Minimal extension for the α-Manhattan norm
Full text
This is a sel -a chi ed e sion o an o iginal a icle. This e sion
may di e om he o iginal in pagina ion and ypog aphic de ails.
Au ho (s):
Ti le:
Yea :
Ve sion:
Copy igh :
Righ s:
Righ s u l:
Please ci e he o iginal e sion:
CC BY 4.0
h ps://c ea i ecommons.o g/licenses/by/4.0/
Minimal ex ension o he α-Manha an no m
© 2024 Accademia Nazionale dei Lincei
Published e sion
Campbell, Daniel; Kau anen, Aapo; Radici, Emanuela
Campbell, D., Kau anen, A., & Radici, E. (2023). Minimal ex ension o he α-Manha an no m.
Rendicon i Lincei: Ma ema ica e Applicazioni, 34(4), 773-807. h ps://doi.o g/10.4171/ lm/1027
2023
Rend. Lincei Ma . Appl. 34 (2023), 773–807
DOI 10.4171/RLM/1027
©2024 Accademia Nazionale dei Lincei
Published by EMS P ess
This wo k licensed unde a CC BY 4.0 license
Calculus o Va ia ions.
–
Minimal ex ension o he
˛
-Manha an no m, by Daniel
Campbell, Aapo Kau anen and Emanuela Radici, communica ed on 10 No embe
2023.
Abs ac .
–
Le
@Q
be he bounda y o a con ex polygon in
R2
,
e˛D.cos ˛; sin ˛/
and
e?
˛D.sin ˛; cos ˛/
a basis o
R2
o some
˛2Œ0; 2/
and
'W@Q!R2
a con inuous,
ini ely piecewise linea injec i e map. We cons uc a ini ely piecewise a ine homeomo phism
WQ!R2
coinciding wi h
'
on
@Q
such ha he ollowing p ope y holds:
jhD ; e˛ij.Q/
( esp.,
hD ; e?
˛ij.Q/
) is as close as we wan o
in jhDu; e˛ij.Q/
( esp.,
in jhDu; e?
˛ij.Q/
)
whe e he in imum is mean o e he class o all
BV
homeomo phisms
u
ex ending
'
inside
Q
.
This esul ex ends ha al eady p o en by P a elli and he hi d au ho in [A i Accad. Naz. Lincei
Rend. Lincei Ma . Appl. 29 (2018), no. 3, 511–555] in he shape o he domain.
Keywo ds. – Homeomo phic ex ension, BV homeomo phisms, s ic app oxima ion in BV.
Ma hema ics Subjec Classi ica ion 2020.
–
46E35 (p ima y); 30E10, 58E20 (seconda y).
1. In oduc ion
In his pape , we a e in e es ed in he p oblem o ex ending injec i e con inuous
and piecewise linea bounda y alues om a con ex polygon by piecewise a ine
homeomo phisms. The mo i a ion o such a s udy a ises in he con ex o app oxima ion
p oblems ound in egula i y heo y o non-linea elas ici y. The e is al eady a plu ali y
o ex ension esul s in a a ie y o con ex s, which ha e been applied o sol e a ious
app oxima ion p oblems. Le us now gi e an o e iew o some examples.
In gene al, we a e in e es ed in he app oxima ion o a weakly di e en iable homeo-
mo phism, which we would like o app oxima e by
C1
homeomo phisms o by locally
ini e piecewise a ine homeomo phisms. The app oxima ion o a plana
W1;p
home-
omo phism
1 < p < 1
in [9,10] elies hea ily on he injec i i y o he ha monic
ex ension o con ex bounda y alues. In [5], he au ho s we e also able o app oxima e
a bi-Lipschi z map and i s in e se simul aneously in he
.p; p/
-bi-Sobole se ing
and o do so used he ex ension esul in [6]. In o de o sol e he
W1;1
case in [8],
he au ho s had o de elop an independen ex ension esul in ha pape which was
u he examined and imp o ed in [1,15]. The ex ension esul was also u ilized in
d. campbell, a. kau anen and e. adici 774
he
.1; 1/
-bi-Sobole se ing in [12]. Finally, le us men ion ha he au ho s o [14]
app oxima e plana
BV
homeomo phisms using an ex ension esul hey p o ed in [13].
Mo e han jus he app oxima ion o weakly di e en iable homeomo phisms by
di eomo phisms, hese ex ension esul s ha e been key in examining he beha io o
weak and s ong limi s o homeomo phisms in hei espec i e classes. Such esul s
include a ca ego iza ion o he closu e o
Hom W1;p
,
p2
, in [11], a ca ego iza ion
o he closu e o Hom W1;p,1<p<2, in [7] and pa ial BV esul in [2,4].
I was demons a ed in [13] ha hei main ex ension esul can be “ o a ed” o
app oxima e a BV homeomo phism s ic ly and simila ly in [14] o he a ea-s ic case.
Ne e heless, his app oach makes he applica ion o he ex ension esul somewha
cumbe some and echnical. The main esul o he p esen pape is a piecewise a ine
homeomo phic ex ension ha imp o es on ha o [13]. Mo e p ecisely, we conside
ex ensions o piecewise linea bounda y alues de ined on a bounda y o con ex
quad ila e als (and no only ec angles pa allel o he coo dina e axes as in [13]) which
a e op imal in a pa icula BV sense. We emphasize ha he gene ali y o he class
o con ex quad ila e als includes he “ o a ed” e sion o he ex ension esul o [13].
Also, ou Theo em 1.2 is s onge han he ex ension heo em he e (no only because
o he shape o
Q
) in he sense ha i immedia ely implies hei ex ension heo em bu
he opposi e is no ue (see Rema k 1.3). Ne e heless, his imp o emen is a case o
sepa a ing es ima es al eady conduc ed in [13].
The mo i a ion o ou ex ension heo em is he ull ca ego iza ion esul in [3],
whe e we iden i y a condi ion which gua an ees ha a map is a s ic o a ea-s ic
limi o BV homeomo phisms. In he cou se o he app oxima ion, we wan o wo k on
g ids ha a e no only made up o ec angles, and we p e e o no ha e o o a e he
ec angles. In ha sense, we need he cu en esul , which we p esen below, a e we
se some necessa y no a ion.
Le
QR2
be a con ex polygon, le
˛2Œ0;2/
be ixed and call
e˛D.cos˛;sin˛/
and e?
˛D.sin ˛; cos ˛/.
We de ine he ollowing numbe s (see Figu e 1):
(1.1) aWD in ®hx; e?
˛i W x2Q¯aCWD sup ®hx; e?
˛i W x2Q¯;
bWD in ®hx; e˛i W x2Q¯; bCWD sup ®hx; e˛i W x2Q¯:
Fo each s2.a; aC/, we de ine V1
s; V 2
suniquely by he condi ions
(1.2) V1
s; V 2
s2@Q;hV1
s; e?
˛iDhV2
s; e?
˛i D s; hV1
s; e˛i<hV2
s; e˛i:
Simila ly, o e e y 2.b; bC/, we de ine H1
; H2
uniquely by
(1.3) H1
; H2
2@Q;hH1
; e˛iDhH2
; e˛i D ; hH1
; e?
˛i<hH2
; e?
˛i:
minimal ex ension o he ˛-manha an no m 775
minimal ex ension o he 𝛼-manha an no m 3
𝑒𝛼
{⟨𝑥, 𝑒𝛼⟩=𝑏+}
{⟨𝑥, 𝑒𝛼⟩=𝑎−}
{⟨𝑥, 𝑒𝛼⟩=𝑏−}
{⟨𝑥, 𝑒𝛼⟩=𝑎+}
𝐻1
𝑡
𝑉1
𝑠
𝐻2
𝑡
𝑉2
𝑠
Q
Figu e 1. Polygon Qwi h 𝐻1
𝑡, 𝐻2
𝑡, 𝑉1
𝑠and 𝑉2
𝑠.
by
𝜌P(A,B)
he geodesic dis ance be ween
A
and
B
inside
P
. We de ine he quan i y
Ψ𝛼(𝜑):=∫𝑎+
𝑎−
𝜌P(𝜑(𝑉1
𝑠), 𝜑(𝑉2
𝑠))𝑑𝑠 +∫𝑏+
𝑏−
𝜌P(𝜑(𝐻1
𝑡), 𝜑(𝐻2
𝑡))𝑑𝑡.
