of the Cas­son handle (H(a1,,an,0),H) in such a fash­ion that the two last con­di­tions (6) and (7) are sat­is­fied. To define i(a1,,ak,0), it is con­veni­ent to graft, onto ik(a1,,ak), a present­a­tion the near part of the Cas­son handle (H(a1,,ak,0),H), to know the Cas­son mul­ti­handle (H(a1,,ak,0)H˚k(a1,,ak,0),δHk(a1,,ak,0)), where ex­cep­tion­ally  ˚  and δ de­note the in­teri­or and the fron­ti­er in H(a1,,ak,0) rather than in N^. The graft­ing is done with the help of Lemma 2.1. The last con­di­tion (7) is as­sured af­ter­wards by an iso­topy in X˚k(a1,,ak). Hav­ing (1) to (7), the read­er will know how to ar­range that (8) is also sat­is­fied.  ◻

Figure 11.
Re­mark: If (a1,a2,) is an in­fin­ite se­quence in {0,1}, the uni­on H(a1,a2,)=kH(a1,a2,,ak) gives a Cas­son handle with an ob­vi­ous present­a­tion. Moreover, the clos­ure H(a1,a2,) is the en­d­point com­pac­ti­fic­a­tion (ex­er­cise). Thus, we have a vast col­lec­tion of Cas­son handles in N, con­veni­ently nes­ted.

Of the sys­tem of handles (H(a1,,ak),H), we es­pe­cially use their skins +H(a1,,ak). The uni­on P3=+H(a1,,ak) of the skins is what one calls a branched man­i­fold in N4, since near every point P3H, the pair (N4,P3) is C1-iso­morph­ic (same as C-iso­morph­ic, after some work that we leave to the read­er) to the product of R2 with the mod­el of branch­ing (R2,Y1) where Y1 is the uni­on of two smooth curves (iso­morph­ic to R1), prop­erly em­bed­ded in R2 and which have in com­mon ex­actly one closed half-line. One ob­serves without dif­fi­culty that the clos­ure P of P in N^ is the en­d­point com­pac­ti­fic­a­tion of P.

The branched man­i­fold P splits along the sin­gu­lar points in­to com­pact man­i­folds: Pk(a1,,ak)=+Ek(a1,,ak)=Ek(a1,,ak)+H(a1,,ak). Thus, Pk(a1,,ak) is the skin of the k-th stage of (H(a1,,ak),H).

3.2. Construction of the design G4 (see Figure 11)

For P3, we con­struct a neigh­bour­hood G4 in N4 called the design, which has a de­com­pos­i­tion I of G4 in­to dis­joint in­ter­vals, sat­is­fy­ing the fol­low­ing.

Figure 12.
  1. For every in­ter­val Iα of I, the in­ter­sec­tion IαN is Iα or the empty set. A neigh­bour­hood of Iα in (G4,P3;I) is iso­morph­ic to the product of R2 with an open 2-di­men­sion­al mod­el (G2,P1;I) as in Fig­ure 12.
  2. The clos­ure G of G in N^ is its en­d­point com­pac­ti­fic­a­tion, and hence co­in­cides with GP.

It fol­lows by com­bin­ing, quite na­ively, two bicol­lars of genu­ine sub­man­i­folds of P3. On the oth­er hand, we clearly are per­mit­ted to sup­pose that G4 con­tains the col­lar NH˚ of +N.

The design (G4,I) de­com­posed in­to in­ter­vals splits in a ca­non­ic­al fash­ion (along the 3-man­i­fold formed by the ex­cep­tion­al in­ter­vals of I hav­ing in­teri­or points on G4) in­to genu­ine trivi­al I-bundles I(a1,,ak)×Pk(a1,,ak), where I(a1,,ak) is a 1-sim­plex and (its centre)×Pk(a1,,ak)G4 is nearly the nat­ur­al in­clu­sion Pk(a1,,ak)G4. More pre­cisely, the two em­bed­dings are iso­top­ic in G4 by an iso­topy which moves only a col­lar of the bound­ary of Pk(a1,,ak). It is con­veni­ent to give a nor­mal ori­ent­a­tion to P3 in N4 (to­wards the ex­ter­i­or), to de­duce from it the ori­ent­a­tion of the 1-sim­plices I(a1,,ak).

3.3. Construction of g:G4B2×D2

This g will be a smooth em­bed­ding which will re­veal the struc­ture of G4. We choose, by re­cur­rence, lin­ear em­bed­dings I(a1,,ak)(0,1] con­serving the ori­ent­a­tion. To start, I()(0,1] ends at 1. Sup­pose now these em­bed­dings have been defined for all se­quences of length k. Then, we em­bed I(a1,,ak,0) and I(a1,,ak,1) re­spect­ively on the ini­tial third and the fi­nal third of the in­ter­val I(a1,,ak)(0,1].

The cent­ral third of I(a1,,ak) is a closed in­ter­val that we may call J(a1,,ak). The com­ple­ment in I() of all the open in­ter­vals J˚(a1,,ak) is then a com­pact Can­tor set in (0,1].

On the oth­er hand, we claim that the em­bed­dings ik(a1,,ak)|:+Hk(a1,,ak)B2×D2 define to­geth­er a smooth map i:PB2×D2. Let φ:(0,1]×B2×D2B2×D2 be the em­bed­ding (t,x,y)(x,ty). We will have the tend­ency to identi­fy do­main and codo­main by φ.

We define g:G4B2×D2 on I(a1,,ak)×Pk(a1,,ak) by the rule that (t,x)φ(t,i(x)). For that defin­i­tion to make sense, we have to first ad­just, by iso­topy, the trivi­al­isa­tion giv­en by the I-fibres I(a1,,ak)×Pk(a1,,ak) in (G4,I), a routine task that is left to the read­er.

3.4. Construction of g0:G04B2×D2
Let G04 be the uni­on of G4 and a small col­lar neigh­bour­hood C4 of N in N that re­spects δG4 (see Fig­ure 11 for Sec­tion 3.2). Let us ex­tend g to an em­bed­ding g0:G04B2×D2. By unique­ness of col­lars, we can ar­range g and g0 so that g0 sends C4G˚4 to (B2λB2)×μD2, where λ(0,1] is near to 1 and μ to the ini­tial point of I(). This com­pletes the con­struc­tion of g0:G04B2×D2. Look­ing near g0 and its im­age, we will claim that we have com­pletely de­scribed the clos­ure G04 of G04 in N^4.
3.5. The image g0(G04)B2×D2
Some nota­tion again (see Fig­ure 13).
  • T(a1,,ak)Tk(a1,,ak)=+X(a1,,ak), a multisol­id tor­us B2×D2.
  • T(a1,,ak)=φ(J(a1,,ak)×T(a1,,ak))B2×D˚2, a ra­di­ally thickened copy of T(a1,,ak), called a hole.
  • B=λB2×μD2 (see defin­i­tion of g0), called the cent­ral hole.
  • Fk={φ(I(a1,,ak1)×T(a1,,ak))k fixed}; the fron­ti­ers δFk, k2, are in­dic­ated in dashed lines in the right-hand fig­ure be­low.
  • (B2×D2)0=(B2×D2B˚){T˚(a1,,ak)}, called the holed stand­ard handle.
  • W0=kFk, a com­pactum in (B2×D2)0.

With this nota­tion, we claim that the im­age g0(G04) is (B2×D2)0W0.

Figure 13.

3.6. The main diagram
The com­mut­at­ive dia­gram to the right gives an over­view of the con­struc­tion to come. The ele­ments will be con­struc­ted in the or­der W0, g1, D, g2, D, g3, D+, f. The proof that (B2×D˚2)/D+ is homeo­morph­ic to B2×D˚2 (by the meth­ods of Bing) will ap­pear in Sec­tion 4. The proof that f is ap­prox­im­able by homeo­morph­isms is post­poned to Sec­tion 5.
3.7. Construction of W0 and g1
W0 is the de­com­pos­i­tion of the com­pact set (B2×D2)0, where nonde­gen­er­ated ele­ments are the con­nec­ted com­pon­ents W of the com­pact set W0(B2×D2)0. Each WW0 is a White­head com­pactum in a single level φ(t×B2×D2). We check na­ively that the in­clu­sion (B2×D2)0W0(B2×D2)0/W0 in­duces a homeo­morph­ism ((B2×D2)0W0)(B2×D2)0/W0. We already know that G^0 is iden­ti­fied with G0N^. We define the homeo­morph­ism g1 as a com­pos­i­tion of homeo­morph­isms: g1:G0G^0g^0((B2×D2)0W0)(B2×D2)0/W0.
3.8. Construction of D and g2
Let D be the de­com­pos­i­tion of B2×D2 giv­en by the B, T(α) (α can be any fi­nite dy­ad­ic se­quence), and the ele­ments of W which are dis­joint from them. To define g2:N^(B2×D2)/D, we must ex­tend q1g1:G0(B2×D2)/D to each con­nec­ted com­pon­ent Y of N^G0. Its fron­ti­er δY is iden­ti­fied by g1 to the quo­tient in (B2×D2)0/W0, either of B, or of a bound­ary of a con­nec­ted com­pon­ent of a hole T(a1,,ak). By defin­i­tion, g2(Y) is the im­age in (B2×D2)/D of this bound­ary. It is easy to check the con­tinu­ity of g2.

Next, g3 and D in the main dia­gram are defined by re­stric­tion. The design G4 has led us in­ex­or­ably to define g3:M4B2×D˚2/D, which com­pares the open Cas­son handle M4 with a very ex­pli­cit quo­tient of the open handle B2×D˚2.

The de­com­pos­i­tion D which spe­cifies this quo­tient has non­cel­lu­lar ele­ments, that is, the holes T(a1,,ak), each of which has the ho­mo­topy type of a circle. There­fore the quo­tient map B2×D˚2B2×D˚2/D is cer­tainly not ap­prox­im­able by homeo­morph­isms. One can also check that the Čech co­homo­logy Hˇ2 of the quo­tient is of in­fin­ite type.

The con­struc­tion of D+ be­low re­pairs this ter­rible de­fect; it will be con­struc­ted by hand; D+ will be less fine than D, which will en­able us to define f=q3g3 without ef­fort.

3.9. Construction of D+
We set W=W0(B2×D˚2)=W0(B2×D2). Its con­nec­ted com­pon­ents define a de­com­pos­i­tion W of B2×D˚2. We have known since the 1950s how to show that B2×D˚2/W is homeo­morph­ic to B2×D˚2, see Sec­tion 4.

For the re­quire­ments of the next para­graph, the quo­tient (B2×D˚2)/D+ must be a quo­tient of B2×D˚2/W by a de­com­pos­i­tion whose ele­ments are the con­nec­ted com­pon­ents of {q(T(α))E(α)α a finite dyadic sequence}. Here {E(α)} is a col­lec­tion of dis­joint, to­po­lo­gic­ally flat multi-2-discs such that for each fi­nite dy­ad­ic se­quence α, the in­ter­sec­tion E(α)(αq(T(α))) is

  1. the bound­ary E(α); and
  2. a mul­ti­lon­git­ude of T(α) far from W (each con­nec­ted com­pon­ent of q(T(α))E(α) is then con­tract­ible).

Moreover, we want that the dia­met­er of the con­nec­ted com­pon­ents of E(a1,,ak) tends to­wards 0 (on each com­pact set) as k. Sec­tion 4 does not de­mand any more than this and vis­ibly, {E(α)} spe­cifies D+.

The spe­cific­a­tion of {E(α)} is un­for­tu­nately te­di­ous. E(α) will be the faith­ful im­age q(D(α)) of a mul­tidisc in B2×D˚2. For fun­da­ment­al group reas­ons, the mul­tidisc D(α) is ob­liged to meet W, but, to as­sure flat­ness of q(D(α)) (proved in Sec­tion 4), it must be a well be­haved meet­ing, per­mit­ted by (7) and (8) of Con­struc­tion 3.1.

We have Tk=αTk(α); con­di­tions (6) and (7) of Con­struc­tion 3.1 as­sure that Tk is a multisol­id tor­us of which cer­tain con­nec­ted com­pon­ents con­sti­tute Tk(α). We have kTk=p(W), which is a rami­fied White­head com­pactum in B2×D2.

To start, we spe­cify (sim­ul­tan­eously and in­de­pend­ently) in B2×D2, (to­po­lo­gic­ally) im­mersed, loc­ally flat discs D(α) which will be the pro­jec­tion p(D(α))=D(α). We as­sume eas­ily the two prop­er­ties (a) and (b), where (b) uses (8) of Con­struc­tion 3.1.

Figure 14.

(a) D(a1,,ak) is a dis­joint uni­on of im­mersed discs in Tk1, with as their only sin­gu­lar­it­ies, an arc of double points for each, above Tk(a1,,ak). The bound­ary D(a1,,ak) is formed from one lon­git­ude of each con­nec­ted com­pon­ent of Tk(a1,,ak). The double points of D(a1,,ak) are out­side T˚k(a1,,ak).

(b) For each lk, the in­ter­sec­tion D˚(a1,,ak)Tl is a mul­tidisc (em­bed­ded in Tk(a1,,ak)) of which each con­nec­ted com­pon­ent D0 is a me­ri­di­on­al disc of Tl that meets the sol­id tori of the next gen­er­at­or (Tl+1/6 with our re­vised in­dex­ing of Change of Nota­tion 2.4) ideally (see the left-hand fig­ure of Fig­ure 10).

By resolv­ing the double points of D(α), which we have to em­bed in (0,1)×B2×D2B2×D2, spe­cify­ing the first co­ordin­ate by a con­veni­ent func­tion ρ(α):D(α)(0,1).

We will em­bed a single D(α) at a time (fol­low­ing some chosen or­der). We em­bed first D(a1,,ak) closer and closer (by a sec­ond­ary in­duc­tion). Some nota­tion: T+(a1,,al)=J(a1,,al)×T(a1,,al1),Fl=p1(p(Fl))=(0,1)×Tl,W+=p1(p(W))=(0,1)×(kTk).

One can eas­ily check that, for D(a1,,ak), the prop­er­ties (c) and (d) for l>k, of which (d) for l is only pro­vi­sion­al.

(c) D(a1,,ak) is em­bed­ded, is con­tained in I(a1,,ak1)×T(a1,,ak1), and is dis­joint from B and from {T+(α)α(a1,,ak)}. The bound­ary D(a1,,ak) is in a single level t×B2×D2, where tJ˚(a1,,ak).

(d) Each con­nec­ted com­pon­ent of the mul­tidisc Fl+D(a1,,ak) is in a single level t×B2×D2; this level is dis­joint from each box T(α), and does not con­tain any oth­er con­nec­ted com­pon­ent of Fl+D(a1,,ak).

For l=k and k+1, here are the il­lus­tra­tions of the graph of ρ in a simple case.