Loosely speaking, he quan i y
Ψ𝛼(𝜑)
accoun s o he leng h o all he geodesics
inside
P
connec ing pai s o poin s on
𝜑(𝜕Q)
whose p eimage in
𝜑
is a pai o poin s
in
𝜕Q
lying on a line pa allel o ei he
𝛼
o
𝛼⊥
. Fu he o
𝑢∈𝐵𝑉 (Ω,R2)
we deno e
he 𝛼-Manha an no m o 𝐷𝑢 as ∥·∥𝛼which we de ine as
∥𝐷𝑢∥𝛼(Q) :=|⟨𝐷𝑢, 𝑒𝛼⟩|(Q) + |⟨𝐷𝑢, 𝑒⊥
𝛼⟩|(Q).
The main esul s o he pape is a e he ollwoing.
Theo em 1.1.Le
𝛼∈ [0,2𝜋)
be ixed,
Q ⊂ R2
be a con ex polygon and
𝜑:𝜕Q → R2
be a con inuous piecewise linea injec i e map. Then o e e y
𝜀 > 0
he e exis s a
ini ely piecewise a ine homeomo phism 𝑣:Q → R2ex ending 𝜑, such ha
(1.4) ∥𝐷𝑣∥𝛼(Q) ≤ Ψ𝛼(𝜑) + 𝜀.
Figu e 1. Polygon Qwi h H1
,H2
,V1
sand V2
s.
Le
'W@Q!R2
be con inuous, injec i e and piecewise linea . We deno e
P
as
he bounded componen o
R2n'.@Q/
. Fo e e y pai o poin s
A;B2x
P
, we deno e
by
P.A;B/
he geodesic dis ance be ween
A
and
B
inside
x
P
. We de ine he quan i y
‰˛.'/ WD ZaC
a
P'.V 1
s/; '.V 2
s/ds CZbC
b
P'.H1
/; '.H2
/d :
Loosely speaking, he quan i y
‰˛.'/
accoun s o he leng h o all he geodesics inside
P
connec ing pai s o poin s on
'.@Q/
whose p eimage in
'
is a pai o poin s in
@Q
lying on a line pa allel o ei he
˛
o
˛?
. Fu he , o
u2BV.; R2/
, we deno e he
˛-Manha an no m o Du as kk˛which we de ine as
kDuk˛.Q/WD ˇˇhDu; e˛iˇˇ.Q/CˇˇhDu; e?
˛iˇˇ.Q/:
The main esul s o he pape a e he ollowing.
Theo em 1.1.Le
˛2Œ0; 2/
be ixed,
QR2
a con ex polygon and
'W@Q!R2
a
con inuous piecewise linea injec i e map. Then, o e e y
">0
, he e exis s a ini ely
piecewise a ine homeomo phism WQ!R2ex ending ', such ha
(1.4) kD k˛.Q/‰˛.'/ C":
Theo em 1.2.Le ">0and le be he ex ension om Theo em 1.1. Then,
(1.5) ˇˇhD ; e˛iˇˇ.Q/ZbC
b
P'.H1
/; '.H2
/d C";
ˇˇhD ; e?
˛iˇˇ.Q/ZaC
a
P'.V 1
s/; '.V 2
s/ds C":
d. campbell, a. kau anen and e. adici 776
Rema k 1.3.We obse e ha Theo em 1.1 is s onge han he esul o [13] as i
p o ides he almos op imal ex ension wi h espec o any o a ed Manha an no m and
no jus o he canonical one (whe e ˛D0).
Le us also ema k ha Theo em 1.2 immedia ely implies Theo em 1.1 bu he
a gumen used o cons uc is exac ly he same. Also, i is immedia e ha
ZbC
b
P'.H1
/; '.H2
/d
in ®ˇˇhDu; e˛iˇˇ.x
Q/Wu2Hom BV.x
Q;R2/; u D'on @Q¯
and
ZaC
a
P'.V 1
s/; '.V 2
s/ds
in ®ˇˇhDu; e?
˛iˇˇ.x
Q/Wu2Hom BV.x
Q;R2/; u D'on @Q¯;
and ou esul in ac shows ha he e is a sequence o homeomo phisms achie ing he
in imum and ha ing a ia ion con e ging o he le -hand side in he sense o
(1.5)
. This
ac is ac ually a di ec consequence o he p oo s in [13], hough i was no explici ly
ema ked he e. The key a gumen is in Theo em 2.9.
1.1. Ske ch o he p oo
Be o e expounding he p oo in de ail, le us look a an o e iew o he p oo . We s a
wi h a con ex polygon
Q
. Up o a o a ion o
˛
, we may assume ha
˛D0
. Ei he
( he o a ed)
Q
has ho izon al sides, o a e emo ing a iny iangle called
T1
close o
he lowes poin o
Q
and a iangle called
T2
close o he highes poin o
Q
we ge a
con ex
ha has a pai o ho izon al sides (see Figu e 2). We ex end
'
on
@T1; @T2
so ha i is con inuous injec i e and piecewise linea . By making he iangles small
enough, we gua an ee ha
‰0.'.// C‰0.'.T1// C‰0.'.T2// ‰0.'.Q// C"
.
He e, ou new 'ex ends he o iginal ' om @Q. This s ep is Lemma 3.2.
Now, we sepa a e
in o hin ho izon al s ips
Si
(see Figu e 3), de ining a con inuous
injec i e piecewise linea
'
on
@Si
so ha
PM
iD1‰0.'.@Si// ‰0.'.@// C"
which
ex ends he o iginal ' om @ [@T1[@T2. This s ep is Lemma 3.3.
We sepa a e each
Si
in o a cen al ec angle and a pai o igh -angle iangles a
each end. On he ec angula domains
Ri
, we can use P oposi ion 2.8 o ex end he
bounda y alues and ge a piecewise a ine homeomo phism
wi
on he
Ri
sa is ying an
es ima e on
jDwij.Ri/
. In Lemma 4.2, we show how we ex end he bounda y alues
o ge a piecewise a ine homeomo phism on he iangles a he ends o he s ips; see
Figu e 6. We do his by u he sepa a ing hem in o e en hinne ec angles whe e
minimal ex ension o he ˛-manha an no m 777
we can ex end and es ima e as abo e. The emaining pa o he se is so small ha i s
con ibu ion o he no m is bounded by 2i".
The inal pa o he p oo is colla ing he es ima es and summing o es ima e ha
ou mapping sa is ies (1.4).
2. P elimina ies
In his sec ion, we ecall a lis o de ini ions and known geome ical esul s which a e
al eady a ailable in he li e a u e. Mos o hem a e aken om [13,14].
No a ion 2.1.Th oughou he pape , we endea o o keep o he ollowing no ms o
no a ion:
•Qis a con ex polygon,
•P
is a 2-dimensional polygon wi h bounda y
@P
. I
'W@Q!R2
is injec i e and
piecewise linea con inuous, hen
P
is he polygon co esponding o he bounded
componen o R2n'.@Q/and PD'.@Q/,
•˛2Œ0; 2/
is a gi en angle and he ec o
e˛WD .cos ˛; sin ˛/
. Also, we deno e
e?
˛WD .cos.˛ C=2/; sin.˛ C=2//,
•uand a e plana BV mappings,
•a;aC; b; bC
a e he numbe s om
(1.1)
, ypically
s2.a;aC/
and
2.b; bC/
and `DaCa,hDbCb,
•is a con ex polygon wi h 2 sides pa allel o ˛,
•T; T1; T2;z
T ; T 1
i; T 2
ia e all iangles,
•V1
s; V 2
s; H1
; H2
a e he poin s sa is ying he condi ions in
(1.2)
and
(1.3)
al hough
we may eplace Qwi h ano he con ex polygon, o example, o T,
•
by
RQDŒa; aCe˛CŒb; bCe?
˛
we deno e he smalles ec angle wi h sides
pa allel o e˛and e?