We ob­served that in push­ing D(a1,,ak) ver­tic­ally, as small as we want, and only on F˚l+D(a1,,ak), we can pass from (d) for l to (d) for l+1, without los­ing (c). There­fore, without los­ing (c), we can pass to the next prop­erty.

(e) For each in­teger l>k, the con­nec­ted com­pon­ents of the mul­tidisc Fl+D(a1,,ak) pro­ject onto as many dis­joint in­ter­vals of ra­di­us in (0,1).

This con­di­tion as­sures that, for all WW, the in­ter­sec­tion WD(a1,,ak) is an in­ter­sec­tion of discs (and so cel­lu­lar). There­fore q(D(a1,,ak)) is cer­tainly a disc (com­pare The­or­em 4.4). In Sec­tion 4, we will prove by hand that it is a flat disc. If, be­fore D(a1,,ak), we have already defined (for the main in­duc­tion) a fi­nite col­lec­tion of discs D(α1),,D(αn), we fol­low the same con­struc­tion as above, al­ways stay­ing in a neigh­bour­hood of T+(a1,,ak) (guar­an­teed by (c)), dis­joint from D(α1)D(αn) and for all ele­ments of W that touch D(α1)D(αn).

Thus the fam­ily {D(α)} of dis­joint 2-discs is defined by a double in­duc­tion and sat­is­fies the prop­er­ties (a), (b), (c) and (e) with p(D(α))=D(α). Next {D(α)} defines D+ as already in­dic­ated. One eas­ily checks all the prop­er­ties wanted for q(D(α))=E(α)in(B2×D˚2)/W, ex­cept loc­al flat­ness of E(α) which is post­poned to Sec­tion 4.

3.10. End of the proof that M is homeomorphic to B2×D˚2 (modulo Sections 4 and 5)
Ac­cept­ing from Sec­tion 4 that (B2×D˚2)/D+ is homeo­morph­ic to B2×D˚2, we show mod­ulo Sec­tion 5 the ap­prox­im­ab­il­ity by homeo­morph­isms of f:M4(B2×D˚2)/D+ in the fol­low­ing fash­ion. We form the com­mut­at­ive dia­gram
where the in­clu­sion IntMS4 ex­ists since M em­beds in B2×D2 (the ex­perts also know that IntM is dif­feo­morph­ic to R4 [e38]), and where f(S4IntM4)=. There­fore, S(f)={yS4f1(y) a point} is vis­ibly a con­tract­ible set.

Also S(f) is nowhere dense. [Here is a proof. The re­stric­tion f| is the same as q3q1g1|:MG04(B2×D˚2)/D+, which is already sur­ject­ive and f1(S(f)) is con­tained in the nowhere dense set of MG04 giv­en by (G0)(ends of G04)g11(αE(α)).]

There­fore, ac­cord­ing to The­or­em 5.1, the map f is ap­prox­im­able by homeo­morph­isms. Next, by Pro­pos­i­tion 4.2 (loc­al­isa­tion prin­ciple), the re­stric­tion IntM4S4{} is also ap­prox­im­able by homeo­morph­ism. Fi­nally, by Pro­pos­i­tion 4.3 (glob­al­isa­tion prin­ciple), the map f:M(B2×D˚2)/D+ is ap­prox­im­able by homeo­morph­isms. Thus The­or­em 2.2 is proved mod­ulo Sec­tions 4 and 5.

Re­mark: S(f)S4 is in fact a com­pactum of di­men­sion 1, be­cause it is the uni­on of a con­tract­ible set S(f) with a set of di­men­sion 0, that is, the ends of G04 which are not in the fron­ti­er of a con­nec­ted com­pon­ent Y of M4G04. For reas­ons of co­homo­logy, dimS(f)1. There­fore it is a com­pactum of di­men­sion ex­actly 1.

4. Bing shrinking

We need to show that the space B2×D˚2/D+ defined in Sec­tion 3 is homeo­morph­ic to B2×D˚2. The ne­ces­sary tech­niques come from a series of art­icles of RH Bing from the 1950s (see es­pe­cially [e3], [e5], [e6]), which made his repu­ta­tion as a great vir­tu­oso of geo­met­ric to­po­logy.

We con­sider a prop­er sur­ject­ive map f:XY between met­ris­able, loc­ally com­pact spaces X, Y. Let D={f1(y)yY} be the de­com­pos­i­tion as­so­ci­ated with f. When is f (strongly) ap­prox­im­able by homeo­morph­isms, in the sense that for all open cov­er­ings V of Y, the V-neigh­bour­hood N(f,V)={g:XYfor all xX, there exists VV such that f(x),g(x)V} con­tains a homeo­morph­ism?

Since f in­duces a homeo­morph­ism φ:X/DY, we see eas­ily that f is ap­prox­im­able by homeo­morph­isms if and only if one can find maps g:XX such that D={g1(x)xX} and that fg ap­prox­im­ates f (in ef­fect, φ trans­lates g in­to a homeo­morph­ism g:YX). This ob­ser­va­tion makes the fol­low­ing the­or­em plaus­ible.

The­or­em 4.1: (Bing shrinking criterion.)f is ap­prox­im­able by homeo­morph­ism if and only if, for every cov­er­ing U of X and V of Y, there ex­ists a homeo­morph­ism h:XX such that h(D)<U, and for all com­pact DD, D and h(D) are f1(V)-near in the sense that there ex­ists an f1(V)f1(V) that con­tains Dh(D).

We then say that D is shrink­able. We can show a proof by hand [e28], or by Baire cat­egory [e35], [e11] (the idea is to find a homeo­morph­ism h:XY that con­verges to­wards g that de­term­ines D). The proof also gives:

Re­mark: In The­or­em 4.1, if h re­spects (or fixes) a closed set AX, then f is ap­prox­im­able by homeo­morph­isms that send A on f(A) (or which co­in­cide on A with f), and re­cip­roc­ally.
Pro­pos­i­tion 4.2: (Localisation principle.)  If f:XY is ap­prox­im­able by homeo­morph­isms and Y is a man­i­fold (or Y sat­is­fies the prin­ciple of de­form­ab­il­ity by homeo­morph­isms com­ing from [e20], D1 of [e23]), then, for each open set V of Y, the re­stric­tion fV:f1(V)V of f is ap­prox­im­able by homeo­morph­isms.

Proof (in­dic­a­tion).  To ap­prox­im­ate fV, we com­bine (by the prin­ciple D1) a series of ap­prox­im­a­tions of f; com­pare ([e23], Sec­tion 3.5). I be­lieve that this lemma is not in the lit­er­at­ure be­cause, for di­men­sion 4, we have stronger res­ults [e22], [e32]. However, upon re­flec­tion, the com­plic­ated ar­gu­ment of [e22] works. In each case that in­terests us, the read­er will be able to find an ad hoc proof that is easi­er.  ◻

Counter­example: This prin­ciple is false if X and Y are Cantor×[0,1]=2N×[0,1], and f=g×Id[0,1], where g(1,a2,a3,)=(a2,a3,),g(0,a2,a3,)=(0,0,0,).
Pro­pos­i­tion 4.3: (Globalisation principle.)  Let f:XY be a prop­er map such that, for an open set VY, the re­stric­tion fV:f1(V)V is ap­prox­im­able by homeo­morph­isms. Then, f is ap­prox­im­able by prop­er maps g such that
  1. g1(V)=f1(V),
  2. gV:g1(V)V is a homeo­morph­ism, and
  3. g=f on Xf1(V).

This prin­ciple is easy to es­tab­lish, be­cause if V is the cov­er­ing of V by open balls centred on yV and of ra­di­us inf{d(y,z)zYV}, then every map γ:f1(V)V that is in N(fV,V), ex­tends by f to a map g:XY. In the very spe­cial case that D is π0(K) for a com­pact set KX, the Bing shrink­ing cri­terion sim­pli­fies as fol­lows. (Then, D con­sists of con­nec­ted com­pon­ents of K and the im­age of K in X/D is 0-di­men­sion­al and is iden­ti­fied with π0(K).)

The­or­em 4.4: (Criterion.)  Un­der these con­di­tions, D is shrink­able if for all ε>0 and for all open D-sat­ur­ated U of X such that UK is com­pact, there is a homeo­morph­ism h:XX with sup­port in U (re­spect­ively AX) such that h(KU) lies in a fi­nite dis­joint uni­on of com­pact sets, each of dia­met­er <ε.

This con­di­tion, mod­ulo loc­al­isa­tion prin­ciple (Pro­pos­i­tion 4.2), is clearly ne­ces­sary.

For all ε>0, one can con­sider Dε={DDdiamDε}. We say that DDεD is a closed sub­set of X. Here is a re­mark­able but dis­turb­ing ex­ample where D is null, Dε is shrink­able for any ε>0, but D is not shrink­able. The ele­ments of D are the con­nec­ted com­pon­ents of a com­pact set X=nFn where F0 and F1 are as il­lus­trated. This im­age is suit­ably rep­lic­ated in each sol­id tor­us; Fn is then 2n sol­id tori. Each DD is clearly cel­lu­lar, hence Dε is shrink­able by Lemma 5.2. But, with the help of cyc­lic cov­ers, one can check that D is not shrink­able (see [e12], [e1]).

Figure 15.

There are thank­fully prop­er­ties of in­di­vidu­al ele­ments, a little stronger than cel­lu­lar­ity, which dis­cards this sort of ex­ample. For a com­pact AX, we con­sider the prop­erty R(X,A): for each ε>0, for every null de­com­pos­i­tion D of X con­tain­ing A, and for all neigh­bour­hoods U of A, there is a map f:XX with sup­port in U that shrinks at least A, (that is, f(A) is a point and f|U:UU is ap­prox­im­able by homeo­morph­isms), such that, for all DD, diamf(D)max(diamD,ε). If D is fixed in ad­vance, we call the (weak­er) prop­erty R(X,A;D).

Ob­ser­va­tion: For every neigh­bour­hood U of A, we have R(X,A) is equi­val­ent to R(U,A). Moreover, R(X,A) is in­de­pend­ent of the met­ric.
Pro­pos­i­tion 4.5: If D is null, and R(X,D;D) is sat­is­fied for all DD, then D is shrink­able.

Proof.  The proof is an edi­fy­ing ex­er­cise.  ◻

Pro­pos­i­tion 4.6: R(X,A) is sat­is­fied if A is a to­po­lo­gic­al flat disc of any codi­men­sion in the in­teri­or of the man­i­fold.

Proof of Pro­pos­i­tion 4.6.  This is R(Rn,Bk) for kn. The proof of R(R2,B1) which is in­dic­ated by Fig­ure 16.

Figure 16.

In (a), every ele­ment of D that meets the big rect­angle has already dia­met­er <ε/4; if DD meets a gap between suc­cess­ive rect­angles, it is dis­joint from the rect­angle after. We set f(B1)=0, and f=Id out­side the biggest rect­angle (which is in U); f is lin­ear on each ver­tic­al in­ter­val in a rect­angle of (b) and also lin­ear on each 1-cell of the rect­an­gu­lar cel­lu­la­tion in (b) of (big rect­angleB1). Moreover, pf=p where p is the pro­jec­tion to the y-ax­is (the Rnk nor­mal to Bk). Fi­nally, the size of the im­age of each of the ver­tic­al rect­angle is <ε/4.  ◻

Figure 17.

We con­sider the White­head pair (B2×S1,j(B2×S1))=(T,T), and the thickened pair (R×T,[0,1]×T).

Lemma 4.7: For ε>0, there ex­ists an iso­topy ht (t[0,1]) of Id|R×T with com­pact sup­port in (ε,1+ε)×IntT such that we have diam(h1(t×T))<εandh1(t×T)[tε,t+ε]×T for all t[0,1].

Idea of proof.  It is sug­ges­ted by Fig­ure 18.  ◻

Figure 18.

By this lemma, one can shrink many de­com­pos­i­tions re­lated to White­head com­pacta. For ex­ample, let WR3 be a White­head com­pactum and let D={tWt[0,1],WW} be the de­com­pos­i­tion I×W of R×R3=R4. Then D is shrink­able by Lemma 4.7 ap­plied to the sol­id tori T, T, T, …whose in­ter­sec­tion is W. There­fore R4/D is homeo­morph­ic to R4. Moreover, by Pro­pos­i­tion 4.2 (loc­al­isa­tion prin­ciple), we have that (0,1)×R3/W is homeo­morph­ic to (0,1)×R3. Hence we have the fol­low­ing cel­eb­rated fact.

The­or­em 4.8: (Celebrated fact [e16].)R×(R3/W)=R4.
Figure 19.

This is a res­ult of An­drews and Ru­bin [e16] in 1965, proved after ana­log­ous res­ults, but more dif­fi­cult, of Bing [e5] in 1959, which is a curi­ous ana­chron­ism. There is a good ex­plan­a­tion! A. Sha­piro, at the time when he suc­ceeded in turn­ing S2 in­side out in S3 by a reg­u­lar ho­mo­topy, com­pare [e31], had also es­tab­lished The­or­em 4.8. In any case, Bing tells me that D. Mont­gomery had com­mu­nic­ated to him this claim without be­ing able him­self to jus­ti­fy it ex­cept by giv­ing an easi­er ar­gu­ment (see Lemma 4.9) show­ing that R×(S3W) is homeo­morph­ic to R4, com­pare [e6]. Did the proof of Sha­piro from the 50s dis­ap­pear without a trace?

To es­tab­lish the flat­ness of the discs {E(α)} con­struc­ted in Sec­tion 3.9, we will also need a lemma that is easi­er than Lemma 4.7, treat­ing again the White­head pair (T,T). Let D be a me­ri­di­on­al disc of T that cuts T trans­vers­ally in two discs.

Lemma 4.9: With this data, we can find in R×T a to­po­lo­gic­al 4-ball B, such that IntB[0,1]×T and B(R×D) is an equat­ori­al 3-ball of the form (interval)×D0R×D.

Proof of Lemma 4.9.  This has noth­ing to do with the proof of Lemma 4.7! We find B eas­ily from a 2-disc im­mersed in T like in Fig­ure 20 (com­pare Sec­tion 3.9).  ◻

Figure 20.

To es­tab­lish that (B2×D˚2)/D+ is homeo­morph­ic to B2×D˚2, we will now use the con­struc­tion of Sec­tion 3.

Pro­pos­i­tion 4.10: The de­com­pos­i­tion W of B2×D˚2 is shrink­able.

Proof of Pro­pos­i­tion 4.10.  We ap­ply The­or­em 4.4, Lemma 4.7 (or Lemma 4.9, without ex­ploit­ing the last con­di­tion of Lemma 4.9). For this, it is con­veni­ent to re­mark first that for all open W-sat­ur­ated U in B2×D2, WU is con­tained in an open sub­set of U that is a dis­joint uni­on of open sets of the form I˚×T˚(a1,,ak), where I is an in­ter­val.  ◻

Our next goal is the flat­ness of the discs E(α)=q(D(α))(B2×D˚2)/W. Let W(α)={wWwD(α)}, and let W(α)=W(α).