˛con aining Q,
•c1; c2; d1; d22Ra e o dina es,
•
by
z
C
we deno e a gene ic cons an whose p ecise alue may a y be ween es ima es,
• he poin s in he p eimage a e A; B; C; D; E; F; G; P; Q,1
•
i
'
is a piecewise linea injec i e map de ined on he iangle
ABC
, we call
dWD j'.A/ '.C /j he leng h o he image o he hypo enuse h ough ',
(1)
We do no need o u ilize he no a ion
B.x; / D ¹yW jyxj< º
so he e is no dange
o con usion when using B o deno e a poin .
d. campbell, a. kau anen and e. adici 778
•ˇ2.0;
2/is he angle a Ain he iangle ABC ,
•>0is a small chosen pa ame e ,
• he poin s in he image a e w i en in bold on , e.g., A;B;C;D;X;Y;Z,
•P;P0;PC
a e polygons in he image, ypically he piecewise a ine image o a
polygon in he p eimage, e.g., PD'.@Q/,
•'
,
a e con inuous injec i e piecewise linea maps om one-dimensional “skele-
ons” (i.e., a ini e union o segmen s) in he p eimage,
•
gi en a se
AR2
and a unc ion
'WA!R2
, we deno e by
'eB
he es ic ion
o ' o a subse BA,
•A;B
is he geodesic cu e om
A
o
B
in
P
and
P.A;B/
is he leng h o ha
cu e.
De ini ion 2.2 (Geodesics and modi ied geodesics).Le
PR2
be a polygon, and
le
A
and
B
be any wo dis inc poin s in
P
. We de ine
AB
as he unique geodesic (i.e.,
cu e o minimal leng h) connec ing hem, lying inside
P
. No ice ha
AB
is a piecewise
linea cu e, whose e ices a e only
A;B
and some e ices o
@P
whose in e nal
angles ha e size a leas
. Assume now ha
A;B2@P
, and le
W1;W2; : : : ; WK
be
all he e ices o Pme by AB, so ha
AB DAW1;W2;:::;WKB:
Fix now any
ı > 0
. Fo e e y
1iK
, le
z
Wi¤Wi
be some a bi a y poin in he
in e nal bisec o o he angle a
Wi
ha ing dis ance om
Wi
smalle han
ı
. The piece-
wise linea cu e zAB DAz
W1;z
W2;:::; z
WKBis hen called a ı-modi ica ion o AB.
No ice ha he e exis s a cons an
x
ı.P/>0
, depending on
P
bu no on
A
and
B
,
such ha he in e io o
zAB
is con ained in he in e io o
P
i
ı < x
ı.P/
, unless he
segmen
AB
is al eady con ained in
@P
, in which case
KD0
and
zAB DAB @P
.
Lemma 2.3 ([13, Lemma 2.4]).Le
A;B;C
and
D
be ou dis inc poin s in a polygon
P
.
Then, he in e sec ion
AB CD
is ei he emp y o closed and connec ed. Assume now
also ha
A;B;C;D2@P
and call
@P1; @P2
he wo componen s o
@Pn ¹C;Dº
. I
A2@P1
and
B2@P2
, hen
AB CD ¤ ;
. I
A;B2@P1
and
AB CD ¤ ;
, hen
he i s and las poin o his in e sec ion mus ei he be e ices o
P
o coincide wi h
one o he poin s Ao B.
We ema k ha in he e e ence he lemma is s a ed wi hou closedness o he
in e sec ion bu i ollows om he ollowing simple obse a ion. I
A;B;C
and
D
a e
ou dis inc poin s in
P
and he in e sec ion
AB CD
is no emp y, hen i is ei he a
poin (hence a closed se ) o a piecewise linea cu e which s a s and ends a co ne s
o @P( hus being he ini e union o closed connec ed segmen s).
minimal ex ension o he ˛-manha an no m 779
Lemma 2.4 ([13, Lemma 2.5]).Le
P
be a polygon, le
A;B2@P
be wo poin s
such ha he segmen
AB
is no con ained in
@P
, hen le
ı < x
ı.P/
and le
zAB
be a
modi ied geodesic in he sense o De ini ion 2.2. Le also
P1
and
P2
be he wo polygons
in which
P
is di ided by
zAB
, and le
">0
be a gi en cons an . I
ı
is small enough,
depending only on "and P, hen he ollowing is ue.
Fo any wo poin s C;D2Pi o i2 ¹1; 2º, one has
(2.1) Pi.C;D/<P.C;D/C":
I
C2P1
,
D2P2
and
E2@P1 @P2
is any poin wi h dis ance a mos
ı
om
CD, hen
(2.2) P1.C;E/CP2.E;D/<P.C;D/C":
De ini ion 2.5 (Se o e ices o a geodesic cu e).Le
PR2
be a polygon. Fo
e e y
A;B2@P
, he e is a unique o de ed se
X.A;B/D ¹X1; : : : ; XNº
such ha
he geodesic
AB
is exac ly he piecewise linea cu e
AX1;:::;XNB
, and he poin s
Xj
a e all he e ices o
P
me by he geodesic
AB
(excep
A
and
B
hemsel es, in
case hey a e al eady e ices). The se X.A;B/is called he se o e ices o AB.
De ini ion 2.6 (
ı
-linea iza ion o a Jo dan cu e).Le
be a Jo dan cu e wi h ini e
leng h, and le
ı > 0
be much smalle han he diame e o he bounded componen
o
R2n
. Le
_
A1B1;_
A2B2;:::; _
ANBN
be ini ely many essen ially disjoin a cs
con ained in
. Le hen
'
be he closed cu e ob ained by eplacing each a c
_
AiBi
wi h he segmen AiBi. We say ha 'is a ı-linea iza ion o i
•'is injec i e,
•e e y a c _
AiBiis such ha H1._
AiBi/<ı,
•_
AiBi 'AiBi.
The
ı
-linea iza ion is said comple e i he union o he a cs
_
AiBi
is he whole cu e
;
hence, 'is piecewise linea .
Lemma 2.7 ([13, Co olla y 4.3]).Le
R2
be a con ex polygon, and le
W@ !
R2
be a pa ame ized Jo dan cu e wi h ini e leng h and le
'W@ !R2
be a
ı-linea iza ion o . Then, o e e y P; Q 2@, one has
'.@/'.P /; '.Q/ .@/ .P /; .Q/C2ı:
In pa icula , o e e y 2Œ0; 2/, we deduce
‰.'/ ‰. / C2ıH1.@/:
d. campbell, a. kau anen and e. adici 780
We conclude his sec ion ecalling wo ex ension esul s ha will be use ul in he
sequel. The nex p oposi ion (P oposi ion 2.8) is p o ed in [13, Theo em A], and
Co olla y 2.10 is a s aigh o wa d consequence o P oposi ion 2.8.
P oposi ion 2.8 (Minimal ex ension o s anda d Manha an no m).Le
RR2
be
a ec angle o he o m
Œa; aCŒb; bC
, and le
'W@R!R2
be a con inuous
injec i e map. Then, o e e y
">0
, he e exis s a piecewise a ine homeomo phism
WR!R2coinciding wi h 'on @Rsuch ha
kD k0.R/‰0.'/ C":
Mo eo e , i
'
is piecewise linea , hen he map
can be chosen ini ely piecewise
a ine.
Theo em 2.9 (Minimal ex ension o s anda d Manha an no m).Le
">0
, and le
be he mapping om P oposi ion 2.8. Then,
jD1 j.Q/ZbC
b
P'.H1
/; '.H2
/d C";
jD2 j.Q/ZaC
a
P'.V 1
s/; '.V 2
s/ds C":
P oo .
The ini ely piecewise a ine homeomo phisms om a ec angle o a polygon
in [13] used in he p oo o P oposi ion 2.8 a e cons uc ed in [13, Lemma 2.12]. The
key es ima es we need o ex ac a e he las wo unnumbe ed equa ions o he p oo ,
ound in [13, p. 543]. They say exac ly ha
jD1 j.Q/ZbC
b
P'.H1
/; '.H2
/d C";
jD2 j.Q/ZaC
a
P'.V 1
s/; '.V 2
s/ds C":
Co olla y 2.10 (
W1;1
ex ension wi h non-op imal bound).The e exis s
z
C > 0
such
ha he ollowing holds. Le
RR2
be a ec angle, le
@R
be i s bounda y and le
'W@R!R2
be a con inuous, piecewise linea and injec i e map. Then, he e exis s
a ini ely piecewise a ine homeomo phism WR!R2ex ending 'such ha
kD kL1.R/z
CH1.@R/H1'.@R/:
P oo . The conclusion ollows by applying P oposi ion 2.8 wi h
"DH1.@R/H1'.@R/
minimal ex ension o he ˛-manha an no m 787
C
S0
P
PC
P0
x1
H2
1
H1
1H1
1
H2
1
V1
s
V3
s
V2
s
V3
s
V2
s
V1
s
Figu e 3. The igu e shows he slicing o he se
in o
C[S0
by he ho izon al line
R ¹ 1º
and Pin o PC[P0by he modi ied geodesic called x1.
numbe
M
, whe e
MM1CM2CM3
. We de ine he s ips
SiD.RŒ i; iC1/
.