Pro­pos­i­tion 4.11: W(α) is shrink­able re­spect­ing D(α). There­fore, the quo­tient qα(D(α)) of D(α) is flat in (B2×D˚2)/W(α).

Proof of Pro­pos­i­tion 4.11.  We ap­ply Lemma 4.9 and the re­l­at­ive cri­ter­ia (The­or­em 4.4). For every open Wα-sat­ur­ated U of B2×D˚2, the in­ter­sec­tion WαU is trivi­ally con­tained in an open set which, for some in­teger l, is a dis­joint uni­on of open sets of the form I˚×T˚U, where T is a con­nec­ted com­pon­ent of mul­tiple sol­id tori Tl(b1,,bl) and i is an in­ter­val.

Con­di­tion (d) of Sec­tion 3.9 al­lows us to choose these sets so that in ad­di­tion, for each:

  • D(α)(I×T) is a single 2-disc, which is pro­jec­ted onto a me­ri­di­on­al disc D of T which is also a con­nec­ted com­pon­ent of D(α)T; see Sec­tion 3.9.

By con­di­tion (b) of Sec­tion 3.9 the me­ri­di­on­al disc D ideally chopped off Tl+1/6T, so Lemma 4.9 gives us dis­joint 4-balls B1,,Bs in I˚×T˚, such that

  1. each in­ter­sec­tion BiD(α) is a dia­metral 2-disc and not knot­ted in Bi, and
  2. B˚1B˚s con­tains the com­pact set W+(I˚×T˚)Wα(I˚×T˚).

For all com­pact K in B˚i and all ε>0, we can eas­ily find a homeo­morph­ism h:BiBi with com­pact sup­port which re­spects B˚iD(α) and such that diamh(K)<ε. The cri­ter­ia of The­or­em 4.4 (re­spect­ing D(α)) is there­fore sat­is­fied.  ◻

Pro­pos­i­tion 4.12: q(D(α))=E(α) is flat in (B2×D˚2)/W.

Proof of Pro­pos­i­tion 4.12.  The open set Uα=(B2×D˚2)(WαD(α)) is clearly homeo­morph­ic to (B2×D˚2)/Wαqα(D(α)) by qα. There­fore, by Pro­pos­i­tions 4.2 and 4.3, the quo­tient map qα:(B2×D˚2)/Wα(B2×D˚2)/W is ap­prox­im­able by homeo­morph­isms fix­ing qα on the flat disc qα(D(α)). There­fore, q(D(α))=q(α)q(α)(D(α)) is flat.  ◻

We now pro­pose to fin­ish by show­ing that the quo­tient maps B2×D˚2(B2×D˚2)/Wp1((B2×D˚2)/W)/{E(α)}p2(B2×D˚2)/D+ are ap­prox­im­able by homeo­morph­isms.

Pro­pos­i­tion 4.13: p1 is ap­prox­im­able by homeo­morph­isms.

Proof of Pro­pos­i­tion 4.13.  This fol­lows from Pro­pos­i­tions 4.12, 4.6 and 4.5.  ◻

To ap­prox­im­ate p2 by homeo­morph­isms, we need a little pre­par­a­tion. Ac­cord­ing to Pro­pos­i­tions 4.13 and 4.10, there is a shrink­ing map r:B2×D˚2B2×D˚2 in­du­cing the same de­com­pos­i­tion as the quo­tient map ((B2×D˚2)/W)/{E(α)}; we can identi­fy the do­main of p2 with B2×D˚2 by r.

The de­com­pos­i­tion P con­sti­tuted of the preim­ages p21(y)={a point} is the count­able col­lec­tion of nat­ur­al quo­tients of con­nec­ted com­pon­ents of holes T(α) and B, which now identi­fy r(T(α)) and r(B)B2×D˚2. We ob­serve that P is null. The quo­tient map λB2×μD2=BrB shrinks the White­head com­pactum W(B)={wWwB}, and these com­pact sets lie in λB2×μD2B.

Pro­pos­i­tion 4.14: r(B) has a bicol­lar neigh­bour­hood V in B2×D˚2, that is, (V,r(B)) is homeo­morph­ic to (R×r(B),0×r(B)).

This will res­ult in the fol­low­ing pro­pos­i­tion.

Pro­pos­i­tion 4.15: The quo­tient of B in (B2×D˚2)/W(B) ad­mits a bicol­lar neigh­bour­hood.

Proof of Pro­pos­i­tion 4.15.  This is equi­val­ent to the ex­ist­ence of a bicol­lar neigh­bour­hood in (R×B)/(0×W(B)). However, by (slightly gen­er­al­ised) The­or­em 4.8 and Pro­pos­i­tions 4.2 and 4.3, the quo­tient map of the lat­ter space on (R×B)/(R×W(B)) is ap­prox­im­able by homeo­morph­isms, fix­ing the quo­tient of 0×B.  ◻

Proof of Pro­pos­i­tion 4.14.  The map r fac­tor­ises in­to rr where r factors through W(B). However, Pro­pos­i­tion 4.15 en­sures a bicol­lar neigh­bour­hood of r(B) in (B2×D˚2)/W(B). Pro­pos­i­tions 4.2 and 4.3 en­sure that r is ap­prox­im­able by homeo­morph­isms fix­ing r(B). There­fore, the pair ((B2×D˚2)/W(B),r(B)) (with the bicol­lar) is homeo­morph­ic to (B2×D˚2,r(B)).  ◻

Pro­pos­i­tion 4.16: R(B2×D˚2,r(B);P) is sat­is­fied.

Proof of Pro­pos­i­tion 4.16.  Giv­en an open neigh­bour­hood U of r(B), there ex­ists, by Pro­pos­i­tion 4.14, a homeo­morph­ism h:B2×D˚2B2×D˚2 with com­pact sup­port in a bicol­lar V of r(B) in U, such that h(r(B))r(B˚). Since r(B˚) is homeo­morph­ic to R4, there ex­ists a map g with sup­port in r(B˚) and ap­prox­im­able by homeo­morph­isms such that ghr(B) is a point in r(B˚). Let f=gh:B2×D˚2B2×D˚2. By uni­form con­tinu­ity on the com­pact sup­port Fr(B)V of f, we know that, for a giv­en ε>0, there ex­ists δ>0 such that for all sets EB2×D˚2 of dia­met­er less than δ, the dia­met­er of f(E) is less than ε. By Lemma 4.17, there ex­ists a stretch homeo­morph­ism θ:B2×D˚2 fix­ing r(B) and with sup­port in V such that, for all PP dis­tinct from B such that θ(P)F, we have diamθ(P)<δ. Then f=f0θ sat­is­fies R(B2×D˚2,B;P).  ◻

Lemma 4.17: (Stretch lemma.)  Let l be a null de­com­pos­i­tion X×[0,) where X is com­pact and all ele­ments of l is dis­joint from X×0. For all ε>0, there ex­ists a homeo­morph­ism with com­pact sup­port φ:[0,)[0,) such that Φ=φ×IdX sat­is­fies that, for all El such that Φ(E)(X×[0,1]), we have diam(Φ(E))<ε.

Proof of Lemma 4.17 (In­dic­a­tions.)  Fig­ure 21 com­pletes the proof. □

Figure 21.

All ele­ments of P dis­tinct from r(B) are of the form r(T(α)) where T(α) is a con­nec­ted com­pon­ent of a tor­us T(α). Fol­low­ing the meth­od of the proof of Pro­pos­i­tion 4.16, we es­tab­lish sim­il­arly the fol­low­ing pro­pos­i­tion.

Pro­pos­i­tion 4.18: R(B2×D˚2,r(T(α));P) is sat­is­fied.

Proof of Pro­pos­i­tion 4.18 (in­dic­a­tions).  The quo­tient of T(α)=J(α)×T(α), by the lon­git­ude l(α) that is in D(α), is a cone whose centre is the quo­tient of l(α), and the base is a sol­id tor­us. δr(T(α))r(l(α)) has a bicol­lar neigh­bour­hood in B2×D˚2, com­pare Pro­pos­i­tion 4.13. The ac­cu­mu­la­tion points of ele­ments Pa point of P are the centre r(l(α)) and a com­pact set r(WT(α)) far from r(l(α)).  ◻

Pro­pos­i­tion 4.19: p2 is ap­prox­im­able by homeo­morph­isms and hence B2×D˚2/D+B2×D˚2.

Proof.  Ap­ply Pro­pos­i­tions 4.18, 4.16 and 4.5.  ◻

5. Freedman’s approximation theorem

The­or­em 5.1: (Freedman’s approximation theorem.)  Sup­pose that X and Y are homeo­morph­ic to the n-sphere. Let f:XY be a sur­ject­ive, con­tinu­ous map such that the sin­gu­lar set S(f)={yYf1(y)a point} is nowhere dense and at most count­able. Then, f can be ap­prox­im­ated by homeo­morph­isms.
Re­mark: For all di­men­sions 4, there ex­ist much stronger ap­prox­im­a­tion the­or­ems [e19], [e18], [e32]. There­fore, in di­men­sion 4, the prob­lem of gen­er­al­ising The­or­em 5.1 re­mains open.

In the case that S(f) is fi­nite, this the­or­em is well known since it con­sti­tutes the es­sen­tial part of the cel­eb­rated Schönflies the­or­em which was es­tab­lished around 1960 by B. Mazur, M. Brown and M. Morse.

Re­call that a com­pact set A in a to­po­lo­gic­al n-man­i­fold M (without bound­ary) is cel­lu­lar if each neigh­bour­hood of A con­tains a neigh­bour­hood which is homeo­morph­ic to Bn.

Lemma 5.2: Let A be a com­pact, cel­lu­lar set in the in­teri­or IntM of a man­i­fold M. Then, the quo­tient map q:MM/A can be ap­prox­im­ated by homeo­morph­isms which are sup­por­ted in an ar­bit­rar­ily giv­en neigh­bour­hood of A.

Com­pare the Bing shrink­ing cri­terion, The­or­em 4.1 [e8]. A dir­ect proof shrinks A gradu­ally to a point.

Proof of The­or­em 5.1 if S(f) is a point.  Let y0=S(f) and A=f1(y0), we have that XA is homeo­morph­ic to Rn. Since X is homeo­morph­ic to Sn, it fol­lows that A is cel­lu­lar in X (ex­er­cise). Then, we ob­tain ap­prox­im­a­tions by ap­ply­ing Lemma 5.2.  ◻

In the set­ting of Freed­man’s ideas, the case where S(f) is n points, n2, is already as dif­fi­cult as The­or­em 5.1. However one can con­sult [e8], [e10] for an easy proof. We re­call the Schönflies the­or­em.

The­or­em 5.3: (Schönflies theorem.)  Let Σn1 be a to­po­lo­gic­ally em­bed­ded (n1)-sphere in Sn such that there is a bicol­lar neigh­bour­hood N of Σ in Sn, that is, (N,Σ) is homeo­morph­ic to (Σ×[1,1],Σ×0). Then the clos­ure of each of the two com­pon­ents of SnΣ is homeo­morph­ic to the n-ball Bn.

Proof of The­or­em 5.3 (start­ing from The­or­em 5.1 for S(f) con­sist­ing of two points).  Let X1 and X2 be two con­nec­ted com­pon­ents of SnN˚ and W1 and W2 be the clos­ures of con­nec­ted com­pon­ents of SnΣn1 con­tain­ing X1 and X2, re­spect­ively. It is ne­ces­sary to show that W1 and W2 are homeo­morph­ic to Bn.

Shrink­ing X1 and X2, we ob­tain a quo­tient map f:SnSn/{X1,X2}(Σ×[1,1])/{Σ×0,Σ×1}Sn that is ap­prox­im­able by homeo­morph­isms from The­or­em 5.1 (the case of S(f) is two points). So X1 and X2 are cel­lu­lar in Sn. Ap­ply Lemma 5.2 to XiW˚i, we de­duce that WiWi/XiΣ×[0,1]/{Σ×1}Bn is ap­prox­im­able by homeo­morph­isms.  ◻

Ob­ser­va­tion: The case of The­or­em 5.3, where we know in ad­vance that Σ bounds an n-ball in Sn, already arises from the case of The­or­em 5.1 where S(f)={1 point} proved above. Freed­man uses this case.

To prove The­or­em 5.1, Freed­man in­tro­duced a nice trick of it­er­ated rep­lic­a­tion of the ap­prox­im­a­tion map, which vaguely re­minds me of the ar­gu­ments of Mazur [e7]. This trick leads us to leave the cat­egory of con­tinu­ous maps and to in­stead work in the less fa­mil­i­ar realm of closed re­la­tions. It was dur­ing the sev­en­ties that closed re­la­tions im­posed them­selves for the first time on geo­met­ric to­po­logy; they sur­faced im­pli­citly in a very ori­gin­al art­icle by M. A. Stanko [e24] and have be­come es­sen­tial since: I be­lieve that it would be a her­culean task to prove, without closed re­la­tions, the sub­sequent the­or­em of An­cel and Can­non [e30] that any to­po­lo­gic­al em­bed­ding Sn1Sn, n5, can be ap­prox­im­ated by loc­ally flat em­bed­dings.

Defin­i­tion: A closed re­la­tion R:XY between met­ris­able spaces X and Y is a closed sub­set R of X×Y. If S:YZ is a closed re­la­tion, the com­pos­i­tion SR:XZ is SR={(x,z)X×Zthere is yY such that (x,y)R and (y,z)S}, which is also closed if Y is com­pact. There­fore the col­lec­tion of closed re­la­tions between com­pact spaces is a cat­egory.

A con­tinu­ous map f:XY gives a closed re­la­tion {(x,f(x))xX} (the graph of f) which we still call f. Re­cip­roc­ally, provided that Y is com­pact, a closed re­la­tion R:XY is the graph of a con­tinu­ous func­tion (which is uniquely de­term­ined) if Rx×Y is a point for all xX.

Re­mark: The nat­ur­al func­tion [0,1)R/Z is con­tinu­ous and biject­ive; the in­verse is dis­con­tinu­ous, but the graphs of both are closed.

By ex­tend­ing usu­al no­tions for con­tinu­ous func­tions, for AX and BY, we have

  1. the im­age R(A)={yYthere exists xA such that (x,y)R},
  2. the re­stric­tion R|A:AY is the closed sub­set RA×Y in A×Y,
  3. the in­verse R1:YX such that {(y,x)Y×X(x,y)R}.
Re­mark: R1 is the in­verse of R in the cat­egor­ic­al sense if and only if R is the graph of a bijec­tion func­tion (if and only if the cat­egor­ic­al in­verse ex­ists).

To ex­ploit an ana­logy between a func­tion XY and a re­la­tion R:XY, we will at any time as­sim­il­ate R to the func­tion that as­so­ci­ates for each point xX to a sub­set R(x)Y.