They a e all con ex quad ila e als wi h wo ho izon al sides.
S ep II. De ini ion o he cu e
x1D x'. .R ¹ 1º//
and he polygons
C[S0D
and
PC[P0DP
.The goal o his s ep is o de ine he piecewise linea cu e
x1
,
in e nal o
P
, which will be he image o he segmen
H1
1H2
1
in a map
x'
ex ending
'
.
The p ecise pa ame iza ion o x'will be p esen ed in he nex s ep; he e we only aim
o de ine he cu e x1P.
Ou a gumen is ecu si e and so we deal wi h he i s cu e
x1
de ined on
H1
1H2
1
sepa a ing
in o
S0
and
RŒ 1; h DC
(see Figu e 3). Simila ly, he cu e
x1
means di iding he polygon
P
in o wo u he polygons: a polygon
P0
(which will
be he image o
S0
) con aining he cu e
'.H1
0H2
0/
and ano he polygon
PC
(which
will be he image o C) (see Figu e 3).
Since
P
is a non-degene a e polygon, le
x
ı.P/>0
be he pa ame e o De ini ion 2.2
and le ı1> 0 be so small ha
ı1<min ²x
ı.P/; ". 1 0/
8hH1.@P/;h
23;"
2³
and Lemma 2.4 applies wi h ı1 o Pand ". 1 0/
8h.` Ch/:
We de ine
x1as a ı1-modi ica ion o he geodesic in Pconnec ing H1
1and H2
1:
S ep III. De ini ion o
x'
on
@S0
.In his s ep, we ca e abou he de ini ion o
x'
on
@S0
.
Mo e p ecisely, we le x'D'on @ and we speci y he pa ame iza ion
x'W@C R ¹ 1º! x1
so ha x'is con inuous, injec i e and piecewise linea , and
(3.8) ‰0.x'e@S0/C‰0.x'e@C/‰0.'/ C"
2h. 1 0/:
d. campbell, a. kau anen and e. adici 788
Le us obse e ha hanks o Lemma 2.7 i is enough o look o a con inuous and
injec i e pa ame iza ion
W@C[@S0!R2
coinciding wi h
'
on
@
such ha
(3.8) holds o wi h e o "
4h . 1 0/; namely,
‰0. e@S0/C‰0. e@C/<‰0.'/ C"
4h. 1 0/:
Indeed, he co ec
x'
can be ound as a
ı
-linea iza ion o
o some
ı
small enough
depending on "
4h . 1 0/such ha
‰0.x'e@S0/C‰0.x'e@C/<‰0. e@S0/C‰0. e@C/C"
4h. 1 0/:
Thanks o ou choice o
ı1
and he ac ha
x1
is a
ı1
-modi ica ion wi h a iable
endpoin s o he geodesic connec ing
H1
1
and
H2
1
, hence spli ing
P
in o he wo
polygons P0and PC, we can apply Lemma 2.4 o ge ha
(3.9)
P0H1
;H2
PH1
;H2
C". 1 0/
8h.` Ch/ o any 0< < 1;
PCH1
;H2
PH1
;H2
C". 1 0/
8h.` Ch/ o any 1< < M:
Fo sho , deno e
c1D.H1
1/1
and
c2D.H2
1/1
. Then,
0c1< c2`
. Fo e e y
0 < s < `
, we call
s
he geodesic inside
P
connec ing
V1
s
and
V2
s
. Mo eo e , whene e
c1< s < c2
, we also se
V3
sWD .s; 1/
he poin in he in e sec ion o
H1
1H2
1
wi h
V1
sV2
s
. Fo e e y
s2.0; c1/[.c2; `/
, we ha e ha ei he
V1
s; V 2
s2S0
and using
Lemma 2.4,
P0.V1
s;V2
s/P.V1
s;V2
s/C". 1 0/
8h.` Ch/;
o V1
s; V 2
s2Cand by Lemma 2.4,
PC.V1
s;V2
s/P.V1
s;V2
s/C". 1 0/
8h.` Ch/:
The wo equa ions abo e can be exp essed simul aneously as
(3.10) max ®P0.V1
s;V2
s/; PC.V1
s;V2
s/¯P.V1
s;V2
s/C". 1 0/
8h.` Ch/
o all s2.0; c1/[.c2; `/.
On he o he hand, whene e
s2.c1;c2/
, he poin s
V1
s2P0
and
V2
s2PC
; hus, he
geodesic
s
necessa ily in e sec s
x1
. Le
be he (injec i e and con inuous) cons an -
speed pa ame iza ion o
x1
om
Œ0; H1.x1/
,
.0/ DH1
1
and
.H1.x1// DH2
1
. Fo
e e y s2.c1; c2/, we hen le X.s/ be he poin in s x1such ha
X.s/ Dmax ®x20; H1.x1/W.x/ 2s x1¯:
minimal ex ension o he ˛-manha an no m 789
Then, hanks o Lemma 2.3, i is easy o see ha he map
s!1.X.s//
is non-
dec easing; he e o e, i c1< s < s0< c2, hen
X.s0/21.X.s//; H1.x1/:
No ice ha , in gene al, he unc ion
s!1ıX.s/
is no con inuous, no injec i e
no su jec i e. Howe e , any one-dimensional mono one unc ion can be app oxima ed
uni o mly by s ic ly mono one unc ions. Fu he , i is always possible o sligh ly modi y
hese s ic ly mono one app oxima ions in such a way ha hey become con inuous
and he p ice o his is loosing he con ol on he uni o m dis ance om he o iginal
unc ion on a subse whose measu e can be made as small as desi ed. Then, o e e y
> 0
, i is always possible o ind a con inuous bijec ion
X
o
Œc1; c2
on o
x1
such
ha
(3.11) H1.J/< whe e JWD ®s2.c1; c2/WˇˇX.s/ X.s/ˇˇ> ¯:
We can hen ix Dı1
2and apply Lemma 2.4 o ge ha
(3.12) P0V1
s;Xı1
2
.s/CPCXı1
2
.s/; V2
sP.V1
s;V2
s/C". 1 0/
8h.` Ch/
o all s2.c1; c2/nJı1
2
. On he o he hand, we ha e he i ial es ima e
(3.13) P0V1
s;Xı1
2
.s/CPCXı1
2
.s/; V2
sH1.@P/CH1.x1/<2H1.@P/
o all s2Jı1
2
.
We de ine
W@C[@S0!R2
as
D'
on
@
and
.V 3
s/DXı1
2
.s/
o e e y
s2.c1; c2/
. In pa icula ,
is con inuous and injec i e and ails o be piecewise linea
only on he segmen
H1
1H2
1
. Mo eo e , ga he ing oge he
(3.9)
,
(3.10)
,
(3.12)
and
(3.13), we deduce ha
‰0. e@S0/C‰0. e@C/
DZ 1
0
P0 .H1
/; .H2
/d CZh
1
PC .H1
/; .H2
/d
CZ.0;c1/[.c2;`/
min ®P0.V1
s;V2
s/CPC.V1
s;V2
s/¯
CZ.c1;c2/nJı1
2
P0 .V 1
s/; .V 3
s/CPC .V 3
s/; .V 2
s/
CZJı1
2
P0 .V 2
s/; .V 3
s/CPC .V 3
s/; .V 2
s/
d. campbell, a. kau anen and e. adici 790
Zh
0
P'.H1
/; '.H2
/d CZ.0;`/nJı1
2
P'.V 1
s/; '.V 2
s/ds
C". 1 0/
8h.` Ch/`ChH1Jı1
2CH1Jı1
22H1.@P/
‰0.'/ C"
4h. 1 0/;
whe e in he las inequali y we used (3.11) and he ac ha ı1<". 1 0/
8hH1.@P/.