Proof of The­or­em 5.1.  Any sub­man­i­fold of codi­men­sion 0 that is in­tro­duced will be as­sumed to be to­po­lo­gic­al and loc­ally flat. Let N be a neigh­bour­hood of f in X×Y. The the­or­em as­serts that there ex­ists a homeo­morph­ism H:XY such that HN.

By re­mov­ing a small n-ball DYS(f) from Y and re­mov­ing its preim­age f1(D) from X, we see that it is per­miss­ible to ad­opt the fol­low­ing.

The­or­em 5.4: (Change of data.)  Sup­pose X and Y are homeo­morph­ic to Bn rather than Sn. Let f:XY be a sur­ject­ive, con­tinu­ous map such that the sin­gu­lar set S(f)={yYf1(y)a point} is nowhere dense and at most count­able and S(f)IntY. Then f can be ap­prox­im­ated by homeo­morph­isms.

It is easy to see that The­or­em 5.4 im­plies The­or­em 5.1 us­ing the spe­cial case of The­or­em 5.3 (Schönflies the­or­em) where Σn1 bounds a ball (see ob­ser­va­tion after The­or­em 5.3).

The first step of an in­duct­ive con­struc­tion of H is to ap­ply the fol­low­ing pro­pos­i­tion to the tri­angle shown to the right. Moreover, the neigh­bour­hood N of Pro­pos­i­tion 5.5 be­comes N the above; and L be­comes Y.

Sup­pose that X and Y are homeo­morph­ic to Bn. A re­la­tion R:XY is called good if it is closed, and sat­is­fy­ing the fol­low­ing con­di­tions:

  1. RX×Y pro­jects onto X and onto Y.
  2. R(x) is not a singleton set for at most count­ably many points in X and these ex­cep­tion­al points con­sti­tute a nowhere dense set con­tained in IntX. The same holds for R1.

It is said that a good re­la­tion R:XY is finer than R if RRX×Y.

Pro­pos­i­tion 5.5: Giv­en the tri­angle of good re­la­tions (which is pos­sibly com­mut­at­ive) shown to the right, where X, Y and Z are homeo­morph­ic to Bn, and f, g are in ad­di­tion con­tinu­ous func­tions; a neigh­bour­hood N of R in X×Y; and LZ an open sub­set (called the gap). We im­pose the fol­low­ing con­di­tions:
  1. R(f1(L)×g1(L))(f1(ZL)×g1(ZL)); it is in­ev­it­able if the tri­angle switches.
  2. R=g1f on f1(L).
  3. R is giv­en by the in­ter­sec­tion graph of a homeo­morph­ism f1(ZL)g1(ZL).
  4. The sin­gu­lar sets S(f) and S(g) are sep­ar­ated on L, that is, there are two open dis­joint sets U and V which con­tain S(f)L and S(g)L, re­spect­ively.

Then, for all ε>0, we can modi­fy the three data g, R, L to g, R, L so that in ad­di­tion to the same con­di­tions above (with g, R, L in­stead of g, R, L), we have R=R on f1(ZL), LL, and for all yY, diamR1(y)<ε.

Ad­dendum: There ex­ists a neigh­bour­hood NN of R in X×Y such that diam(N1(y))<ε for all yY.

Proof of Ad­dendum.  If the con­clu­sion is false, then there are two se­quences of points of X×Y, say (xk,yk), (xk,yk), k=1,2,3,, which con­verge in com­pact R and such that d(yk,yk)ε. By com­pact­ness of X×Y, we can ar­range that the se­quences xk, yk and yk con­verge to x, y and y, re­spect­ively. Then, (x,y) and (x,y) be­long to com­pact R, but d(y,y)ε, which is a con­tra­dic­tion.  ◻

Pro­pos­i­tion 5.5 (with Ad­dendum) will be used as a ma­chine that swal­lows the data f, g, R, L, N, ε and man­u­fac­tures f, g, R, L, N.

Let us con­tin­ue con­struct­ing the homeo­morph­ism H, as­sum­ing Pro­pos­i­tion 5.5. For k1, the k-th step con­structs a tri­angle shown to the right (where Z is a copy of Y); a sub­man­i­fold LkZ and a neigh­bour­hood Nk of Rk in X×Y such that fk, gk, Rk, Lk, Nk sat­is­fy the con­di­tions im­posed on f, g, R, L, N in Pro­pos­i­tion 5.5. The first step is already spe­cified: Pro­pos­i­tion 5.5 cre­ates f1, g1, R1, L1, N1 from f, Id, f, Y, N, 1.

Sup­pose that the k-th tri­angle is con­struc­ted and we con­struct the (k+1)-th tri­angle.

  1. If k is odd, then Pro­pos­i­tion 5.5 gives gk+1, fk+1, Rk+11, Lk+1, Nk+11 from gk, fk, Rk1, Lk, Nk1, 1/k. In brief, we ap­ply Pro­pos­i­tion 5.5 to the re­verse tri­angle shown to the right.
  2. If k is even, then it is same as the first step: Pro­pos­i­tion 5.5 gives fk+1, gk+1, Rk+1, Lk+1, Nk+1 from fk, gk, Rk, Lk, Nk, 1/k.

By in­duc­tion, we have NN1N2. We define H=kNk. Then, H is a homeo­morph­ism since, for all x, we have diamH(x)diamNk(x)1/k, for all even k, and diamH1(x)diamNk1(x)1/k, for all odd k. This homeo­morph­ism H in the neigh­bour­hood N of f com­pletes the proof of The­or­em 5.1 as­sum­ing Pro­pos­i­tion 5.5.  ◻

Proof of Pro­pos­i­tion 5.5.  To ex­plain the es­sen­tial idea of Freed­man, the read­er should read the proof with a view to (re)prov­ing that a sur­jec­tion f:BnBn such that S(f)={a point}IntBn is ap­prox­im­able by homeo­morph­isms (for this, we set f=R and g=Id). Then, it should be noted that as soon as S(f)={k points}IntBn, the same ar­gu­ment leads us to ap­prox­im­ate f by re­la­tions which crush noth­ing, but which blow up k(k1) points.

Con­sider the preim­ages R1(y), yY, of dia­met­er ε, that we want to elim­in­ate. Ac­cord­ing to (a), (b) and (c), these sets con­sti­tute the preim­age by f of the set (Sε(f)L)Z, where Sε(f)={zZdiamf1(z)ε}, which will al­low us to fol­low the case in Z. Note that Sε(f) is com­pact al­though, typ­ic­ally, S(f) is not. For ex­ample, Sε(f) is fi­nite in the case of in­terest to Freed­man (see Sec­tion 4).

Lemma 5.6: (General position.)  In the in­teri­or of a com­pact to­po­lo­gic­al man­i­fold M, let A and B be two count­able sets and nowhere dense. Then there ex­ists a small auto­morph­ism θ of M fix­ing all points of M, such that θ(A) and B are sep­ar­ated, that is, con­tained in dis­joint open sets.

Proof of Lemma 5.6  Con­sider the space Aut(M,M) of auto­morph­isms of M fix­ing M, provided with the com­plete met­ric sup(d(f,g),d(f1,g1)) where d is the uni­form con­ver­gence met­ric. In Aut(M,M), the set of auto­morph­isms θ, such that the first k points Ak of A and Bk of B sat­is­fy­ing θ(Ak)B==θ(A¯)Bk, con­sti­tute an open sub­set UkAut(M,M) every­where dense in Aut(M,M), be­cause A¯ and B are closed, nowhere dense in M.