Finally, hanks o Lemma 2.7 and he conside a ions o he i s pa o he s ep, we
can ind a unc ion
x'W@C[@S0!R2
ha is con inuous, injec i e and piecewise
linea and such ha (3.8) holds.
S ep IV. De ini ion o
z'
on
@T 1
0[@R0[@T 2
0
.In his s ep, we u he subdi ide he
s ip
S0
in he essen ially disjoin union o wo iangles
T1
0
,
T2
0
and a ec angle
R0
wi h he ollowing p ope ies. The ec angle
R0
is he bigges ec angle wi h ho izon al
and e ical sides inside
S0
, such ha he ho izon al sides a e con ained in
@S0
, while
T1
0,T2
0a e he wo disjoin igh -angle iangles con aining I1
0,I2
0, espec i ely.
We con inue o de ine some new
z'W@C[@T 1
0[@R0[@T 2
0!R2
coinciding wi h
x'
on
@C[@S0
such ha
z'
is injec i e, con inuous and piecewise
linea and sa is ies he ollowing es ima e:
(3.14) ‰0.z'e@T 1
0/C‰0.z'e@R0/C‰0.z'e@T 2
0/‰0.x'e@S0/C"
2h. 1 0/:
Le us emphasize ha z'will be de ined so ha z'.@T 1
0[@R0[@T 2
0/P0.
We deno e by
d1
and
d2
he wo alues such ha he p ojec ion o
@S0
on o
R ¹0º
is
exac ly
Œd1;d2 ¹0º
, and we call
x1
and
x2
hose alues o which he p ojec ion o
@R0
on o
R ¹0º
is he segmen
Œx1; x2 ¹0º
. No ice ha
d1c1x1< x2c2d2
.
Since
P0
is a non-degene a e polygon, we le
x
ı.P0/
be he pa ame e o De ini-
ion 2.2 and ake
ı0
1<min ²x
ı.P0/; ". 1 0/
32hH1.@P0/³
and Lemma 2.4 applies wi h ı0
1 o P0and ". 1 0/
16h.`Ch/ .
Le now
x1
be he geodesic inside
P0
connec ing
x'.V 1
x1/
and
x'.V 3
x1/
and le
xx1
be i s
ı0
1
-modi ica ion in he sense o De ini ion 2.2. In pa icula ,
xx1
spli s
P0
in o
wo non-degene a e polygons
P1
0
and
U
, whe e
P1
0
con ains
x'.I 1
0/
and
U
con ains
x'.I 2
0/
, whe e
I1;2
0
a e he wo non-ho izon al segmen s o
@S0 @
. This si ua ion is
depic ed in Figu e 4.
minimal ex ension o he ˛-manha an no m 791
20 d. campbell, a. kau anen and e. adici
𝑇1
0
𝑇2
0
𝑅0
𝐻1
𝑡1
𝐻2
𝑡1
𝐻1
𝑡
𝐻2
𝑡
𝐻3
𝑡
U
P1
0
𝑉1
𝑥1
𝑉3
𝑥1
¯𝜈𝑥1
H1
𝑡
H2
𝑡
Y(𝑡)
Figu e 4. The se s 𝑇1
0, 𝑇2
0and 𝑅0, and P1
0and Uin S ep IV.
while o he emaining 𝑡∈𝐽𝑙
𝛿′
1
2
we ge
(3.19)
𝜌P1
0𝜓1(𝐻1
𝑡), 𝜓1(𝐻3
𝑡)+𝜌U𝜓1(𝐻3
𝑡), 𝜓1(𝐻2
𝑡),≤ H1(¯𝜈𝑥1) + H1(𝜕P0)
≤2H1(𝜕P0).
Then (3.16), (3.17), (3.18) and (3.19) gi e
Ψ0(𝜓1
⌉𝜕𝑇1
0
)+Ψ0(𝜓1
⌉𝜕(𝑆0 𝑇1
0))
≤∫𝑐1
𝑑1
𝜌P0¯𝜑(𝑉1
𝑠),¯𝜑(𝑉2
𝑠)+𝜀(𝑡1−𝑡0)
16ℎ(ℓ+ℎ)𝑑𝑠
+∫𝑐2
𝑐1
𝜌P0¯𝜑(𝑉1
𝑠),¯𝜑(𝑉3
𝑠)+𝜀(𝑡1−𝑡0)
16ℎ(ℓ+ℎ)𝑑𝑠
+∫𝑑2
𝑐2
𝜌P0¯𝜑(𝑉1
𝑠),¯𝜑(𝑉2
𝑠)+𝜀(𝑡1−𝑡0)
16ℎ(ℓ+ℎ)𝑑𝑠
+∫(0,𝑡1) 𝐽1
𝛿′
1
2
𝜌P0¯𝜑(𝐻1
𝑡),¯𝜑(𝐻2
𝑡)+𝜀(𝑡1−𝑡0)
16ℎ(ℓ+ℎ)𝑑𝑡
+2H1𝐽1
𝛿′
1
2H1(𝜕P0)
≤Ψ0(¯𝜑⌉ P0) + 𝜀
8ℎ(𝑡1−𝑡0),
and, as in s ep III, Lemma 2.7 ensu es ha he e is some con inuous, injec i e and
piecewise linea map ˜𝜑1:𝜕𝑇1
0∪𝜕𝑆0→R2such ha ˜𝜑1=𝜓1=¯𝜑on 𝜕𝑆0and
Ψ0(˜𝜑1
⌉𝜕𝑇1
0
) + Ψ0(˜𝜑1
⌉𝜕(𝑆0 𝑇1
0)) ≤ Ψ0(¯𝜑⌉ P0) + 𝜀
4ℎ(𝑡1−𝑡0).
To conclude he s ep we need o epea he e y same a gumen on
˜𝜑1
⌉𝜕(𝑆0 𝑇1
0)
by
eplacing
P0
wi h
U
and conside ing a
𝛿′′
1
-modi ica ion o
𝜈𝑥2
, whe e
𝛿′′
1
is chosen so
Figu e 4. The se s T1
0,T2
0,R0,P1
0and Uin S ep IV.
Thanks o Lemma 2.4, we ha e
(3.15)
Ux'.V 1
s/; x'.V 3
s/P0x'.V 1
s/; x'.V 3
s/C". 1 0/
16h.`Ch/; o all s2.x1; x2/;
Ux'.V 1
s/; x'.V 2
s/P0x'.V 1
s/; x'.V 2
s/C". 1 0/
16h.`Ch/; o all s2.x2; d2/;
while o all s2.d1; x1/,
(3.16) P1
0x'.V 1
s/; x'.V 2
s/P0x'.V 1
s/; x'.V 2
s/C". 1 0/
16h.` Ch/:
We con inue simila ly as desc ibed in S ep III. Fo e e y
2.0; 1/
, we deno e he
poin
x'.H1
/DH1
2P1
0
and
x'.H2
/DH2
2U
; hus, he geodesic
connec ing
H1
and
H2
inside
P0
mus in e sec
xx1
. So also in his case, o e e y
2.0; 1/
, we can
ind a map
Y. /
iden i ying he las poin o he in e sec ion
xx1
unning
xx1
om
x'.V 1
x1/
o
x'.V 3
x1/
. Mo eo e , exac ly as explained in S ep III, we can ind a con inuous
and injec i e app oxima ion Yı0
1
2
such ha
ˇˇY. / Yı0
1
2
. /ˇˇ<ı0
1
2 o all 2.0; 1/nJ1
ı0
1
2
and H1J1
ı0
1
2<ı0
1
2:
So i we now call
H3
WD . ; x1/
, hen we can de ine
1W@T 1
0[@S0!R2
in his
way: 1D x'on @S0and
1.H3
/DYı0
1
2
. / o all 2.0; 1/:
Then, he map
1
is con inuous and injec i e and ails o be piecewise linea only on
he segmen
¹x1º Œ0; 1
. Using Lemma 2.4, o all
2.0; 1/nJ1
ı0
1
2
, we can es ima e
P1
0 1.H1
/; 1.H3
/CU 1.H3
/; 1.H2
/
P0x'.H1
/; x'.H2
/C". 1 0/
16h.` Ch/;
(3.17)
d. campbell, a. kau anen and e. adici 792
while o he emaining 2Jl
ı0
1
2
, we ge
P1
0 1.H1
/; 1.H3
/CU 1.H3
/; 1.H2
/
H1.xx1/CH1.@P0/2H1.@P0/:
(3.18)
Then, (3.15), (3.16), (3.17) and (3.18) gi e
‰0 1
e@T 1
0C‰0 1
[email protected]
0/
Zc1
d1
P0x'.V 1
s/; x'.V 2
s/C". 1 0/
16h.` Ch/ds
CZc2
c1
P0x'.V 1
s/; x'.V 3
s/C". 1 0/
16h.` Ch/ds
CZd2
c2
P0x'.V 1
s/; x'.V 2
s/C". 1 0/
16h.` Ch/ds
CZ.0; 1/nJ1
ı0
1
2
P0x'.H1
/; x'.H2
/C". 1 0/
16h.` Ch/d C2H1J1
ı0
1
2H1.@P0/
‰0.x'eP0/C"
8h. 1 0/;
and, as in S ep III, Lemma 2.7 ensu es ha he e is some con inuous, injec i e and
piecewise linea map z'1W@T 1
0[@S0!R2such ha z'1D 1D x'on @S0and
‰0z'1
e@T 1
0C‰0z'1
[email protected]
0/‰0.x'eP0/C"
4h. 1 0/:
To conclude he s ep, we need o epea he e y same a gumen on
z'1
[email protected]
0/
by
eplacing
P0
wi h
U
and conside ing a
ı00
1
-modi ica ion o
x2
, whe e
ı00
1
is chosen so
ha
ı00
1<min ²x
ı.U/; ". 1 0/
32`H1.@U/³
and Lemma 2.4 applies wi h ı00
1 o Uand ". 1 0/
16h.`Ch/ .
This would p o ide a con inuous, injec i e and piecewise linea map
z'W@T 1
0[
@R0[@T 2
0!R2ex ending z'1(hence, ul ima ely, x') such ha
‰0.z'e@T 1
0/C‰0.z'e@R0/C‰0.z'e@T 2
0/‰0z'1
[email protected]
0/C"
4h. 1 0/
‰0.x'eP0/C"
2h. 1 0/
hus p o ing (3.14) and concluding he s ep.
minimal ex ension o he ˛-manha an no m 793
S ep V. Recu sion and conclusion. In his inal s ep, we wan o conclude ou cons uc ion
by ecu sion. In S eps III and IV, we di ided
in o a new con ex polygon wi h wo
ho izon al sides
C
and a ho izon al s ip
S0
gi en by a ec angle
R0
and wo iangles
T1
0,T2
0.
Then, we de ined con inuous, injec i e and piecewise linea unc ions
x'; z'
such
ha x'D z'on @Csa is ies (by (3.8), (3.14) and ou choice o ı1)
‰0.x'e@S0/C‰0.x'e@C/‰0.'/ C"
2h. 1 0C2ı1/
‰0.'/ C"
2h. 1 0/C"
4
1
2;
‰0.z'e@T 1
0/C‰0.z'e@R0/C‰0.z'e@T 2
0/‰0.x'e@S0/C"
2h. 1 0C2ı1/
‰0.x'e@S0/C"
2h. 1 0/C"
4
1
2:
I e a ing he cons uc ion and choosing o e e y
i
he pa ame e
ıi
sui ably small
depending on i1,h
2iC1and he polygon PCDPnSi1
jD0Pj, we hen ind
M1
X
iD0
‰0.x'e@Si/‰0.'/ C"
2h
M1
X
iD0
. iC1 i/C"
4
M1
X
iD0
1
2i
and
M1
X
iD0
‰0.z'e@T 1
i/C‰0.z'e@Ri/C‰0.z'e@T 2
i/
M1
X
iD0
‰0.x'e@Si/C"
2h
M1
X
iD0
. iC1 i/C"
4
M1
X
iD0
1
2i;
which inally imply (3.6) and (3.7), espec i ely.
4. Piecewise a ine ex ension
In his sec ion, we in es iga e wo possible ini ely piecewise a ine homeomo phic
ex ension inside iangles.
Lemma 4.1 (Ex ension-di ec ).Le
TR2
be a iangle o co ne s
A; B; C
such ha
BC
is ho izon al and
A
is he in e sec ion o
BC
wi h he bisec o o he angle a
A
. I
'W@T !R2
is con inuous, injec i e and linea on each o he segmen s
AB; AC; BA
and
AC
, hen he e exis s a bi-a ine homeomo phism
WT!R2
such ha
D'
on @T and
kD k0.T / H1'.@T /H1.@T /:
d. campbell, a. kau anen and e. adici 794
P oo .
The p oo is immedia e, and indeed i is enough o conside he con inuous
map which is a ine on each o he iangles T1WD ABA,T2WD BAC. Then, we
can compu e
D1 eT1D'.A/'.B/
.A/1.B/1
L2; D1 eT2D'.C / '.A/
.C /1.A/1
L2
and
D2 eT1D'.A/ '.A/D1 eT1.A/1.A/1
.A/2.A/2
L2;
D2 eT2D'.A/ '.A/D1 eT2.A/1.A/1
.A/2.A/2
L2:
In pa icula , one can es ima e
kD k0.T / D jD1 j.T1/C jD2 j.T1/C jD1 j.T2/C jD2 j.T2/
1
2ˇˇ'.A/'.B/ˇˇˇˇ.A/2.A/2ˇˇCˇˇ.A/1.A/1ˇˇ
C1
2ˇˇ'.A/'.A/ˇˇˇˇ.B/1.A/1ˇˇ
C1
2ˇˇ'.C / '.A/ˇˇˇˇ.A/2.A/2ˇˇCˇˇ.A/1.A/1ˇˇ
C1
2ˇˇ'.A/'.A/ˇˇˇˇ.C /1.A/1ˇˇ
H1'.@T /1
24jAAjCjBCj
H1'.@T /1
22jABj C 2jACj C 2jBCj
H1'.@T /H1.@T /:
Lemma 4.2 (Ex ension-indi ec ).Le
TR2
be a iangle o co ne s
A; B; C
such
ha
AB
is ho izon al,
BC
is e ical and le
'W@T !R2
be a con inuous, piecewise
linea , injec i e map such ha
'
is linea on he hypo enuse
AC
. Fo e e y
">0
, he e
exis s a ini ely piecewise a ine homeomo phism
WT!R2
such ha
D'
on
@T
and
kD k0.T / ‰0.'/ C242H1.@T /H1'.AC /C":
P oo . Le ">0be ixed a bi a y small.
Fo simplici y o no a ion, h ough he p oo we e e o
ˇ
as he in e nal angle
o he co ne
AD.0; 0/
and we will deno e
dWD j'.A/ '.C /j D H1.'.AC //
and
dD jACj. Clea ly, since he in e nal angle in Bis =2, hen ˇ2.0; =2/.
minimal ex ension o he ˛-manha an no m 795
X
Y
Z
A B
C
'.A/
'.C /
D
E
F
z
T
G
H1
H2
H3
V3
s
V2
s
V1
s
PADEC P
Pz
T
Q
z'.D/'.E/
P
'.D/
'.F /
ˇ
Figu e 5. The decomposi ion o
T
in o
TD[z
T[ADEC
, and he co esponding decom-
posi ion o in '.@T / in o P[Pz
T[PADEC .
Since he polygon o bounda y
'.@T /
is non-degene a e, hen De ini ion 2.2 p o ides
some cons an x
ı > 0, and hen we conside
(4.1) < ²1; d
2;
x
ı
4;d
14;"
4;1C1
2 an ˇ1
;kD'k1
1³:
The basic idea o he p oo is o ind a sui able one-dimensional skele on
‡
inside
T
,
cons uc a con inuous, piecewise linea and injec i e map
z'W‡!R2
coinciding
wi h
'
on
@T
and inally pe o m a sui able piecewise a ine ex ension inside each
componen o he pa i ion o Tiden i ied by ‡.
Fo cla i y, we p esen he p oo in h ee sepa a e s eps.
S ep I. De ini ion o a i s skele on
„
and a con inuous piecewise linea injec i e map
'1W„!R2
.In his s ep, we would like o cons uc a one-dimensional skele on o
he o m
„D@T [DE [F G
, o some sui ably chosen poin s
D; E; F; G
, and we
will de ine an ex ension
'1
o
'
on
DE [F G
ha is s ill con inuous, piecewise linea
and injec i e. See Figu e 5 o an illus a ion.