Then, the fam­ous Baire cat­egory the­or­em as­serts that the count­able in­ter­sec­tion kUk is every­where dense in Aut(M,M). Note that kUk is the set of θ in Aut(M,M) such that θ(A)B==θ(A¯)B. But, for X1, X2 in a met­ris­able M, the con­di­tion that X1X2==X1X2 leads to the sep­ar­a­tion of X1 and X2 in M. In ef­fect, seen in the open sub­set M(X1X2) of M, the sets X1(X1X2)and(X1X2) are al­ways dis­joint, closed and hence sep­ar­ated. The men­tioned con­di­tion en­sures that they con­tain re­spect­ively X1 and X2.  ◻

Claim 5.7: (Trivial if Sε(f) is finite.)  There ex­ists a fi­nite uni­on B+ of dis­joint n-balls in L sat­is­fy­ing the fol­low­ing con­di­tions:
  1. Sε(f)LB˚+.
  2. S(g)B+=.
  3. Each con­nec­ted com­pon­ent B+ of B+ is small in the sense that (f1(B+))×(g1(B+))N, and stand­ard in the sense that ZIntB+ is homeo­morph­ic to Sn1×[0,1].

Proof of Claim 5.7.  Identi­fy Z with BnRn to give L an af­fine lin­ear struc­ture. Let K be a com­pact neigh­bour­hood of the com­pact set Sε(f)L which is a sub­poly­hedra of L and dis­joint from S(g), see Pro­pos­i­tion 5.5(c). We sub­divide K in­to a sim­pli­cial com­plex of which each sim­plex L is lin­ear in L and so small such that f1(Δ)g1(Δ)N. Then (com­pare, the proof of Lemma 5.6), by a small per­turb­a­tion (a trans­la­tion if we want) of K in L, we dis­en­gage the (n1)-skel­et­on K(n1) from the com­pact count­able Sε(f), without harm­ing the prop­er­ties of K already es­tab­lished. Fi­nally, B+ is defined as K minus a small δ open neigh­bour­hood of K(n1) in Rn. Each com­pon­ent B+ of B+ is con­vex and in IntZ=B˚n; there­fore ZB˚+ is homeo­morph­ic to Sn1×[0,1], by an ele­ment­ary ar­gu­ment.  ◻

In IntB+, we choose now a uni­on B of balls (one in each con­nec­ted com­pon­ent of B+), which still sat­is­fies (1), (2), (3) and also

(4) S(f)B=.

We set L=LB. For each con­nec­ted com­pon­ent B+ of B+, we are now modi­fy­ing g and R above B+ to define g and R. These changes for the vari­ous con­nec­ted com­pon­ents B+ are dis­joint and in­de­pend­ent. There­fore, it is enough to spe­cify one. Moreover, in or­der to sim­pli­fy the nota­tion, we al­low ourselves to spe­cify this change only in the case that B+ is con­nec­ted.

Figure 22.

Let c:ZB+ be a homeo­morph­ism, called the com­pres­sion, which fixes all points of B. (We re­mem­ber that ZB˚ is homeo­morph­ic to Sn1×[0,1] and B+B˚.) We should modi­fy c by com­pos­ing with a homeo­morph­ism θ of B+B˚ fix­ing B+B giv­en by Lemma 5.6, to as­sure that S(f) and c(S(f)) are sep­ar­ated on the open B˚+B.

Since g1(B+) is a ball in Y (in fact, g is a homeo­morph­ism over B+), we can also choose i so that i|X is (g1cf)|X. We set g={gon g1(ZB˚+),cfi1on g1(B+). On g1(B+), g is well-defined since g=cf(g1cf)1 on g1(B+). We set R={Ron f1(ZB˚+),g1f=(if1c1)fon f1(B+). More pre­cisely, on f1(B+), we spe­cify R={(if1c1)fon f1(B+B˚),ion f1(B). On f1(B), R is well-defined since c fixes all points of B, and S(f)B=.

We have now spe­cified the modi­fic­a­tion L, g, R of L, g, R claimed by Pro­pos­i­tion 5.5. (We re­mark that if B+ is a uni­on of k balls, rather than one ball, the modi­fic­a­tion is done in k dis­joint and in­de­pend­ent steps, each sim­il­ar to the one just spe­cified for con­nec­ted B+.)

Veri­fy­ing the claimed prop­er­ties for L, g, R is dir­ect. (There are already manuscripts [1], [2] which of­fer more de­tails.)  ◻

Re­mark 1: The sys­tem of the above for­mula, spe­cify­ing g and R, hides the geo­metry. We now try to re­veal it by look­ing f and g re­spect­ively as fibra­tions φ and γ, of base Z, and vari­able fibre, which al­lows us to use the no­tion of fibre re­stric­tion. Let γ0=γ(γ|B˚+). We form γ of φγ0 by identi­fy­ing, via c|Z, the sub­fibres (whose fibres are points) φ|Z and γ0|B+. Then, we can identi­fy the total space and the base of γ=φγ0 to those of γ by an ex­ten­sion of γ0γ. More pre­cisely, we use (Id|ZB˚+)c between bases, which is the iden­tity on BB+. Then, φandγ:YgZ are fibra­tions over Z nat­ur­ally iso­morph­ic on BL, which defines a re­la­tion R finer than the simple cor­res­pond­ence of fibres g1f.
Re­mark 2: In the proof of The­or­em 5.1, we can eas­ily en­sure that fn and gn con­verge to­wards f and g, and that gH=f. Thus, as fibres, f and g are iso­morph­ic. Moreover, each fibre of f or g is homeo­morph­ic to a fibre of f. I point out that f and g re­mind me of the two in­fin­ite products of Mazur [e7] and that H re­minds me of the fam­ous Eilenberg–Mazur swindle, which com­pletes the proof of The­or­em 5.3 (weakened ver­sion) giv­en in [e7].
Re­mark 3: (Following Remark 2.)  If we want to avoid un­ne­ces­sary com­plic­a­tions in the struc­ture of f and g, it should be noted that in the defin­i­tion of g above, we have the right to re­place the map f which oc­curs by any good map f:XY such that f=f on f1(B). Then, for every use of Pro­pos­i­tion 5.5 in the proof of The­or­em 5.1, we find that one al­ways has the pos­sib­il­ity of choos­ing for an al­tern­at­ive f a map iso­morph­ic to a map f giv­en in The­or­em 5.1. With this little re­fine­ment, the proof of The­or­em 5.1 in the case S(f)={2 points} is close to the ar­gu­ment of [e7]. In par­tic­u­lar, S(f) and S(g) can be homeo­morph­ic to Z{,}.

Works

[1] article M. H. Freed­man: “A fake S3×R,” Ann. of Math. (2) 110 : 1 (1979), pp. 177–​201. MR 0541336 Zbl 0442.​57014

[2]F. An­cel: “Nowhere dense tame 0-di­men­sion­al de­com­pos­i­tions of S4 (an ex­pos­i­tion of a the­or­em of Mike Freed­man)”. Un­pub­lished manuscript, 1981. This work was ori­gi­nally (in­cor­rec­tly) cited by Sie­ben­mann in his list of re­fe­ren­ces as “No­where dense tame 0-di­men­sio­nal de­com­po­si­tions of S4; an ex­po­si­tion of a the­orem of Mike Freed­man”. Ric Ancel notes the cor­rect title (email to the editor, dated 28 May 2020) and has made the PDF avail­able to rea­ders of CM. techreport

[3]L. Sieben­mann: “La con­jec­ture de Poin­caré to­po­lo­gique en di­men­sion 4 (d’après M. H. Freed­man),” pp. 219–​248 in Bourbaki Sem­in­ar, vol. 1981/1982. Astérisque 92. Soc. Math. France, Par­is, 1982. MR 689532 incollection