We will selec D; E; F; G so ha
D2AB; E; F 2BC; G 2DE; F G kAB and DE kAC;
sa is ying he ollowing es ima es:
(4.2)
jADj< and jCEj< ;
H1'1.@ z
T /< 4;
‰0.'1
e@/‰0.'/ C.dC14/H1.@T /;
whe e
z
TT
is he iangle o co ne s
E; F; G
and
T
is he apezoid o co ne s
D; B; F; G.
d. campbell, a. kau anen and e. adici 796
Ha ing ixed
, by he assump ions on
'
, we can choose
D2AB and E2BC so ha
(i) jADj< and jCEj< ;
(ii) j'.A/ '.D/j< and j'.C / '.E/j< ;
(iii) he es ic ion o 'is linea on AD and CE;
(i ) DE is pa allel o AC ;
( )
he poin
X
is on he in e nal bisec o o
'.A/
and
Y
on he in e nal bisec o
o
'.C /
such ha
j'.A/ Xj< 2
and
j'.C / Yj< 2
and he piecewise
linea pa h
z'.D/'.E/ WD '.D/XY'.E/
lies in he in e io o
'.@T /
and is a
x
ı=2-modi ica ion o he geodesic '.D/'.E/ in he sense o De ini ion 2.2.
Obse e ha (ii) and ( ) imply ha j'.D/ Xj;j'.E/ Yj< 3 and also
H1.z'.D/'.E// < j'.D/ XjCjXYjCj'.E/ Yj
< 6 C j'.A/ XjCj'.A/ '.C /jCj'.C / Yj
<dC10:
(4.3)
We ind a poin
F
on he segmen
EB
, a poin
G2DE
and a poin
Z2ŒY'.E/
such ha
( i) jFEj< and j'.F / '.E/j< ;
( ii) 'is linea on EF ;
( iii) j'.E/ Zj< and Œ'.F /Zlies in in '.@T /, and as a consequence,
j'.F / Zj< 2;
(ix) G2DF2,jGEj< .sin ˇ/1and jGFj< . an ˇ/1.
This concludes he de ini ion o
„
. Indeed, (i) ensu es he i s equa ion o
(4.2)
,
hen z
Tis a igh -angle iangle and is a apezoid inside T.
We now p oceed o cons uc a unc ion
'1W„!R2
ex ending
'
such ha he
second and hi d es ima es o
(4.2)
a e sa is ied. In o de o do ha , we conside wo
u he auxilia y poin s P; Q 2DG so ha
(x) jPDj< and jQGj< ;
and we se
'1.P / WD X; '1.Q/ WD Yand '1.G/ WD Z:
We hen de ine
'1W„!R2
so ha
'1D'
on
@T
,
'1
eDP
is he pa ame iza ion a
cons an speed o he segmen
'.D/X
,
'1
ePQ
is he pa ame iza ion a cons an speed
o he segmen
XY
,
'1
eQG
is he pa ame iza ion a cons an speed o he segmen
YZ
,
'1
eGE
is he pa ame iza ion a cons an speed o he segmen
Z'.E/
and, inally,
'1
eGF
is he pa ame iza ion a cons an speed o he segmen
Z'.F /
. A ske ch o he
si ua ion is p esen ed in Figu e 5.
minimal ex ension o he ˛-manha an no m 803
We obse e ha
is con inuous and injec i e, hence a homeomo phism, since
D z'
on
‡
. F om he same obse a ion, we also deduce ha
D'
on
@T
because
z'D'
he e. Ga he ing oge he (4.10), (4.11), (4.12) and (4.7), we ind
kD k0.T / D kD!k0.P/C kDwk0.z
T / C
M1
X
iD0
kD ik0.Ri/C kD k0.y
T /
z
C ddC21C1
2 an ˇC
M1
X
iD0‰0.z'e@Ri/C"
2iCd"
z
C ddC21C1
2 an ˇC‰0.'/ CCdH1.@T / Cz
C "
‰0.'/ Cz
CH1.@T /H1'.AC /Cz
C ";
whe e in he las inequali y, we used ha
dD jACj H1.@T / and dDˇˇ'A'.C /ˇˇ:
5. P oo o Theo ems 1.1 and 1.2
P oo o Theo em 1.1.Le ">0be a bi a y ixed and le
(5.1) 0 < < min ²1; "
z
CCz
CH1.Q/³
o some la ge app op ia e bu ixed geome ic cons an
z
C
. We desc ibe in de ail he
p oo when
Q
is o class (iii) in he sense o Rema k 3.1, while he o he cases a e an
ob ious modi ica ion o he cu en a gumen .
Applying Lemma 3.2 o
Q; '; ˛
and he pa ame e
, we can pa i ion
Q
in wo
iangles
T1; T2
and a con ex polygon
and ind a con inuous, piecewise linea ,
injec i e map
x'W@T1[@T2[@ !R2
wi h he p ope ies lis ed in he s a emen o
Lemma 3.2. In pa icula , om (3.3), i ollows ha
(5.2) ‰˛.x'e@/‰˛.'/ C;
whe e (3.1) and (3.2) ensu e ha
(5.3) H1.@T1/CH1.@T2/ < ; H1x'.@T1/CH1x'.@T2/< :
Fu he mo e, since
is a con ex polygon wi h wo pa allel sides in di ec ion
˛
, we
can apply he
˛
- o a ed e sion o Lemma 3.3 o
,
x'1
and he pa ame e
o ind
M
inc easing alues
. i/M1
iD0
and
˛
- o a ed s ips
Si
, which can be seen as he union o a
d. campbell, a. kau anen and e. adici 804
ec angle
Ri
and wo iangles
T1
i
and
T2
i
, and a con inuous piecewise linea injec i e
map
O'WSM1
iD0@T 1
i[@Ri[@T 2
i!R2
coinciding wi h
x'
on
@
wi h he p ope ies
o Lemma 3.3. In pa icula , hanks o (3.7), we deduce
(5.4)
M1
X
iD0‰˛O'e@T 1
iC‰˛O'e@RiC‰˛O'e@T 2
i‰˛x'1
e@C;
while (3.5) ensu es ha
(5.5) H1O'.I 1
i/CH1O'.I 2
i/<
whe e I1
iD@T 1
i @ and I2
iD@T 2
i @.
We a e inally in posi ion o de ine a unc ion
z'W@T1[ M1
[
iD0
@T 1
i[@Ri[@T 2
i![@T2!R2
ha is con inuous, injec i e, ini ely piecewise linea and such ha
z'D x'1
on
@T1[@T2
and z'D O'on SM1
iD0@T 1
i[@Ri[@T 2
i.
Once he e we will pe o m he homeomo phic piecewise a ine ex ension on
T1; T2; T 1
i; T 2
i
and
Ri
independen ly o e e y
iD0; : : : ; M 1
. We i s ocus
on he ex ension inside he s ips
Si
. Le
i2 ¹0;:::;M 1º
be ixed; we hen apply
he
˛
- o a ed e sion o P oposi ion 2.8 o
Ri;z'e@Ri
and pa ame e
. iC1 i/
o ind
ini ely piecewise a ine homeomo phisms
iWRi!R2
coinciding wi h
z'
on
@Ri
such ha
(5.6) kD ik˛.Ri/‰˛.z'e@Ri/C. iC1 i/:
By cons uc ion, we ha e ha
T1
i; T 2
i
a e igh -angle iangles whose hypo enuse is
con ained in
@ @Q
, and we can apply he
˛
- o a ed e sion o Lemma 4.2 o
T1;2
i
,
z'e@T 1;2
i
and pa ame e . iC1 i/ o ind ini ely piecewise a ine homeomo phisms
w1;2
iWT1;2
i!R2
coinciding wi h z'on @T 1;2
isuch ha
kDw1;2
ik˛.T 1;2
i/‰˛z'e@T 1;2
iCz
CH1.@T 1;2
i/H1z'.I1;2
i/Cz
C . iC1 i/;
which hanks o (5.5) implies
kDw1
ik˛.T 1
i/C kDw2
ik˛.T 2
i/
‰˛z'e@T 1
iC‰˛z'e@T 2
iCz
CH1.@T 1
i/CH1.@T 2
i/Cz
C . iC1 i/:
(5.7)
minimal ex ension o he ˛-manha an no m 805
Le us also no ice ha , o u u e need, since he iangles
T1;2
i
ha e one angle
equal o
=2
and, by cons uc ion, hei hypo enuse is
I1;2
i
, hen one has
H1.@T 1;2
i/
3H1.I1;2
i/
o e e y
i
. Mo eo e , ha ing
I1;2
i.@ @Q/
and he iangles pai wise
essen ially disjoin , we deduce
(5.8)
M1
X
iD0H1.@T 1
i/CH1.@T 2
i/3
M1
X
iD0H1.I1
i/CH1.I2
i/3H1.@Q/:
Le us now conside he ex ension inside he iangles
T1; T2
. In his case, by
cons uc ion, we a e in posi ion o apply he
˛
- o a ed e sion o Lemma 4.1 o
T1;2
and z'e@T1;2 o ind bi-a ine homeomo phisms w1;2 WT1;2 !R2such ha
kDw1;2k˛.T1;2/H1z'.@T1;2/H1@T1;2
which, hanks o (5.3), gi es
(5.9) kDw1k˛.T1/C kDw2k˛.T2/ < 22:
We can inally de ine WQ!R2 o be he piecewise a ine unc ion such ha
Dw1;2 on T1;2; Dw1;2
ion T1;2
iand D ion Ri o e e y iD0;: : :;M 1:
By cons uc ion,
is con inuous because
D z'
on he one-dimensional skele on
@T1[.SM1
iD0@T 1
i[@Ri[@T 2
i/[@T2
, and, mo eo e ,
coincides wi h
z'D'
on
@Q
. To conclude, i is only le o e i y he alidi y o
(1.4)
, bu his is now a
s aigh o wa d consequence o
(5.6)
,
(5.7)
,
(5.9)
,
(5.4)
,
(5.8)
and
(5.2)
. Indeed, we ge
kD k˛.Q/D kDw1k˛.T1/C kDw2k˛.T2/
C
M1
X
iD0kDw1
ik˛.T 1
i/C kD ik˛.Ri/C kDw2
ik˛.T 2
i/
M1
X
iD0‰˛O'e@T 1
iC‰˛O'e@RiC‰˛O'e@T 2
i
C42Cz
C
M1
X
iD0H1.@T 1
i/CH1.@T 2
i/C. iC1 i/
‰˛.x'1
e@/CC42Cz
C H1.@Q/Cdiam Q
‰˛.x'e@/C6 Cz
C H1.@Q/;
‰˛.'/ Cz
C 1CH1.@Q/;
and hen es ima e (1.4) ollows since has been chosen as in (5.1).
d. campbell, a. kau anen and e. adici 806
5.1. P oo o Theo em 1.2
To p o e he claim, i su ices o epea he p oo o he abo e lemmas as be o e bu using
he es ima es om Theo em 2.9 ins ead o om P oposi ion 2.8. In all o ou calcula ions,
we es ima e
RP.'1.H2
/;'1.H 3
//d
and
RB1
D1P.'1.V 1
s/;'1.V 2
s//ds
sepa a ely.
Now i su ices o keep hem sepa a e ins ead o summing hem.
The key es ima es in Lemma 3.2 a e
(3.4)
, in Lemma 3.3 hey a e
(3.12)
,
(3.13)
and
he calcula ion ollowing, and in Lemma 4.2 hey a e (4.4), (4.5), (4.6).
We can hen epea he p oo o Theo em 1.1 wi h he di e ence ha in
(5.4)
,
(5.6)
and so on we use he sepa a e es ima es, a he han he summed es ima es exp essed
using ‰˛.
Funding. – The i s au ho was suppo ed by g an GACR 20-19018Y.
Re e ences
[1]
D. Campbell,Di eomo phic app oxima ion o plana Sobole homeomo phisms in O licz–
Sobole spaces.J. Func . Anal. 273 (2017), no. 1, 125–205. Zbl 1377.46022
MR 3646299
[2]
D. Campbell
–
S. Hencl
–
A. Kau anen
–
E. Radici,S ic limi s o plana BV homeo-
mo phisms.Nonlinea Anal. 177 (2018), 209–237. Zbl 1404.30026 MR 3865195
[3]
D. Campbell
–
A. Kau anen
–
E. Radici, Classi ica ion o a ea-s ic limi s o plana
BV homeomo phisms. 2022, a Xi :2212.08394.
[4]
D. Campbell
–
A. Kau anen
–
E. Radici,Classi ica ion o s ic limi s o plana
BV
homeomo phisms.J. Func . Anal. 285 (2023), no. 3, a icle no. 109953. Zbl 07683487
MR 4579917
[5]
S. Dane i
–
A. P a elli,Smoo h app oxima ion o bi-Lipschi z o ien a ion-p ese ing
homeomo phisms.Ann. Ins . H. Poinca é C Anal. Non Linéai e 31 (2014), no. 3, 567–589.
Zbl 1348.37071 MR 3208455
[6]
S. Dane i
–
A. P a elli,A plana bi-Lipschi z ex ension heo em.Ad . Calc. Va . 8
(2015), no. 3, 221–266. Zbl 1331.26020 MR 3365742
[7]
G. De Philippis
–
A. P a elli,The closu e o plana di eomo phisms in Sobole spaces.
Ann. Ins . H. Poinca é C Anal. Non Linéai e 37 (2020), no. 1, 181–224. Zbl 1446.46018
MR 4049920
[8]
S. Hencl
–
A. P a elli,Di eomo phic app oxima ion o
W1;1
plana Sobole homeo-
mo phisms.J. Eu . Ma h. Soc. (JEMS) 20 (2018), no. 3, 597–656. Zbl 1393.26013
MR 3776275
[9]
T. Iwaniec
–
L. V. Ko ale
–
J. Onninen,Di eomo phic app oxima ion o Sobole
homeomo phisms.A ch. Ra ion. Mech. Anal. 201 (2011), no. 3, 1047–1067.
Zbl 1260.46023 MR 2824471
minimal ex ension o he ˛-manha an no m 807
[10]
T. Iwaniec
–
L. V. Ko ale
–
J. Onninen,Hop di e en ials and smoo hing Sobole
homeomo phisms.In . Ma h. Res. No . IMRN 2012 (2012), no. 14, 3256–3277.
Zbl 1248.49052 MR 2946225
[11]
T. Iwaniec
–
J. Onninen,Limi s o Sobole homeomo phisms.J. Eu . Ma h. Soc. (JEMS)
19 (2017), no. 2, 473–505. Zbl 1364.30041 MR 3605022
[12]
A. P a elli,On he bi-Sobole plana homeomo phisms and hei app oxima ion.Nonlin-
ea Anal. 154 (2017), 258–268. Zbl 1372.46029 MR 3614654
[13]
A. P a elli
–
E. Radici,On he plana minimal BV ex ension p oblem.A i Accad. Naz.
Lincei Rend. Lincei Ma . Appl. 29 (2018), no. 3, 511–555. Zbl 1407.46030
MR 3819102
[14]
A. P a elli
–
E. Radici,App oxima ion o plana BV homeomo phisms by di eomo -
phisms.J. Func . Anal. 276 (2019), no. 3, 659–686. Zbl 1426.41019 MR 3906288
[15]
E. Radici,A plana Sobole ex ension heo em o piecewise linea homeomo phisms.
Paci ic J. Ma h. 283 (2016), no. 2, 405–418. Zbl 1359.46036 MR 3519110
Recei ed 20 Janua y 2023,
and in e ised o m 30 Augus 2023
Daniel Campbell
Depa men o Ma hema ical Analysis, Cha les Uni e si y
Sokolo ská 83, 186 00 P ague 8, Czech Republic
[email p o ec ed]
Aapo Kau anen
Depa men o Ma hema ics and S a is ics, Uni e si y o Jy äskylä
PL 35, 40014 Jy äsklyän yliopis o, Finland
[email p o ec ed]
Emanuela Radici
Depa men o In o ma ion Enginee ing, Compu e Science and Ma hema ics (DISIM),
Uni e si y o L’Aquila
Via Ve oio 1 (Coppi o), 67100 L’Aquila, I aly
emanuela. adici@uni aq.i